Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Handle stale file handles on packed-refs file
On a local filesystem the packed-refs file will be orphaned if it is
replaced by another client while the current client is reading the old
one. However, since NFS servers do not keep track of open files, instead
of orphaning the old packed-refs file, such a replacement will cause the
old file to be garbage collected instead. A stale file handle exception
will be raised on NFS servers if the file is garbage collected (deleted)
on the server while it is being read. Since we no longer have access to
the old file in these cases, the previous code would just fail. However,
in these cases, reopening the file and rereading it will succeed (since
it will reopen the new replacement file). So retrying the read is a
viable strategy to deal with stale file handles on the packed-refs file,
implement such a strategy.
Since it is possible that the packed-refs file could be replaced again
while rereading it (multiple consecutive updates can easily occur with
ref deletions), loop on stale file handle exceptions, up to 5 extra
times, trying to read the packed-refs file again, until we either read
the new file, or find that the file no longer exists. The limit of 5 is
arbitrary, and provides a safe upper bounds to prevent infinite loops
consuming resources in a potential unforeseen persistent error
condition.
Change-Id: I085c472bafa6e2f32f610a33ddc8368bb4ab1814
Signed-off-by: Martin Fick<mfick@codeaurora.org>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com> pirms 8 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Handle stale file handles on packed-refs file
On a local filesystem the packed-refs file will be orphaned if it is
replaced by another client while the current client is reading the old
one. However, since NFS servers do not keep track of open files, instead
of orphaning the old packed-refs file, such a replacement will cause the
old file to be garbage collected instead. A stale file handle exception
will be raised on NFS servers if the file is garbage collected (deleted)
on the server while it is being read. Since we no longer have access to
the old file in these cases, the previous code would just fail. However,
in these cases, reopening the file and rereading it will succeed (since
it will reopen the new replacement file). So retrying the read is a
viable strategy to deal with stale file handles on the packed-refs file,
implement such a strategy.
Since it is possible that the packed-refs file could be replaced again
while rereading it (multiple consecutive updates can easily occur with
ref deletions), loop on stale file handle exceptions, up to 5 extra
times, trying to read the packed-refs file again, until we either read
the new file, or find that the file no longer exists. The limit of 5 is
arbitrary, and provides a safe upper bounds to prevent infinite loops
consuming resources in a potential unforeseen persistent error
condition.
Change-Id: I085c472bafa6e2f32f610a33ddc8368bb4ab1814
Signed-off-by: Martin Fick<mfick@codeaurora.org>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com> pirms 8 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Do not fake a SymbolicRef as an ObjectIdRef
When doing a detaching operation, JGit fakes a SymbolicRef as an
ObjectIdRef. This is because RefUpdate#updateImpl dereferences the
SymbolicRef when updating it. For example, assume that HEAD is
pointing to refs/heads/master. If I try to make a detached HEAD
pointing to a commit c0ffee, RefUpdate dereferences HEAD as
refs/heads/master first and changes refs/heads/master to c0ffee. The
detach argument of RefDatabase#newUpdate avoids this dereference by
faking HEAD as ObjectIdRef.
This faking is problematic for the linking operation of
DfsRefDatabase. It does a compare-and-swap operation on every
reference change because of its distributed systems nature. If a
SymbolicRef is faked as an ObjectRef, it thinks that there is a
racing change in the reference and rejects the update. Because of
this, DFS based repositories cannot change the link target of symbolic
refs. This has not been a problem for file-based repositories because
they have a file-lock based semantics instead of the CAS based one.
The reference implementation, InMemoryRepository, is not affected
because it only compares ObjectIds.
When [1] introduced this faking code, there was no way for RefUpdate
to distinguish the detaching operation. When [2] fixed the detaching
operation, it introduced a detachingSymbolicRef flag. This commit uses
this flag to control whether it needs to dereference the symbolic refs
by calling Ref#getLeaf. The same flag is used in the reflog update
operation.
This commit does not affect any operation that succeeds currently. In
some DFS repository implementations, this fixes a ref linking
operation, which is currently failing.
[1]: https://eclipse.googlesource.com/jgit/jgit/+/01b5392cdbc12ce2e21fd1d1afbd61fdf97e1c38
[2]: https://eclipse.googlesource.com/jgit/jgit/+/3a86868c0883d2a564db88bf9ae4a5fe235bb63f
Change-Id: I118f85f0414dbfad02250944e28d74dddd59469b
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
pirms 7 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Handle stale file handles on packed-refs file
On a local filesystem the packed-refs file will be orphaned if it is
replaced by another client while the current client is reading the old
one. However, since NFS servers do not keep track of open files, instead
of orphaning the old packed-refs file, such a replacement will cause the
old file to be garbage collected instead. A stale file handle exception
will be raised on NFS servers if the file is garbage collected (deleted)
on the server while it is being read. Since we no longer have access to
the old file in these cases, the previous code would just fail. However,
in these cases, reopening the file and rereading it will succeed (since
it will reopen the new replacement file). So retrying the read is a
viable strategy to deal with stale file handles on the packed-refs file,
implement such a strategy.
Since it is possible that the packed-refs file could be replaced again
while rereading it (multiple consecutive updates can easily occur with
ref deletions), loop on stale file handle exceptions, up to 5 extra
times, trying to read the packed-refs file again, until we either read
the new file, or find that the file no longer exists. The limit of 5 is
arbitrary, and provides a safe upper bounds to prevent infinite loops
consuming resources in a potential unforeseen persistent error
condition.
Change-Id: I085c472bafa6e2f32f610a33ddc8368bb4ab1814
Signed-off-by: Martin Fick<mfick@codeaurora.org>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com> pirms 8 gadiem Handle stale file handles on packed-refs file
On a local filesystem the packed-refs file will be orphaned if it is
replaced by another client while the current client is reading the old
one. However, since NFS servers do not keep track of open files, instead
of orphaning the old packed-refs file, such a replacement will cause the
old file to be garbage collected instead. A stale file handle exception
will be raised on NFS servers if the file is garbage collected (deleted)
on the server while it is being read. Since we no longer have access to
the old file in these cases, the previous code would just fail. However,
in these cases, reopening the file and rereading it will succeed (since
it will reopen the new replacement file). So retrying the read is a
viable strategy to deal with stale file handles on the packed-refs file,
implement such a strategy.
Since it is possible that the packed-refs file could be replaced again
while rereading it (multiple consecutive updates can easily occur with
ref deletions), loop on stale file handle exceptions, up to 5 extra
times, trying to read the packed-refs file again, until we either read
the new file, or find that the file no longer exists. The limit of 5 is
arbitrary, and provides a safe upper bounds to prevent infinite loops
consuming resources in a potential unforeseen persistent error
condition.
Change-Id: I085c472bafa6e2f32f610a33ddc8368bb4ab1814
Signed-off-by: Martin Fick<mfick@codeaurora.org>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com> pirms 8 gadiem Handle stale file handles on packed-refs file
On a local filesystem the packed-refs file will be orphaned if it is
replaced by another client while the current client is reading the old
one. However, since NFS servers do not keep track of open files, instead
of orphaning the old packed-refs file, such a replacement will cause the
old file to be garbage collected instead. A stale file handle exception
will be raised on NFS servers if the file is garbage collected (deleted)
on the server while it is being read. Since we no longer have access to
the old file in these cases, the previous code would just fail. However,
in these cases, reopening the file and rereading it will succeed (since
it will reopen the new replacement file). So retrying the read is a
viable strategy to deal with stale file handles on the packed-refs file,
implement such a strategy.
Since it is possible that the packed-refs file could be replaced again
while rereading it (multiple consecutive updates can easily occur with
ref deletions), loop on stale file handle exceptions, up to 5 extra
times, trying to read the packed-refs file again, until we either read
the new file, or find that the file no longer exists. The limit of 5 is
arbitrary, and provides a safe upper bounds to prevent infinite loops
consuming resources in a potential unforeseen persistent error
condition.
