177

How can I make git log only show commits that changed files other than the ones I specify?

With git log, I can filter the commits I see to those that touch a given set of paths. What I want is to invert that filter so that only commits that touch paths other than the specified ones will be listed.

I can get what I want with

git log --format="%n/%n%H" --name-only | ~/filter-log.pl | git log --stdin --no-walk

where filter-log.pl is:

#!/usr/bin/perl
use strict;
use warnings;

$/ = "\n/\n";
<>;

while (<>) {
    my ($commit, @files) = split /\n/, $_;

    if (grep { $_ && $_ !~ m[^(/$|.etckeeper$|lvm/(archive|backup)/)] } @files) {
        print "$commit\n";
    }
}

except I want something somewhat more elegant than that.

Note that I am not asking how to make git ignore the files. These files should be tracked and committed. It's just that, most of the time, I'm not interested in seeing them.

Related question: How to invert git log --grep=<pattern> or How to show Git logs that don't match a pattern. It's the same question except for commit messages rather than paths.

Forum discussion on this subject from 2008: Re: Excluding files from git-diff. This looked promising but the thread seems to have dried up.

Michael
  • 8,362
  • 6
  • 61
  • 88
Anonymoose
  • 5,662
  • 4
  • 33
  • 41
  • I'm not sure if there is a built-in way, and your perl solution looks pretty decent. If you modify it to accept the paths as command-line arguments, you could just create an alias something like `!f() { git log ... | path/to/filter-log.pl "$@" | git log --stdin --no-walk; f`, or even wrap that pipeline part up into the script as well. – Cascabel Apr 16 '11 at 17:25
  • As a workaround, I use `find` to filter out directories whose commits I do not want to see. If I wanted to ignore log entries from commits made to the root-level directory `SiteConfig` then I would say: `git log \`find . -type d -mindepth 1 -maxdepth 1 ! -name *SiteConfig\`` – Noah Sussman Mar 10 '12 at 16:30
  • For Git 1.9/2.0 (Q1 2014), see [my answer below](http://stackoverflow.com/a/21079437/6309): `git log --oneline --format=%s -- . ":!sub"` will work (with the **pathspec magic `:(exclude)` and its short form `:!`**) – VonC Jan 12 '14 at 19:48

3 Answers3

329

It is implemented now (git 1.9/2.0, Q1 2014) with the introduction pathspec magic :(exclude) and its short form :! in commit ef79b1f and commit 1649612, by Nguyễn Thái Ngọc Duy (pclouds), documentation can be found here.

You now can log everything except a sub-folder content:

git log -- . ':(exclude)sub'
git log -- . ':!sub'

Or you can exclude specific elements within that sub-folder

  • a specific file:

      git log -- . ':(exclude)sub/sub/file'
      git log -- . ':!sub/sub/file'
    
  • any given file within sub:

      git log -- . ':(exclude)sub/*file'
      git log -- . ':!sub/*file'
      git log -- . ':(exclude,glob)sub/*/file'
    

You can make that exclusion case insensitive!

git log -- . ':(exclude,icase)SUB'

As Kenny Evitt noted

Don't forget to use single quotes or proper escaping in double quotes if you're running git in a bash shell, e.g. ':!sub' or ":\!sub". Otherwise you will run into bash: ... event not found errors


Note: Git 2.13 (Q2 2017) adds a synonym ^ to !

See commit 859b7f1, commit 42ebeb9 (08 Feb 2017) by Linus Torvalds (torvalds).
(Merged by Junio C Hamano -- gitster -- in commit 015fba3, 27 Feb 2017)

pathspec magic: add '^' as alias for '!'

The choice of '!' for a negative pathspec ends up not only not matching what we do for revisions, it's also a horrible character for shell expansion since it needs quoting.

So add '^' as an alternative alias for an excluding pathspec entry.


Note that, before Git 2.28 (Q3 2020), the use of negative pathspec, while collecting paths including untracked ones in the working tree, was broken.

See commit f1f061e (05 Jun 2020) by Elijah Newren (newren).
(Merged by Junio C Hamano -- gitster -- in commit 64efa11, 18 Jun 2020)

dir: fix treatment of negated pathspecs

Reported-by: John Millikin
Signed-off-by: Elijah Newren

do_match_pathspec() started life as match_pathspec_depth_1() and for correctness was only supposed to be called from match_pathspec_depth(). match_pathspec_depth() was later renamed to match_pathspec(), so the invariant we expect today is that do_match_pathspec() has no direct callers outside of match_pathspec().

