Mapping-Aware Megamodeling: Design Patterns and Laws

Megamodeling is the activity of specifying systems of models and mappings, their properties, and operations over them. The latter functionality is the most important for applications, and megamodels are often used as an abstract workflow language for model processing. To be independent of a particular modeling language, typical megamodels reduce relationships between models to unstructured edges encoding nothing but a labeled pair of models, thus creating a significant gap between megamodels and code implementing them. To bridge the gap, we propose mapping-aware megamodels, which treat edges as model mappings: structured sets of links (pairs of model elements) rather than pairs of models. The workflow can then be represented as an algebraic term built from elementary operations with models and model mappings.