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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago 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>
14 years ago |
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341 |
- /*
- * Copyright (C) 2007, Dave Watson <dwatson@mimvista.com>
- * Copyright (C) 2008-2010, Google Inc.
- * Copyright (C) 2006-2010, Robin Rosenberg <robin.rosenberg@dewire.com>
- * Copyright (C) 2006-2008, Shawn O. Pearce <spearce@spearce.org>
- * and other copyright owners as documented in the project's IP log.
- *
- * This program and the accompanying materials are made available
- * under the terms of the Eclipse Distribution License v1.0 which
- * accompanies this distribution, is reproduced below, and is
- * available at http://www.eclipse.org/org/documents/edl-v10.php
- *
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or
- * without modification, are permitted provided that the following
- * conditions are met:
- *
- * - Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- *
- * - Redistributions in binary form must reproduce the above
- * copyright notice, this list of conditions and the following
- * disclaimer in the documentation and/or other materials provided
- * with the distribution.
- *
- * - Neither the name of the Eclipse Foundation, Inc. nor the
- * names of its contributors may be used to endorse or promote
- * products derived from this software without specific prior
- * written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
- * CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
- * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
- * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
- * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
- * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
- * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
- * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
- * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
- * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
- * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
- package org.eclipse.jgit.lib;
-
- import java.io.File;
- import java.io.IOException;
- import java.util.ArrayList;
- import java.util.Collection;
- import java.util.Collections;
- import java.util.HashMap;
- import java.util.HashSet;
- import java.util.LinkedList;
- import java.util.List;
- import java.util.Map;
- import java.util.Set;
- import java.util.Vector;
- import java.util.concurrent.atomic.AtomicInteger;
-
- import org.eclipse.jgit.dircache.DirCache;
- import org.eclipse.jgit.errors.ConfigInvalidException;
- import org.eclipse.jgit.errors.IncorrectObjectTypeException;
- import org.eclipse.jgit.errors.RevisionSyntaxException;
- import org.eclipse.jgit.util.FS;
- import org.eclipse.jgit.util.SystemReader;
-
- /**
- * Represents a Git repository. A repository holds all objects and refs used for
- * managing source code (could by any type of file, but source code is what
- * SCM's are typically used for).
- *
- * In Git terms all data is stored in GIT_DIR, typically a directory called
- * .git. A work tree is maintained unless the repository is a bare repository.
- * Typically the .git directory is located at the root of the work dir.
- *
- * <ul>
- * <li>GIT_DIR
- * <ul>
- * <li>objects/ - objects</li>
- * <li>refs/ - tags and heads</li>
- * <li>config - configuration</li>
- * <li>info/ - more configurations</li>
- * </ul>
- * </li>
- * </ul>
- * <p>
- * This class is thread-safe.
- * <p>
- * This implementation only handles a subtly undocumented subset of git features.
- *
- */
- public class Repository {
- private final AtomicInteger useCnt = new AtomicInteger(1);
-
- private final File gitDir;
-
- private final FileBasedConfig userConfig;
-
- private final RepositoryConfig config;
-
- private final RefDatabase refs;
-
- private final ObjectDirectory objectDatabase;
-
- private GitIndex index;
-
- private final List<RepositoryListener> listeners = new Vector<RepositoryListener>(); // thread safe
- static private final List<RepositoryListener> allListeners = new Vector<RepositoryListener>(); // thread safe
-
- private File workDir;
-
- private File indexFile;
-
- /**
- * Construct a representation of a Git repository.
- *
- * The work tree, object directory, alternate object directories and index
- * file locations are deduced from the given git directory and the default
- * rules.
- *
- * @param d
- * GIT_DIR (the location of the repository metadata).
- * @throws IOException
- * the repository appears to already exist but cannot be
- * accessed.
- */
- public Repository(final File d) throws IOException {
- this(d, null, null, null, null); // go figure it out
- }
-
- /**
- * Construct a representation of a Git repository.
- *
- * The work tree, object directory, alternate object directories and index
- * file locations are deduced from the given git directory and the default
- * rules.
- *
- * @param d
- * GIT_DIR (the location of the repository metadata). May be
- * null work workTree is set
- * @param workTree
- * GIT_WORK_TREE (the root of the checkout). May be null for
- * default value.
- * @throws IOException
- * the repository appears to already exist but cannot be
- * accessed.
- */
- public Repository(final File d, final File workTree) throws IOException {
- this(d, workTree, null, null, null); // go figure it out
- }
-
- /**
- * Construct a representation of a Git repository using the given parameters
- * possibly overriding default conventions.
- *
- * @param d
- * GIT_DIR (the location of the repository metadata). May be null
- * for default value in which case it depends on GIT_WORK_TREE.
