Achieving UML and OCL Model Quality by Utilizing Metamodeling

Abstract

Model-driven engineering (MDE) is a system development methodology, which helps to abstract systems architecture and provides a promising means for addressing architecture complexity and design quality. Models are the backbone of the MDE approach. Therefore, an efficient means of exploring, querying, and assuring the quality of models play essential roles in a successful project that applies model-based techniques. In this thesis, we initially propose an approach for extending a two-level modeling style to a three-level one by adding a meta-model at the topmost level. Standard OCL does not support reflective constraints, e.g., constraints concerning properties of the model, such as the depth of inheritance. By adding an auto-generated metamodel instantiation that reflects the model under consideration to the middle level, we can offer an option for writing reflective constraints and queries. Metric measurement and smell detection are efficient mechanisms for evaluating the internal quality of models. The metrics and design smell employed in this thesis are defined at a metamodel level, and their evaluation is performed in an automatically generated metamodel instantiation. The employed metrics cover both the class scope and the model scope, and software designers can define their own metrics and design smells as well. We also introduce a complete process for model quality assessment with pre-defined metrics. In addition, a metrics configuration defined by an experienced chief designer can be translated into OCL invariants and then evaluated to provide quality feedback to software designers relieving them from detailed OCL expertise. Another contribution of this thesis is an evaluation of metric measurement and smell detection on a large dataset of Unified Modeling Language (UML) models collected from practice. This evaluation not only illustrates the feasibility and usefulness of our approaches but also helps to answer the research questions regarding the characteristics and quality of models in practice.