Solving Disjunctive Constraints for Interactive Graphical Applications

Kim Marriott, Peter Moulder, Peter J. Stuckey, Alan Borning

Abstract: In interactive graphical applications we often require that objects do not overlap. Such non-overlap constraints can be modelled as disjunctions of arithmetic inequalities. Unfortunately, disjunctions are typically not handled by constraint solvers that support direct manipulation, in part because solving such problems is NP-hard. We showhere that is in fact possible to (re-)solve systems of disjunctive constraints representing non-overlap constraints suficiently fast to support direct manipulation in interactive graphical applications. The key insight behind our algorithms is that the disjuncts in a non-overlap constraint are not disjoint: during direct manipulation we need only move between disjuncts that are adjacent in the sense that they share the current solution. We give both a generic algorithm, and a version specialised for linear arithmetic constraints that makes use of the Cassowary constraint solving algorithm.