socketpair(at)gmail(dot)com writes:
> query:
> select * from tablename order by id limit 1
> and query:
> select distinct * from tablename order by id limit 1
> should be planned exactly (as if distinct was not specified)
Seems like a waste of planning cycles to check for this.
The uniquification step isn't going to take enough time to be
an issue, if only one input row arrives. On the other hand,
the added planning logic would contribute some small overhead
to *every* query using DISTINCT.
regards, tom lane