Unfortunately, this intention was lost with the renames of the two functions, and additional calls to do_match_pathspec() were added in commits

  • 75a6315f74 ("ls-files: add pathspec matching for submodules", 2016-10-07, Git v2.11.0-rc0 -- merge listed in batch #11)
  • 89a1f4aaf7 ("dir: if our pathspec might match files under a dir, recurse into it", 2019-09-17, Git v2.24.0-rc0).

Of course, do_match_pathspec() had an important advantge over match_pathspec() -- match_pathspec() would hardcode flags to one of two values, and these new callers needed to pass some other value for flags.

Also, although calling do_match_pathspec() directly was incorrect, there likely wasn't any difference in the observable end output, because the bug just meant that fill_diretory() would recurse into unneeded directories.

Since subsequent does-this-path-match checks on individual paths under the directory would cause those extra paths to be filtered out, the only difference from using the wrong function was unnecessary computation.

The second of those bad calls to do_match_pathspec() was involved -- via either direct movement or via copying+editing -- into a number of later refactors.

See commits

  • 777b420347 ("dir: synchronize treat_leading_path() and read_directory_recursive()", 2019-12-19, Git v2.25.0-rc0 -- merge)
  • 8d92fb2927 ("dir: replace exponential algorithm with a linear one", 2020-04-01, Git v2.27.0-rc0 -- merge listed in batch #5)
  • 95c11ecc73 ("Fix error-prone fill_directory() API; make it only return matches", 2020-04-01, Git v2.27.0-rc0 -- merge listed in batch #5)

The last of those introduced the usage of do_match_pathspec() on an individual file, and thus resulted in individual paths being returned that shouldn't be.

The problem with calling do_match_pathspec() instead of match_pathspec() is that any negated patterns such as :!unwanted_path will be ignored.

Add a new match_pathspec_with_flags() function to fulfill the needs of specifying special flags while still correctly checking negated patterns, add a big comment above do_match_pathspec() to prevent others from misusing it, and correct current callers of do_match_pathspec() to instead use either match_pathspec() or match_pathspec_with_flags().

One final note is that DO_MATCH_LEADING_PATHSPEC needs special consideration when working with DO_MATCH_EXCLUDE.

The point of DO_MATCH_LEADING_PATHSPEC is that if we have a pathspec like

*/Makefile

and we are checking a directory path like

src/module/component

that we want to consider it a match so that we recurse into the directory because it might have a file named Makefile somewhere below.

However, when we are using an exclusion pattern, i.e. we have a pathspec like

:(exclude)*/Makefile

we do NOT want to say that a directory path like

src/module/component

is a (negative) match.

While there might be a file named Makefile somewhere below that directory, there could also be other files and we cannot pre-emptively rule all the files under that directory out; we need to recurse and then check individual files.

Adjust the DO_MATCH_LEADING_PATHSPEC logic to only get activated for positive pathspecs.

