Re: [HACKERS] Phantom row from aggregate in self-join in 6.5

Thomas Lockhart <lockhart(at)alumni(dot)caltech(dot)edu> writes:
>> IIRC, you were the main advocate of the position that the code's
>> existing behavior is correct. Does that mean I can go change it? ;-)

> Yes, after you slap me around a bit for being so wrong. Do you
> remember when we were discussing it? I want to go back and see why I
> thought this was right. I'm guessing that the example was not phrased
> in exactly this way, and that there may be some other behavior we need
> to maintain. (Otherwise, I might have used up my "one wrong idea per
> year" ;)

Actually, it may be my recollection that's wrong. The only discussion
of the point that I can find right now is the thread "SUM() and GROUP
BY" from around 1/12/99 in pghackers, and it seems to be mostly focused
on arguments about whether you should get NULL or 0 from a no-input
SUM...

I would've sworn I remember a couple of other related threads in the
past year or so, but I cannot find them now.

Anyway, unless someone speaks up in favor of the way the code currently
works, I will see about changing the results for the GROUP-BY-with-no-
input-rows case. I got a few other things to do first though.

regards, tom lane