On Mon, 2006-07-31 at 17:26 +0200, Peter Eisentraut wrote:
> Am Mittwoch, 26. Juli 2006 22:58 schrieb Tom Lane:
> > The reason people want this syntax is that they expect to be
> > able to write, say,
> >
> > UPDATE mytab SET (foo, bar, baz) =
> > (SELECT alpha, beta, gamma FROM othertab WHERE key = mytab.key);
>
> I don't find any derivation in the standard that would permit this. The only
> thing I could find are variations on
>
> SET (a) = x -- no parentheses
> SET (a, b) = (x, y)
> SET (a, b) = ROW (x, y)
>
> where x and y are some sort of value expression. I would have expected the
> sort of thing that you describe, but if you know how to derive that, I'd like
> to see it.
I believe <contextually typed row value constructor element list> can be
one or more <value expressions> which includes a <row value expression>.
<row value expression> gives us the <row subquery> option.
For that matter the below portion of <contextually typed row value
constructor> gives us:
| <left paren> <contextually typed row value constructor element>
<comma>
<contextually typed row value constructor element list> <right
paren>
This breaks down into one or more comma separated <row subquery>s.
UPDATE tab SET (...) = ((SELECT foo, bar from a), (select bif,baz from
b));
--
---------------------------(end of broadcast)---------------------------
TIP 3: Have you checked our extensive FAQ?
http://www.postgresql.org/docs/faq
