IMPLEMENTING CONSTRAINT SOLVERS IN B-PROLOG

摘要

Constraint Logic Programming (CLP) defines a family of programming languages that extend Prolog to support constraint solving over certain domains. Current state-of-the-art CLP systems are based on abstract machines or language constructs that are good for certain domains and propagation algorithms but are either not sufficiently expressive or inefficient for some other domains and propagation algorithms. B-Prolog provides a construct, called action rules, for programming interactive agents such as constraint propagators. As far as constraint propagation is concerned, an agent maintains dynamically a certain level of consistency for a constraint. This paper presents constraint solvers implemented in action rules for six domains, namely, finite-domains, Boolean, trees, lists, sets, and floating-point intervals. Some of the solvers such as the finite-domain and set solvers are competitive in performance with the fastest solvers available now.