Re: comparing NEW and OLD (any good this way?)

On Thu, Aug 13, 2009 at 11:53:49PM +0100, Greg Stark wrote:
> On Thu, Aug 13, 2009 at 11:44 PM, Daniel Verite<daniel(at)manitou-mail(dot)org> wrote:
> >> In other discussions about similar issues I've said that the expression:
> >>
> >> ROW(NULL,NULL) IS DISTINCT FROM NULL
> >>
> >> should evaluate to FALSE. I still think this is correct and generally
> >> useful behavior.
> >
> > I see no reason to disagree with this. Besides, the fact that
> > ROW(NULL,NULL) IS DISTINCT FROM NULL evaluates to true
> > while ROW(NULL,NULL) IS NULL also evaluates to true
> > looks quite puzzling to me.
>
> Why is this thread still going on?

Because I'm a stickler for details and people keep replying!

> What does the spec say we should be
> doing and are we violating it in any of these cases?

Whenever I've looked through I've not found anything definite either
way. I think my interests here are more pedagogical that anything else,
but PG's behavior is somewhat inconsistent and it could be nice to
figure out what the "best" way of fixing these inconsistencies are.

--
Sam http://samason.me.uk/