Re: Logical equivalence of simple and complex types under the relational model?

From: Paul <paul_at_test.com>
Date: Wed, 01 Dec 2004 22:47:09 +0000
Message-ID: <41ae49ed$0$52991$ed2619ec@ptn-nntp-reader03.plus.net>

Rene de Visser wrote:

>> The way I see it, DBMSs consist of two (orthogonal) components: a
>> type system and a relational system. If the DBMS can decompose a
>> value using the relational system alone, it's a complex value. If
>> it needs the help of the type system to decompose the value, it's a
>> simple value.

>
> What exactly do you mean by "if it needs the help of the type system
> to decompose the value, its a simple value"?
>
> How do you know if the relational system is "using the help of the
> type system"?

Well to decompose the value, you could either use a relational operator (such as JOIN, UNION, etc.) or a type operator (such as SUBSTRING, MONTH, etc.). But the standard relational operators can't be used to decompose values, thus all values are simple.

If you are using some kind of amended relational model with some relational operator like UNPACK or something that can decompose values, the values are complex.

> In some senses the relational system is always using the type system
> for everything. i.e. you can't have tuples without relational
> variables and domains, and without types you don't have these.

Agreed, so from this I deduce that all values are simple (in the standard relational model).

Paul. Received on Wed Dec 01 2004 - 16:47:09 CST