From: Junio C Hamano <gitster@pobox.com>
Date: Wed, 17 Sep 2008 20:44:02 GMT
Subject: Re: [PATCH] Teach git diff about Objective-C syntax
Message-ID: <7vljxq4i5p.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <57518fd10809171214u3b5b3b96yc432c1c410faf8b4@mail.gmail.com>

"Jonathan del Strother" <maillist@steelskies.com> writes:

> Johannes already convinced me to do it as a separate patch.  Andreas
> seems to think that even if that change were in a separate patch, it
> is pure nonsense.  I think it's pretty subjective - I was just making
> things consistent.

I think the discussion is getting into personal preference, but I do agree
making things consistent is good.

Short and simple ones can stay the same as before, but I'd actually prefer
doing ones on multiple lines like this:

	{ "java",
	  /* Do not match these keywords */
	  "!^[ \t]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
		"new\\|return\\|switch\\|throw\\|while\\)\n"

	  /* possibly indented "foo bar(..." */
	  "^[ \t]*\\(\\([ \t]*[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}[ \t]*([^;]*\\)$"
	},

	{ "objc",
	  /* Do not match with these C statements */
	  "!^[ \t]*\\(do\\|for\\|if\\|else\\|return\\|switch\\|while\\)\n"

	  /* Objective-C methods "-(fo*&^)%*&%$^" or "+(bar&^)%$#$%" */
	  "^[ \t]*\\([-+][ \t]*([ \t]*[A-Za-z_][A-Za-z_0-9]*.*).*\\)$"

	  "\\|" /* C functions "foo bar(..." */
	  "^[ \t]*\\(\\([ \t]*[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}[ \t]*([^;]*\\)$"

	  "\\|" /* Objective-C class/protocol definitions */
	  "^@\\(implementation\\|interface\\|protocol\\).*"
	},

That is:

 - The pattern string indented to align with the pattern name;
 - Indent second and later lines of a single alternatives list that is
   split over multiple physical lines;

Are you sure the regexp you have for Objective-C methods quotes a dot "."
correctly, by the way?  It appears to match almost anything enclosed in a
pair of parentheses, as long as you have two alpha after open paren.

Also I am not sure if you can do the pattern alternates the way you did.
If you have this:

	"...\\(A\\|B\\)$"
        "\\|"
        "...\\(C\\|D\\)$"

A and B will be captured as $1 but wouldn't C or D captured as $2?