Explications pour comprendre la trace d'un solveur de contraintes sur domaines finis

Some works in progress on finite domain constraint solvers concern the implementation of a XML trace of the computation according to the OADymPPaC DTD (for example in GNU-Prolog, PaLM, CHIP). Because of the large size of traces, even for small toy problems, some tools are needed to understand this trace. Explanations of value withdrawal (or nogoods) are used during domain reduction by some solvers. In this paper, we use a formalization of explanations by proof trees in a fixpoint framework based on iteration of monotonic local consistency operators. Proof trees provide a declarative view of the computations by constraint propagation. We show how explanations may be naturally extracted from the OADymPPaC trace format. Explanations allow a better understanding of the domain reductions in the trace. MOTS-CLÉS : Explication, trace, contrainte, reduction de domaines