log: add option to search for header or body

[vegerot]

cc: Phillip Wood phillip.wood123@gmail.com
cc: strager strager.nds@gmail.com
cc: me mchcopl@gmail.com

[vegerot]

Note to reviewer: please help me come up with a better name than --header-or. I want this option to be easily discoverable by someone like me, who is looking for commits either authored or co-authored by someone

[vegerot]

/preview

[gitgitgadget]

Preview email sent as pull.1710.git.1712353473406.gitgitgadget@gmail.com

[vegerot]

/submit

[gitgitgadget]

Submitted as pull.1710.git.1712353687464.gitgitgadget@gmail.com

To fetch this version into FETCH_HEAD:

git fetch https://github.com/gitgitgadget/git/ pr-1710/vegerot/header-or-v1

To fetch this version to local tag pr-1710/vegerot/header-or-v1:

git fetch --no-tags https://github.com/gitgitgadget/git/ tag pr-1710/vegerot/header-or-v1

[gitgitgadget]

On the Git mailing list, Phillip Wood wrote (reply to this):

Hi Max

On 05/04/2024 22:48, Max Coplan via GitGitGadget wrote:
> From: =?UTF-8?q?Max=20=F0=9F=91=A8=F0=9F=8F=BD=E2=80=8D=F0=9F=92=BB=20Copl?=
>   =?UTF-8?q?an?= <mchcopl@gmail.com>
> > Note to reviewer: I hate the name `--header-or`!  Please help me come up
> with a better name.
> > Summary:
> This change adds a new option to `git log` that allows users to search
> for commits that match either the author or the commit message. This is
> useful for finding commits that were either authored or co-authored by a
> specific person.

I think that having a way to find all the commits that are authored by, or have a specific trailer naming someone is a good idea. I'm not sure about combining "--author" with "--grep" to do it though. Perhaps we could design a new option (tentatively called "--indent") that allows the user to specify a name and which trailers and commit headers (author and/or committer) to match.

A couple of house keeping notes. We do not use the "feat(...)" style of commit messages in this project. The subject line should just start with "log: " the rest of the subject looks fine. I think everything under the "Summary:" line above does a good job of explaining why the change is useful which is great. We tend to avoid emoji's in peoples names as well.

> [...]
> +test_expect_success 'log --grep --author --header-or uses union' '
> +	# grep matches only third and fourth
> +	# author matches only initial and third

I'm a bit confused by this comment as we seem to only expect "third" and "fourth" below - what happened to "initial"?

> +	git log --author="A U Thor" --grep=r --header-or --format=%s >actual &&
> +	{
> +	    echo fourth && echo third
> +	} >expect &&

We have a test helper to help with this

    test_write_lines fourth third >expect

Best Wishes

Phillip

[gitgitgadget]

User Phillip Wood <phillip.wood123@gmail.com> has been added to the cc: list.

[gitgitgadget]

There are issues in commit f38e52f:
log: add option to search for header or body
Lines in the body of the commit messages should be wrapped between 60 and 76 characters.
Indented lines, and lines without whitespace, are exempt

[vegerot]

/submit

[gitgitgadget]

Submitted as pull.1710.v2.git.1712460247516.gitgitgadget@gmail.com

To fetch this version into FETCH_HEAD:

git fetch https://github.com/gitgitgadget/git/ pr-1710/vegerot/header-or-v2

To fetch this version to local tag pr-1710/vegerot/header-or-v2:

git fetch --no-tags https://github.com/gitgitgadget/git/ tag pr-1710/vegerot/header-or-v2

[gitgitgadget]

On the Git mailing list, Junio C Hamano wrote (reply to this):

"Max Coplan via GitGitGadget" <gitgitgadget@gmail.com> writes:

> This change adds a new option to `git log` that allows users to search
> for commits that match either the author or the commit message. This is
> useful for finding commits that were either authored or co-authored by a
> specific person.

I have this feeling that the "solution" presented is not quite
addressing the use case in a more useful and direct way than it
could be.  When I designed how the --author/--committer restriction
and --grep in the body of the message interact, I made a concious
decision that "among those commits that were authored by person X,
find the ones that mention Y" is far more useful than "done by X, or
done by anybody that mention Y", especially when Y is just a text
search in the free form.  There was nothing that limits the mention
of Y to those specifically involved in the commit---the mention could
just have been part of text, like "earlier Max Coplan sent a patch,
but this commit is not related to it".

But these days, we have a more established "convention" that lists
people at the end in the form of "trailers", and that changes the
picture quite a lot from how the world order was back then.

In other words, if the true objective is to find commits that
involved person X, Y or Z (which is very common and would be a lot
more useful than finding those that involve all of them), shouldn't
we be limiting the --grep side even further so that a random mention
of person Y is excluded and hit is counted only when person Y is
mentioned on a trailer (while loosening the --author side so that it
is OR'ed instead of AND'ed)?

I am imagining a pair of new options to name people (all OR'ed) and
to name places the names of these people should appear (again, all
OR'ed).  I am not good at naming, so the option names in the example
is not more than illustration of an idea and not my recommendation,
but a command:

    git log --by="Max Coplan" --by="Junio C Hamano" \
	    --by-where=author,Signed-off-by,Co-authored-by

would find a commit that has one (or more) of the given names
in one (or more) of the places that are specified, where the places
can be either "author", "committer" to specify these headers in the
commit object, or random other string to specify trailer lines with
given keys.

Hmm?

[gitgitgadget]

On the Git mailing list, Phillip Wood wrote (reply to this):

On 07/04/2024 07:08, Junio C Hamano wrote:
> "Max Coplan via GitGitGadget" <gitgitgadget@gmail.com> writes:
> I am imagining a pair of new options to name people (all OR'ed) and
> to name places the names of these people should appear (again, all
> OR'ed).  I am not good at naming, so the option names in the example
> is not more than illustration of an idea and not my recommendation,
> but a command:
> >      git log --by="Max Coplan" --by="Junio C Hamano" \
> 	    --by-where=author,Signed-off-by,Co-authored-by
> > would find a commit that has one (or more) of the given names
> in one (or more) of the places that are specified, where the places
> can be either "author", "committer" to specify these headers in the
> commit object, or random other string to specify trailer lines with
> given keys.

I like this, yesterday[1] I didn't have a clear idea in my head of how such an option should work but I think passing the names and the fields to match those names against as two separate options is a good idea.

Best Wishes

Phillip

[1] https://lore.kernel.org/git/c93817ba-5945-4ec0-9775-5621481b972c@gmail.com/