Implementation Alternatives for Bottom-Up Evaluation

Abstract

Bottom-up evaluation is a central part of query evaluation / program execution in deductive databases. It is used after a source code optimization like magic sets or SLDmagic that ensures that only facts relevant for the query can be derived. Then bottom-up evaluation simply performs the iteration of the standard TP -operator to compute the minimal model. However, there are different ways to implement bottom-up evaluation efficiently. Since this is most critical for the performance of a deductive database system, and since performance is critical for the acceptance of deductive database technology, this question deserves a thorough analysis. In this paper we start this work by discussing several different implementation alternatives. Especially, we propose a new implementation of bottom-up evaluation called "Push-Method".