- * @param workTree
- * GIT_WORK_TREE (the root of the checkout). May be null for
- * default value if GIT_DIR is
- * @param objectDir
- * GIT_OBJECT_DIRECTORY (where objects and are stored). May be
- * null for default value. Relative names ares resolved against
- * GIT_WORK_TREE
- * @param alternateObjectDir
- * GIT_ALTERNATE_OBJECT_DIRECTORIES (where more objects are read
- * from). May be null for default value. Relative names ares
- * resolved against GIT_WORK_TREE
- * @param indexFile
- * GIT_INDEX_FILE (the location of the index file). May be null
- * for default value. Relative names ares resolved against
- * GIT_WORK_TREE.
- * @throws IOException
- * the repository appears to already exist but cannot be
- * accessed.
- */
- public Repository(final File d, final File workTree, final File objectDir,
- final File[] alternateObjectDir, final File indexFile) throws IOException {
-
- if (workTree != null) {
- workDir = workTree;
- if (d == null)
- gitDir = new File(workTree, Constants.DOT_GIT);
- else
- gitDir = d;
- } else {
- if (d != null)
- gitDir = d;
- else
- throw new IllegalArgumentException("Either GIT_DIR or GIT_WORK_TREE must be passed to Repository constructor");
- }
-
- userConfig = SystemReader.getInstance().openUserConfig();
- config = new RepositoryConfig(userConfig, FS.resolve(gitDir, "config"));
-
- loadUserConfig();
- loadConfig();
-
- if (workDir == null) {
- String workTreeConfig = getConfig().getString("core", null, "worktree");
- if (workTreeConfig != null) {
- workDir = FS.resolve(d, workTreeConfig);
- } else {
- workDir = gitDir.getParentFile();
- }
- }
-
- refs = new RefDirectory(this);
- if (objectDir != null)
- objectDatabase = new ObjectDirectory(FS.resolve(objectDir, ""),
- alternateObjectDir);
- else
- objectDatabase = new ObjectDirectory(FS.resolve(gitDir, "objects"),
- alternateObjectDir);
-
- if (indexFile != null)
- this.indexFile = indexFile;
- else
- this.indexFile = new File(gitDir, "index");
-
- if (objectDatabase.exists()) {
- final String repositoryFormatVersion = getConfig().getString(
- "core", null, "repositoryFormatVersion");
- if (!"0".equals(repositoryFormatVersion)) {
- throw new IOException("Unknown repository format \""
- + repositoryFormatVersion + "\"; expected \"0\".");
- }
- }
- }
-
- private void loadUserConfig() throws IOException {
- try {
- userConfig.load();
- } catch (ConfigInvalidException e1) {
- IOException e2 = new IOException("User config file "
- + userConfig.getFile().getAbsolutePath() + " invalid: "
- + e1);
- e2.initCause(e1);
- throw e2;
- }
- }
-
- private void loadConfig() throws IOException {
- try {
- config.load();
- } catch (ConfigInvalidException e1) {
- IOException e2 = new IOException("Unknown repository format");
- e2.initCause(e1);
- throw e2;
- }
- }
-
-
- /**
- * Create a new Git repository initializing the necessary files and
- * directories. Repository with working tree is created using this method.
- *
- * @throws IOException
- * @see #create(boolean)
- */
- public synchronized void create() throws IOException {
- create(false);
- }
-
- /**
- * Create a new Git repository initializing the necessary files and
- * directories.
- *
- * @param bare
- * if true, a bare repository is created.
- *
- * @throws IOException
- * in case of IO problem
- */
- public void create(boolean bare) throws IOException {
- final RepositoryConfig cfg = getConfig();
- if (cfg.getFile().exists()) {
- throw new IllegalStateException("Repository already exists: "
- + gitDir);
- }
- gitDir.mkdirs();
- refs.create();
- objectDatabase.create();
-
- new File(gitDir, "branches").mkdir();
-
- RefUpdate head = updateRef(Constants.HEAD);
- head.disableRefLog();
- head.link(Constants.R_HEADS + Constants.MASTER);
-
- cfg.setInt("core", null, "repositoryformatversion", 0);
- cfg.setBoolean("core", null, "filemode", true);
- if (bare)
- cfg.setBoolean("core", null, "bare", true);
- cfg.setBoolean("core", null, "logallrefupdates", !bare);
- cfg.setBoolean("core", null, "autocrlf", false);
- cfg.save();
- }
-
- /**
- * @return GIT_DIR
- */
- public File getDirectory() {
- return gitDir;
- }
-
- /**
- * @return the directory containing the objects owned by this repository.
- */
- public File getObjectsDirectory() {
- return objectDatabase.getDirectory();
- }
-
- /**
- * @return the object database which stores this repository's data.
