Frank van Bortel wrote:
> Deltones schreef:
> > I posted a question last week where people where some of the guys who
> > replied recommanded that I use "create global temporary table" instead
> > of "create table". I tried it with the script I've been asked to modify
> > that contained the latter construction.
> >
> > I get my results every time with the "create table" version, but if I'm
> > lucky, I get results maybe 50% of the time with the "create global
> > temporary table". Take note that it's the same queries that are run in
> > both cases and that I also tried the "on commit delete rows", "on
> > commit preserve rows" clause, and that the tables are dropped at the
> > end. Why does it do this?
> >
> > Are tables created with "global temporary" different that regular
> > tables?
> >
> > Thanks
> >
> > Denis
> >
> Yes - they exist for the duration of the session, if not shorter.
> And you do not "use the create global temporary table everytime",
> you just use it once.
>
> Your style reminds me of dBaseIII and Foxbase I used to program.
>
> And if you get the wrong results, you might run into a bug,
> in which case versions do matter enormously!

It's the style of the script I have to modify, not mine. The main query
to create the first temp table is an 800 lines abomination

As for the duration of the session, I get that, but here what I mean. I
run the script, I get my results. Then I run the same script, let's say
15 seconds later, I don't get anything.

But with the regular "create table" version, I get my results every
time. That's what confuses me.