Increase performance up to 15000%

[ossinkine]

The basic idea is to use an array of characters instead of string processing.

Benchmark
For test I took the current css-file from github.com: https://assets-cdn.github.com/assets/github-a15e59d993eb0844d736a04652f63d465d909127e84cf7cd27073076df530f05.css

$source = file_get_contents('https://assets-cdn.github.com/assets/github-a15e59d993eb0844d736a04652f63d465d909127e84cf7cd27073076df530f05.css');
$time = microtime(true);
$parser = new Sabberworm\CSS\Parser($source);
$css = $parser->parse();
var_dump(microtime(true) - $time);

On the my machine I received the following results:
before float(180.93577218056)
after float(1.1720938682556)

[sabberworm]

Why 150000000% while you’re at it…

Just kidding. Great work, thanks a lot!

[sabberworm]

I see a problem here: splitting should only happen after the charset is known (because it depends on the charset used).

[sabberworm]

Also, the @charset declaration will call this method during parsing. But it should trigger a re-parse if the count is different in the new charset. Which it has never done. So, again, this is due for a rework that hasn’t necessarily got anything to do with your changeset.

[ossinkine]

If it so, why is bMultibyteSupport flag needed? Can I pass bMultibyteSupport = false and @charset "UTF-8"?