- */
- public ObjectDatabase getObjectDatabase() {
- return objectDatabase;
- }
-
- /** @return the reference database which stores the reference namespace. */
- public RefDatabase getRefDatabase() {
- return refs;
- }
-
- /**
- * @return the configuration of this repository
- */
- public RepositoryConfig getConfig() {
- if (userConfig.isOutdated()) {
- try {
- loadUserConfig();
- } catch (IOException e) {
- throw new RuntimeException(e);
- }
- }
- if (config.isOutdated()) {
- try {
- loadConfig();
- } catch (IOException e) {
- throw new RuntimeException(e);
- }
- }
- return config;
- }
-
- /**
- * Construct a filename where the loose object having a specified SHA-1
- * should be stored. If the object is stored in a shared repository the path
- * to the alternative repo will be returned. If the object is not yet store
- * a usable path in this repo will be returned. It is assumed that callers
- * will look for objects in a pack first.
- *
- * @param objectId
- * @return suggested file name
- */
- public File toFile(final AnyObjectId objectId) {
- return objectDatabase.fileFor(objectId);
- }
-
- /**
- * @param objectId
- * @return true if the specified object is stored in this repo or any of the
- * known shared repositories.
- */
- public boolean hasObject(final AnyObjectId objectId) {
- return objectDatabase.hasObject(objectId);
- }
-
- /**
- * @param id
- * SHA-1 of an object.
- *
- * @return a {@link ObjectLoader} for accessing the data of the named
- * object, or null if the object does not exist.
- * @throws IOException
- */
- public ObjectLoader openObject(final AnyObjectId id)
- throws IOException {
- final WindowCursor wc = new WindowCursor();
- try {
- return openObject(wc, id);
- } finally {
- wc.release();
- }
- }
-
- /**
- * @param curs
- * temporary working space associated with the calling thread.
- * @param id
- * SHA-1 of an object.
- *
- * @return a {@link ObjectLoader} for accessing the data of the named
- * object, or null if the object does not exist.
- * @throws IOException
- */
- public ObjectLoader openObject(final WindowCursor curs, final AnyObjectId id)
- throws IOException {
- return objectDatabase.openObject(curs, id);
- }
-
- /**
- * Open object in all packs containing specified object.
- *
- * @param objectId
- * id of object to search for
- * @param curs
- * temporary working space associated with the calling thread.
- * @return collection of loaders for this object, from all packs containing
- * this object
- * @throws IOException
- */
- public Collection<PackedObjectLoader> openObjectInAllPacks(
- final AnyObjectId objectId, final WindowCursor curs)
- throws IOException {
- Collection<PackedObjectLoader> result = new LinkedList<PackedObjectLoader>();
- openObjectInAllPacks(objectId, result, curs);
- return result;
- }
-
- /**
- * Open object in all packs containing specified object.
- *
- * @param objectId
- * id of object to search for
- * @param resultLoaders
- * result collection of loaders for this object, filled with
- * loaders from all packs containing specified object
- * @param curs
- * temporary working space associated with the calling thread.
- * @throws IOException
- */
- void openObjectInAllPacks(final AnyObjectId objectId,
- final Collection<PackedObjectLoader> resultLoaders,
- final WindowCursor curs) throws IOException {
- objectDatabase.openObjectInAllPacks(resultLoaders, curs, objectId);
- }
-
- /**
- * @param id
- * SHA'1 of a blob
- * @return an {@link ObjectLoader} for accessing the data of a named blob
- * @throws IOException
- */
- public ObjectLoader openBlob(final ObjectId id) throws IOException {
- return openObject(id);
- }
-
- /**
- * @param id
- * SHA'1 of a tree
- * @return an {@link ObjectLoader} for accessing the data of a named tree
- * @throws IOException
- */
- public ObjectLoader openTree(final ObjectId id) throws IOException {
- return openObject(id);
- }
-
- /**
- * Access a Commit object using a symbolic reference. This reference may
- * be a SHA-1 or ref in combination with a number of symbols translating
- * from one ref or SHA1-1 to another, such as HEAD^ etc.
- *
- * @param revstr a reference to a git commit object
- * @return a Commit named by the specified string
- * @throws IOException for I/O error or unexpected object type.
- *
- * @see #resolve(String)
- */
- public Commit mapCommit(final String revstr) throws IOException {
- final ObjectId id = resolve(revstr);
- return id != null ? mapCommit(id) : null;
- }
-
- /**
- * Access any type of Git object by id and
- *
- * @param id
- * SHA-1 of object to read
- * @param refName optional, only relevant for simple tags
- * @return The Git object if found or null
- * @throws IOException
- */
- public Object mapObject(final ObjectId id, final String refName) throws IOException {
- final ObjectLoader or = openObject(id);
- if (or == null)
- return null;
- final byte[] raw = or.getBytes();
- switch (or.getType()) {
- case Constants.OBJ_TREE:
- return makeTree(id, raw);
-
- case Constants.OBJ_COMMIT:
- return makeCommit(id, raw);
-
- case Constants.OBJ_TAG:
- return makeTag(id, refName, raw);
-
- case Constants.OBJ_BLOB:
- return raw;
-
- default:
- throw new IncorrectObjectTypeException(id,
- "COMMIT nor TREE nor BLOB nor TAG");
- }
- }
-
- /**
- * Access a Commit by SHA'1 id.