Change-Id: I085c472bafa6e2f32f610a33ddc8368bb4ab1814
Signed-off-by: Martin Fick<mfick@codeaurora.org>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com> pirms 8 gadiem Handle stale file handles on packed-refs file
On a local filesystem the packed-refs file will be orphaned if it is
replaced by another client while the current client is reading the old
one. However, since NFS servers do not keep track of open files, instead
of orphaning the old packed-refs file, such a replacement will cause the
old file to be garbage collected instead. A stale file handle exception
will be raised on NFS servers if the file is garbage collected (deleted)
on the server while it is being read. Since we no longer have access to
the old file in these cases, the previous code would just fail. However,
in these cases, reopening the file and rereading it will succeed (since
it will reopen the new replacement file). So retrying the read is a
viable strategy to deal with stale file handles on the packed-refs file,
implement such a strategy.
Since it is possible that the packed-refs file could be replaced again
while rereading it (multiple consecutive updates can easily occur with
ref deletions), loop on stale file handle exceptions, up to 5 extra
times, trying to read the packed-refs file again, until we either read
the new file, or find that the file no longer exists. The limit of 5 is
arbitrary, and provides a safe upper bounds to prevent infinite loops
consuming resources in a potential unforeseen persistent error
condition.
Change-Id: I085c472bafa6e2f32f610a33ddc8368bb4ab1814
Signed-off-by: Martin Fick<mfick@codeaurora.org>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com> pirms 8 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem RefDirectory: Throw exception if CAS of packed ref list fails
The contents of the packedRefList AtomicReference should never differ
from what we expect prior to writing, because this segment of the code
is protected by the packed-refs lock file on disk. If it does happen,
whether due to programmer error or a rogue process not respecting the
locking protocol, it's better to let the caller know than to silently
drop the whole commit operation on the floor.
The existing concurrentOnlyOneWritesPackedRefs test is inherently
nondeterministic as written, and was already about 6% flaky as measured
by bazel:
$ bazel test --runs_per_test=200 //org.eclipse.jgit.test:org_eclipse_jgit_internal_storage_file_GcPackRefsTest
...
INFO: Elapsed time: 42.608s, Critical Path: 10.35s
//org.eclipse.jgit.test:org_eclipse_jgit_internal_storage_file_GcPackRefsTest FAILED in 12 out of 200 in 1.6s
Stats over 200 runs: max = 1.6s, min = 1.1s, avg = 1.3s, dev = 0.1s
This flakiness was caused by the assumption that exactly one of the 2
threads would fail, when both might actually succeed in practice due to
racing on the compare-and-swap.
For whatever reason, this change affected the interleaving behavior in
such a way that the flakiness jumped to around 50%. Making the
interleaving of the test fully deterministic is beyond the scope of this
change, but a simple tweak to the assertion is enough to make it pass
consistently 200+ times both before and after this change.
Change-Id: I5ff4dc39ee05bda88d47909acb70118f3d0c8f74
pirms 6 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem Rewrite reference handling to be abstract and accurate
This commit actually does three major changes to the way references
are handled within JGit. Unfortunately they were easier to do as
a single massive commit than to break them up into smaller units.
Disambiguate symbolic references:
---------------------------------
Reporting a symbolic reference such as HEAD as though it were
any other normal reference like refs/heads/master causes subtle
programming errors. We have been bitten by this error on several
occasions, as have some downstream applications written by myself.
Instead of reporting HEAD as a reference whose name differs from
its "original name", report it as an actual SymbolicRef object
that the application can test the type and examine the target of.
With this change, Ref is now an abstract type with different
subclasses for the different types.
In the classical example of "HEAD" being a symbolic reference to
branch "refs/heads/master", the Repository.getAllRefs() method
will now return:
Map<String, Ref> all = repository.getAllRefs();
SymbolicRef HEAD = (SymbolicRef) all.get("HEAD");
ObjectIdRef master = (ObjectIdRef) all.get("refs/heads/master");
assertSame(master, HEAD.getTarget());
assertSame(master.getObjectId(), HEAD.getObjectId());
assertEquals("HEAD", HEAD.getName());
assertEquals("refs/heads/master", master.getName());
A nice side-effect of this change is the storage type of the
symbolic reference is no longer ambiguous with the storge type
of the underlying reference it targets. In the above example,
if master was only available in the packed-refs file, then the
following is also true:
assertSame(Ref.Storage.LOOSE, HEAD.getStorage());
assertSame(Ref.Storage.PACKED, master.getStorage());
(Prior to this change we returned the ambiguous storage of
LOOSE_PACKED for HEAD, which was confusing since it wasn't
actually true on disk).
Another nice side-effect of this change is all intermediate
symbolic references are preserved, and are therefore visible
to the application when they walk the target chain. We can
now correctly inspect chains of symbolic references.
As a result of this change the Ref.getOrigName() method has been
removed from the API. Applications should identify a symbolic
reference by testing for isSymbolic() and not by using an arcane
string comparsion between properties.
Abstract the RefDatabase storage:
---------------------------------
RefDatabase is now abstract, similar to ObjectDatabase, and a
new concrete implementation called RefDirectory is used for the
traditional on-disk storage layout. In the future we plan to
support additional implementations, such as a pure in-memory
RefDatabase for unit testing purposes.
Optimize RefDirectory:
----------------------
The implementation of the in-memory reference cache, reading, and
update routines has been completely rewritten. Much of the code
was heavily borrowed or cribbed from the prior implementation,
so copyright notices have been left intact as much as possible.
The RefDirectory cache no longer confuses symbolic references
with normal references. This permits the cache to resolve the
value of a symbolic reference as late as possible, ensuring it
is always current, without needing to maintain reverse pointers.
The cache is now 2 sorted RefLists, rather than 3 HashMaps.
Using sorted lists allows the implementation to reduce the
in-memory footprint when storing many refs. Using specialized
types for the elements allows the code to avoid additional map
lookups for auxiliary stat information.
To improve scan time during getRefs(), the lists are returned via
a copy-on-write contract. Most callers of getRefs() do not modify
the returned collections, so the copy-on-write semantics improves
access on repositories with a large number of packed references.
Iterator traversals of the returned Map<String,Ref> are performed
using a simple merge-join of the two cache lists, ensuring we can
perform the entire traversal in linear time as a function of the
number of references: O(PackedRefs + LooseRefs).
Scans of the loose reference space to update the cache run in
O(LooseRefs log LooseRefs) time, as the directory contents
are sorted before being merged against the in-memory cache.
Since the majority of stable references are kept packed, there
typically are only a handful of reference names to be sorted,
so the sorting cost should not be very high.
Locking is reduced during getRefs() by taking advantage of the
copy-on-write semantics of the improved cache data structure.
This permits concurrent readers to pull back references without
blocking each other. If there is contention updating the cache
during a scan, one or more updates are simply skipped and will
get picked up again in a future scan.
Writing to the $GIT_DIR/packed-refs during reference delete is
now fully atomic. The file is locked, reparsed fresh, and written
back out if a change is necessary. This avoids all race conditions
with concurrent external updates of the packed-refs file.
The RefLogWriter class has been fully folded into RefDirectory
and is therefore deleted. Maintaining the reference's log is
the responsiblity of the database implementation, and not all
implementations will use java.io for access.
Future work still remains to be done to abstract the ReflogReader
class away from local disk IO.
Change-Id: I26b9287c45a4b2d2be35ba2849daa316f5eec85d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
pirms 14 gadiem |
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448 |
- /*
- * Copyright (C) 2007, Dave Watson <dwatson@mimvista.com>
- * Copyright (C) 2009-2010, Google Inc.
