From State-Based to Delta-Based Bidirectional Model Transformation

Existing bidirectional model transformation languages are mainly state-based: a transformation is considered composed from functions whose inputs and outputs only consist of original and updated models, but alignment relationships between the models are not specified. In the paper we identify and discuss three major problems caused by this under-specification. We then propose a novel formal framework based on a graphical language: models are nodes and updates are arrows, and show how the three problems can be fixed.