Automatic Generation of Local-First, Collaborative Modeling Languages

Abstract

Modeling languages provide engineers with shared, domain-specific concepts for collaboratively structuring domain knowledge. Although modern language workbenches can generate much of the language-specific tooling from a metamodel, collaborative editing usually relies on external infrastructure that requires additional integration. In particular, existing solutions provide insufficient support for hybrid collaborative modeling, in which modelers alternate seamlessly between online and offline editing and their changes are automatically reconciled upon reconnection. This article presents an approach for automatically generating metamodel-specific replicated model runtimes directly from Ecore metamodels. We define a translation from a practically relevant subset of EMF/Ecore to compositions of Conflict-free Replicated Data Types (CRDTs), each encapsulating a well-defined conflict-resolution policy. The resulting runtimes expose model APIs for creating, updating, and querying replicated model states, while their constituent CRDTs define how concurrent updates are reconciled. The approach is implemented as a configurable code generator and evaluated through three case-study runtimes and all 5,475 Ecore files in ModelSet v0.9.4. The results quantify the generator’s end-to-end coverage and provide complementary formal and empirical evidence for convergence within the stated scope and assumptions.