> On Mon, Mar 21, 2011 at 11:32 AM, <tv@fuzzy.cz> wrote:
>>> Incredible! Setting enable_nestloop off temporarily for the run of
>>> this
>>> script made it run in less than a minute (had been running in 10 or 11
>>> minutes). I think you have found a solution for many of my slow
>>> running
>>> scripts that use these same type of joins. Thanks again.
>>> Julie
>>
>> Nice. Can you post EXPLAIN ANALYZE again, so that we can see why this
>> plan
>> was evaluated as as more expensive before disabling nested loops?
>
> well the problem is obvious -- the planner is estimating ~ 250 loops,
> when it in fact has to do ~ 60k. That's a two orders of magnitude
> miss.
Yeah, you're right, although I think the estimate is 1 loop vs. 2882 loops
in reality. The 250 vs. 60k is related to the result set.
Tomas