Re: idea for concurrent seqscans

On Fri, Feb 25, 2005 at 04:30:17PM -0800, Jeff Davis wrote:
> On Fri, 2005-02-25 at 18:03 -0600, Jim C. Nasby wrote:
> > On Fri, Feb 25, 2005 at 01:30:57PM -0500, Tom Lane wrote:
> > > Jeff Davis <jdavis-pgsql(at)empires(dot)org> writes:
> > > > I didn't consider that. Is there a reason the regression tests assume
> > > > the results will be returned in a certain order (or a consistent order)?
> > >
> > > We use diff as the checking tool.
> >
> > Doesn't the SQL spec specifically state that the only time you'll get
> > results in a deterministic order is if you use ORDER BY? Assuming
> > otherwise seems a bad idea (though at least in the case of testing it
> > makes the test more strenuous rather than less...)
>
> True, that was my reasoning when I proposed synchronized scanning.
>
> Keep in mind that this is a criticism of only the regression tests, not
> the RDBMS itself.
>
> I don't know much about the regression tests, so maybe it's impractical
> to not assume consistent order. I'm sure the developers will vote one
> way or the other. I hate to throw away a potential performance boost,
> but I also hate to burden the developers with rewriting a lot of
> regression tests when their time could be better spent elsewhere.

Certainly, but I suspect it's just a matter of adding ORDER BY to
everything, which just about anyone (even myself!) should be able to do.
--
Jim C. Nasby, Database Consultant decibel(at)decibel(dot)org
Give your computer some brain candy! www.distributed.net Team #1828

Windows: "Where do you want to go today?"
Linux: "Where do you want to go tomorrow?"
FreeBSD: "Are you guys coming, or what?"