Resolution and Datalog Rewriting Under Value Invention and Equality Constraints

Abstract: This paper present several refinements of the Datalog +/- framework based on resolution and Datalog-rewriting. We first present a resolution algorithm which is complete for arbitrary sets of tgds and egds. We then show that a technique of saturation can be used to achieve completeness with respect to First-Order (FO) query rewriting. We then investigate the class of guarded tgds (with a loose definition of guardedness), and show that every set of tgds in this class can be rewritten into an equivalent set of standard Datalog rules. On the negative side, this implies that Datalog +/- has (only) the same expressive power as standard Datalog in terms of query answering. On the positive side however, this mean that known results and existing optimization techniques (such as Magic-Set) may be applied in the context of Datalog +/- despite its richer syntax.