- * Copyright (C) 2007, Robin Rosenberg <robin.rosenberg@dewire.com>
- * Copyright (C) 2006, Shawn O. Pearce <spearce@spearce.org> and others
- *
- * This program and the accompanying materials are made available under the
- * terms of the Eclipse Distribution License v. 1.0 which is available at
- * https://www.eclipse.org/org/documents/edl-v10.php.
- *
- * SPDX-License-Identifier: BSD-3-Clause
- */
-
- package org.eclipse.jgit.internal.storage.file;
-
- import static java.nio.charset.StandardCharsets.UTF_8;
- import static org.eclipse.jgit.lib.Constants.HEAD;
- import static org.eclipse.jgit.lib.Constants.LOGS;
- import static org.eclipse.jgit.lib.Constants.OBJECT_ID_STRING_LENGTH;
- import static org.eclipse.jgit.lib.Constants.PACKED_REFS;
- import static org.eclipse.jgit.lib.Constants.R_HEADS;
- import static org.eclipse.jgit.lib.Constants.R_REFS;
- import static org.eclipse.jgit.lib.Constants.R_TAGS;
- import static org.eclipse.jgit.lib.Ref.Storage.LOOSE;
- import static org.eclipse.jgit.lib.Ref.Storage.NEW;
- import static org.eclipse.jgit.lib.Ref.Storage.PACKED;
-
- import java.io.BufferedReader;
- import java.io.File;
- import java.io.FileInputStream;
- import java.io.FileNotFoundException;
- import java.io.IOException;
- import java.io.InputStreamReader;
- import java.io.InterruptedIOException;
- import java.nio.file.DirectoryNotEmptyException;
- import java.nio.file.Files;
- import java.nio.file.Path;
- import java.security.DigestInputStream;
- import java.security.MessageDigest;
- import java.text.MessageFormat;
- import java.util.Arrays;
- import java.util.Collection;
- import java.util.Collections;
- import java.util.HashMap;
- import java.util.LinkedList;
- import java.util.List;
- import java.util.Map;
- import java.util.concurrent.atomic.AtomicInteger;
- import java.util.concurrent.atomic.AtomicReference;
- import java.util.concurrent.locks.ReentrantLock;
- import java.util.stream.Stream;
-
- import org.eclipse.jgit.annotations.NonNull;
- import org.eclipse.jgit.annotations.Nullable;
- import org.eclipse.jgit.errors.InvalidObjectIdException;
- import org.eclipse.jgit.errors.LockFailedException;
- import org.eclipse.jgit.errors.MissingObjectException;
- import org.eclipse.jgit.errors.ObjectWritingException;
- import org.eclipse.jgit.events.RefsChangedEvent;
- import org.eclipse.jgit.internal.JGitText;
- import org.eclipse.jgit.lib.ConfigConstants;
- import org.eclipse.jgit.lib.Constants;
- import org.eclipse.jgit.lib.ObjectId;
- import org.eclipse.jgit.lib.ObjectIdRef;
- import org.eclipse.jgit.lib.Ref;
- import org.eclipse.jgit.lib.RefComparator;
- import org.eclipse.jgit.lib.RefDatabase;
- import org.eclipse.jgit.lib.RefUpdate;
- import org.eclipse.jgit.lib.RefWriter;
- import org.eclipse.jgit.lib.Repository;
- import org.eclipse.jgit.lib.SymbolicRef;
- import org.eclipse.jgit.revwalk.RevObject;
- import org.eclipse.jgit.revwalk.RevTag;
- import org.eclipse.jgit.revwalk.RevWalk;
- import org.eclipse.jgit.util.FS;
- import org.eclipse.jgit.util.FileUtils;
- import org.eclipse.jgit.util.IO;
- import org.eclipse.jgit.util.RawParseUtils;
- import org.eclipse.jgit.util.RefList;
- import org.eclipse.jgit.util.RefMap;
- import org.slf4j.Logger;
- import org.slf4j.LoggerFactory;
-
- /**
- * Traditional file system based {@link org.eclipse.jgit.lib.RefDatabase}.
- * <p>
- * This is the classical reference database representation for a Git repository.
- * References are stored in two formats: loose, and packed.
- * <p>
- * Loose references are stored as individual files within the {@code refs/}
- * directory. The file name matches the reference name and the file contents is
- * the current {@link org.eclipse.jgit.lib.ObjectId} in string form.
- * <p>
- * Packed references are stored in a single text file named {@code packed-refs}.
- * In the packed format, each reference is stored on its own line. This file
- * reduces the number of files needed for large reference spaces, reducing the
- * overall size of a Git repository on disk.
- */
- public class RefDirectory extends RefDatabase {
- private static final Logger LOG = LoggerFactory
- .getLogger(RefDirectory.class);
-
- /** Magic string denoting the start of a symbolic reference file. */
- public static final String SYMREF = "ref: "; //$NON-NLS-1$
-
- /** Magic string denoting the header of a packed-refs file. */
- public static final String PACKED_REFS_HEADER = "# pack-refs with:"; //$NON-NLS-1$
-
- /** If in the header, denotes the file has peeled data. */
- public static final String PACKED_REFS_PEELED = " peeled"; //$NON-NLS-1$
-
- /** The names of the additional refs supported by this class */
- private static final String[] additionalRefsNames = new String[] {
- Constants.MERGE_HEAD, Constants.FETCH_HEAD, Constants.ORIG_HEAD,
- Constants.CHERRY_PICK_HEAD };
-
- @SuppressWarnings("boxing")
- private static final List<Integer> RETRY_SLEEP_MS =
- Collections.unmodifiableList(Arrays.asList(0, 100, 200, 400, 800, 1600));
-
- private final FileRepository parent;
-
- private final File gitDir;
-
- final File refsDir;
-
- final File packedRefsFile;
-
- final File logsDir;
-
- final File logsRefsDir;
-
- /**
- * Immutable sorted list of loose references.
- * <p>
- * Symbolic references in this collection are stored unresolved, that is
- * their target appears to be a new reference with no ObjectId. These are
- * converted into resolved references during a get operation, ensuring the
- * live value is always returned.
- */
- private final AtomicReference<RefList<LooseRef>> looseRefs = new AtomicReference<>();
-
- /** Immutable sorted list of packed references. */
- final AtomicReference<PackedRefList> packedRefs = new AtomicReference<>();
-
- /**
- * Lock for coordinating operations within a single process that may contend
- * on the {@code packed-refs} file.
- * <p>
- * All operations that write {@code packed-refs} must still acquire a
- * {@link LockFile} on {@link #packedRefsFile}, even after they have acquired
- * this lock, since there may be multiple {@link RefDirectory} instances or
- * other processes operating on the same repo on disk.
- * <p>
- * This lock exists so multiple threads in the same process can wait in a fair
- * queue without trying, failing, and retrying to acquire the on-disk lock. If
- * {@code RepositoryCache} is used, this lock instance will be used by all
- * threads.
- */
- final ReentrantLock inProcessPackedRefsLock = new ReentrantLock(true);
-
- /**
- * Number of modifications made to this database.
- * <p>
- * This counter is incremented when a change is made, or detected from the
- * filesystem during a read operation.
- */
- private final AtomicInteger modCnt = new AtomicInteger();
-
- /**
- * Last {@link #modCnt} that we sent to listeners.
- * <p>
- * This value is compared to {@link #modCnt}, and a notification is sent to
- * the listeners only when it differs.
- */
- private final AtomicInteger lastNotifiedModCnt = new AtomicInteger();
-
- private List<Integer> retrySleepMs = RETRY_SLEEP_MS;
-
- RefDirectory(FileRepository db) {
- final FS fs = db.getFS();
- parent = db;
- gitDir = db.getDirectory();
- refsDir = fs.resolve(gitDir, R_REFS);
- logsDir = fs.resolve(gitDir, LOGS);
- logsRefsDir = fs.resolve(gitDir, LOGS + '/' + R_REFS);
- packedRefsFile = fs.resolve(gitDir, PACKED_REFS);
-
- looseRefs.set(RefList.<LooseRef> emptyList());
- packedRefs.set(NO_PACKED_REFS);
- }
-
- Repository getRepository() {
- return parent;
- }
-
- ReflogWriter newLogWriter(boolean force) {
- return new ReflogWriter(this, force);
- }
-
- /**
- * Locate the log file on disk for a single reference name.
