RE: Speaking of New Features (named pairs of values)
Date: Thu, 10 Sep 2009 22:25:50 -0400
without being either an advocate or against the syntax, I don't think he is suggesting eliminating any existing valid syntax.
But even if he was, then
insert into t select ( a => x, b => y) from r;
would seem to meet the formalism.
If you prefer this syntax for source code there really is no impediment to hauling out something like yacc and writing it. You'll just have to pass over your "source" code before you shove it at a SQL parser.
Heh, you could probably write it in PERL. and start using it in a few days.
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org]
On Behalf Of Michael Moore
Sent: Thursday, September 10, 2009 7:28 PM To: oracle-l
Subject: Re: Speaking of New Features
If it was as chet suggested, then how could you do
insert into t (a,b) select x,y from r;
It would mean you would need to have two valid syntaxes for the INSERT.
1.) insert into t(a,b) values ('a','b');
2) insert into t (a => 'a', b=>'b');
both would need to be valid. Not that that is a terrible thing but if you are going to have both 'named' and 'positional' for INSERT, then you would probably want both forms available for UPDATE as well. MikeReceived on Thu Sep 10 2009 - 21:25:50 CDT