I've tested this under 7.3, and it works beautifully for the cases I've built over the last 2 days. I can no longer bugger a plan up mearly by reordering the WHERE clauses. Note that 2 of the five parts won't patch in (involving constantqual). Looks to be code refactoring between here and planmain.c on the 7.4 branch? I tried to hand-patch it in, and gave up. it _seems_ to work without it, but I probably haven't covered that codepath.
Ross On Thu, Nov 14, 2002 at 01:33:05PM -0500, Tom Lane wrote: > I said: > > Well, we could define it as a bug ;-) --- that is, a performance regression. > > I'd be happier about adding a dozen lines of code to sort quals by > > whether or not they contain a subplan than about flip-flopping on the > > original patch. That would actually solve the class of problem you > > exhibited, whereas the other is just a band-aid that happens to work for > > your particular example. > > The attached patch does the above. I think it's a very low-risk change, > but am tossing it out on the list to see if anyone objects to applying > it in the 7.3 branch. (I intend to put it in 7.4devel in any case.) > > regards, tom lane ---------------------------(end of broadcast)--------------------------- TIP 2: you can get off all lists at once with the unregister command (send "unregister YourEmailAddressHere" to [EMAIL PROTECTED])