Slavisa Markovic
Modern software is reaching levels of complexity encountered in biological systems; sometimes comprising systems of systems each of which may include tens of millions of lines of code. Model Driven Engineering (MDE) advocates raising the level of abstraction as an instrument to deal with software complexity. It promotes usage of software models as primary artifacts in a software development process. Traditionally, these MDE models are specified by Unified Modeling Language (UML) or by a modeling language created for a specific domain. However, in the vast area of software engineering there are other techniques used to improve quality of software under development. One of such techniques is refactoring which represents introducing structured changes in software in order to improve its readability, extensibility, and maintainability, while preserving behavior of the software. The main application area for refactorings is still programming code, despite the fact that modeling languages and techniques has significantly gained in popularity, in recent years. The main topic of this thesis is making an alliance between the two virtually orthogonal techniques: software modeling and refactoring. In this thesis we have investigated how to raise the level of abstraction of programming code refactorings to the modeling level. This resulted in a catalog of model refactorings each specified as a model transformation rule. In addition, we have investigated synchronization problems between different models used to describe one software system, i.e. when one model is refactored what is the impact on all dependent models and how this impact can be formalized. We have concentrated on UML class diagrams as domain of refactorings. As models dependent on class diagrams, we have selected Object Constraint Language (OCL) annotations, and object diagrams. This thesis formalizes the most important refactoring rules for UML class diagrams and classifies them with respect to their impact on object diagrams and annotated OCL constraints. For refactoring rules that have an impact on dependent artifacts we formalize the necessary changes of these artifacts. Moreover, in this thesis, we present a simple criterion and a proof technique for the semantic preservation of refactoring rules that are defined for UML class and object diagrams, and OCL constraints. In order to be able to prove semantic preservation, we propose a model transformation approach to specify the semantics of constraint languages.
EPFL2008
