CB Bailey <cb@xxxxxxxxxxxxx> writes:
> From: CB Bailey <cbailey32@xxxxxxxxxxxxx>
>
> shortlog always respects the mailmap in its output. Pass the mailmap
> into the revision walker to allow the mailmap to be used with revision
> limiting options such as '--author'.
I am a bit torn on this.
If an author of interest has entries in the mailmap, with the same
name but three or more different e-mail addresses, you could do
$ git shortlog --author=junio@xxxxxxxxxx --author=junio@xxxxxxxxxxx
to find commits by the author under these two addresses, and exclude
commits by the same author made under other addresses. With this
patch applied, it becomes no longer possible, and you can only look
for commits written under all of my identities, or none of them,
which is a minor annoyance.
But more importantly, the user now needs to know which one is the
"canonical" spelling of the author's ident [*1*]. Asking for
commits written by junio@xxxxxxxxxx will not yield any result with
the patch applied, no?
For the author in the above example, luckily, AUTHOR_NAME is the
same and unique, so without your patch, you could still do
$ git shortlog --author="Junio C"
to grab all of them, but there may be authors for whom there is no
catch-all substring that would match all of the idents they ever
used but still unique enough to match only that author, and I do
agree that the new feature proposed by the patch in question would
have uses in such a case.
Which nudges me to say that there needs a way to ask for the
original behaviour, disabling the rewriting of commit author
identity before --author filter kicks in.
There may also need a way to ask omitting mailmap processing even at
the output phase. I do not think it is a particularly interesting
feature to be able to ask for my commits under only two of my
identities [*2*], but as long as that "feature" exists, it also
should be possible to see the result of that "feature" more clearly
by not merging them into a single list. "shortlog --no-use-mailmap"
may be a way to do so, but unlike "log", the command currently does
not take the "--[no-]use-mailmap" option.
[Footnote]
*1* Passing the query string given as --author=... through the
mailmap machinery _might_ be a possible ingredient to solve "the
user needs to know the caonical spelling of the author ident"
issue this patch has, but I do not think it generally is
possible (e.g. how would you rewrite --author='.*@kernel.org').
*2* In other words, I think the loss of "find only my @kernel.org
and @twinsun.com commits" is mere annoyance and people can live
with it. But I think "you must look for gitster@xxxxxxxxx, and
looking for junio@xxxxxxxxxx will not find my commits made under
that identity" is a show stopper. That is why I use a very weak
"may also need" to this "optionally not using mailmap at all"
feature.
