Re: [PATCH 4/6] [GSOC] ref-filter: add %(rest) atom and --rest option

From: Junio C Hamano <gitster@pobox.com>
Date: Jun 8, 2021, 06:50 UTC
Message-ID: <xmqqo8cgq28e.fsf@gitster.g>
In-Reply-To: <CAOLTT8S+5m+-XF-AcQi9t8njTvyDYzHt=BU+4OPcvTT27RP6dw@mail.gmail.com>

ZheNing Hu <adlternative@gmail.com> writes:

> First of all, although %(rest) is meaningless in ordinary circumstances,
> ref-filter must learn %(rest), it is impossible for us to leave the parsing
> of %(rest) in cat-file.c alone.

Oh, there is no question about that.

> Then, `--rest` is a strategy that make %(rest) can use in `git for-each-ref`
> or `git branch -l`.

If there is no need to expose %(rest) to the users who write --format for these two commands, it would be much better to detect attempted use of %(rest) and error out.

> This sounds like it might help `cat-file` to reject some useless atoms
> like %(refname).

Yes.

Show 6 quoted lines

> So something like:
>
> $ git for-each-ref --format="%(objectname) %(objectsize)"
> --refject-atoms="%(objectsize) %(objectname)"
>
> will fail.

I don't understand. Why do you even need to add --reject? Why would any user would want to use it with for-each-ref?

Without any end-user input, %(rest) for for-each-ref would not make sense, and %(refname) for cat-file --batch would not make sense, I would imagine, so there is no need to be able to tell --reject=rest to for-each-ref. It is not like giving --no-reject=rest to for-each-ref and make it interpolate to an empty string is a useful feature anyway, so I do not see a need for such an option.