Demystifying Model Transformations: An Approach Based on Automated Rule Inference

摘要

Model-driven development (MDD) is widely used to develop modern business applications. MDD involves creating models at different levels of abstractions. Starting with models of domain concepts, these abstractions are successively refined, using transforms, to design-level models and, eventually, code-level artifacts. Although many tools exist that support transform creation and verification, tools that help users in understanding and using transforms are rare. In this paper, we present an approach for assisting users in understanding model transformations and debugging their input models. We use automated program-analysis techniques to analyze the transform code and compute constraints under which a transformation may fail or be incomplete. These code-level constraints are mapped to the input model elements to generate model-level rules. The rules can be used to validate whether an input model violates transform constraints, and to support general user queries about a transformation. We have implemented the analysis in a tool called XYLEM. We present empirical results, which indicate that (1) our approach can be effective in inferring useful rules, and (2) the rules let users efficiently diagnose a failing transformation without examining the transform source code.