Re: [RFC PATCH v1 0/1] ls-remote: inconsistency from the order of args and opts

From: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Date: Jan 14, 2022, 20:57 UTC
Message-ID: <220114.86y23iqbbk.gmgdl@evledraar.gmail.com>
In-Reply-To: <xmqqbl0eaw2e.fsf@gitster.g>

nOn Fri, Jan 14 2022, Junio C Hamano wrote:

Show 26 quoted lines

> Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes:
>
>> We have multiple commands that are in GNU-fashion loose about whether
>> you provide options first before no-option args, or after. E.g. we
>> accept both of:
>>
>>     git push --dry-run <remote> <ref>
>>
>> And:
>>
>>     git push <remote> <ref> --dry-run
>
> Yes, but I consider that a bug that we cannot fix due to backward
> compatibility issues.
>
> That is why my preference is to encourage users to stick to the
> POSIX way in gltcli, just like we recommend "stuck" form of options
> its parameter.
>
>> But when GNU came around its option parser was generally happy to accept
>> options and args in either order. E.g. these both work with GNU
>> coreutils, but the latter will fail on FreeBSD and various other
>> capital-U UNIX-es:
>>
>>     touch foo; rm -v foo
>>     touch foo; rm foo -v

This is only an approximate list, but:
    
    $ git grep -C3 'parse_options' -- 'builtin/*.c'|grep -c PARSE_OPT_STOP_AT_NON_OPTION
    16
    $ git grep -C3 'parse_options' -- 'builtin/*.c'|grep -c -F ', 0);'
    101

The GNU-like behavior is far more common in our codebase, and I think it's less surprising if commands work the same way for consistency.

I manually looked through the PARSE_OPT_STOP_AT_NON_OPTION cases, and I think this is the only one that's using it for no good reason. The others (e.g. "git config") would become ambiguous or error out as a result.

Show 7 quoted lines

> Yes, among the harm GNU has done on mankind, this is one of the
> biggest ones.  We shouldn't waste our engineering time to support
> more of them in our tools.
>
> As long as users stick to the recommended "dashed options first and
> then args, among which revs come first and then pathspecs", they
> will be fine.

I find it quite useful. E.g. if you typo a command or forget/want to remove an option:

    git push origin HEAD --dry-run

You can just (under readline) do C-p M-DEL, instead of the equivalent navigating back a few words, or having to use more advanced readline features like ^--dry-run^^ or whatever.

Anecdotally, I've been surprised by the amount of regular terminal users whose readline skills pretty much and at using the arrow keys to make command corrections. I think this GNU UX decision has probably saved several accumulated man-lifetimes by now :)