A new approach to the semantics of model diagrams


Sometimes, a diagram can say more than a thousand lines of code. But, sadly, most of the time, software engineers give up on diagrams after the design phase, and all real work is done in code. The supremacy of code over diagrams would be leveled if diagrams were code. This paper suggests that model and instance diagrams, or, which amounts to the same, class and object diagrams, become first level entities in a suitably expressive programming language, viz., type theory.

The proposed semantics of diagrams is compositional and self-describing, i.e., reflexive, or metacircular. Moreover, it is well suited for metamodelling and model driven engineering, as it is possible to prove model transformations correct in type theory. The encoding into type theory has the additional benefit of making diagrams immediately useful, given an implementation of type theory.