On 1/14/14, 6:15 PM, Tom Lane wrote:
> We don't actually implement this in PG yet, except for trivial cases, but
> it will certainly happen eventually. I think your sketch above deviates
> unnecessarily from what the standard says for UPDATE. In particular
> I think it'd be better to write things like
>
> (a, b) = ROW(1, 2);
> (a, b, c) = (SELECT x, y, z FROM foo WHERE id = 42);
>
> which would exactly match what you'd write in a multiple-assignment
> UPDATE, and it has the same rejects-multiple-rows semantics too.
Just in case someone's interested: I won't be working on this for 9.5.
If someone feels like picking this patch up, be my guest.
.marko
