Automated inference of likely metamorphic relations for model transformations

Abstract

Model transformations play a cornerstone role in Model-Driven Engineering (MDE) as they provide the essential mechanisms for manipulating and transforming models. Checking whether the output of a model transformation is correct is a manual and errorprone task, referred to as the oracle problem. Metamorphic testing alleviates the oracle problem by exploiting the relations among di erent inputs and outputs of the program under test, so-called metamorphic relations (MRs). One of the main challenges in metamorphic testing is the automated inference of likely MRs. This paper proposes an approach to automatically infer likely MRs for ATL model transformations, where the tester does not need to have any knowledge of the transformation. The inferred MRs aim at detecting faults in model transformations in three application scenarios, namely regression testing, incremental transformations and migrations among transformation languages. In the experiments performed, the inferred likely MRs have proved to be quite accurate, with a precision of 96.4% from a total of 4101 true positives out of 4254 MRs inferred. Furthermore, they have been useful for identifying mutants in regression testing scenarios, with a mutation score of 93.3%. Finally, our approach can be used in conjunction with current approaches for the automatic generation of test cases.