Martijn van Oosterhout <kleptog(at)svana(dot)org> writes:
> On Fri, Nov 11, 2005 at 11:05:45AM +0100, Matteo Beccati wrote:
>> It seems that the plan outputted is not the optimized one (available
>> since 8.1) that is really used when running the plain query.
> It may also be that the overhead of calling gettimeofday() several
> times per tuple is blowing the time out. What platform is this?
Martijn's explanation is by far the more probable. The high overhead
of EXPLAIN ANALYZE has been documented before.
regards, tom lane