-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
I've checked in a few quick tweaks to cvs, just fyi: * wfCheckLimits() function factors out some of the $offset/$limit-checking common to a lot of special pages * some old str_replace("$1"...wfMsg()) sequences replaced with wfMsg("...","...") * Preliminary support for mysql4 fulltext boolean mode search (not done yet: need to split up terms for highlighting and make sure intl conversion is done) * Very preliminary patches for adding license fields to the image table. Not even checked for syntax, and not supported by any code yet.
Just a quick though also: it _would_ be nice to be able to search for eg "C++" or "C#". Mysql's fulltext search I don't think will preserve these, but in theory we can armor them the same way we do for utf-8 high characters.
Also: http://www.mysql.com/doc/en/Fulltext_Search.html scroll down for the boolean mode notes. For people used to google-style searches (or our current search method), the default behavior being 'or' instead of 'and' may be confusing. We could force adding +s to the search terms to change the behavior, but this could be tricky to do right, and we'd probably want a way to undo it.
Also, the phrase search ("foo bar") may be problematic with our handling of links, where "biz [[foo|bar]] bax" is indexed as "biz foo bar bax" and "biz [[foo]]bar bax" as "biz foo foobar bax" or some such.
Current cvs running on test.wikipedia.org.
- -- brion vibber (brion @ pobox.com)
wikitech-l@lists.wikimedia.org