- *
- * @param name
- * name of the ref, relative to the Git repository top level
- * directory (so typically starts with refs/).
- * @return the log file location.
- */
- public File logFor(String name) {
- if (name.startsWith(R_REFS)) {
- name = name.substring(R_REFS.length());
- return new File(logsRefsDir, name);
- }
- return new File(logsDir, name);
- }
-
- /** {@inheritDoc} */
- @Override
- public void create() throws IOException {
- FileUtils.mkdir(refsDir);
- FileUtils.mkdir(new File(refsDir, R_HEADS.substring(R_REFS.length())));
- FileUtils.mkdir(new File(refsDir, R_TAGS.substring(R_REFS.length())));
- newLogWriter(false).create();
- }
-
- /** {@inheritDoc} */
- @Override
- public void close() {
- clearReferences();
- }
-
- private void clearReferences() {
- looseRefs.set(RefList.<LooseRef> emptyList());
- packedRefs.set(NO_PACKED_REFS);
- }
-
- /** {@inheritDoc} */
- @Override
- public void refresh() {
- super.refresh();
- clearReferences();
- }
-
- /** {@inheritDoc} */
- @Override
- public boolean isNameConflicting(String name) throws IOException {
- // Cannot be nested within an existing reference.
- int lastSlash = name.lastIndexOf('/');
- while (0 < lastSlash) {
- String needle = name.substring(0, lastSlash);
- if (exactRef(needle) != null) {
- return true;
- }
- lastSlash = name.lastIndexOf('/', lastSlash - 1);
- }
-
- // Cannot be the container of an existing reference.
- return !getRefsByPrefix(name + '/').isEmpty();
- }
-
- @Nullable
- private Ref readAndResolve(String name, RefList<Ref> packed) throws IOException {
- try {
- Ref ref = readRef(name, packed);
- if (ref != null) {
- ref = resolve(ref, 0, null, null, packed);
- }
- return ref;
- } catch (IOException e) {
- if (name.contains("/") //$NON-NLS-1$
- || !(e.getCause() instanceof InvalidObjectIdException)) {
- throw e;
- }
-
- // While looking for a ref outside of refs/ (e.g., 'config'), we
- // found a non-ref file (e.g., a config file) instead. Treat this
- // as a ref-not-found condition.
- return null;
- }
- }
-
- /** {@inheritDoc} */
- @Override
- public Ref exactRef(String name) throws IOException {
- try {
- return readAndResolve(name, getPackedRefs());
- } finally {
- fireRefsChanged();
- }
- }
-
- /** {@inheritDoc} */
- @Override
- @NonNull
- public Map<String, Ref> exactRef(String... refs) throws IOException {
- try {
- RefList<Ref> packed = getPackedRefs();
- Map<String, Ref> result = new HashMap<>(refs.length);
- for (String name : refs) {
- Ref ref = readAndResolve(name, packed);
- if (ref != null) {
- result.put(name, ref);
- }
- }
- return result;
- } finally {
- fireRefsChanged();
- }
- }
-
- /** {@inheritDoc} */
- @Override
- @Nullable
- public Ref firstExactRef(String... refs) throws IOException {
- try {
- RefList<Ref> packed = getPackedRefs();
- for (String name : refs) {
- Ref ref = readAndResolve(name, packed);
- if (ref != null) {
- return ref;
- }
- }
- return null;
- } finally {
- fireRefsChanged();
- }
- }
-
- /** {@inheritDoc} */
- @Override
- public Map<String, Ref> getRefs(String prefix) throws IOException {
- final RefList<LooseRef> oldLoose = looseRefs.get();
- LooseScanner scan = new LooseScanner(oldLoose);
- scan.scan(prefix);
- final RefList<Ref> packed = getPackedRefs();
-
- RefList<LooseRef> loose;
- if (scan.newLoose != null) {
- scan.newLoose.sort();
- loose = scan.newLoose.toRefList();
- if (looseRefs.compareAndSet(oldLoose, loose))
- modCnt.incrementAndGet();
- } else
- loose = oldLoose;
- fireRefsChanged();
-
- RefList.Builder<Ref> symbolic = scan.symbolic;
- for (int idx = 0; idx < symbolic.size();) {
- final Ref symbolicRef = symbolic.get(idx);
- final Ref resolvedRef = resolve(symbolicRef, 0, prefix, loose, packed);
- if (resolvedRef != null && resolvedRef.getObjectId() != null) {
- symbolic.set(idx, resolvedRef);
- idx++;
- } else {
- // A broken symbolic reference, we have to drop it from the
- // collections the client is about to receive. Should be a
- // rare occurrence so pay a copy penalty.
- symbolic.remove(idx);
- final int toRemove = loose.find(symbolicRef.getName());
- if (0 <= toRemove)
- loose = loose.remove(toRemove);
- }
- }
- symbolic.sort();
-
- return new RefMap(prefix, packed, upcast(loose), symbolic.toRefList());
- }
-
- /** {@inheritDoc} */
- @Override
- public List<Ref> getAdditionalRefs() throws IOException {
- List<Ref> ret = new LinkedList<>();
- for (String name : additionalRefsNames) {
- Ref r = exactRef(name);
- if (r != null)
- ret.add(r);
- }
- return ret;
- }
-
- @SuppressWarnings("unchecked")
- private RefList<Ref> upcast(RefList<? extends Ref> loose) {
- return (RefList<Ref>) loose;
- }
-
- private class LooseScanner {
- private final RefList<LooseRef> curLoose;
-
- private int curIdx;
-
- final RefList.Builder<Ref> symbolic = new RefList.Builder<>(4);
-
- RefList.Builder<LooseRef> newLoose;
-
- LooseScanner(RefList<LooseRef> curLoose) {
- this.curLoose = curLoose;
- }
-
- void scan(String prefix) {
- if (ALL.equals(prefix)) {
- scanOne(HEAD);
- scanTree(R_REFS, refsDir);
-
- // If any entries remain, they are deleted, drop them.
- if (newLoose == null && curIdx < curLoose.size())
- newLoose = curLoose.copy(curIdx);
-
- } else if (prefix.startsWith(R_REFS) && prefix.endsWith("/")) { //$NON-NLS-1$
- curIdx = -(curLoose.find(prefix) + 1);
- File dir = new File(refsDir, prefix.substring(R_REFS.length()));
- scanTree(prefix, dir);
-
- // Skip over entries still within the prefix; these have
- // been removed from the directory.
- while (curIdx < curLoose.size()) {
- if (!curLoose.get(curIdx).getName().startsWith(prefix))
- break;
- if (newLoose == null)
- newLoose = curLoose.copy(curIdx);
- curIdx++;
- }
-
- // Keep any entries outside of the prefix space, we
- // do not know anything about their status.
- if (newLoose != null) {
- while (curIdx < curLoose.size())
- newLoose.add(curLoose.get(curIdx++));
- }
- }
- }
-
- private boolean scanTree(String prefix, File dir) {
- final String[] entries = dir.list(LockFile.FILTER);
- if (entries == null) // not a directory or an I/O error
- return false;
- if (0 < entries.length) {
- for (int i = 0; i < entries.length; ++i) {
- String e = entries[i];
- File f = new File(dir, e);
- if (f.isDirectory())
- entries[i] += '/';
- }
- Arrays.sort(entries);
- for (String name : entries) {
- if (name.charAt(name.length() - 1) == '/')
- scanTree(prefix + name, new File(dir, name));
- else
- scanOne(prefix + name);
- }
- }
- return true;
- }
-
- private void scanOne(String name) {
- LooseRef cur;
-
- if (curIdx < curLoose.size()) {
- do {
- cur = curLoose.get(curIdx);
- int cmp = RefComparator.compareTo(cur, name);
- if (cmp < 0) {
- // Reference is not loose anymore, its been deleted.
