Validation in model-driven engineering: testing model transformations

TLDR

The Model‑Driven Architecture (MDA) uses models linked by transformations and promises improved reliability, consistency, and maintainability, but requires robust testing to support software engineering processes. This paper explores how testing can be applied to model transformations, the most novel aspect of MDA. The authors review testing roles across model‑driven development, identify adequacy criteria, oracles, and data generation challenges for model transformations, and propose adapting functional and structural testing techniques to address these tasks.

Abstract

The OMG's model-driven architecture is quickly attracting attention as a method of constructing systems that offers advantages over traditional approaches in terms of reliability, consistency, and maintainability. The key concepts in the MDA are models that are related by model transformations. However, for the MDA to provide an adequate alternative to existing approaches, it must offer comparable support for software engineering processes such as requirements analysis, design and testing. This paper attempts to explore the application of the last of these processes, testing, to the most novel part of the MDA, that of model transformation. We present a general view of the roles of testing in the different stages of model-driven development, and a more detailed exploration of approaches to testing model transformations. Based on this, we highlight the particular issues for the different testing tasks, including adequacy criteria, test oracles and automatic test data generation. We also propose possible approaches for the testing tasks, and show how existing functional and structural testing techniques can be adapted for use in this new development context.