Abstract
The replacement of nonlinear recursions with equivalent linear recursions is a potentially useful query optimization strategy, since it permits the use of efficient algorithms for the evaluation of linear logic programs. We show that a member of a certain class of bilinear recursions is linearizable in a strong sense if and only if a specific partial proof tree derived from this recursion is contained in a bounded number of partial proof trees generated by the recursion. Further, while each such test of containment between proof trees involves an exponential number of conjunctive-query containment tests, we present syntactic conditions on the recursion that are necessary and sufficient for the containment and verifiable in polynomial time.

This publication has 10 references indexed in Scilit: