[MacRuby-devel] StringScanner Performance

[Sylvester Keil]

Dear all,

while debugging performance issues for a gem (bibtex-ruby) I noticed that MacRuby's StringScanner implementation creates new Regular Expression objects every time #scan is called; as I am dealing with a lexical analyzer based on StringScanner the #scan method is quite crucial and the current implementation performs so slowly that it is basically unusable on MacRuby.

This is the problematic method in MacRuby:

https://github.com/MacRuby/MacRuby/blob/master/lib/strscan.rb#L638

Both MRI and Rubinius work around this by using a feature of Oniguruma patterns to match the pattern at the beginning of a string only. Here are the corresponding sections:

https://github.com/ruby/ruby/blob/trunk/ext/strscan/strscan.c#L437
https://github.com/rubinius/rubinius/blob/master/lib/strscan.rb#L264

Do regular expressions in MacRuby expose similar functionality to either Ruby or C extensions? I'd be happy to help resolving this issue, but have no experience with MacRuby so any pointers on how to best approach this are much appreciated.

Thanks,

Sylvester