Re: Git tag output order is incorrect (IMHO)

From: Junio C Hamano <gitster@pobox.com>
Date: Sep 11, 2013, 17:54 UTC
Message-ID: <xmqqioy7tfba.fsf@gitster.dls.corp.google.com>
In-Reply-To: <CABURp0o5tOswiv_avfAQOZwGREpX3v7MwcdVzt7dkubY2Y0O6g@mail.gmail.com>

Phil Hord <phil.hord@gmail.com> writes:

Show 10 quoted lines

> Someone at $work asked me this week how to find the current and
> previous tags on his branch so he could generate release notes.  I
> just need "last two tags on head in topo-order". I was surprised by
> how complicated this turned out to be. I ended up with this:
>
>   git log --decorate=full --pretty=format:'%d' HEAD |
>     sed -n -e 's-^.* refs/tags/\(.*\)[ )].*$-\1-p' |
>     head -2
>
> Surely there's a cleaner way, right?

That looks clean enough (I would have used "head -n 2" though) and in line with the way how you can exercise the flexibility of the system, at least to me ;-).

Joking aside, I agree that a "--merged X" primitive, i.e. "what refs can be reachable from commit X?", in the listing mode of "git tag" or "git for-each-ref" would have helped. As the sorting and formatting primitives are already there in for-each-ref, it would have been

	git for-each-ref \
            --format='%(refname:short)' \
            --sort='-*committerdate' \
            --count=2 \
            --merged my-branch \
            refs/tags/

or something like that.