Greg Stark <gsstark(at)mit(dot)edu> writes:
> I think this is just a bad case of "nested loops are costed wrong".

Looks to me like a statistical failure.  Why does it think there will be
4000 rows out of that join when there are only 93?

			regards, tom lane

In response to

• Bad planner results at 2004-08-24 07:41:06 from Greg Stark

pgsql-general by date

Next:	From: Tom Lane	Date: 2004-08-24 15:04:49
	Subject: Re: pg_dump in stand alone backend
Previous:	From: Jeff Amiel	Date: 2004-08-24 14:56:54
	Subject: pg_stat_activity versus ps