Model Management with Relation Types by Michalis Famelis

Model Management with Relation Types Michalis Famelis Master of Science Graduate Department of Computer Science University of Toronto 2010 Software development in application domains where model-based techniques are employed faces challenges related to managing sets of inter-dependent models of various types. In such settings, capturing the semantic structure of sets of models, as well as being able to check the validity of relationships between models that constitute this structure, can be of significant importance. Additionally, as inconsistencies can arise between models that are not obviously related, the ability to infer implicit relationships between models can help to further enforce the semantic cohesiveness of sets of related models. To this end, we propose an approach to model management centered around the declaration and definition of relation types. We describe how hierarchies of such relation types can be created to express their purpose, application scenarios and well-formedness rules. Such hierarchies consist of abstract relation types via which the purpose for relating models can be expressed. Concrete relation types for specific application scenarios can then be created by extending the abstract ones. Such concrete relationships are tied to particular metamodels, constrained by well-formedness rules, which can be used to verify the proper application of relation types. We showcase our approach by applying it to the domain of automotive software engineering, a domain where model-based techniques are becoming increasingly prominent, to demonstrate how typed relationships can be employed in multi-model environments to check for consistency between models and to compose relationships to infer new ones.