Re: (domain support) Re: theory and practice: ying and yang

Really good question.

Now, let's look at this query another way.

 (Adopting the frowned upon practice of relying on the order of the attribute declarations to substitute for names.)

  Domains: { INTEGER }

  Relations:    tbl             ( INTEGER );
                     sqr           ( INTEGER, INTEGER );
                     less_than  ( INTEGER, INTEGER );

  Operators: Let's just include JOIN and RENAME

  Query: ( tbl ( X, X as _ ) sqr ( X, Y ) less_than ( Y, 0 ) );

 Now, if the system was aware of the following re-write rule / constraint:

  Constraint: (sqr ( _, X ) less_than ( X, 0) ) -> ();

 Then the query could be reduced to () analytically.

 Supporting this class of re-writes turns an important part of query processing into a magic-theorem-proving-hat-dance, but what it shows is that there is no need for anything except domains, relations, constraints and (relational) operators. No need for domain operators, or inheritance, or any other blasphemy. Received on Wed Jun 01 2005 - 00:40:59 CEST