Marc Cousin <cousinmarc(at)gmail(dot)com> writes:
> Anyway here is the problem : when using constraint_exclusion=partition, a
> delete query scans all partitions, when the same query rewritten as a select
> is removing partitions as expected. When constraint_exclusion=on, the
> partition removal works as expected with the delete too. I've tested it on 8.4
> and 9.0
Hmm, this seems like a shortcoming in the constraint_exclusion=partition
feature. The reason it doesn't work is that inheritance expansion of a
DELETE/UPDATE target rel is handled entirely differently from expansion
of a SELECT source. But it seems like it would be desirable if it did
work.
regards, tom lane