sszabo(at)bigpanda(dot)com writes:
> This seems to generally work in postgres for the simple cases I tried:
> select x from foo group by x order by min(y)
> Now I don't know if there are any hidden gotchas in that (or wierdness
> with the spec), but it also feels better to me than using distinct in this
> case as well, because it seems to explicitly describe how you want y
> ordered.
Yes, I like that better too.
I wonder if we could/should rewrite all uses of DISTINCT into GROUP
BY under-the-hood...
regards, tom lane