Re: [PATCH] userdiff: add built-in pattern for CSS

[William Duclot <william.duclot@ensimag.grenoble-inp.fr>]

On Fri, Jun 03, 2016 at 07:52:45AM +0200, Johannes Sixt wrote:
> Am 03.06.2016 um 00:48 schrieb William Duclot:
>>Logic behind the "pattern" regex is:
> 
> The name of the macro parameter is "pattern", but the actual meaning
> is "function name" regex.
> 
>>diff --git a/Documentation/gitattributes.txt b/Documentation/gitattributes.txt
>>index e3b1de8..81f60ad 100644
>>--- a/Documentation/gitattributes.txt
>>+++ b/Documentation/gitattributes.txt
>>@@ -525,6 +525,8 @@ patterns are available:
>>
>>  - `csharp` suitable for source code in the C# language.
>>
>>+- `css` suitable for source code in the CSS language.
> 
> CSS is not so much source code. How about "suitable for cascaded
> style sheets"?

Technically correct yes
 
>>diff --git a/t/t4018/css-common b/t/t4018/css-common
>>new file mode 100644
>>index 0000000..84ed754
>>--- /dev/null
>>+++ b/t/t4018/css-common
>>@@ -0,0 +1,4 @@
>>+RIGHT label.control-label {
>>+    margin-top: 10px!important;
>>+    border : 10px ChangeMe #C6C6C6;
>>+}
> 
>>diff --git a/t/t4018/css-rule b/t/t4018/css-rule
>>new file mode 100644
>>index 0000000..84ed754
>>--- /dev/null
>>+++ b/t/t4018/css-rule
>>@@ -0,0 +1,4 @@
>>+RIGHT label.control-label {
>>+    margin-top: 10px!important;
>>+    border : 10px ChangeMe #C6C6C6;
>>+}
> 
> These two are the same. Please pick only one. I propose the name
> "common" because it is how CSS rules are written most commonly.

Right, I was distracted

>>+IPATTERN("css",
>>+	 "!^.*;\n"
> 
> Is there a difference between this and "!;\n"? Is it necessary to
> anchor the pattern at the beginning of the line?
> 
> In the commit message you talk about colon (':'), but you actually
> use a semicolon (';'). Thinking a bit more about it, rejecting lines
> with either one would be even better. Consider this case (without
> the indentation):
> 
>    h1 {
>    color:
>    red;
>    }
> 
> (New test case, hint, hint!) Therefore, it could be: "![:;]\n".

You're right! (plus the potential trailing spaces)

>>+	 "^[_a-z0-9].*$",
>>+	 /* -- */
>>+	 /*
>>+	  * This regex comes from W3C CSS specs. Should theoretically also
>>+	  * allow ISO 10646 characters U+00A0 and higher,
>>+	  * but they are not handled in this regex.
>>+	  */
>>+	 "-?[_a-zA-F][-_a-zA-F0-9]*" /* identifiers */
> 
> Drop A-F.
> 
>>+	 "|-?[0-9]+|\\#[0-9a-fA-F]+" /* numbers */
> 
> Here, too: it is an IPATTERN.

Here I have to disagree (with you and Junio): the IPATTERN is
case-insensitive only on the "pattern" regex, not the "word_regex"
regex. It can be seen in the macro definition, and a quick test confirm
that (and we can see that the fortran word_regex, for example, bother
with uppercase and lowercase even thought it use IPATTERN).
On the identifier line, I have "A-F" instead of "A-Z" though