Thomas Hallgren wrote:
> Steven,
>
> > That assumes that developers will implement queries in their code
> > without testing them. Unfortunately, that's probably not too far from
> > reality. I've thought of it as a nice "debugging" feature while I'm
> > trying to hammer out a complicated query for the first time.
>
> I don't see how that makes a difference really. As a developer, I'd
> rather prefer if I get an explanatory error result rather than a notice
> (often invisible) and an incorrect result when testing. If I don't test
> at all (God forbid) I want the same thing to happen the first time the
> code is deployed.
This particular error may be less than obvious even during crude testing
because the number of tuples in the relation in question may be zero or
one, so the cross-product wouldn't produce anything unusual.
Mike Mascari