Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

How to search through all Git and Mercurial commits in the repository for a certain string?

Tags:

git

mercurial

People also ask

How do I search for text in a git repository?

To search within a particular repository or organization, navigate to the repository or organization page, type what you're looking for into the search field at the top of the page, and press Enter.

How do I search for a string in git log?

Simple! Drop the 'log' argument. This will show the commit ID and line number of all instances of your string. (The string can also be a regular expression, which is handy if you know regex, too.)

How do I see all commits on a repository?

On GitHub, you can see the commit history of a repository by: Navigating directly to the commits page of a repository. Clicking on a file, then clicking History, to get to the commit history for a specific file.


You can see dangling commits with git log -g.

-g, --walk-reflogs
 Instead of walking the commit ancestry chain, walk reflog entries from
 the most recent one to older ones. 

So you could do this to find a particular string in a commit message that is dangling:

git log -g --grep=search_for_this

Alternatively, if you want to search the changes for a particular string, you could use the pickaxe search option, "-S":

git log -g -Ssearch_for_this
# this also works but may be slower, it only shows text-added results
git grep search_for_this $(git log -g --pretty=format:%h)

Git 1.7.4 will add the -G option, allowing you to pass -G<regexp> to find when a line containing <regexp> was moved, which -S cannot do. -S will only tell you when the total number of lines containing the string changed (i.e. adding/removing the string).

Finally, you could use gitk to visualise the dangling commits with:

gitk --all $(git log -g --pretty=format:%h)

And then use its search features to look for the misplaced file. All these work assuming the missing commit has not "expired" and been garbage collected, which may happen if it is dangling for 30 days and you expire reflogs or run a command that expires them.


In Mercurial you use hg log --keyword to search for keywords in the commit messages and hg log --user to search for a particular user. See hg help log for other ways to limit the log.


In addition to richq answer of using git log -g --grep=<regexp> or git grep -e <regexp> $(git log -g --pretty=format:%h): take a look at the following blog posts by Junio C Hamano, current git maintainer

  • Fun with "git log --grep"
  • Fun with "git grep"

Summary

Both git grep and git log --grep are line oriented, in that they look for lines that match specified pattern.

You can use git log --grep=<foo> --grep=<bar> (or git log --author=<foo> --grep=<bar> that internally translates to two --grep) to find commits that match either of patterns (implicit OR semantic).

Because of being line-oriented, the useful AND semantic is to use git log --all-match --grep=<foo> --grep=<bar> to find commit that has both line matching first and line matching second somewhere.

With git grep you can combine multiple patterns (all which must use the -e <regexp> form) with --or (which is the default), --and, --not, ( and ). For grep --all-match means that file must have lines that match each of alternatives.


Building on rq's answer, I found this line does what I want:

git grep "search for something" $(git log -g --pretty=format:%h -S"search for something")

Which will report the commit ID, filename, and display the matching line, like this:

91ba969:testFile:this is a test

... Does anyone agree that this would be a nice option to be included in the standard git grep command?


Any command that takes references as arguments will accept the --all option documented in the man page for git rev-list as follows:

   --all
       Pretend as if all the refs in $GIT_DIR/refs/ are listed on the
       command line as <commit>.

So for instance git log -Sstring --all will display all commits that mention string and that are accessible from a branch or from a tag (I'm assuming that your dangling commits are at least named with a tag).


With Mercurial you do a

$ hg grep "search for this" [file...]

There are other options that narrow down the range of revisions that are searched.


Don't know about git, but in Mercurial I'd just pipe the output of hg log to some sed/perl/whatever script to search for whatever it is you're looking for. You can customize the output of hg log using a template or a style to make it easier to search on, if you wish.

This will include all named branches in the repo. Mercurial does not have something like dangling blobs afaik.