- // Skip the name in the new result list.
- if (newLoose == null)
- newLoose = curLoose.copy(curIdx);
- curIdx++;
- cur = null;
- continue;
- }
-
- if (cmp > 0) // Newly discovered loose reference.
- cur = null;
- break;
- } while (curIdx < curLoose.size());
- } else
- cur = null; // Newly discovered loose reference.
-
- LooseRef n;
- try {
- n = scanRef(cur, name);
- } catch (IOException notValid) {
- n = null;
- }
-
- if (n != null) {
- if (cur != n && newLoose == null)
- newLoose = curLoose.copy(curIdx);
- if (newLoose != null)
- newLoose.add(n);
- if (n.isSymbolic())
- symbolic.add(n);
- } else if (cur != null) {
- // Tragically, this file is no longer a loose reference.
- // Kill our cached entry of it.
- if (newLoose == null)
- newLoose = curLoose.copy(curIdx);
- }
-
- if (cur != null)
- curIdx++;
- }
- }
-
- /** {@inheritDoc} */
- @Override
- public Ref peel(Ref ref) throws IOException {
- final Ref leaf = ref.getLeaf();
- if (leaf.isPeeled() || leaf.getObjectId() == null)
- return ref;
-
- ObjectIdRef newLeaf = doPeel(leaf);
-
- // Try to remember this peeling in the cache, so we don't have to do
- // it again in the future, but only if the reference is unchanged.
- if (leaf.getStorage().isLoose()) {
- RefList<LooseRef> curList = looseRefs.get();
- int idx = curList.find(leaf.getName());
- if (0 <= idx && curList.get(idx) == leaf) {
- LooseRef asPeeled = ((LooseRef) leaf).peel(newLeaf);
- RefList<LooseRef> newList = curList.set(idx, asPeeled);
- looseRefs.compareAndSet(curList, newList);
- }
- }
-
- return recreate(ref, newLeaf);
- }
-
- private ObjectIdRef doPeel(Ref leaf) throws MissingObjectException,
- IOException {
- try (RevWalk rw = new RevWalk(getRepository())) {
- RevObject obj = rw.parseAny(leaf.getObjectId());
- if (obj instanceof RevTag) {
- return new ObjectIdRef.PeeledTag(leaf.getStorage(), leaf
- .getName(), leaf.getObjectId(), rw.peel(obj).copy());
- }
- return new ObjectIdRef.PeeledNonTag(leaf.getStorage(),
- leaf.getName(), leaf.getObjectId());
- }
- }
-
- private static Ref recreate(Ref old, ObjectIdRef leaf) {
- if (old.isSymbolic()) {
- Ref dst = recreate(old.getTarget(), leaf);
- return new SymbolicRef(old.getName(), dst);
- }
- return leaf;
- }
-
- void storedSymbolicRef(RefDirectoryUpdate u, FileSnapshot snapshot,
- String target) {
- putLooseRef(newSymbolicRef(snapshot, u.getRef().getName(), target));
- fireRefsChanged();
- }
-
- /** {@inheritDoc} */
- @Override
- public RefDirectoryUpdate newUpdate(String name, boolean detach)
- throws IOException {
- boolean detachingSymbolicRef = false;
- final RefList<Ref> packed = getPackedRefs();
- Ref ref = readRef(name, packed);
- if (ref != null)
- ref = resolve(ref, 0, null, null, packed);
- if (ref == null)
- ref = new ObjectIdRef.Unpeeled(NEW, name, null);
- else {
- detachingSymbolicRef = detach && ref.isSymbolic();
- }
- RefDirectoryUpdate refDirUpdate = new RefDirectoryUpdate(this, ref);
- if (detachingSymbolicRef)
- refDirUpdate.setDetachingSymbolicRef();
- return refDirUpdate;
- }
-
- /** {@inheritDoc} */
- @Override
- public RefDirectoryRename newRename(String fromName, String toName)
- throws IOException {
- RefDirectoryUpdate from = newUpdate(fromName, false);
- RefDirectoryUpdate to = newUpdate(toName, false);
- return new RefDirectoryRename(from, to);
- }
-
- /** {@inheritDoc} */
- @Override
- public PackedBatchRefUpdate newBatchUpdate() {
- return new PackedBatchRefUpdate(this);
- }
-
- /** {@inheritDoc} */
- @Override
- public boolean performsAtomicTransactions() {
- return true;
- }
-
- void stored(RefDirectoryUpdate update, FileSnapshot snapshot) {
- final ObjectId target = update.getNewObjectId().copy();
- final Ref leaf = update.getRef().getLeaf();
- putLooseRef(new LooseUnpeeled(snapshot, leaf.getName(), target));
- }
-
- private void putLooseRef(LooseRef ref) {
- RefList<LooseRef> cList, nList;
- do {
- cList = looseRefs.get();
- nList = cList.put(ref);
- } while (!looseRefs.compareAndSet(cList, nList));
- modCnt.incrementAndGet();
- fireRefsChanged();
- }
-
- void delete(RefDirectoryUpdate update) throws IOException {
- Ref dst = update.getRef();
- if (!update.isDetachingSymbolicRef()) {
- dst = dst.getLeaf();
- }
- String name = dst.getName();
-
- // Write the packed-refs file using an atomic update. We might
- // wind up reading it twice, before and after the lock, to ensure
- // we don't miss an edit made externally.
- final PackedRefList packed = getPackedRefs();
- if (packed.contains(name)) {
- inProcessPackedRefsLock.lock();
- try {
- LockFile lck = lockPackedRefsOrThrow();
- try {
- PackedRefList cur = readPackedRefs();
- int idx = cur.find(name);
- if (0 <= idx) {
- commitPackedRefs(lck, cur.remove(idx), packed, true);
- }
- } finally {
- lck.unlock();
- }
- } finally {
- inProcessPackedRefsLock.unlock();
- }
- }
-
- RefList<LooseRef> curLoose, newLoose;
- do {
- curLoose = looseRefs.get();
- int idx = curLoose.find(name);
- if (idx < 0)
- break;
- newLoose = curLoose.remove(idx);
- } while (!looseRefs.compareAndSet(curLoose, newLoose));
-
- int levels = levelsIn(name) - 2;
- delete(logFor(name), levels);
- if (dst.getStorage().isLoose()) {
- update.unlock();
- delete(fileFor(name), levels);
- }
-
- modCnt.incrementAndGet();
- fireRefsChanged();
- }
-
- /**
- * Adds a set of refs to the set of packed-refs. Only non-symbolic refs are
- * added. If a ref with the given name already existed in packed-refs it is
- * updated with the new value. Each loose ref which was added to the
- * packed-ref file is deleted. If a given ref can't be locked it will not be
- * added to the pack file.
- *
- * @param refs
- * the refs to be added. Must be fully qualified.
- * @throws java.io.IOException
- */
- public void pack(List<String> refs) throws IOException {
- pack(refs, Collections.emptyMap());
- }
-
- PackedRefList pack(Map<String, LockFile> heldLocks) throws IOException {
- return pack(heldLocks.keySet(), heldLocks);
- }
-
- private PackedRefList pack(Collection<String> refs,
- Map<String, LockFile> heldLocks) throws IOException {
- for (LockFile ol : heldLocks.values()) {
- ol.requireLock();
- }
- if (refs.isEmpty()) {
- return null;
- }
- FS fs = parent.getFS();
-
- // Lock the packed refs file and read the content
- inProcessPackedRefsLock.lock();
- try {
- LockFile lck = lockPackedRefsOrThrow();
- try {
- final PackedRefList packed = getPackedRefs();
- RefList<Ref> cur = readPackedRefs();
-
- // Iterate over all refs to be packed
- boolean dirty = false;
- for (String refName : refs) {
- Ref oldRef = readRef(refName, cur);
- if (oldRef == null) {
- continue; // A non-existent ref is already correctly packed.
