Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com> writes:
> On 05/31/15 13:00, Peter J. Holzer wrote:
>> (There was no analyze on facttable_stat_fta4 (automatic or manual) on
>> facttable_stat_fta4 between those two tests, so the statistics on
>> facttable_stat_fta4 shouldn't have changed - only those for term.)
> So maybe there was autoanalyze, because otherwise it really should be
> the same in both plans ...
No, because that's the inside of a nestloop with significantly different
outer-side rowcount estimates. The first case gets a benefit from the
expectation that it will be re-executed many times (see the impact of
loop_count on cost_index).
regards, tom lane