On Thu, May 02, 2002 at 12:32:06PM +0100, Patrick Welche wrote:
> On Wed, May 01, 2002 at 10:21:48PM -0400, Tom Lane wrote:
> > Patrick Welche <prlw1@newn.cam.ac.uk> writes:
> > > ... the select appears to take much longer to complete if the WHERE clause
> > > matches zero rows, than if it matches some rows.
> >
> > Doesn't make any sense to me. Are you sure the same plan is being
> > chosen in both cases? If so, could you post a complete example?
>
> Yes, same plan both cases. Thanks for the answer: as it doesn't make any
> sense to anyone, it must mean there is some corruption somewhere as a result
> of Monday morning's powercut => I'll look elsewhere.
And indeed a recompile/dump/initdb/restore cured this strange behaviour. The only
sign of a database problem though, apart from the speed, was a single occurance of
"Tuple could not find parent"...
Cheers,
Patrick