Re: Onto a potential relational manipulation language

vadimtro_at_gmail.com wrote:
...
> There is established Logic <-> Algebra correspondence. For
> propositional calculus we have boolean algebra. What algebra do we
> have for predicate calculus? None. I'd suggest that RL is predicate
> calculus without quantifiers and relation attributes.
> ...

Hold on, Vadim! Regarding quantification, I thought Codd's algebra included analogies for Exists and Forall in the projection (fundamental, can't be defined in terms of the other fundamental ops, ie., REMOVE, NAND or NOR, and TCLOSE) and division (defineable in terms of the other ops). Same must be so of D&D A-algebra. If so, RL must at least have quantification since it has a form of projection in its lattice union. If I've got all that right, there must be a way to express Forall in RL with some syntax or other. Am I distorting the situation?

(BTW, the thing I like (given my small knowleCone names the attributes to be projected and the ones that are 'removed' are implicitly the header minus the named ones. But if projection has two operands, it opens the door for perhaps more exotic structures, such as the "multi-relations" Darwen has written about lately (note I'm not saying that he advocates them just because he's written about them), where tuples in the same structure can have different attributes. I gather part of the motivation behind multi-relations is to help deal with so-called "missing information", whereas my attitude so far is that it could just as easily be a way for one structure to allow multiple predicates, which might give some programming leverage, eg., allowing multiple predicate references, even updates in a single structure reference.) Received on Fri Dec 12 2008 - 20:58:34 CET