- }
- if (oldRef.isSymbolic()) {
- continue; // can't pack symbolic refs
- }
- // Add/Update it to packed-refs
- Ref newRef = peeledPackedRef(oldRef);
- if (newRef == oldRef) {
- // No-op; peeledPackedRef returns the input ref only if it's already
- // packed, and readRef returns a packed ref only if there is no
- // loose ref.
- continue;
- }
-
- dirty = true;
- int idx = cur.find(refName);
- if (idx >= 0) {
- cur = cur.set(idx, newRef);
- } else {
- cur = cur.add(idx, newRef);
- }
- }
- if (!dirty) {
- // All requested refs were already packed accurately
- return packed;
- }
-
- // The new content for packed-refs is collected. Persist it.
- PackedRefList result = commitPackedRefs(lck, cur, packed,
- false);
-
- // Now delete the loose refs which are now packed
- for (String refName : refs) {
- // Lock the loose ref
- File refFile = fileFor(refName);
- if (!fs.exists(refFile)) {
- continue;
- }
-
- LockFile rLck = heldLocks.get(refName);
- boolean shouldUnlock;
- if (rLck == null) {
- rLck = new LockFile(refFile);
- if (!rLck.lock()) {
- continue;
- }
- shouldUnlock = true;
- } else {
- shouldUnlock = false;
- }
-
- try {
- LooseRef currentLooseRef = scanRef(null, refName);
- if (currentLooseRef == null || currentLooseRef.isSymbolic()) {
- continue;
- }
- Ref packedRef = cur.get(refName);
- ObjectId clr_oid = currentLooseRef.getObjectId();
- if (clr_oid != null
- && clr_oid.equals(packedRef.getObjectId())) {
- RefList<LooseRef> curLoose, newLoose;
- do {
- curLoose = looseRefs.get();
- int idx = curLoose.find(refName);
- if (idx < 0) {
- break;
- }
- newLoose = curLoose.remove(idx);
- } while (!looseRefs.compareAndSet(curLoose, newLoose));
- int levels = levelsIn(refName) - 2;
- delete(refFile, levels, rLck);
- }
- } finally {
- if (shouldUnlock) {
- rLck.unlock();
- }
- }
- }
- // Don't fire refsChanged. The refs have not change, only their
- // storage.
- return result;
- } finally {
- lck.unlock();
- }
- } finally {
- inProcessPackedRefsLock.unlock();
- }
- }
-
- @Nullable
- LockFile lockPackedRefs() throws IOException {
- LockFile lck = new LockFile(packedRefsFile);
- for (int ms : getRetrySleepMs()) {
- sleep(ms);
- if (lck.lock()) {
- return lck;
- }
- }
- return null;
- }
-
- private LockFile lockPackedRefsOrThrow() throws IOException {
- LockFile lck = lockPackedRefs();
- if (lck == null) {
- throw new LockFailedException(packedRefsFile);
- }
- return lck;
- }
-
- /**
- * Make sure a ref is peeled and has the Storage PACKED. If the given ref
- * has this attributes simply return it. Otherwise create a new peeled
- * {@link ObjectIdRef} where Storage is set to PACKED.
- *
- * @param f
- * @return a ref for Storage PACKED having the same name, id, peeledId as f
- * @throws MissingObjectException
- * @throws IOException
- */
- private Ref peeledPackedRef(Ref f)
- throws MissingObjectException, IOException {
- if (f.getStorage().isPacked() && f.isPeeled()) {
- return f;
- }
- if (!f.isPeeled()) {
- f = peel(f);
- }
- ObjectId peeledObjectId = f.getPeeledObjectId();
- if (peeledObjectId != null) {
- return new ObjectIdRef.PeeledTag(PACKED, f.getName(),
- f.getObjectId(), peeledObjectId);
- }
- return new ObjectIdRef.PeeledNonTag(PACKED, f.getName(),
- f.getObjectId());
- }
-
- void log(boolean force, RefUpdate update, String msg, boolean deref)
- throws IOException {
- newLogWriter(force).log(update, msg, deref);
- }
-
- private Ref resolve(final Ref ref, int depth, String prefix,
- RefList<LooseRef> loose, RefList<Ref> packed) throws IOException {
- if (ref.isSymbolic()) {
- Ref dst = ref.getTarget();
-
- if (MAX_SYMBOLIC_REF_DEPTH <= depth)
- return null; // claim it doesn't exist
-
- // If the cached value can be assumed to be current due to a
- // recent scan of the loose directory, use it.
- if (loose != null && dst.getName().startsWith(prefix)) {
- int idx;
- if (0 <= (idx = loose.find(dst.getName())))
- dst = loose.get(idx);
- else if (0 <= (idx = packed.find(dst.getName())))
- dst = packed.get(idx);
- else
- return ref;
- } else {
- dst = readRef(dst.getName(), packed);
- if (dst == null)
- return ref;
- }
-
- dst = resolve(dst, depth + 1, prefix, loose, packed);
- if (dst == null)
- return null;
- return new SymbolicRef(ref.getName(), dst);
- }
- return ref;
- }
-
- PackedRefList getPackedRefs() throws IOException {
- boolean trustFolderStat = getRepository().getConfig().getBoolean(
- ConfigConstants.CONFIG_CORE_SECTION,
- ConfigConstants.CONFIG_KEY_TRUSTFOLDERSTAT, true);
-
- final PackedRefList curList = packedRefs.get();
- if (trustFolderStat && !curList.snapshot.isModified(packedRefsFile)) {
- return curList;
- }
-
- final PackedRefList newList = readPackedRefs();
- if (packedRefs.compareAndSet(curList, newList)
- && !curList.id.equals(newList.id)) {
- modCnt.incrementAndGet();
- }
- return newList;
- }
-
- private PackedRefList readPackedRefs() throws IOException {
- int maxStaleRetries = 5;
- int retries = 0;
- while (true) {
- final FileSnapshot snapshot = FileSnapshot.save(packedRefsFile);
- final MessageDigest digest = Constants.newMessageDigest();
- try (BufferedReader br = new BufferedReader(new InputStreamReader(
- new DigestInputStream(new FileInputStream(packedRefsFile),
- digest),
- UTF_8))) {
- try {
- return new PackedRefList(parsePackedRefs(br), snapshot,
- ObjectId.fromRaw(digest.digest()));
- } catch (IOException e) {
- if (FileUtils.isStaleFileHandleInCausalChain(e)
- && retries < maxStaleRetries) {
- if (LOG.isDebugEnabled()) {
- LOG.debug(MessageFormat.format(
- JGitText.get().packedRefsHandleIsStale,
- Integer.valueOf(retries)), e);
- }
- retries++;
- continue;
- }
- throw e;
- }
- } catch (FileNotFoundException noPackedRefs) {
- if (packedRefsFile.exists()) {
- throw noPackedRefs;
- }
- // Ignore it and leave the new list empty.
