Re: [PATCH] patch-ids.c: cache patch IDs in a notes tree

From: John Keeping <john@keeping.me.uk>
Date: May 13, 2013, 15:52 UTC
Message-ID: <20130513155241.GS2299@serenity.lan>
In-Reply-To: <7vd2suopda.fsf@alter.siamese.dyndns.org>

On Mon, May 13, 2013 at 08:45:21AM -0700, Junio C Hamano wrote:

Show 10 quoted lines

> John Keeping <john@keeping.me.uk> writes:
> 
> > This has the advantage that you get the benefit of the cache if you run
> > "git log --cherry-mark" with the same paths more than once.  In my
> > testing the cache is beneficial as soon as you examine more than one
> > similar range (e.g. master...feature-A and then master...feature-B).
> 
> OK, so perhaps the notes that are keyed with commit ID will record
> multiple entries, one for each invocation pattern (i.e. all pathspec
> given, possibly with nonstandard options)?

That would be possible, but I didn't do it in the current version of the patch.

> "git diff -- t Documentation" and "git diff -- Docu\* t", even
> though they use different pathspec, would produce the same diff;
> instead of pathspec you may need to key with the actual list of
> paths in the patch, though.

Maybe, but I think that would be overkill.

I'm interested to see how much of a benefit we could get by not calculating the patch ID of any commits on the larger side of a symmetric range that touch paths outside the set touched by the smaller side. (revision.c::cherry_pick_list() remembers patch IDs for the smaller side of a symmetric range and then checks if anything on the larger side matches so this fits in naturally.)

In my usage I generally compare a relatively small topic branch against whatever has happened in some upstream branch so I think this could give a big improvement but I haven't had time to try it yet.