Re: Complete axiomatization of relational algebra

David Cressey wrote:

> "Neil Nelson" <n_nelson_at_pacbell.net> wrote in message
> news:3A8DDF4E.CCEEC550_at_pacbell.net...
> > elements each having a uniform set of properties. I.e., we should be able
 to
> > `denormalize' to a single table. The primary operation on this table is
> to ask
>
> I've never carried out this 'denormalization', so what follows is
> speculative.
>
> I think that, in order to completely denormalize a schema to a single table,
> you need to allow outer joins as well as inner joins and unions. When you
> carry out outer joins, you can get some NULLS in some columns.
>
> An interesting exercise, occasionally, is to take a column that permits
> NULLS and discover a pair of decomposed tables
> and an outer join that would have resulted in the single table with optional
> column we were given.
>
> I did this exercise a few times. It resulting in my discovery of more and
> more about less and less, until I finally knew a lot about nothing at all!
> ;) ;) ;)

I expect that if you are looking for nothing (nulls) you may have a good chance of finding it. But of course if one wants to avoid having the logic operate on nulls by accepting a row because of a null satisfying a negated monadic predicate (assuming one does not want nulls) then the default correction is immediate.

Clearly the resulting denormalized single table is not efficient nor elegant, and it would be rare that someone would care to use such a table. However the point was that there exists a fairly simple common model that identifies what we typically expect of data structures. The model is of such simplicity that it would be unlikely that some algebra or calculus has not been developed, or if in the very unlikely case that one has not been developed, it could be expected that such a calculus could be easily developed.

Regards,

Neil Nelson Received on Sat Feb 17 2001 - 16:07:29 CET