- return NO_PACKED_REFS;
- }
- }
- }
-
- private RefList<Ref> parsePackedRefs(BufferedReader br)
- throws IOException {
- RefList.Builder<Ref> all = new RefList.Builder<>();
- Ref last = null;
- boolean peeled = false;
- boolean needSort = false;
-
- String p;
- while ((p = br.readLine()) != null) {
- if (p.charAt(0) == '#') {
- if (p.startsWith(PACKED_REFS_HEADER)) {
- p = p.substring(PACKED_REFS_HEADER.length());
- peeled = p.contains(PACKED_REFS_PEELED);
- }
- continue;
- }
-
- if (p.charAt(0) == '^') {
- if (last == null)
- throw new IOException(JGitText.get().peeledLineBeforeRef);
-
- ObjectId id = ObjectId.fromString(p.substring(1));
- last = new ObjectIdRef.PeeledTag(PACKED, last.getName(), last
- .getObjectId(), id);
- all.set(all.size() - 1, last);
- continue;
- }
-
- int sp = p.indexOf(' ');
- if (sp < 0) {
- throw new IOException(MessageFormat.format(
- JGitText.get().packedRefsCorruptionDetected,
- packedRefsFile.getAbsolutePath()));
- }
- ObjectId id = ObjectId.fromString(p.substring(0, sp));
- String name = copy(p, sp + 1, p.length());
- ObjectIdRef cur;
- if (peeled)
- cur = new ObjectIdRef.PeeledNonTag(PACKED, name, id);
- else
- cur = new ObjectIdRef.Unpeeled(PACKED, name, id);
- if (last != null && RefComparator.compareTo(last, cur) > 0)
- needSort = true;
- all.add(cur);
- last = cur;
- }
-
- if (needSort)
- all.sort();
- return all.toRefList();
- }
-
- private static String copy(String src, int off, int end) {
- // Don't use substring since it could leave a reference to the much
- // larger existing string. Force construction of a full new object.
- return new StringBuilder(end - off).append(src, off, end).toString();
- }
-
- PackedRefList commitPackedRefs(final LockFile lck, final RefList<Ref> refs,
- final PackedRefList oldPackedList, boolean changed)
- throws IOException {
- // Can't just return packedRefs.get() from this method; it might have been
- // updated again after writePackedRefs() returns.
- AtomicReference<PackedRefList> result = new AtomicReference<>();
- new RefWriter(refs) {
- @Override
- protected void writeFile(String name, byte[] content)
- throws IOException {
- lck.setFSync(true);
- lck.setNeedSnapshot(true);
- try {
- lck.write(content);
- } catch (IOException ioe) {
- throw new ObjectWritingException(MessageFormat.format(JGitText.get().unableToWrite, name), ioe);
- }
- try {
- lck.waitForStatChange();
- } catch (InterruptedException e) {
- lck.unlock();
- throw new ObjectWritingException(
- MessageFormat.format(
- JGitText.get().interruptedWriting, name),
- e);
- }
- if (!lck.commit())
- throw new ObjectWritingException(MessageFormat.format(JGitText.get().unableToWrite, name));
-
- byte[] digest = Constants.newMessageDigest().digest(content);
- PackedRefList newPackedList = new PackedRefList(
- refs, lck.getCommitSnapshot(), ObjectId.fromRaw(digest));
-
- // This thread holds the file lock, so no other thread or process should
- // be able to modify the packed-refs file on disk. If the list changed,
- // it means something is very wrong, so throw an exception.
- //
- // However, we can't use a naive compareAndSet to check whether the
- // update was successful, because another thread might _read_ the
- // packed refs file that was written out by this thread while holding
- // the lock, and update the packedRefs reference to point to that. So
- // compare the actual contents instead.
- PackedRefList afterUpdate = packedRefs.updateAndGet(
- p -> p.id.equals(oldPackedList.id) ? newPackedList : p);
- if (!afterUpdate.id.equals(newPackedList.id)) {
- throw new ObjectWritingException(
- MessageFormat.format(JGitText.get().unableToWrite, name));
- }
- if (changed) {
- modCnt.incrementAndGet();
- }
- result.set(newPackedList);
- }
- }.writePackedRefs();
- return result.get();
- }
-
- private Ref readRef(String name, RefList<Ref> packed) throws IOException {
- final RefList<LooseRef> curList = looseRefs.get();
- final int idx = curList.find(name);
- if (0 <= idx) {
- final LooseRef o = curList.get(idx);
- final LooseRef n = scanRef(o, name);
- if (n == null) {
- if (looseRefs.compareAndSet(curList, curList.remove(idx)))
- modCnt.incrementAndGet();
- return packed.get(name);
- }
-
- if (o == n)
- return n;
- if (looseRefs.compareAndSet(curList, curList.set(idx, n)))
- modCnt.incrementAndGet();
- return n;
- }
-
- final LooseRef n = scanRef(null, name);
- if (n == null)
- return packed.get(name);
-
- // check whether the found new ref is the an additional ref. These refs
- // should not go into looseRefs
- for (String additionalRefsName : additionalRefsNames) {
- if (name.equals(additionalRefsName)) {
- return n;
- }
- }
-
- if (looseRefs.compareAndSet(curList, curList.add(idx, n)))
- modCnt.incrementAndGet();
- return n;
- }
-
- LooseRef scanRef(LooseRef ref, String name) throws IOException {
- final File path = fileFor(name);
- FileSnapshot currentSnapshot = null;
-
- if (ref != null) {
- currentSnapshot = ref.getSnapShot();
- if (!currentSnapshot.isModified(path))
- return ref;
- name = ref.getName();
- }
-
- final int limit = 4096;
- final byte[] buf;
- FileSnapshot otherSnapshot = FileSnapshot.save(path);
- try {
- buf = IO.readSome(path, limit);
- } catch (FileNotFoundException noFile) {
- if (path.exists() && path.isFile()) {
- throw noFile;
- }
- return null; // doesn't exist or no file; not a reference.
- }
-
- int n = buf.length;
- if (n == 0)
- return null; // empty file; not a reference.
-
- if (isSymRef(buf, n)) {
- if (n == limit)
- return null; // possibly truncated ref
-
- // trim trailing whitespace
- while (0 < n && Character.isWhitespace(buf[n - 1]))
- n--;
- if (n < 6) {
- String content = RawParseUtils.decode(buf, 0, n);
- throw new IOException(MessageFormat.format(JGitText.get().notARef, name, content));
- }
- final String target = RawParseUtils.decode(buf, 5, n);
- if (ref != null && ref.isSymbolic()
- && ref.getTarget().getName().equals(target)) {
- assert(currentSnapshot != null);
- currentSnapshot.setClean(otherSnapshot);
- return ref;
- }
- return newSymbolicRef(otherSnapshot, name, target);
- }
-
- if (n < OBJECT_ID_STRING_LENGTH)
- return null; // impossibly short object identifier; not a reference.
-
- final ObjectId id;
- try {
- id = ObjectId.fromString(buf, 0);
- if (ref != null && !ref.isSymbolic()
- && id.equals(ref.getTarget().getObjectId())) {
- assert(currentSnapshot != null);
- currentSnapshot.setClean(otherSnapshot);
- return ref;
- }
-
- } catch (IllegalArgumentException notRef) {
- while (0 < n && Character.isWhitespace(buf[n - 1]))
- n--;
- String content = RawParseUtils.decode(buf, 0, n);
-
- throw new IOException(MessageFormat.format(JGitText.get().notARef,
- name, content), notRef);
- }
- return new LooseUnpeeled(otherSnapshot, name, id);
- }
-
- private static boolean isSymRef(byte[] buf, int n) {
- if (n < 6)
- return false;
- return /**/buf[0] == 'r' //
- && buf[1] == 'e' //
- && buf[2] == 'f' //
- && buf[3] == ':' //
- && buf[4] == ' ';
- }
-
- /**
- * Detect if we are in a clone command execution
- *
- * @return {@code true} if we are currently cloning a repository
- * @throws IOException
- */
- boolean isInClone() throws IOException {
- return hasDanglingHead() && !packedRefsFile.exists() && !hasLooseRef();
- }
-
- private boolean hasDanglingHead() throws IOException {
- Ref head = exactRef(Constants.HEAD);
- if (head != null) {
- ObjectId id = head.getObjectId();
- return id == null || id.equals(ObjectId.zeroId());
- }
- return false;
- }
-
- private boolean hasLooseRef() throws IOException {
- try (Stream<Path> stream = Files.walk(refsDir.toPath())) {
- return stream.anyMatch(Files::isRegularFile);
- }
- }
-
- /** If the parent should fire listeners, fires them. */
- void fireRefsChanged() {
- final int last = lastNotifiedModCnt.get();
- final int curr = modCnt.get();
- if (last != curr && lastNotifiedModCnt.compareAndSet(last, curr) && last != 0)
- parent.fireEvent(new RefsChangedEvent());
- }
-
- /**
- * Create a reference update to write a temporary reference.
