Re: apply_scanjoin_target_to_paths and partitionwise join

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> I'm obviously missing something here, because I'm sure Jakub is quite
> right when he says that this actually happened and actually hosed an
> EDB customer. But I don't understand HOW it happened, and I think if
> we're going to change the code we really ought to understand that and
> write some code comments about it. In general, I think that it's very
> reasonable to expect that a bunch of small joins will beat one big
> join, which is why the code does what it currently does.

I am wondering if the problem is not that the plan is slower, it's
that for some reason the planner took a lot longer to create it.
It's very plausible that partitionwise planning takes longer, and
maybe we have some corner cases where the time is O(N^2) or worse.

However, this is pure speculation without a test case, and any
proposed fix would be even more speculative. I concur with your
bottom line: we should insist on a public test case before deciding
what to do about it.

regards, tom lane