From: Marcin Raczkowski <swistak@...>
Date: 2007-04-11T16:43:47+09:00
Subject: Re: Slow ruby regexes

On Tuesday 10 April 2007 19:14, MenTaLguY wrote:
> On Wed, 11 Apr 2007 03:56:28 +0900, "Robert Dober" <robert.dober@gmail.com> 
wrote:
> > The bad news is that the power of regexs will just not let you escape
> > with bad design all the time, so knowing a lot about regexes is a
> > necessary thing to use them well.
>
> No, the bad news is that the bad design of most "regex" implementations
> will not let you escape with using "regexes" as if they were true regular
> expressions; knowing a lot about the implementation of the regex engine is
> a necessary thing to use them well.
>
> > But you are right of course that Ruby's current Regex implementation
> > is not state of the art, useful, very useful nevertheless.
>
> NFAs were state-of-the-art in the 1960s.  Backreferences are nice, but in
> other respects we've actually regressed since then.
>
> I'm not suggesting we abandon backreferences and so forth (which are useful
> sometimes), but rather that NFAs should be used for evaluating the majority
> of "regexes" which don't use non-regular features.
>
> -mental

Well topic is rather interesting, i'll try to check it soon, and i'll try to 
write ruby interface for c library that author of article reference, i was 
looking for good case to test including c into ruby.
I'm not promising anything tho :)

regexps are great tool for text processing, but serious log analyzers (like 
one my company use) are written in c and base on lots of low-level string 
manipulation, raw binary access to DB et cecera, not to mention that ruby is 
much much slower then c anyway.

Greets
Marcin Raczkowski