On 30-12-2011 22:29, Tom Lane wrote:
I poked at this a little bit. AFAICS the only potentially relevant
planner change between 9.0.4 and 9.0.5 was the removal of eqjoinsel's
ndistinct-clamping heuristic,
http://git.postgresql.org/gitweb/?p=postgresql.git&a=commitdiff&h=3505862a8d3e3b389ab9263460
On 30-12-2011 19:35, Merlin Moncure wrote:
try this (curious):
create table pos as select n from generate_series(1,32) n;
and swap that for the in-query generate series call. your statistics
in the query are completely off (not 100% sure why), so I'm thinking
to replace that since it lies to th
