Optimization Properties for Classes of Conjunctive Regular Path Queries

Abstract

We are interested in the theoretical foundations of the optimization of conjunctive regular path queries (CRPQs). The basic problem here is deciding query containment both in the absence and presence of constraints. Containment without constraints for CRPQs is EXPSPACE-complete, as opposed to only NP-complete for relational conjunctive queries. Our past experience with implementing similar algorithms suggests that staying in PSPACE might still be useful. Therefore we investigate the complexity of containment for a hierarchy of fragments of the CRPQ language. The classifying principle of the fragments is the expressivity of the regular path expressions allowed in the query atoms. For most of these fragments, we give matching lower and upper bounds for containment in the absence of constraints. We also introduce for every fragment a naturally corresponding class of constraints in whose presence we show both decidability and undecidability results for containment in various fragments. Finally, we apply our results to give a complete algorithm for rewriting with views in the presence of constraints for a fragment that contains Kleene-star and disjunction.