🤖 AI Summary
This paper addresses the decidability of termination for the restricted chase under linear multi-head tuple-generating dependencies (TGDs)—a long-standing open problem in database theory. While termination is known to be decidable for linear TGDs under the standard (oblivious) chase, its decidability under the order-sensitive restricted chase—particularly for multi-head extensions—has remained unresolved. We establish decidability for this setting by devising a novel algorithm that combines dependency graph analysis with finite model construction to precisely characterize cyclic constraints on multi-head derivation paths. Our method identifies and bounds all possible infinite derivations via structural properties of rule interactions, ensuring soundness and completeness. This result resolves a fundamental theoretical challenge for linear multi-head TGDs and provides a foundational framework for analyzing termination in broader classes of dependencies, including guarded TGDs and their variants.
📝 Abstract
The chase is a ubiquitous algorithm in database theory. However, for existential rules (aka tuple-generating dependencies), its termination is not guaranteed, and even undecidable in general. The problem of termination becomes particularly difficult for the restricted (or standard) chase, for which the order of rule application matters. Thus, decidability of restricted chase termination is still open for many well-behaved classes such as linear or guarded multi-headed rules. We make a step forward by showing that all-instances restricted chase termination is decidable in the linear multi-headed case.