On Thu, 20 Feb 2003, Tom Lane wrote:
> Hannu Krosing <hannu@tm.ee> writes:
> > Are you against it just on grounds of cleanliness and ANSI compliance,
> > or do you see more serious problems in letting it in ?
>
> At this point it seems there are two different things being tossed
> about. I originally understood Dave to be asking for parens to be
> allowed around individual target column names, which seems a useless
> frammish to me. What Bruce has pointed out is that a syntax that lets
> you assign multiple columns from a single rowsource would be an actual
> improvement in functionality, or at least in convenience and efficiency.
> (It would also be a substantial bit of work, which is why I think this
> isn't what Dave was offering a quick patch to do...) What I'd like to
> know right now is which interpretation Informix actually implements.
>
> I don't like adding nonstandard syntaxes that add no functionality ---
> but if Informix has done what Bruce is talking about, that's a different
> matter altogether.
Tom, I was purusing the wild and wonderfully exciting new SQL
(found here:
ftp://sqlstandards.org/SC32/WG3/Progression_Documents/FCD/4FCD1-01-Framework-2002-01.pdf)
ANSI TC NCITS H2
ISO/IEC JTC 1/SC 32/WG 3
Database
document to see what it had to say, and on this subject, and it looks like
update is going to be supporing this same style we're discussing here.
Look on or around p. 858 in that doc.)