Improve listing performance by using go-git (#6478)

* Use go-git for tree reading and commit info lookup.

Signed-off-by: Filip Navara <navara@emclient.com>

* Use TreeEntry.IsRegular() instead of ObjectType that was removed.

Signed-off-by: Filip Navara <navara@emclient.com>

* Use the treePath to optimize commit info search.

Signed-off-by: Filip Navara <navara@emclient.com>

* Extract the latest commit at treePath along with the other commits.

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix listing commit info for a directory that was created in one commit and never modified after.

Signed-off-by: Filip Navara <navara@emclient.com>

* Avoid nearly all external 'git' invocations when doing directory listing (.editorconfig code path is still hit).

Signed-off-by: Filip Navara <navara@emclient.com>

* Use go-git for reading blobs.

Signed-off-by: Filip Navara <navara@emclient.com>

* Make SHA1 type alias for plumbing.Hash in go-git.

Signed-off-by: Filip Navara <navara@emclient.com>

* Make Signature type alias for object.Signature in go-git.

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix GetCommitsInfo for repository with only one commit.

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix PGP signature verification.

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix issues with walking commit graph across merges.

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix typo in condition.

Signed-off-by: Filip Navara <navara@emclient.com>

* Speed up loading branch list by keeping the repository reference (and thus all the loaded packfile indexes).

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix lising submodules.

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix build

Signed-off-by: Filip Navara <navara@emclient.com>

* Add back commit cache because of name-rev

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix tests

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix code style

* Fix spelling

* Address PR feedback

Signed-off-by: Filip Navara <navara@emclient.com>

* Update vendor module list

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix getting trees by commit id

Signed-off-by: Filip Navara <navara@emclient.com>

* Fix remaining unit test failures

* Fix GetTreeBySHA

* Avoid running `git name-rev` if not necessary

Signed-off-by: Filip Navara <navara@emclient.com>

* Move Branch code to git module

* Clean up GPG signature verification and fix it for tagged commits

* Address PR feedback (import formatting, copyright headers)

* Make blob lookup by SHA working

* Update tests to use public API

* Allow getting content from any type of object through the blob interface

* Change test to actually expect the object content that is in the GIT repository

* Change one more test to actually expect the object content that is in the GIT repository

* Add comments

1 files changed, 15 insertions, 51 deletions

diff --git a/modules/git/blob.go b/modules/git/blob.go
index e194b973db..171b4a1010 100644
--- a/modules/git/blob.go
+++ b/modules/git/blob.go
@@ -1,76 +1,40 @@
 // Copyright 2015 The Gogs Authors. All rights reserved.
+// Copyright 2019 The Gitea Authors. All rights reserved.
 // Use of this source code is governed by a MIT-style
 // license that can be found in the LICENSE file.
 
 package git
 
 import (
-	"bytes"
 	"encoding/base64"
-	"fmt"
 	"io"
 	"io/ioutil"
-	"os"
-	"os/exec"
+
+	"gopkg.in/src-d/go-git.v4/plumbing"
 )
 
 // Blob represents a Git object.
 type Blob struct {
-	repo *Repository
-	*TreeEntry
-}
-
-// Data gets content of blob all at once and wrap it as io.Reader.
-// This can be very slow and memory consuming for huge content.
-func (b *Blob) Data() (io.Reader, error) {
-	stdout := new(bytes.Buffer)
-	stderr := new(bytes.Buffer)
-
-	// Preallocate memory to save ~50% memory usage on big files.
-	stdout.Grow(int(b.Size() + 2048))
-
-	if err := b.DataPipeline(stdout, stderr); err != nil {
-		return nil, concatenateError(err, stderr.String())
-	}
-	return stdout, nil
-}
+	ID SHA1
 
-// DataPipeline gets content of blob and write the result or error to stdout or stderr
-func (b *Blob) DataPipeline(stdout, stderr io.Writer) error {
-	return NewCommand("show", b.ID.String()).RunInDirPipeline(b.repo.Path, stdout, stderr)
-}
-
-type cmdReadCloser struct {
-	cmd    *exec.Cmd
-	stdout io.Reader
-}
-
-func (c cmdReadCloser) Read(p []byte) (int, error) {
-	return c.stdout.Read(p)
-}
-
-func (c cmdReadCloser) Close() error {
-	io.Copy(ioutil.Discard, c.stdout)
-	return c.cmd.Wait()
+	gogitEncodedObj plumbing.EncodedObject
+	name            string
 }
 
 // DataAsync gets a ReadCloser for the contents of a blob without reading it all.
 // Calling the Close function on the result will discard all unread output.
 func (b *Blob) DataAsync() (io.ReadCloser, error) {
-	cmd := exec.Command("git", "show", b.ID.String())
-	cmd.Dir = b.repo.Path
-	cmd.Stderr = os.Stderr
-
-	stdout, err := cmd.StdoutPipe()
-	if err != nil {
-		return nil, fmt.Errorf("StdoutPipe: %v", err)
-	}
+	return b.gogitEncodedObj.Reader()
+}
 
-	if err = cmd.Start(); err != nil {
-		return nil, fmt.Errorf("Start: %v", err)
-	}
+// Size returns the uncompressed size of the blob
+func (b *Blob) Size() int64 {
+	return b.gogitEncodedObj.Size()
+}
 
-	return cmdReadCloser{stdout: stdout, cmd: cmd}, nil
+// Name returns name of the tree entry this blob object was created from (or empty string)
+func (b *Blob) Name() string {
+	return b.name
 }
 
 // GetBlobContentBase64 Reads the content of the blob with a base64 encode and returns the encoded string