Michael
  • 8,362
  • 6
  • 61
  • 88
VonC
  • 1,262,500
  • 529
  • 4,410
  • 5,250
  • 9
    Can you do multiple files? – Justin Thomas Mar 15 '14 at 14:07
  • 16
    @JustinThomas I believe (not tested yet) that you can repeat that path exclusion pattern multiple time `":(exclude)pathPattern1" ":(exclude)pathPattern2"`, hence ignoring multiple folders/files. – VonC Mar 15 '14 at 15:16
  • Yeah, that worked, I just didn't have 1.9 yet. Super awesome feature for a repo with minified files checked in. – Justin Thomas Mar 15 '14 at 16:38
  • 8
    If you're running Git in a Bash shell, use `':!sub'` instead to [avoid `bash: ... event not found` errors](http://unix.stackexchange.com/questions/3051/how-to-echo-a-bang). `":\!sub"` doesn't work. – Kenny Evitt Feb 02 '17 at 15:10
  • 2
    @KennyEvitt Thank for your edit and comment. I have included the latter in the answer for more visibility. – VonC Feb 02 '17 at 15:14
  • Is it possible to configure it s.t. I can permanently exclude, say, the `po` directory? – Frederick Nord Jul 12 '17 at 18:57
  • @FrederickNord would you like to keep the `po` directory content tracked? https://stackoverflow.com/a/18111571/6309. If the content is not yet tracked, all you need to do is add `po/` to a `.gitignore`. – VonC Jul 12 '17 at 19:42
  • @VonC I have tested repeating with multiple path exclusion patterns, but it does not quite do what @Anonymoose asked for, which is to exclude commits which touch `.etckeeper` *or* `lvm/archive` *or* `lvm/backup`, *even* if they *also* touch other paths. Instead it only excludes commits which *only* touch those paths. If you have a commit which touches (say) `.etckeeper` and `foobar`, it's still displayed. So while this is a great answer which taught me something new, sadly I think it doesn't technically answer the question. – Adam Spiers Feb 15 '18 at 20:57
  • @AdamSpiers OK. And I suppose you tested that with the latest Git 2.16.x? – VonC Feb 15 '18 at 20:59
  • 1
    @VonC Yes, tested with both 2.16.1.194.gb2e45c695d and an older 2.14.x. – Adam Spiers Feb 15 '18 at 21:06
  • 2
    For those wondering where the official documentation on this functionality is, see `git help glossary` (which I found listed in `git help -g` [which I found suggested in `git help`]). – ravron Nov 06 '18 at 01:49
  • Does it work for `--no-index`? – Michael Feb 23 '23 at 18:24
  • 1
    @Michael Do you mean [`git diff --no-index`](https://git-scm.com/docs/git-diff#Documentation/git-diff.txt-emgitdiffemltoptionsgt--no-index--ltpathgtltpathgt)? I do not see a `git log --no-index` option. – VonC Feb 23 '23 at 18:39
  • Yes, sorry. I meant `git diff --no-index`. I was linked here from a question about that and didn't pay attention to the context. – Michael Feb 23 '23 at 19:39
  • @Michael OK. You wouldn't have the URL of the question you were linked from, by any chance? That would help for understanding the context of this inquiry. – VonC Feb 23 '23 at 22:14
  • Yes. It looks like I was the one who [linked it](https://stackoverflow.com/q/49969016/241211#comment107218225_49969016) with "Possibly related?" Probably as a reminder to come back and see whether `:!` syntax worked there. – Michael Feb 24 '23 at 15:09
  • 1
    @Michael OK. I don't think that syntax can work with `git diff --no-index`, since there is no "tree" to walk. As opposed to [`git diff`, where you can use a negative pathspec](https://stackoverflow.com/q/4380945/6309). As stated as much in "[`git diff --no-index` exclude files](https://stackoverflow.com/a/53475515/6309)". – VonC Feb 24 '23 at 16:50
4

tl;dr: shopt -s extglob && git log !(unwanted/glob|another/unwanted/glob)

If you're using Bash you should be able to use the extended globbing feature to get only the files you need:

$ cd -- "$(mktemp --directory)" 
$ git init
Initialized empty Git repository in /tmp/tmp.cJm8k38G9y/.git/
$ mkdir aye bee
$ echo foo > aye/foo
$ git add aye/foo
$ git commit -m "First commit"
[master (root-commit) 46a028c] First commit
 0 files changed
 create mode 100644 aye/foo
$ echo foo > bee/foo
$ git add bee/foo
$ git commit -m "Second commit"
[master 30b3af2] Second commit
 1 file changed, 1 insertion(+)
 create mode 100644 bee/foo
$ shopt -s extglob
$ git log !(bee)
commit ec660acdb38ee288a9e771a2685fe3389bed01dd
Author: My Name <jdoe@example.org>
Date:   Wed Jun 5 10:58:45 2013 +0200

    First commit

You can combine this with globstar for recursive action.

l0b0
  • 55,365
  • 30
  • 138
  • 223
-2

You can temporarily ignore the changes in a file with:

git update-index --skip-worktree path/to/file

Going forward, all changes to those files will be ignored by git status, git commit -a, etc. When you're ready to commit those files, just reverse it:

git update-index --no-skip-worktree path/to/file

and commit as normal.

rubysolo
  • 533
  • 2
  • 8
  • 9
    This appears to address a slightly different situation. `git update-index --skip-worktree` does not cause `git log` to filter commits that have already been made. – Anonymoose Jul 07 '11 at 19:58