Re: Non-standard feature request

Rocco Altier wrote:
> 
> On Fri, 14 Jun 2002, Mike Mascari wrote:
> 
> > That is what I want to do, except by extending the grammar. I must admit
> > to actually being surprised that a TEMP table created inside a
> > transaction lived after the transaction completed. That's when I looked
> > at the standard and saw that PostgreSQL's implementation was correct. I
> > would think for most people session-long temp tables are more the
> > exception than the rule. But I guess SQL92 doesn't think so. Regardless,
> > a couple of other people have shown some interest in the idea. I'll post
> > it to general as well as Tom suggests...
> >
> Actually, we needed to use temp tables that live beyond the transaction,
> because there are no session variables in postgres.  So I did an
> implementation that used temp tables instead.
> 
> Having the temp table not live for the life of the session would be a big
> problem for me.

Sure, which is why I'm proposing to extend the grammar. Only if you
created the temporary table with

CREATE TEMPORARY TABLE
...
ON COMMIT DROP;

would it drop the temporary table at transaction commit. It should be
100% compatible with existing code. 

Mike Mascari
mascarm@mascari.com