A Search Based Test Data Generation Approach for Model Transformations

Model transformations are a fundamental part of Model Driven Engineering. Automated testing of model transformation is challenging due to the complexity of generating test models as test data. In the case of model transformations, the test model is an instance of a meta-model. Generating input models manually is a laborious and error prone task. Test cases are typically generated to satisfy a coverage criterion. Test data generation corresponding to various structural testing coverage criteria requires solving a number of predicates. For model transformation, these predicates typically consist of constraints on the source meta-model elements. In this paper, we propose an automated search-based test data generation approach for model transformations. The proposed approach is based on calculating approach level and branch distances to guide the search. For this purpose, we have developed specialized heuristics for calculating branch distances of model transformations. The approach allows test data generation corresponding to various coverage criteria, including statement coverage, branch coverage, and multiple condition/decision coverage. Our approach is generic and can be applied to various model transformation languages. Our developed tool, MOTTER, works with Atlas Transformation Language (ATL) as a proof of concept. We have successfully applied our approach on a well-known case study from ATL Zoo to generate test data.