threads / patch / 49548

patchsubtree: performance improvement for finding unexpected parent commits

Subject: [PATCH] subtree: performance improvement for finding unexpected parent commits

## tl;dr

3 messages between Oct 12, 2018 and Oct 12, 2018. Diffs are folded; open one to read it.

replies: 2people: 3as markdown or json

Roger Strain· Oct 12, 2018, 13:52 UTC · lore

After testing a previous patch at larger scale, a performance issue was detected when using git show to locate parent revisions, with a single run of the git show command taking 2 seconds or longer in a complex repo. When the command is required tens or hundreds of times in a run of the script, the additional wait time is unaccepatable. Replacing the command with git rev-parse resulted in significantly increased performance, with the command in question returning instantly.

Signed-off-by: Roger Strain <rstrain@swri.org>
Thanks-to: Junio C Hamano <gitster@pobox.com>
---
 contrib/subtree/git-subtree.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Show changes to contrib/subtree/git-subtree.sh +1 −1

diff --git a/contrib/subtree/git-subtree.sh b/contrib/subtree/git-subtree.sh
index 1c157dbd9..147201dc6 100755
--- a/contrib/subtree/git-subtree.sh
+++ b/contrib/subtree/git-subtree.sh
@@ -633,7 +633,7 @@ process_split_commit () {
 	else
 		# processing commit without normal parent information;
 		# fetch from repo
-		parents=$(git show -s --pretty=%P "$rev")
+		parents=$(git rev-parse "$rev^@")
 		extracount=$(($extracount + 1))
 	fi

-- 
2.19.1

Strain, Roger L.· Oct 12, 2018, 13:55 UTC · re: Roger Strain · lore

RE: [PATCH] subtree: performance improvement for finding unexpected parent commits

Original failed to include the note that this patch is for the 'next' branch, my apologies.

Show 37 quoted lines

> -----Original Message-----
> From: Roger Strain <rstrain@swri.org>
> Sent: Friday, October 12, 2018 8:52 AM
> To: git@vger.kernel.org
> Cc: Strain, Roger L. <roger.strain@swri.org>
> Subject: [PATCH] subtree: performance improvement for finding
> unexpected parent commits
> 
> After testing a previous patch at larger scale, a performance issue was
> detected when using git show to locate parent revisions, with a single run of
> the git show command taking 2 seconds or longer in a complex repo.
> When the command is required tens or hundreds of times in a run of the
> script, the additional wait time is unaccepatable. Replacing the command
> with git rev-parse resulted in significantly increased performance, with the
> command in question returning instantly.
> 
> Signed-off-by: Roger Strain <rstrain@swri.org>
> Thanks-to: Junio C Hamano <gitster@pobox.com>
> ---
>  contrib/subtree/git-subtree.sh | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/contrib/subtree/git-subtree.sh b/contrib/subtree/git-subtree.sh
> index 1c157dbd9..147201dc6 100755
> --- a/contrib/subtree/git-subtree.sh
> +++ b/contrib/subtree/git-subtree.sh
> @@ -633,7 +633,7 @@ process_split_commit () {
>  	else
>  		# processing commit without normal parent information;
>  		# fetch from repo
> -		parents=$(git show -s --pretty=%P "$rev")
> +		parents=$(git rev-parse "$rev^@")
>  		extracount=$(($extracount + 1))
>  	fi
> 
> --
> 2.19.1

Junio C Hamano· Oct 12, 2018, 14:22 UTC · re: Roger Strain · lore

Re: [PATCH] subtree: performance improvement for finding unexpected parent commits

Roger Strain <rstrain@swri.org> writes:

Show 10 quoted lines

> After testing a previous patch at larger scale, a performance issue was
> detected when using git show to locate parent revisions, with a single
> run of the git show command taking 2 seconds or longer in a complex repo.
> When the command is required tens or hundreds of times in a run of the
> script, the additional wait time is unaccepatable. Replacing the command
> with git rev-parse resulted in significantly increased performance, with
> the command in question returning instantly.
>
> Signed-off-by: Roger Strain <rstrain@swri.org>
> Thanks-to: Junio C Hamano <gitster@pobox.com>

That usually is spelled as "Helped-by:".

Will queue.  Thanks.

I still find it disturbing not to know why "show -s --format=..." takes measurable time, though. "-s" means "we do not need any diff output", so it ought to be comparable to "git cat-file commit $rev" with some formatting, but apparently your repository is making Git spend a lot more than that. Puzzled...

Show 16 quoted lines

> ---
>  contrib/subtree/git-subtree.sh | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/contrib/subtree/git-subtree.sh b/contrib/subtree/git-subtree.sh
> index 1c157dbd9..147201dc6 100755
> --- a/contrib/subtree/git-subtree.sh
> +++ b/contrib/subtree/git-subtree.sh
> @@ -633,7 +633,7 @@ process_split_commit () {
>  	else
>  		# processing commit without normal parent information;
>  		# fetch from repo
> -		parents=$(git show -s --pretty=%P "$rev")
> +		parents=$(git rev-parse "$rev^@")
>  		extracount=$(($extracount + 1))
>  	fi