- * @param id
- * @return Commit or null
- * @throws IOException for I/O error or unexpected object type.
- */
- public Commit mapCommit(final ObjectId id) throws IOException {
- final ObjectLoader or = openObject(id);
- if (or == null)
- return null;
- final byte[] raw = or.getBytes();
- if (Constants.OBJ_COMMIT == or.getType())
- return new Commit(this, id, raw);
- throw new IncorrectObjectTypeException(id, Constants.TYPE_COMMIT);
- }
-
- private Commit makeCommit(final ObjectId id, final byte[] raw) {
- Commit ret = new Commit(this, id, raw);
- return ret;
- }
-
- /**
- * Access a Tree object using a symbolic reference. This reference may
- * be a SHA-1 or ref in combination with a number of symbols translating
- * from one ref or SHA1-1 to another, such as HEAD^{tree} etc.
- *
- * @param revstr a reference to a git commit object
- * @return a Tree named by the specified string
- * @throws IOException
- *
- * @see #resolve(String)
- */
- public Tree mapTree(final String revstr) throws IOException {
- final ObjectId id = resolve(revstr);
- return id != null ? mapTree(id) : null;
- }
-
- /**
- * Access a Tree by SHA'1 id.
- * @param id
- * @return Tree or null
- * @throws IOException for I/O error or unexpected object type.
- */
- public Tree mapTree(final ObjectId id) throws IOException {
- final ObjectLoader or = openObject(id);
- if (or == null)
- return null;
- final byte[] raw = or.getBytes();
- switch (or.getType()) {
- case Constants.OBJ_TREE:
- return new Tree(this, id, raw);
-
- case Constants.OBJ_COMMIT:
- return mapTree(ObjectId.fromString(raw, 5));
-
- default:
- throw new IncorrectObjectTypeException(id, Constants.TYPE_TREE);
- }
- }
-
- private Tree makeTree(final ObjectId id, final byte[] raw) throws IOException {
- Tree ret = new Tree(this, id, raw);
- return ret;
- }
-
- private Tag makeTag(final ObjectId id, final String refName, final byte[] raw) {
- Tag ret = new Tag(this, id, refName, raw);
- return ret;
- }
-
- /**
- * Access a tag by symbolic name.
- *
- * @param revstr
- * @return a Tag or null
- * @throws IOException on I/O error or unexpected type
- */
- public Tag mapTag(String revstr) throws IOException {
- final ObjectId id = resolve(revstr);
- return id != null ? mapTag(revstr, id) : null;
- }
-
- /**
- * Access a Tag by SHA'1 id
- * @param refName
- * @param id
- * @return Commit or null
- * @throws IOException for I/O error or unexpected object type.
- */
- public Tag mapTag(final String refName, final ObjectId id) throws IOException {
- final ObjectLoader or = openObject(id);
- if (or == null)
- return null;
- final byte[] raw = or.getBytes();
- if (Constants.OBJ_TAG == or.getType())
- return new Tag(this, id, refName, raw);
- return new Tag(this, id, refName, null);
- }
-
- /**
- * Create a command to update, create or delete a ref in this repository.
- *
- * @param ref
- * name of the ref the caller wants to modify.
- * @return an update command. The caller must finish populating this command
- * and then invoke one of the update methods to actually make a
- * change.
- * @throws IOException
- * a symbolic ref was passed in and could not be resolved back
- * to the base ref, as the symbolic ref could not be read.
- */
- public RefUpdate updateRef(final String ref) throws IOException {
- return updateRef(ref, false);
- }
-
- /**
- * Create a command to update, create or delete a ref in this repository.
- *
- * @param ref
- * name of the ref the caller wants to modify.
- * @param detach
- * true to create a detached head
- * @return an update command. The caller must finish populating this command
- * and then invoke one of the update methods to actually make a
- * change.
- * @throws IOException
- * a symbolic ref was passed in and could not be resolved back
- * to the base ref, as the symbolic ref could not be read.
- */
- public RefUpdate updateRef(final String ref, final boolean detach) throws IOException {
- return refs.newUpdate(ref, detach);
- }
-
- /**
- * Create a command to rename a ref in this repository
- *
- * @param fromRef
- * name of ref to rename from
- * @param toRef
- * name of ref to rename to
- * @return an update command that knows how to rename a branch to another.
- * @throws IOException
- * the rename could not be performed.
- *
- */
- public RefRename renameRef(final String fromRef, final String toRef) throws IOException {
- return refs.newRename(fromRef, toRef);
- }
-
- /**
- * Parse a git revision string and return an object id.
- *
- * Currently supported is combinations of these.