- *
- * @return an update for a new temporary reference.
- * @throws IOException
- * a temporary name cannot be allocated.
- */
- RefDirectoryUpdate newTemporaryUpdate() throws IOException {
- File tmp = File.createTempFile("renamed_", "_ref", refsDir); //$NON-NLS-1$ //$NON-NLS-2$
- String name = Constants.R_REFS + tmp.getName();
- Ref ref = new ObjectIdRef.Unpeeled(NEW, name, null);
- return new RefDirectoryUpdate(this, ref);
- }
-
- /**
- * Locate the file on disk for a single reference name.
- *
- * @param name
- * name of the ref, relative to the Git repository top level
- * directory (so typically starts with refs/).
- * @return the loose file location.
- */
- File fileFor(String name) {
- if (name.startsWith(R_REFS)) {
- name = name.substring(R_REFS.length());
- return new File(refsDir, name);
- }
- return new File(gitDir, name);
- }
-
- static int levelsIn(String name) {
- int count = 0;
- for (int p = name.indexOf('/'); p >= 0; p = name.indexOf('/', p + 1))
- count++;
- return count;
- }
-
- static void delete(File file, int depth) throws IOException {
- delete(file, depth, null);
- }
-
- private static void delete(File file, int depth, LockFile rLck)
- throws IOException {
- if (!file.delete() && file.isFile()) {
- throw new IOException(MessageFormat.format(
- JGitText.get().fileCannotBeDeleted, file));
- }
-
- if (rLck != null) {
- rLck.unlock(); // otherwise cannot delete dir below
- }
- File dir = file.getParentFile();
- for (int i = 0; i < depth; ++i) {
- try {
- Files.deleteIfExists(dir.toPath());
- } catch (DirectoryNotEmptyException e) {
- // Don't log; normal case when there are other refs with the
- // same prefix
- break;
- } catch (IOException e) {
- LOG.warn(MessageFormat.format(JGitText.get().unableToRemovePath,
- dir), e);
- break;
- }
- dir = dir.getParentFile();
- }
- }
-
- /**
- * Get times to sleep while retrying a possibly contentious operation.
- * <p>
- * For retrying an operation that might have high contention, such as locking
- * the {@code packed-refs} file, the caller may implement a retry loop using
- * the returned values:
- *
- * <pre>
- * for (int toSleepMs : getRetrySleepMs()) {
- * sleep(toSleepMs);
- * if (isSuccessful(doSomething())) {
- * return success;
- * }
- * }
- * return failure;
- * </pre>
- *
- * The first value in the returned iterable is 0, and the caller should treat
- * a fully-consumed iterator as a timeout.
- *
- * @return iterable of times, in milliseconds, that the caller should sleep
- * before attempting an operation.
- */
- Iterable<Integer> getRetrySleepMs() {
- return retrySleepMs;
- }
-
- void setRetrySleepMs(List<Integer> retrySleepMs) {
- if (retrySleepMs == null || retrySleepMs.isEmpty()
- || retrySleepMs.get(0).intValue() != 0) {
- throw new IllegalArgumentException();
- }
- this.retrySleepMs = retrySleepMs;
- }
-
- /**
- * Sleep with {@link Thread#sleep(long)}, converting {@link
- * InterruptedException} to {@link InterruptedIOException}.
- *
- * @param ms
- * time to sleep, in milliseconds; zero or negative is a no-op.
- * @throws InterruptedIOException
- * if sleeping was interrupted.
- */
- static void sleep(long ms) throws InterruptedIOException {
- if (ms <= 0) {
- return;
- }
- try {
- Thread.sleep(ms);
- } catch (InterruptedException e) {
- InterruptedIOException ie = new InterruptedIOException();
- ie.initCause(e);
- throw ie;
- }
- }
-
- static class PackedRefList extends RefList<Ref> {
-
- private final FileSnapshot snapshot;
-
- private final ObjectId id;
-
- private PackedRefList(RefList<Ref> src, FileSnapshot s, ObjectId i) {
- super(src);
- snapshot = s;
- id = i;
- }
- }
-
- private static final PackedRefList NO_PACKED_REFS = new PackedRefList(
- RefList.emptyList(), FileSnapshot.MISSING_FILE,
- ObjectId.zeroId());
-
- private static LooseSymbolicRef newSymbolicRef(FileSnapshot snapshot,
- String name, String target) {
- Ref dst = new ObjectIdRef.Unpeeled(NEW, target, null);
- return new LooseSymbolicRef(snapshot, name, dst);
- }
-
- private static interface LooseRef extends Ref {
- FileSnapshot getSnapShot();
-
- LooseRef peel(ObjectIdRef newLeaf);
- }
-
- private static final class LoosePeeledTag extends ObjectIdRef.PeeledTag
- implements LooseRef {
- private final FileSnapshot snapShot;
-
- LoosePeeledTag(FileSnapshot snapshot, @NonNull String refName,
- @NonNull ObjectId id, @NonNull ObjectId p) {
- super(LOOSE, refName, id, p);
- this.snapShot = snapshot;
- }
-
- @Override
- public FileSnapshot getSnapShot() {
- return snapShot;
- }
-
- @Override
- public LooseRef peel(ObjectIdRef newLeaf) {
- return this;
- }
- }
-
- private static final class LooseNonTag extends ObjectIdRef.PeeledNonTag
- implements LooseRef {
- private final FileSnapshot snapShot;
-
- LooseNonTag(FileSnapshot snapshot, @NonNull String refName,
- @NonNull ObjectId id) {
- super(LOOSE, refName, id);
- this.snapShot = snapshot;
- }
-
- @Override
- public FileSnapshot getSnapShot() {
- return snapShot;
- }
-
- @Override
- public LooseRef peel(ObjectIdRef newLeaf) {
- return this;
- }
- }
-
- private static final class LooseUnpeeled extends ObjectIdRef.Unpeeled
- implements LooseRef {
- private FileSnapshot snapShot;
-
- LooseUnpeeled(FileSnapshot snapShot, @NonNull String refName,
- @NonNull ObjectId id) {
- super(LOOSE, refName, id);
- this.snapShot = snapShot;
- }
-
- @Override
- public FileSnapshot getSnapShot() {
- return snapShot;
- }
-
- @NonNull
- @Override
- public ObjectId getObjectId() {
- ObjectId id = super.getObjectId();
- assert id != null; // checked in constructor
- return id;
- }
-
- @Override
- public LooseRef peel(ObjectIdRef newLeaf) {
- ObjectId peeledObjectId = newLeaf.getPeeledObjectId();
- ObjectId objectId = getObjectId();
- if (peeledObjectId != null) {
- return new LoosePeeledTag(snapShot, getName(),
- objectId, peeledObjectId);
- }
- return new LooseNonTag(snapShot, getName(), objectId);
- }
- }
-
- private static final class LooseSymbolicRef extends SymbolicRef implements
- LooseRef {
- private final FileSnapshot snapShot;
-
- LooseSymbolicRef(FileSnapshot snapshot, @NonNull String refName,
- @NonNull Ref target) {
- super(refName, target);
- this.snapShot = snapshot;
- }
-
- @Override
- public FileSnapshot getSnapShot() {
- return snapShot;
- }
-
- @Override
- public LooseRef peel(ObjectIdRef newLeaf) {
- // We should never try to peel the symbolic references.
- throw new UnsupportedOperationException();
- }
- }
- }
|