On Tue, Feb 19, 2008 at 1:49 PM, Bryan Oakley <[EMAIL PROTECTED]> wrote: > My code uses fts2, and for the first time today I discovered that no > stemming appears to be going on under the covers. Googling around I > see that I need to add "tokenize porter" when creating my table. > Adding that seems to improve the results I get back. > > However, this seems to be an undocumented feature. Is there a reason > it's undocumented, and should I avoid using it?
I'm not sure how you're qualifying "documented" versus "undocumented", here. fts is pretty poorly documented. > Also, I've seen references to fts3 -- is there a compelling reason for > me to switch to fts3 from fts2? fts2 has a design flaw which can cause corruption of the fts2 index if you run VACUUM against the database. If you never run VACUUM, you'll be fine. fts3 is mostly identical to fts2, with this flaw fixed. I'd link to the email thread about this, but ... I can't find it. It was back in July or so. The current versions of fts1.c and fts2.c throw a #error when you compile, if you look into the code to see why you'll see a comment about the issue. -scott _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
