On Mon, Dec 19, 2005 at 10:08:12AM -0500, Tom Lane wrote: > "Let's not fix it" is really not an acceptable answer, because the > behavior in the affected locales is entirely broken (inconsistent, > etc). And how do you know which locales are affected, anyway? I don't know which locales are affected. It just can't be that widespread because we're not getting similar reports for 99% of the locales out there. I'm not saying not fixing it is an option, it's just that a second compare on equality is going to be a waste in >99% of cases. > (Also, to be blunt, the COLLATE feature is pie in the sky until it > is done and supports indexes --- which I'm not sure you even know > how to do yet. I'm not prepared to assume it will be in 8.2.) Oh, getting indexes to use COLLATE is (relatively) the easy part. Where I'm mostly stuck on is teaching the planner how to decide which index returns the right order for what it's looking for. For example, should I be returning a seperate path for each possible collation order. Obviously not, but there will certainly be several orders returned depending on indexes, maybe others depending on any ORDER BY clauses in the query. Getting *that* right is harder and my planner knowledge isn't that great. Right now I'm updating the code for CVS tip. pg_indent has a great way of making large patches cease to work :(. Some other things seem to have changed also. Oh well... It's not a huge amount of work, it's just lots of interconnected pieces. Have a nice day, -- Martijn van Oosterhout <kleptog@xxxxxxxxx> http://svana.org/kleptog/ > Patent. n. Genius is 5% inspiration and 95% perspiration. A patent is a > tool for doing 5% of the work and then sitting around waiting for someone > else to do the other 95% so you can sue them.
Attachment:
pgpxlrrs2fwKK.pgp
Description: PGP signature