- * <ul>
- * <li>SHA-1 - a SHA-1</li>
- * <li>refs/... - a ref name</li>
- * <li>ref^n - nth parent reference</li>
- * <li>ref~n - distance via parent reference</li>
- * <li>ref@{n} - nth version of ref</li>
- * <li>ref^{tree} - tree references by ref</li>
- * <li>ref^{commit} - commit references by ref</li>
- * </ul>
- *
- * Not supported is
- * <ul>
- * <li>timestamps in reflogs, ref@{full or relative timestamp}</li>
- * <li>abbreviated SHA-1's</li>
- * </ul>
- *
- * @param revstr A git object references expression
- * @return an ObjectId or null if revstr can't be resolved to any ObjectId
- * @throws IOException on serious errors
- */
- public ObjectId resolve(final String revstr) throws IOException {
- char[] rev = revstr.toCharArray();
- Object ref = null;
- ObjectId refId = null;
- for (int i = 0; i < rev.length; ++i) {
- switch (rev[i]) {
- case '^':
- if (refId == null) {
- String refstr = new String(rev,0,i);
- refId = resolveSimple(refstr);
- if (refId == null)
- return null;
- }
- if (i + 1 < rev.length) {
- switch (rev[i + 1]) {
- case '0':
- case '1':
- case '2':
- case '3':
- case '4':
- case '5':
- case '6':
- case '7':
- case '8':
- case '9':
- int j;
- ref = mapObject(refId, null);
- while (ref instanceof Tag) {
- Tag tag = (Tag)ref;
- refId = tag.getObjId();
- ref = mapObject(refId, null);
- }
- if (!(ref instanceof Commit))
- throw new IncorrectObjectTypeException(refId, Constants.TYPE_COMMIT);
- for (j=i+1; j<rev.length; ++j) {
- if (!Character.isDigit(rev[j]))
- break;
- }
- String parentnum = new String(rev, i+1, j-i-1);
- int pnum;
- try {
- pnum = Integer.parseInt(parentnum);
- } catch (NumberFormatException e) {
- throw new RevisionSyntaxException(
- "Invalid commit parent number",
- revstr);
- }
- if (pnum != 0) {
- final ObjectId parents[] = ((Commit) ref)
- .getParentIds();
- if (pnum > parents.length)
- refId = null;
- else
- refId = parents[pnum - 1];
- }
- i = j - 1;
- break;
- case '{':
- int k;
- String item = null;
- for (k=i+2; k<rev.length; ++k) {
- if (rev[k] == '}') {
- item = new String(rev, i+2, k-i-2);
- break;
- }
- }
- i = k;
- if (item != null)
- if (item.equals("tree")) {
- ref = mapObject(refId, null);
- while (ref instanceof Tag) {
- Tag t = (Tag)ref;
- refId = t.getObjId();
- ref = mapObject(refId, null);
- }
- if (ref instanceof Treeish)
- refId = ((Treeish)ref).getTreeId();
- else
- throw new IncorrectObjectTypeException(refId, Constants.TYPE_TREE);
- }
- else if (item.equals("commit")) {
- ref = mapObject(refId, null);
- while (ref instanceof Tag) {
- Tag t = (Tag)ref;
- refId = t.getObjId();
- ref = mapObject(refId, null);
- }
- if (!(ref instanceof Commit))
- throw new IncorrectObjectTypeException(refId, Constants.TYPE_COMMIT);
- }
- else if (item.equals("blob")) {
- ref = mapObject(refId, null);
- while (ref instanceof Tag) {
- Tag t = (Tag)ref;
- refId = t.getObjId();
- ref = mapObject(refId, null);
- }
- if (!(ref instanceof byte[]))
- throw new IncorrectObjectTypeException(refId, Constants.TYPE_BLOB);
- }
- else if (item.equals("")) {
- ref = mapObject(refId, null);
- while (ref instanceof Tag) {
- Tag t = (Tag)ref;
- refId = t.getObjId();
- ref = mapObject(refId, null);
- }
- }
- else
- throw new RevisionSyntaxException(revstr);
- else
- throw new RevisionSyntaxException(revstr);
- break;
- default:
- ref = mapObject(refId, null);
- if (ref instanceof Commit) {
- final ObjectId parents[] = ((Commit) ref)
- .getParentIds();
- if (parents.length == 0)
- refId = null;
- else
- refId = parents[0];
- } else
- throw new IncorrectObjectTypeException(refId, Constants.TYPE_COMMIT);
-
- }
- } else {
- ref = mapObject(refId, null);
- while (ref instanceof Tag) {
- Tag tag = (Tag)ref;
- refId = tag.getObjId();
- ref = mapObject(refId, null);
- }
- if (ref instanceof Commit) {
- final ObjectId parents[] = ((Commit) ref)
- .getParentIds();
- if (parents.length == 0)
- refId = null;
- else
- refId = parents[0];
- } else
- throw new IncorrectObjectTypeException(refId, Constants.TYPE_COMMIT);
- }
- break;
- case '~':
- if (ref == null) {
- String refstr = new String(rev,0,i);
- refId = resolveSimple(refstr);
- if (refId == null)
- return null;
- ref = mapObject(refId, null);
- }
- while (ref instanceof Tag) {
- Tag tag = (Tag)ref;
- refId = tag.getObjId();
- ref = mapObject(refId, null);
- }
- if (!(ref instanceof Commit))
