Robert Haas <> writes:
> I think the only part of this that would be really brutal to try to
> represent is alternative join orders.  I see no reasonable way for
> EXPLAIN to output useful information about what other join orders were
> considered and why they were not chosen; the only thing that seems
> like it would help in that case would be an easy way to force the
> exact join order you want and then see what the plan looks like.

That exists today: write your query as a nest of explicit JOIN syntax
and set join_collapse_limit = 1.  Not sure that it's really all that
useful, though.

                        regards, tom lane

Sent via pgsql-hackers mailing list (
To make changes to your subscription:

Reply via email to