git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
From: Jeff King <peff@peff.net>
To: tytso@mit.edu
Cc: Avery Pennarun <apenwarr@gmail.com>, git@vger.kernel.org
Subject: Re: Why is "git tag --contains" so slow?
Date: Thu, 1 Jul 2010 11:03:31 -0400	[thread overview]
Message-ID: <20100701150331.GA12851@sigill.intra.peff.net> (raw)
In-Reply-To: <20100701121711.GF1333@thunk.org>

On Thu, Jul 01, 2010 at 08:17:11AM -0400, tytso@mit.edu wrote:

> Yeah, I'm not talking about the tag and branch names that show up in
> the top gitk box (and which you can also get via git log --annotate).
> I'm specifically talking about what you get with:
> 
> 	git tags --contains <commit-id>

I think there is room for improving the speed of "git tag --contains".
Consider these two similar operations:

  $ time git name-rev HEAD~200
  HEAD~200 tags/v1.7.1-rc2~2

  real    0m0.060s
  user    0m0.048s
  sys     0m0.012s

  $ time git tag --contains HEAD~200
  v1.7.1
  v1.7.1-rc2
  v1.7.1.1
  v1.7.2-rc0
  v1.7.2-rc1

  real    0m2.179s
  user    0m2.156s
  sys     0m0.020s

Obviously the latter is producing the full list of tags, but both should
have to do a traversal starting at each tag ref.

It looks like "git tag --contains" is implemented as something like:

  for ref in `git tag -l`; do
    if `git merge-base $ref $commit | grep $commit`; then
      echo $ref
    fi
  done

except in C. But you can see we'll do a full merge-base calculation for
each tag. And that merge-base calculation may potentially go quite far
back in history.

The name-rev code has a simple heuristic that it stops traversing when
the timestamp of the traversed commit gets more than a day behind that
of $commit. This doesn't do well with clock skew, but can save a lot of
time in practice. We should perhaps consider something like that here.

I suspect we could do even better by sharing information between
traversals. That is, walk the graph from each ref, but retain marks on
commits between each walk. For a commit c, if walking down each parent
gets us to a root without hitting $commit, then we can mark "c" as not
possibly containing the commit. We can cut short any walk when we hit a
"c" that is already marked. So by traversing from "v1.7.1", if we find
that we do not contain $commit, we will have already marked v1.7.0,
v1.6.*, etc, and will not need to traverse them again.

When you do find one that contains $commit, you can also mark it as
"definitely contains", and you can stop a traversal that definitely
contains. So if we see something is in v1.7.0, then the traversal from
v1.7.1 only needs to go as far back as v1.7.0.

-Peff

  reply	other threads:[~2010-07-01 15:04 UTC|newest]

Thread overview: 46+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-07-01  0:54 Why is "git tag --contains" so slow? Theodore Ts'o
2010-07-01  0:58 ` Shawn O. Pearce
2010-07-03 23:27   ` Sam Vilain
2010-07-01  1:00 ` Avery Pennarun
2010-07-01 12:17   ` tytso
2010-07-01 15:03     ` Jeff King [this message]
2010-07-01 15:38       ` Jeff King
2010-07-02 19:26         ` tytso
2010-07-03  8:06           ` Jeff King
2010-07-04  0:55             ` tytso
2010-07-05 12:27               ` Jeff King
2010-07-05 12:33                 ` [RFC/PATCH 1/4] tag: speed up --contains calculation Jeff King
2010-10-13 22:07                   ` Jonathan Nieder
2010-10-13 22:56                   ` Clemens Buchacher
2011-02-23 15:51                   ` Ævar Arnfjörð Bjarmason
2011-02-23 16:39                     ` Jeff King
2010-07-05 12:34                 ` [RFC/PATCH 2/4] limit "contains" traversals based on commit timestamp Jeff King
2010-10-13 23:21                   ` Jonathan Nieder
2010-07-05 12:35                 ` [RFC/PATCH 3/4] default core.clockskew variable to one day Jeff King
2010-07-05 12:36                 ` [RFC/PATCH 4/4] name-rev: respect core.clockskew Jeff King
2010-07-05 12:39                 ` Why is "git tag --contains" so slow? Jeff King
2010-10-14 18:59                   ` Jonathan Nieder
2010-10-16 14:32                     ` Clemens Buchacher
2010-10-27 17:11                       ` Jeff King
2010-10-28  8:07                         ` Clemens Buchacher
2010-07-05 14:10                 ` tytso
2010-07-06 11:58                   ` Jeff King
2010-07-06 15:31                     ` Will Palmer
2010-07-06 16:53                       ` tytso
2010-07-08 11:28                         ` Jeff King
2010-07-08 13:21                           ` Will Palmer
2010-07-08 13:54                             ` tytso
2010-07-07 17:45                       ` Jeff King
2010-07-08 10:29                         ` Theodore Tso
2010-07-08 11:12                           ` Jakub Narebski
2010-07-08 19:29                             ` Nicolas Pitre
2010-07-08 19:39                               ` Avery Pennarun
2010-07-08 20:13                                 ` Nicolas Pitre
2010-07-08 21:20                                   ` Jakub Narebski
2010-07-08 21:30                                     ` Sverre Rabbelier
2010-07-08 23:10                                       ` Nicolas Pitre
2010-07-08 23:15                                     ` Nicolas Pitre
2010-07-08 11:31                           ` Jeff King
2010-07-08 14:35                           ` Johan Herland
2010-07-08 19:06                           ` Nicolas Pitre
2010-07-07 17:50                       ` Jeff King

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: http://vger.kernel.org/majordomo-info.html

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20100701150331.GA12851@sigill.intra.peff.net \
    --to=peff@peff.net \
    --cc=apenwarr@gmail.com \
    --cc=git@vger.kernel.org \
    --cc=tytso@mit.edu \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this public inbox

	https://80x24.org/mirrors/git.git

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).