A Top-Down Approach to Rewriting Conjunctive Queries Using Views

Abstract

The problem of answering queries using views is concerned with finding answers to a query using only answers to views. In data integration context with the Local-As-Views approach, this problem translates to finding maximally contained rewriting for a given query. Existing solutions follow a bottom-up approach and, for efficiency reason, often require a post-processing phase, which comes at an additional cost.

We propose a solution which follows a top-down approach. For this, we first present a graph-based model for conjunctive queries and views, and identify conditions that if satisfied ensures maximality of a rewriting. Using this model as a basis, we then introduce a novel top-down algorithm, TreeWise, which efficiently generates maximally contained rewritings which are in general less expensive to evaluate, compared to the bottom-up algorithms, without requiring post-processing. The preliminary results of our experiments indicate that while TreeWise has comparable performance, it generally produces better quality rewritings.