Rete Network Slicing for Model Queries

Declarative model queries captured by graph patterns are frequently used in model driven engineering tools for the validation of well-formedness constraint or the calculation of various model metrics. However, their high level nature might make it hard to understand all corner cases of complex queries. When debugging erroneous patterns, a common task is to identify which conditions or constraints of a query caused some model elements to appear in the results. Slicing techniques in traditional programming environments are used to calculate similar dependencies between program statements. Here, we introduce a slicing approach for model queries based on Rete networks, a cache structure applied for the incremental evaluation of model queries. The proposed method reuses the structural information encoded in the Rete networks to calculate and present a trace of operations resulting in some model elements to appear in the result set. The approach is illustrated on a running example of validating well-formedness over UML state machine models using graph patterns as a model query formalism.