After some extensive discussion on irc, berkus, myself and a few others
think we have uncovered a possible bug, or at the least some odd
behavior in > 8.1.1. It centers around my recent post to performance
http://archives.postgresql.org/pgsql-performance/2006-01/msg00248.php
and how I could not seem to get some of the estimates to become
reasonable even after bumping up my stats target to 400 which caused
every row to be analyzed. If you look at the left join and hash join
estimates of the third query you'll note they seem to always get
estimated to 1 for no reason that we could come up with.
Someone else on irc seemed to have a similar problem to this, so we are
wondering if there is some problem here. So the question really is if
someone can deduce the behavior from looking at what was provided in the
email? If not and you have questions let me know, otherwise I can send a
chopped up test database which can reproduce the query issues off list
should someone want to walk through the pg code to investigate. TIA
Robert Treat
--
Build A Brighter Lamp :: Linux Apache {middleware} PostgreSQL