Re: Help! I can't support normalization

Lauri Pietarinen <lauri.pietarinen_at_atbusiness.com> wrote in message news:<3DDA5DFD.6020503_at_atbusiness.com>...
> >
> >
> >I'm often having to programatically create tables containing integers
> >1 to 100 say, it would be good if this was built-in so I could just do
> >"SELECT integer FROM domain_integer WHERE integer BETWEEN 1 AND 100".
...
>
> It proposes a table called TableNum with just one column containing
> integers from 1 to ....
> It also has some proposed applications.
>
> Now, would such a table be a virtual or a real one??

Well, it depends what you man by "real". In a sense all tables are virtual because they are purely logical constructs. But what would it mean say to delete a row from such a "domain relation"? Really it shouldn't be allowed.

I guess what you're meaning is that we would have two categories of table: ones where the rows are listed explicitly and ones where the rows are defined in some kind of induction process. These second sort of tables you could only SELECT from, as INSERT, UPDATE, DELETE might not make sense. Or maybe they would? Say I wanted to change my integer domain to be all integers (within a certain range) excluding "13"? Should I be allowed to do this or should the "built-in" domains remain as they are and I should define a new domain based on, but separate to, the integer domain (type inheritance?)?

I've only briefly read about them but is this moving towards things like constraint databases?

Does this violate Codd's principles? Should all relations be able in theory to be manipulated by any SQL DML statement?

Paul. Received on Wed Nov 20 2002 - 13:07:29 CET