From: Junio C Hamano <gitster@pobox.com>
Date: Mon, 13 May 2013 05:28:41 GMT
Subject: Re: [PATCH v7 03/10] Add colors to interactive git-clean
Message-ID: <7vhai7qwhi.fsf@alter.siamese.dyndns.org>
In-Reply-To: <CANYiYbGqZ6FiHTZgejrA3a-njffQNJyaefFmCWfycuE6PGjwbA@mail.gmail.com>

Jiang Xin <worldhello.net@gmail.com> writes:

> 2013/5/13 Matthieu Moy <Matthieu.Moy@grenoble-inp.fr>:
>> Jiang Xin <worldhello.net@gmail.com> writes:
>>
>>>  * color.interactive.<slot>: Use customized color for interactive
>>>    git-clean output (like git add --interactive). <slot> may be
>>>    prompt, header, help or error.
>>
>> This should go to the documentation (a short summary is welcome in the
>> commit messages in addition, but users won't read this...)
>>
>>> +     if (!prefixcmp(var, "color.interactive.")) {
>>> +             int slot = parse_clean_color_slot(var, 18);
>>
>> For readability and maintainability: please use
>> strlen("color.interactive."), not 18.
>
> Feel like conventional:
>
>     git grep -C2 prefixcmp builtin/apply.c builtin/archive.c
> builtin/branch.c builtin/checkout.c
>
> But maybe 18 characters are too long. ;-)

Why does it even have to know where the prefix ends or how long the
prefix is?

Doesn't it suggest that perhaps the parse_clean_color_slot()'s
external interface is misdesigned?  In the other examples you
showed, e.g.

    builtin/apply.c:	if (!prefixcmp(buffer, "delta ")) {
    builtin/apply.c-		patch_method = BINARY_DELTA_DEFLATED;
    builtin/apply.c-		origlen = strtoul(buffer + 6, NULL, 10);

we are calling external libraries that are designed to take a
pointer to a character, but the parse-clean-color-slot knows that it
is fed the name of a configuration variable and it knows the shape
of its input far better than a generic function like strtoul(), no?