On 13 Mar 2011, at 00:12, Andreas Gohr wrote: >> I profiled using xdebug. >> >> My test data is the Pro Git book. >> http://whoopdedo.org/doku/_media/progit-data.tar.gz >> >> Individual functions were evaluated simply > > Did you try different input texts? AFAIR we had issues where > performance differed substantially on different input lengths... Would > be interested to have numbers for short (<25 chars), medium and large > texts in different encodings (ASCII-only, some UTF-8 chars (like > German), all UTF-8 chars (like Chinese)). > I recall, that some time ago, we did pretty extensive testing on the utf8 functions, before deciding on the functions that were used - although that might just have been utf8_strlen and utf8_substr. I wonder if the real change here is updates to PHP. It would be interesting to run Tom's tests on PHP 4.x and see if they come up with similar or different results. - Chris -- DokuWiki mailing list - more info at http://www.dokuwiki.org/mailinglist