Avoiding cartesian products for multiple joins

Abstract

Computing the natural join of a set of relations is an important operation in relational database systems. The ordering of joins determines to a large extent the computation time of the join. Since the number of possible orderings could be very large, query optimizers first reduce the search space by using various heuristics and then try to select an optimal ordering of joins. Avoiding Cartesian products is a common heuristic for reducing the search space, but it cannot guarantee optimal ordering in its search space, because the cheapest Cartesian-product-free (CPF, for short) ordering could be significantly worse than an optimal non-CPF ordering by a factor of an arbitrarily large number. In this paper, we use programs consisting of joins, semijoins, and projections for computing the join of some relations, and we introduce a novel algorithm that derives programs from CPF orderings of joins. We show that there exists a CPF ordering from which our algorithm derives a program whose cost is within a constant factor of the cost of an optimal ordering. Thus, our result demonstrates the effectiveness of avoiding Cartesian products as a heuristic for restricting the search space of orderings of joins.