=?ISO-8859-1?Q?Kirill_M=FCller?= <kirill.mueller@ivt.baug.ethz.ch> writes:
> On 01/12/2012 02:02 AM, Tom Lane wrote:
>> There were fixes for that in 8.4.9, so I'd be interested to know if you
>> get a better estimate in a more up-to-date version.
> Something weird happened today. The problem vanished into thin air. Plus
> we got our server upgrade to 8.4.9, but definitely after the problem
> disappeared. Unfortunately, I didn't have the time to capture an
> execution plan before the upgrade. The database has changed during the
> day, some hundred more tables with geometry columns were added.
> Anyway, I have attached the new execution plan, as produced by server
> 8.4.9 (squeeze/updates/main postgresql-8.4 amd64 8.4.9-0squeeze1+b1).
This doesn't seem to be for the same query --- I only see one antijoin
in this plan, whereas the plan you posted at
http://archives.postgresql.org/pgsql-general/2012-01/msg00303.php
has two (and therefore doesn't match the query given there, either).
regards, tom lane