Re: NULLs: theoretical problems?

On 25 aug, 01:53, Bob Badour <bbad..._at_pei.sympatico.ca> wrote:
> Jan Hidders wrote:
> > On 24 aug, 16:35, "V.J. Kumar" <vjkm..._at_gmail.com> wrote:
>
> >>Jon Heggland <jon.heggl..._at_idi.ntnu.no> wrote innews:famjo6$i68$1_at_orkan.itea.ntnu.no:
>
> >>>In other words, the DEF query is equivalent to
>
> >>>SELECT * FROM t WHERE t.a IS NOT NULL AND (t.a = 5 OR TRUE)
>
> >>My understanding was that the DEF as written was a shorthand:
>
> >>e.g.
> >>'def(x):((x or y) and x)' -> '((def(x) or y) and def(x))'
>
> >>You may be right, but then why the formula was not written with an
> >>explicit 'and' ?
>
> > Because it does not satisify all the logical laws of an AND, so to
> > avoid confusion another notation is used. Since it's related to an
> > existential quantifier a similar notation was chosen (think of EXISTS
> > x : f).
>
> > And, yes, as Jon correctly remarked it is in fact equivalent with
> > something you can already write in SQL so you could achieve the same
> > with some self-discipline.
>
> Let me suggest the use of 6th normal form and then EXISTS is all one
> would need.

Why does everybody keep thinking that normal forms say something about null values!?(*) ;-) But I of course understand what you mean, and yes, that is all one would need, but that doesn't necessarily mean we cannot think about trying to go a little beyond the bare necessities.

(*) If they do then it is moste likely already 1NF that does so since it defines what a relation is and that definition would normally not include null values.

• Jan Hidders