STypeS: Nonrecursive Datalog Rewriter for Linear TGDs and Conjunctive Queries

Abstract

We present STYPES, a system that rewrites ontology-mediated queries with linear tuple-generating dependencies and conjunctive queries to equivalent nonrecursive datalog (NDL) queries. The main feature of STYPES is that it produces polynomial-size rewritings whenever the treewidth of the input conjunctive queries and the size of the chases for the ontology atoms as well as their arity are bounded; moreover, the rewritings can be constructed and executed in LOGCFL, indicating high parallelisability in theory. We show experimentally that Apache Flink on a cluster of machines with 20 virtual CPUs is indeed able to parallelise execution of a series of NDL-rewritings constructed by STYPES, with the time decreasing proportionally to the number of CPUs available.