diff options
| author | David Symonds <dsymonds@golang.org> | 2019-11-21 09:51:39 +1100 |
|---|---|---|
| committer | David Symonds <dsymonds@golang.org> | 2020-03-10 09:57:31 +1100 |
| commit | d3dd84b7108ed4a062dced6e1d2d5dd4537c9e57 (patch) | |
| tree | 53c87602397a3aef815393294c9322e080d13a51 | |
| parent | 18a858b38ca4197987a6da7058ab0397255fa82c (diff) | |
| download | go-git-d3dd84b7108ed4a062dced6e1d2d5dd4537c9e57.tar.gz | |
plumbing/object: avoid O(N^2) string building when decoding commit message
Most commits have relatively small messages, so this was never
noticeable. However, there are some repositories that have
semi-automated messages that can get very large (e.g.
github.com/riscv/riscv-clang and its riscv-trunk branch), on the order
of 109k lines. Changing from string += to using a bytes.Buffer reduces
the time for Commit.Decode for that specific case from 35s to 74ms.
Signed-off-by: David Symonds <dsymonds@golang.org>
| -rw-r--r-- | plumbing/object/commit.go | 7 |
1 files changed, 5 insertions, 2 deletions
diff --git a/plumbing/object/commit.go b/plumbing/object/commit.go index 6b50934..eb86a01 100644 --- a/plumbing/object/commit.go +++ b/plumbing/object/commit.go @@ -177,6 +177,7 @@ func (c *Commit) Decode(o plumbing.EncodedObject) (err error) { var message bool var pgpsig bool + var msgbuf bytes.Buffer for { line, err := r.ReadBytes('\n') if err != nil && err != io.EOF { @@ -221,13 +222,15 @@ func (c *Commit) Decode(o plumbing.EncodedObject) (err error) { pgpsig = true } } else { - c.Message += string(line) + msgbuf.Write(line) } if err == io.EOF { - return nil + break } } + c.Message = msgbuf.String() + return nil } // Encode transforms a Commit into a plumbing.EncodedObject. |