On Fri, Oct 3, 2014 at 10:41 AM, Marti Raudsepp <marti@juffo.org> wrote:
Hi list,
Attached patch allows semijoin/antijoin/hashed SubPlan optimization when an EXISTS subquery contains a LIMIT clause with a positive constant. It seems to be a fairly common meme to put LIMIT 1 into EXISTS() subqueries, and it even makes sense when you're not aware that the database already does this optimization.
I had a quick look at this, and the code looks fairly simple. Although, I've got mixed feelings about it;
I guess there's not really any real performance penalty in planning time for everyone else who does not put LIMIT clauses into their exists subqueries, so maybe it's worth it as it seems there could still be a few people out there suffering from this, but at the same time, the argument for this would have been much stronger if anti join support had just been added last week. It's been quite a few years now and the argument for this must be getting weaker with every release.
I think I'm leaning towards a +1 on this as it seems a shame for people who have no control over the queries sent to their database to have to be excluded from the benefits of semi join and anti join.
