An Implementation of an Execution Engine for the Foundational UML Subset

This thesis examines an OMG proposal for a tool-chain for working with Executable UML. The proposal defines a subset of UML, called fUML, with well-defined semantics and an Execution Model. The goal of this thesis is to understand whether fUML models can be useful in commercial user interface development. The idea is that an executable model (1) receives input from the data layer, which triggers an execution of the model, (2) transforms the data according to the model, i.e., executes the model on the given input, and (3) pushes the result to the presentation layer, or another part of the program that listens for the transformed data. We use the Eclipse Modeling Framework (EMF) to implement an execution engine for fUML models and show a full example, including constructing and executing a model. We conclude that Executable UML is not mature enough to be used instead of traditional code in commercial quality products. Firstly, the model construction tools we have used cannot compete with modern development environments for code and we find it cumbersome to work with the models. Secondly, UML lacks important primitive types, e.g., real numbers, which forces the execution engine to work with strings, which in turn might cause performance problems. We also find it difficult to mix vendors in the tool-chain, for example, a user might want to buy a modeling tool to do the actual modeling, but use an open-source execution engine. This is difficult because the format on which models are stored is not well-defined. We suggest a few improvements to the Graphical Modeling Framework (GMF) UML2 Editor. These improvements together with an extension of the UML primitive types and the development of a model library, would make fUML more useful in practice