Re: Help! I can't support normalization

From: Lauri Pietarinen <lauri.pietarinen_at_atbusiness.com>
Date: Wed, 20 Nov 2002 23:22:17 +0200
Message-ID: <3DDBFD09.7030903_at_atbusiness.com>

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 wouldn't call it a "domain relation", just a relation like any other one.

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?)?

Well, I think that TableNum is a special case. As for excluding '13' all you have to do
is  'select N from TableNum where N <> 13'

My opinion is that relations and domains are basically different things.  Both are needed but
they suffice.

The "real world" is like a piece of soap:  when you try to completely model it, it just slips from your
hands.  So what is a relation and what is a domain really depends on what angle
of the "real world" you are taking.

The concept of 'color' could be a domain in one application (say a product catalog), but
in a paint-factory it would probably consist of several relations (anybody know,
by the way?)

I quote from Date: Introduction to Databases (7ed) pages 66-67:

  -Types [=domains] are (sets of) things we can talk about
  - Relations are (sets of) things we say about the things we can talk about

[...]

  Types [=domains] are to relations as nouns are to sentences

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?

I wouldn't think so.  Why would it?

regards,
Lauri Pietarinen
Received on Wed Nov 20 2002 - 22:22:17 CET