Dynamic Analysis of Bounds Versus Domain Propagation

Constraint propagation solvers interleave propagation (removing impossible values from variable domains) with search. Previously, Schulte and Stuckey introduced the use of static analysis to determine where in a constraint program domain propagators can be replaced by more efficient bounds propagators and still ensure that the same search space is traversed. This paper introduces a dynamic yet considerably simpler approach to uncover the same information. The information is obtained by a linear time traversal of an analysis graph that straightforwardly reflects the properties of propagators implementing constraints. Experiments confirm that the simple dynamic method is efficient and that it can be used interleaved with search, taking advantage of the simplification of the constraint graph that arises from search.