Re: Is relational theory irrelevant? (was Re: Dreaming About Redesigning SQL)

Mikito Harakiri wrote:
> Ordering is legitimate mathematical construct. It is a question, though, if
> one can have a simple algebra/calculus that would include ordering too.

For what i can remember from my math lectures, ordering is defined by a relation (math meaning) between members of a set. So in relationnal terms, an order on a relation is another relation that lists all couples of the relation that are linked by the order operator. You can choose to implement it either by giving the full list of couples or by defining the operator. Both of those methods are external to the relation.

The queries you are giving are using an order operator defined on the domain of one of the attribute to implicitely define the "order relation" on the relation.

So, the relationnal algebra already permits to express ordering. I agree with Bob's "Relations have no order", but i do not attach it at the physical level.

I think it is also related to Paul Vernon's paragraph in "Database-valued attributes?" thread : interesting data structures are often represented by 2 or more relations, because additionnal relation is frequently needed to define structure/order.

"data" is one relation and "structure" is another, at least in canonical representation, that are as close as possible to the math definition.

Tom. Received on Wed Nov 12 2003 - 03:48:28 CET