- throw new IncorrectObjectTypeException(refId, Constants.TYPE_COMMIT);
- int l;
- for (l = i + 1; l < rev.length; ++l) {
- if (!Character.isDigit(rev[l]))
- break;
- }
- String distnum = new String(rev, i+1, l-i-1);
- int dist;
- try {
- dist = Integer.parseInt(distnum);
- } catch (NumberFormatException e) {
- throw new RevisionSyntaxException(
- "Invalid ancestry length", revstr);
- }
- while (dist > 0) {
- final ObjectId[] parents = ((Commit) ref).getParentIds();
- if (parents.length == 0) {
- refId = null;
- break;
- }
- refId = parents[0];
- ref = mapCommit(refId);
- --dist;
- }
- i = l - 1;
- break;
- case '@':
- int m;
- String time = null;
- for (m=i+2; m<rev.length; ++m) {
- if (rev[m] == '}') {
- time = new String(rev, i+2, m-i-2);
- break;
- }
- }
- if (time != null)
- throw new RevisionSyntaxException("reflogs not yet supported by revision parser", revstr);
- i = m - 1;
- break;
- default:
- if (refId != null)
- throw new RevisionSyntaxException(revstr);
- }
- }
- if (refId == null)
- refId = resolveSimple(revstr);
- return refId;
- }
-
- private ObjectId resolveSimple(final String revstr) throws IOException {
- if (ObjectId.isId(revstr))
- return ObjectId.fromString(revstr);
- final Ref r = refs.getRef(revstr);
- return r != null ? r.getObjectId() : null;
- }
-
- /** Increment the use counter by one, requiring a matched {@link #close()}. */
- public void incrementOpen() {
- useCnt.incrementAndGet();
- }
-
- /**
- * Close all resources used by this repository
- */
- public void close() {
- if (useCnt.decrementAndGet() == 0) {
- objectDatabase.close();
- refs.close();
- }
- }
-
- /**
- * Add a single existing pack to the list of available pack files.
- *
- * @param pack
- * path of the pack file to open.
- * @param idx
- * path of the corresponding index file.
- * @throws IOException
- * index file could not be opened, read, or is not recognized as
- * a Git pack file index.
- */
- public void openPack(final File pack, final File idx) throws IOException {
- objectDatabase.openPack(pack, idx);
- }
-
- public String toString() {
- return "Repository[" + getDirectory() + "]";
- }
-
- /**
- * Get the name of the reference that {@code HEAD} points to.
- * <p>
- * This is essentially the same as doing:
- *
- * <pre>
- * return getRef(Constants.HEAD).getTarget().getName()
- * </pre>
- *
- * Except when HEAD is detached, in which case this method returns the
- * current ObjectId in hexadecimal string format.
- *
- * @return name of current branch (for example {@code refs/heads/master}) or
- * an ObjectId in hex format if the current branch is detached.
- * @throws IOException
- */
- public String getFullBranch() throws IOException {
- Ref head = getRef(Constants.HEAD);
- if (head == null)
- return null;
- if (head.isSymbolic())
- return head.getTarget().getName();
- if (head.getObjectId() != null)
- return head.getObjectId().name();
- return null;
- }
-
- /**
- * Get the short name of the current branch that {@code HEAD} points to.
- * <p>
- * This is essentially the same as {@link #getFullBranch()}, except the
- * leading prefix {@code refs/heads/} is removed from the reference before
- * it is returned to the caller.
- *
- * @return name of current branch (for example {@code master}), or an
- * ObjectId in hex format if the current branch is detached.
- * @throws IOException
- */
- public String getBranch() throws IOException {
- String name = getFullBranch();
- if (name != null)
- return shortenRefName(name);
- return name;
- }
-
- /**
- * Get a ref by name.
- *
- * @param name
- * the name of the ref to lookup. May be a short-hand form, e.g.
- * "master" which is is automatically expanded to
- * "refs/heads/master" if "refs/heads/master" already exists.
- * @return the Ref with the given name, or null if it does not exist
- * @throws IOException
- */
- public Ref getRef(final String name) throws IOException {
- return refs.getRef(name);
- }
-
- /**
- * @return mutable map of all known refs (heads, tags, remotes).
- */
- public Map<String, Ref> getAllRefs() {
- try {
- return refs.getRefs(RefDatabase.ALL);
- } catch (IOException e) {
- return new HashMap<String, Ref>();
- }
- }
-
- /**
- * @return mutable map of all tags; key is short tag name ("v1.0") and value
- * of the entry contains the ref with the full tag name
- * ("refs/tags/v1.0").
