Re: 9.3 Pre-proposal: Range Merge Join

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> On Apr 16, 2012, at 1:40 AM, Jeff Davis <pgsql(at)j-davis(dot)com> wrote:
>> See attached SQL for example. The
>> Problem statement: slow. Nested loops are the only option, although they
>> can benefit from an inner GiST index if available. But if the join is
>> happening up in the plan tree somewhere, then it's impossible for any
>> index to be available.

> Hmm. This sounds like something that Tom's recent work on
> parameterized plans ought to have fixed, or if not, it seems closely
> related.

Not really. It's still going to be a nestloop, and as such not terribly
well suited for queries where there are a lot of matchable rows on both
sides. The work I've been doing is really about making nestloops usable
in cases where join order restrictions formerly prevented it --- but
Jeff's complaint has nothing to do with that. (This thought also makes
me a bit dubious about the nearby suggestions that more indexes will
fix it.)

regards, tom lane