Re: NULLs: theoretical problems?

From: Evan Keel <evankeel_at_sbcglobal.net>
Date: Wed, 22 Aug 2007 19:12:24 -0400
Message-ID: <Ip3zi.4527$LL7.553_at_nlpi069.nbdc.sbc.com>

"Jan Hidders" <hidders_at_gmail.com> wrote in message

news:1187811230.504947.11400_at_x40g2000prg.googlegroups.com...

> On 22 aug, 17:37, "V.J. Kumar" <vjkm..._at_gmail.com> wrote:
> > Jan Hidders <hidd..._at_gmail.com> wrote
innews:1187787815.215103.100820_at_k79g2000hse.googlegroups.com:
> >
> >
> >
> > > On 22 aug, 13:23, "V.J. Kumar" <vjkm..._at_gmail.com> wrote:
> > >> Jan Hidders <hidd..._at_gmail.com> wrote
> > >> innews:1187766113.827952.167510_at_i38g2000prf.googlegroups.com:
> >
> > >> > On 22 aug, 00:06, "V.J. Kumar" <vjkm..._at_gmail.com> wrote:
> > >> >> Jan Hidders <hidd..._at_gmail.com> wrote
> > >> >> innews:1187729150.610272.117790_at_r29g2000hsg.googlegroups.com:
> > >> >> >> I do not understand. You have:
> >
> > >> >> >> DEF y y DEF y:y
> > >> >> >> 1 1 1 (1)
> > >> >> >> 1 0 0 (2)
> > >> >> >> 0 0 (3)
> >
> > >> >> >> So 'DEF y:y' will give the same result when y is either
> > >> >> >> undefined or 'false', rows (2) and (3). How is it not
> > >> >> >> substituting 'false' for undefined ?
> >
> > >> >> > In the way that if y is undefined then "DEF y : f" is not always
> > >> >> > equivalent with "f[y/false]" i.e. "f" with all free occurrence
> > >> >> > of y replaced with "false".
> >
> > >> >> I do not understand. Could you show what you mean with an example
> > >> >> ?
> >
> > >> > If y is undefined then "DEF y : NOT(y)" evaluates to "false".
> >
> > >> Is it not what line(3) shows and what SQL queries do, namely
> > >> substituting 'false' for unknown ?
> >
> > > Yes, it is what line 3 shows, but I would not describe that as
> > > "substituting 'false' for unknown".
> >
> > In your language, the expression 'def y:x AND y' where 'y' is
> > 'undefined' evaluates to 'false'. In SQL, the expression 'x AND y'
where
> > 'y'is 'unknown' evaluates to 'unknown'. The effect of having a
predicate
> > that evaluates to 'unknown' is the same as having a predicate that
> > evaluates to 'false': no rows will be selected. That's what I meant by
> > "substituting 'false' for unknown".

>

> That's not exactly the same because there are formulas f(x) that
> evaluate to 'true' if x is 'unknown'.

>

> -- Jan Hidders

>
You wrote formulas f(x) that
> evaluate to 'true' if x is 'unknown'.

You meant a function, right?

Null not only has theoretical problems, but worse, practical problems:

Null belongs to no domain. So you can do nothing with a null.

Null has no arithmetic operators.

Null cannot be seduced by any logical operators.

I met Ted Codd once and our discussion centered on domains.

I think null as semantic construct makes a lot of sense. But as a data type? Null is a datatype (when you think about it) and not a biz rule. Received on Thu Aug 23 2007 - 01:12:24 CEST