- */
- public Map<String, Ref> getTags() {
- try {
- return refs.getRefs(Constants.R_TAGS);
- } catch (IOException e) {
- return new HashMap<String, Ref>();
- }
- }
-
- /**
- * Peel a possibly unpeeled reference to an annotated tag.
- * <p>
- * If the ref cannot be peeled (as it does not refer to an annotated tag)
- * the peeled id stays null, but {@link Ref#isPeeled()} will be true.
- *
- * @param ref
- * The ref to peel
- * @return <code>ref</code> if <code>ref.isPeeled()</code> is true; else a
- * new Ref object representing the same data as Ref, but isPeeled()
- * will be true and getPeeledObjectId will contain the peeled object
- * (or null).
- */
- public Ref peel(final Ref ref) {
- try {
- return refs.peel(ref);
- } catch (IOException e) {
- // Historical accident; if the reference cannot be peeled due
- // to some sort of repository access problem we claim that the
- // same as if the reference was not an annotated tag.
- return ref;
- }
- }
-
- /**
- * @return a map with all objects referenced by a peeled ref.
- */
- public Map<AnyObjectId, Set<Ref>> getAllRefsByPeeledObjectId() {
- Map<String, Ref> allRefs = getAllRefs();
- Map<AnyObjectId, Set<Ref>> ret = new HashMap<AnyObjectId, Set<Ref>>(allRefs.size());
- for (Ref ref : allRefs.values()) {
- ref = peel(ref);
- AnyObjectId target = ref.getPeeledObjectId();
- if (target == null)
- target = ref.getObjectId();
- // We assume most Sets here are singletons
- Set<Ref> oset = ret.put(target, Collections.singleton(ref));
- if (oset != null) {
- // that was not the case (rare)
- if (oset.size() == 1) {
- // Was a read-only singleton, we must copy to a new Set
- oset = new HashSet<Ref>(oset);
- }
- ret.put(target, oset);
- oset.add(ref);
- }
- }
- return ret;
- }
-
- /**
- * @return a representation of the index associated with this repo
- * @throws IOException
- */
- public GitIndex getIndex() throws IOException {
- if (index == null) {
- index = new GitIndex(this);
- index.read();
- } else {
- index.rereadIfNecessary();
- }
- return index;
- }
-
- /**
- * @return the index file location
- */
- public File getIndexFile() {
- return indexFile;
- }
-
- static byte[] gitInternalSlash(byte[] bytes) {
- if (File.separatorChar == '/')
- return bytes;
- for (int i=0; i<bytes.length; ++i)
- if (bytes[i] == File.separatorChar)
- bytes[i] = '/';
- return bytes;
- }
-
- /**
- * @return an important state
- */
- public RepositoryState getRepositoryState() {
- // Pre Git-1.6 logic
- if (new File(getWorkDir(), ".dotest").exists())
- return RepositoryState.REBASING;
- if (new File(gitDir,".dotest-merge").exists())
- return RepositoryState.REBASING_INTERACTIVE;
-
- // From 1.6 onwards
- if (new File(getDirectory(),"rebase-apply/rebasing").exists())
- return RepositoryState.REBASING_REBASING;
- if (new File(getDirectory(),"rebase-apply/applying").exists())
- return RepositoryState.APPLY;
- if (new File(getDirectory(),"rebase-apply").exists())
- return RepositoryState.REBASING;
-
- if (new File(getDirectory(),"rebase-merge/interactive").exists())
- return RepositoryState.REBASING_INTERACTIVE;
- if (new File(getDirectory(),"rebase-merge").exists())
- return RepositoryState.REBASING_MERGE;
-
- // Both versions
- if (new File(gitDir, "MERGE_HEAD").exists()) {
- // we are merging - now check whether we have unmerged paths
- try {
- if (!DirCache.read(this).hasUnmergedPaths()) {
- // no unmerged paths -> return the MERGING_RESOLVED state
- return RepositoryState.MERGING_RESOLVED;
- }
- } catch (IOException e) {
- // Can't decide whether unmerged paths exists. Return
- // MERGING state to be on the safe side (in state MERGING
- // you are not allow to do anything)
- e.printStackTrace();
- }
- return RepositoryState.MERGING;
- }
-
- if (new File(gitDir,"BISECT_LOG").exists())
- return RepositoryState.BISECTING;
-
- return RepositoryState.SAFE;
- }
-
- /**
- * Check validity of a ref name. It must not contain character that has
- * a special meaning in a Git object reference expression. Some other
- * dangerous characters are also excluded.
