On 17/12/16 23:04, Tom Lane wrote:
> so that you don't need to use
> SELECT DISTINCT? The sort/unique steps needed to do DISTINCT are
> eating a large part of the runtime,
Does a hash join result in a set of buckets that are then read out
in order? It might, unless the sort method takes advantage of
partially-sorted inout, be cheaper (by log(num-buckets)) to sort/uniq
each bucket separately (and it would parallelize, too).