Neil Conway <neilc@samurai.com> writes:
> Interesting -- I wonder if it would be possible for the optimizer to
> detect this and avoid the redundant inner sort ... (/me muses to
> himself)
I think the ability to generate two sort steps is a feature, not a bug.
This has been often requested in connection with user-defined
aggregates, where it's handy to be able to control the order of arrival
of rows at the aggregation function. If the optimizer suppressed the
inner sort then we'd lose that ability.
regards, tom lane