On Sun, 28 Sep 2003, Bruce Momjian wrote:
> Stephan Szabo wrote:
> > Hmm, my initial testing showed that it really was a little slower
> > than a more complicated one with NOT EXISTS so I'd abandoned it. How does
> > it fare for you compared to:
> > select f1, f2 from fk where not exists (select 1 from pk where pk.f1=fk.f1
> > and pk.f2=pk.f2) where fk.f1 is not null and fk.f2 is not null;
> >
> > I believe the above is the appropriate not exists form for match
> > unspecified.
> >
> > I've actually got code (that no longer cleanly applies, but...) that uses
> > the single query version with NOT EXISTS (which could be easily changed to
> > either of the other forms) and was planning to put it together for a patch
> > when 7.5 devel started because I figured it wasn't precisely a bug and
> > wouldn't get accepted for 7.4.
>
> I am a little lost on this point myself --- are we talking 7.4 or 7.5
> for this change?
I'd thought 7.5, but I now see that it's on the 7.4 open items list.
---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
joining column's datatypes do not match
