On Fri, 2009-07-10 at 01:36 +0100, Greg Stark wrote:
> Arguably the missing feature here is skip-scans where we scan the
> index but only pull out one record for each distinct value. I'm not
> sure there's anything particularly stopping Postgres from being able
> to do them, but it might be a lot of code for a narrow use case.
Hypothetically, would something like a "sort distinct" operator be of
any use? I wonder how much work it would save if the sort could save
steps by weeding out duplicate tuples while sorting. That might make
sort into a better plan in cases where don't have a good estimate of the
distinct values.
Regards,
Jeff Davis