Re: [PATCH 1/2] cat-file: force flush of stdout on empty string

[Junio C Hamano <gitster@pobox.com>]

Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes:

>> I think overall this approach is cleaner and makes sense. My only
>> question is, are there more commands in the future that will need some
>> special command syntax? Just wondering whether YAGNI applies here.
>
> An obvious addition is to at least add the ability to set the various
> options on the fly, i.e. now you need to use --batch-check, and then
> kill it and restart if you'd like the content with --batch, ditto for
> --textconv.
>
> E.g. the gitaly backend for gitlab.com keeps two cat-filfe processes
> around just to flip-flop between those two, sometimes you want the
> content, sometimes you're just checking if the object exists.
>
> I'd also like to add something to expose the likes of -e and -t
> directly, i.e. even with --batch-check you often want to just check
> existence, but get the size too, you could supply a format, but like the
> above you sometimes want the size or whatever, and killing/starting a
> new process just for that is a hassle...

Yeah, with "plug" and "unplug" instruction you do not have to keep
issuing "flush" when you want to go interactive, and other things
become easy to do, so even though it would make it a bit more
verbose to require "object " prefix for the kind of lines that were
historically the only ones accepted by the command, I think it is a
good direction to go in.

Thanks.