Deductive Generation of Constraint Propagation Rules

Constraint propagation can often be conveniently expressed by rules. In recent years, a number of techniques for automatic generation of rule-based constraint solvers have been developed, most of them using a generate-and-test approach. We examine a generation method that is based on deduction. A solver (i. e., a set of rules) for a complex constraint is obtained from one or several weaker solvers for simple constraints. We describe incremental solver constructions for several types of constraint modifications, including conjunction, existential and universal quantification.