Approximating Term Rewriting Systems: A Horn Clause Specification and Its Implementation

Abstract

We present a technique for approximating the set of reachable terms of a given term rewriting system starting from a given initial regular set of terms. The technique is based on previous work by other authors with the same goal, and yields a finite tree automaton recognising an over-approximation of the set of reachable terms. Our contributions are, firstly, to use Horn clauses to specify the transitions of a possibly infinite-state tree automaton defining (at least) the reachable terms. Apart from being a clear specification, the Horn clause model is the basis for further automatic approximations using standard logic program analysis techniques, yielding finite-state tree automata. The approximations are applied in two stages: first a regular approximation of the model of the given Horn clauses is constructed, and secondly a more precise relational abstraction is built using the first approximation. The analysis uses efficient representations based on BDDs, leading to more scalable implementations. We report on preliminary experimental results.