- *
- * For portability reasons '\' is excluded
- *
- * @param refName
- *
- * @return true if refName is a valid ref name
- */
- public static boolean isValidRefName(final String refName) {
- final int len = refName.length();
- if (len == 0)
- return false;
- if (refName.endsWith(LockFile.SUFFIX))
- return false;
-
- int components = 1;
- char p = '\0';
- for (int i = 0; i < len; i++) {
- final char c = refName.charAt(i);
- if (c <= ' ')
- return false;
- switch (c) {
- case '.':
- switch (p) {
- case '\0': case '/': case '.':
- return false;
- }
- if (i == len -1)
- return false;
- break;
- case '/':
- if (i == 0 || i == len - 1)
- return false;
- components++;
- break;
- case '{':
- if (p == '@')
- return false;
- break;
- case '~': case '^': case ':':
- case '?': case '[': case '*':
- case '\\':
- return false;
- }
- p = c;
- }
- return components > 1;
- }
-
- /**
- * Strip work dir and return normalized repository path.
- *
- * @param workDir Work dir
- * @param file File whose path shall be stripped of its workdir
- * @return normalized repository relative path or the empty
- * string if the file is not relative to the work directory.
- */
- public static String stripWorkDir(File workDir, File file) {
- final String filePath = file.getPath();
- final String workDirPath = workDir.getPath();
-
- if (filePath.length() <= workDirPath.length() ||
- filePath.charAt(workDirPath.length()) != File.separatorChar ||
- !filePath.startsWith(workDirPath)) {
- File absWd = workDir.isAbsolute() ? workDir : workDir.getAbsoluteFile();
- File absFile = file.isAbsolute() ? file : file.getAbsoluteFile();
- if (absWd == workDir && absFile == file)
- return "";
- return stripWorkDir(absWd, absFile);
- }
-
- String relName = filePath.substring(workDirPath.length() + 1);
- if (File.separatorChar != '/')
- relName = relName.replace(File.separatorChar, '/');
- return relName;
- }
-
- /**
- * @return the workdir file, i.e. where the files are checked out
- */
- public File getWorkDir() {
- return workDir;
- }
-
- /**
- * Override default workdir
- *
- * @param workTree
- * the work tree directory
- */
- public void setWorkDir(File workTree) {
- this.workDir = workTree;
- }
-
- /**
- * Register a {@link RepositoryListener} which will be notified
- * when ref changes are detected.
- *
- * @param l
- */
- public void addRepositoryChangedListener(final RepositoryListener l) {
- listeners.add(l);
- }
-
- /**
- * Remove a registered {@link RepositoryListener}
- * @param l
- */
- public void removeRepositoryChangedListener(final RepositoryListener l) {
- listeners.remove(l);
- }
-
- /**
- * Register a global {@link RepositoryListener} which will be notified
- * when a ref changes in any repository are detected.
- *
- * @param l
- */
- public static void addAnyRepositoryChangedListener(final RepositoryListener l) {
- allListeners.add(l);
- }
-
- /**
- * Remove a globally registered {@link RepositoryListener}
- * @param l
- */
- public static void removeAnyRepositoryChangedListener(final RepositoryListener l) {
- allListeners.remove(l);
- }
-
- void fireRefsChanged() {
- final RefsChangedEvent event = new RefsChangedEvent(this);
- List<RepositoryListener> all;
- synchronized (listeners) {
- all = new ArrayList<RepositoryListener>(listeners);
- }
- synchronized (allListeners) {
- all.addAll(allListeners);
- }
- for (final RepositoryListener l : all) {
- l.refsChanged(event);
- }
- }
-
- void fireIndexChanged() {
- final IndexChangedEvent event = new IndexChangedEvent(this);
- List<RepositoryListener> all;
- synchronized (listeners) {
- all = new ArrayList<RepositoryListener>(listeners);
- }
- synchronized (allListeners) {
- all.addAll(allListeners);
- }
- for (final RepositoryListener l : all) {
- l.indexChanged(event);
- }
- }
-
- /**
- * Force a scan for changed refs.
- *
- * @throws IOException
- */
- public void scanForRepoChanges() throws IOException {
- getAllRefs(); // This will look for changes to refs
- getIndex(); // This will detect changes in the index
- }
-
- /**
- * @param refName
- *
- * @return a more user friendly ref name
- */
- public String shortenRefName(String refName) {
- if (refName.startsWith(Constants.R_HEADS))
- return refName.substring(Constants.R_HEADS.length());
- if (refName.startsWith(Constants.R_TAGS))
- return refName.substring(Constants.R_TAGS.length());
- if (refName.startsWith(Constants.R_REMOTES))
- return refName.substring(Constants.R_REMOTES.length());
- return refName;
- }
-
- /**
- * @param refName
- * @return a {@link ReflogReader} for the supplied refname, or null if the
- * named ref does not exist.
- * @throws IOException the ref could not be accessed.
- */
- public ReflogReader getReflogReader(String refName) throws IOException {
- Ref ref = getRef(refName);
- if (ref != null)
- return new ReflogReader(this, ref.getName());
- return null;
- }
- }
|