Re: [PATCH] grep: correct help string for --exclude-standard

[Jeff King <peff@peff.net>]

On Wed, Mar 04, 2015 at 06:13:32PM +0700, Duy Nguyen wrote:

> On Wed, Mar 4, 2015 at 5:16 PM, Jeff King <peff@peff.net> wrote:
> > Hmm. If the default is "--exclude-standard", then what expect people to
> > use is "--no-exclude-standard". Would it make more sense to list that in
> > the "-h" output?
> 
> I thought about it and actually edited git-grep man page to clarify
> the default, only to find out this.  When --untracked is used,
> --exclude-standard is the default. When --no-index is used,
> --no-exclude-standard is the default. Can't have it both ways. This is
> already mentioned with the subtle phrase "only useful with...".

Yuck. :)

I agree your patch is the right thing to do, then.

> > It might be nice if parseopt had a PARSE_OPT_NEGHELP option or something
> > to show the "--no-" form.
> 
> Regardless, yes it would be nice to have something like this. I think
> there are places that can make use of this.

Grepping around, it looks like the best form would be an OPT_NEGBOOL
that acts like a boolean but negates the truth value, and advertises the
negative form. We have a lot of:

  OPT_BOOL('n', "no-checkout", &option_no_checkout,
           N_("don't create a checkout"))

where countermanding an earlier "--no-checkout" has to be spelled as
"--no-no-checkout", rather than "--checkout". If we could write:

  OPT_NEGBOOL('n', "checkout", ...)

that would be nicer. But the short option is a bit weird. We'd want:

  -n: option_no_checkout=true
  --checkout: option_no_checkout=false
  --no-checkout: option_no_checkout=true

That is, we flip the sense of the long option, but the short option
still yields "true". I think that would be useful, but it sure is weird
to explain.

-Peff