Systematic and Automated Development of Hybrid Graphical-Textual Domain-Specific Languages

Graphical syntaxes are convenient for the simplification of the high-level concepts of a system. However, it can be tedious and error-prone to define and maintain graphically the fine details of a domain, such as precise behaviour and complex expressions. For modelling such aspects, the use of graphical syntaxes may lead to large, crowded diagrams that can hamper readability. Therefore, textual syntaxes are often more appropriate in such cases, as they provide a more concise and effective means of expressing detailed domain logic, leading to improved development experience and maintenance efficiency. As such, hybrid graphical-textual domain-specific languages can deliver the best of both worlds of graphical and textual modelling, due to the synergies obtained when combining both notations.

Hybrid graphical-textual model editors for domain-specific languages described in the literature lack essential facilities, such as consistency enforcement, temporary inconsistency tolerance, uniform error reporting, and integrated rename refactoring. Furthermore, their development is a non-trivial endeavour, as it requires a large amount of hand-written code, which entails substantial manual effort for language engineers.

This thesis proposes a language engineering methodology and contributes automated facilities for the systematic and automated development of feature-rich hybrid graphical-textual model editors for domain-specific languages, through the application of model-based techniques. By leveraging the contributions of this research, language engineers can construct feature-rich hybrid graphical-textual model editors with minimal development effort, requiring only a limited amount of hand-written code. The developed editors include features such as smart textual editors, textual-graphical cross-referencing, temporary inconsistency tolerance, integrated abstract syntax graph, consistency enforcement, uniform error reporting, and integrated rename refactoring.

The proposed approach is evaluated through four case studies, each involving the development of a hybrid graphical-textual domain-specific language and its supporting editor. The evaluation results indicate that the approach reduces development effort by requiring a small amount of hand-written code, while maintaining completeness and correctness. Furthermore, the performance analysis indicates that the approach has scalability implications with respect to loading models, storing models, renaming model elements, and deleting model elements.

Overall, the findings of this research support the hypothesis that feature-rich hybrid graphical-textual model editors for domain-specific languages can be developed in a systematic and automated manner through the application of model-based techniques.