Towards interactive, test-driven development of model transformations

Developing model transformations is a complex task because it requires a deep knowledge of the semantics of the input and output meta-models. Many times this knowledge is refined as the transformation is developed. A simple approach to encode this knowledge is in the form of test cases, consisting of pairs of input models and expected output models. However, creating these test cases is a time consuming and error-prone activity and it is barely used. Moreover, there is little tool support for refining a transformation interactively at the same time that the test suite is created. This paper explores an approach for test-driven development of model transformations based on creating test cases using a model finder in an interactive manner. The synthesis of the input models of the test cases is driven by the results of static analysis with respect to the coverage of the transformation, with the goal of instantiating input models which are not yet handled by the transformation. The expected models are automatically derived using an instrumented transformation, and the user is in charge of inspecting the result to validate them. The approach has been implemented for ATL and integrated into AnATLyzer using USE Model Validator as the backend model finder.