On Fri, 2007-11-02 at 14:45 -0700, Joshua D. Drake wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > On Fri, 02 Nov 2007 13:49:27 -0700 > "Jeffrey W. Baker" <jwbaker@xxxxxxx> wrote: > > > Nested Loop Left Join (cost=13920.16..2257575559347.46 > > rows=3691992705807 width=128) > > > > After a call to ANALYZE, the same query gave me: > > > > Merge Left Join (cost=16382.02..16853.87 rows=126768 width=59) > > > > And runs in 5 seconds. If I had been able to tell pg to reject any > > plan with cost over, say 10E9, that would have saved my server from > > half an hour of nested sequential scans. > > I am confused as to why you would want to do that... seems like a > band aid for lack of maintenance. Well it's not "maintenance" really since all the inputs are temp tables, but I do see your point. Often I have wished for a language which is not SQL which would allow me to simply specify the whole execution plan. That would cut out a lot of ambiguity. Pie in the sky, I know. -jwb ---------------------------(end of broadcast)--------------------------- TIP 2: Don't 'kill -9' the postmaster
