Re: [grep] about 200 times slower than Sed or Perl

[Paolo Bonzini]

> The problem is indeed related to the locale. With non-UTF-8 locale, I
> get the usual speed. With an UTF-8 locale and "-v MAC" I get the
> breakdown described in the original report. But grep -v MÄC (for
> example, an UTF-8 string) is fast, OTOH. Very confusing results.

Confusing but easy to explain.  grep treats MÄC as a multibyte string, perl
ignores the locale completely, and sed (via glibc) has a special optimization
that is specific to the UTF-8 locale that allows it to run at its full speed
(which is slower than grep).  If you put in the ring super-sed as well
(http://sed.sf.net/grabbag) you'll find that it ignores the locale completely
as well, and is a bit faster than perl.

Paolo