Re: Towards a definition of atomic

From: Bob Badour <>
Date: Fri, 01 Feb 2008 12:15:00 -0400
Message-ID: <47a34585$0$4044$>

David Cressey wrote:

> "Roy Hann" <specially_at_processed.almost.meat> wrote in message

>>"David Cressey" <> wrote in message
>>>>I think we could make the meaning of "atomic" more tangible if we can
>>>>define what decompositions of an attribute are valid.
>>>Perhaps the place to start is to define what kinds of compositions a
>>>relational system is capable of.  Once you have that in place,  it

> should
>>>straightforward to define relational decompositions as the inverse of
>>>relational compositions.
>>Why not just understand that relational systems don't care about about
>>composition/decomposition and want nothing to do with the idea?  It is no
>>more relevant than is the concept of colour to Euclidean geometry.

> I disagree. A relation is composed of attributes. (If you prefer, a table
> is composed of columns). Relational systems are clearly concerned with
> composition in some contexts.
> One of the relational operators discussed is the "compose" operator, by
> which a result relation can be composed from given relations. (If you
> prefer, a result table can be composed from given tables (or views)).

Normalization likewise involves decomposition by PROJECT and composition by JOIN.

The important thing in this discussion is: The relational model treats domain values as neither composable nor decomposable. They are simply values upon which one may evaluate defined operations.

Thus, no decompositions of attributes are meaningful from the perspective of the RM. Received on Fri Feb 01 2008 - 17:15:00 CET

Original text of this message