Re: [PATCH v3 05/13] parse the -L options

[Bo Yang <struggleyb.nku@gmail.com>]

Hi Junio,

On Wed, Jul 14, 2010 at 7:03 AM, Junio C Hamano <gitster@pobox.com> wrote:
>>       if (argc == 2 && !strcmp(argv[1], "-h"))
>>               usage(builtin_log_usage);
>> +
>> +     parse_options_start(&ctx, argc, argv, prefix, PARSE_OPT_KEEP_DASHDASH |
>> +                     PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_STOP_AT_NON_OPTION);
>> +     for (;;) {
>> +             switch (parse_options_step(&ctx, options, log_opt_usage)) {
>> +             case PARSE_OPT_HELP:
>> +                     exit(129);
>> +             case PARSE_OPT_DONE:
>> +                     goto parse_done;
>> +             case PARSE_OPT_NON_OPTION:
>> +                     path = parse_options_current(&ctx);
>> +                     pathspec = prefix_path(prefix, prefix ? strlen(prefix) : 0, path);
>> +                     range->spec = alloc_filespec(pathspec);
>> +                     free((void *)pathspec);
>> +                     if (range->nr == 0) {
>> +                             if(range->next) {
>> +                                     die("Path %s need a -L <range> option\n"
>> +                                     "If you want follow the history of the whole file "
>> +                                     "whether to using 'git log' without -L or using "
>> +                                     "'git log -L 1,$ <path>'", range->spec->path);
>> +                             } else {
>> +                                     parse_options_next(&ctx, 1);
>> +                                     continue;
>
> This loop smells bad.
>
> When "-L n,m" appears on the command line, log_line_range_callback() would
> be called and would eat n,m (which is correct), but at that point you
> would not just want to be prepared to accept a non-option ("path" in "-L
> n,m path"), but actually would want to force the user to give a path, no?
> IOW, isn't "git log -L n,m -U20" an error, unless "-U20" is a filename
> that the user wants to track?
>
> I somehow suspect that futzing with STOP_AT_NON_OPTION (done in the first
> two patches in the series) to parse "-L n,m path" is a misguided design
> attempt.  Shouldn't you be instead giving a support for option callback to
> take more than one argument to do this?
>
>> +                             }
>> +                     }
>> +                     struct diff_line_range *r = xmalloc(sizeof(*r));
>
> decl-after-statement, but at this point it may be moot as I am doubting
> the higher-level design of this option parser now.
>

The point is that, the syntax we support is:

-L n1,m1 -L n2,m2 pathspec1  -L n3,m3 pathspec2

There can be multiple -L options for one pathspec, so I think the
STOP_AT_NON_OPTION way is better.

-- 
Regards!
Bo
----------------------------
My blog: http://blog.morebits.org
Why Git: http://www.whygitisbetterthanx.com/