Rule Chains Coverage for Testing QVT-Relations Transformations

Traditional software testing techniques have been adapted to deal with the verification of model transformations. Black-box techniques have the benefit of simplicity as well as the advantage of being independent of the implementation language, and thus compatible with any model transformation language. Although this is important, the inherent complexity of metamodels may result in a significant amount of non-relevant test models. On the contrary the use of white-box techniques allows generating more effective models at a higher cost. In this paper we propose an approach for the verification of QVT-Relations transformations which considers the dependencies between transformation rules and the standard semantics. Test models generation is based on the construction of what we call a rule chain: a set of rule patterns and conditions satisfying a top rule, as well as on an adaptation of other techniques, as grammar testing and partition analysis. We introduce the approach and explain its application using a model transformation devised for the generation of service models from business process models. This approach generates more effective test models than existing approaches, for which we are working on several tests to prove it.