Chapter 18. Model Compilers

This final chapter tells you what to do with your Executable UML models. The short answer is that you compile your Executable UML models into an implementation. No rearranging. No adornment. No elaboration.

A longer answer is that you must choose how to compile your Executable UML models based on the performance requirements and the environment of your application. From this information, you can select a model compiler that meets your needs, compile the models, and deliver the running system.

A programming language compiler generally produces a decorated syntax tree or a stream of tokens from which code can be generated. A model compiler generally accesses a repository that captures the underlying semantic representation of an arbitrary model. The model compiler comprises a set of mechanisms that animate the fundamental constructs of Executable UML, and a set of rules that weave these constructs together with the application.

We describe two complementary approaches to establishing which model compiler to use, one based on understanding the environment into which the application must fit (the rate at which requests arrive, for example), and another based on our knowledge of how software goes together to make a coherent system (the resulting performance of the chosen approach for putting the software together).

You can buy an existing model compiler as a "design-in-a-box." Alternatively, you can modify an existing model compiler or even build your own from scratch.

A model compiler is a realization of the more general idea of an application-independent software architecture. This chapter briefly explores this concept, points to some related literature, and describes an approach for modeling software architectures.