On Structured Model-Driven Transformations

Structural aspects play a key role in the model-driven development of software systems. Effective techniques and tools must therefore be based on suitable representation formalisms that facilitate the specification, manipulation and analysis of the structure of models. Graphical and algebraic approaches have been shown to be very successful for such purposes: 1) graphs offer natural a representation of topological structures, 2) algebras offer a natural representation of compositional structures, 3) both graphs and algebras can be manipulated in a declarative way by means of rule-based techniques, 4) they allow for a layered presentation of models that enables compositional techniques and favours scalability. Most of the existing approaches represent such layering in a plain manner by overlapping the intra- and the inter-layered structure. It has been shown that some layering structures can be conveniently represented by an explicit hierarchical structure enabling then structurally inductive manipulations of the resulting models. Moreover, providing an inductive presentation of the structure facilitates the compositional analysis and verification of models. In this paper we compare and reconcile some recent approaches and synthesise them into an algebraic and graph-based formalism for representing and manipulating models with inductively defined hierarchical structure.