[pLog-svn] db optimization

[Christoph Feddersen]

Mark Wu wrote:
> Wow... That's nice! Thanks for such detail  analysis.
> 
> BTW, there are two reasons that why we did not use fulltext index:
> 
> 1. For mysql version compatability:
> 
> Our original idea is support mysql DB from 3.2.x to 4.0.x, 4.1.x and 5.x. As
> I know Full text index only work in the version that higher then 4.0.x....
> (Maybe we can give up the mysql 3.2.x version, not sure how many people
> still use it?)
There's fulltext support in > 3.2.23
http://dev.mysql.com/doc/refman/4.1/en/fulltext-search.html

> 2. For mulitiple language support:
> 
> The most bad things is, mysql fulltext index does not support CJK (actually,
> all double bytes languages), even I use the utf-8 as charset and collation.
I only have to work with multi byte languages from time to time, so that
 I'm not that experienced with it, but at least MySQL 4.1 should support
it when using UTF-8, BUT with some restrictions:

You need to set the minimum word length to 1, which results in a big index:
set ft_min_word_len = 1

You'll have to parse the search words and convert all UTF-8 (or other MB
spaces, depending on the used charset) spaces to ASCII spaces.