Implementing global constraints as graphs of elementary constraints

Abstract

Global constraints are cardinal concepts of CLP (FD), a constraint programming language. They are means to find a set of integers that satisfy certain relations. The fact that defining global constraints often requires the knowledge of a specification language makes sharing constraints between scientists and programmers difficult. Nicolas Beldiceanu presented a theory that could solve this problem, because it depicts global constraints as graphs: an abstraction that everyone understands. The abstract description language defined by the theory may also be interpreted by a computer program. This paper deals with the problematic issues of putting the theory into practice by implementing such a program. It introduces a concrete syntax of the language and presents three programs understanding that syntax. These case studies represent two different approaches of propagation. One of these offers exhausting pruning with poor efficiency, the other, yet unfinished attempt provides a better alternative at the cost of being a lot more complicated.