Deep, Seamless, Multi-format, Multi-notation Definition and Use of Domain-specific Languages

Domain-specific modeling has gained increased attention in industry and academia in recent years leading to the emergence of numerous highly specialized, domain-specific modeling approaches and tools. However, the tools available today focus on one specific aspect of domain-specific modeling and tend to be isolated and difficult to integrate with each other. For instance, some tools focus exclusively on graphical, diagrammatic languages, while others focus on textual languages. The problems caused by this heterogeneous tool landscape become most evident when attempting to use multiple formats together. A first step towards integrating different modeling formats within one single environment has been taken by the latest projective modeling environments which support the embedding of non-textual formats (e. g. tables) into textual model editors. However, in these modeling environments the textual format dominates the embedded formats. Moreover, all of these tools are based on two-level modeling technologies which are limited to one, hard-wired metamodel describing the modeling language and one instance level describing the user model expressed in terms of the language. In addition to the metamodel, the concrete syntax features offered by today's modeling tools are typically hardwired as well. When multiple notations are available, modelers usually have to decide for one notation before starting. To view the model in an alternative notation, the model has to be opened in a second editor. The approach presented in this thesis for the deep, seamless, multi-format, multi-notation definition and use of domain-specific languages overcomes the aforementioned weaknesses. First, it treats all formats equally, no matter whether text, diagram or some other format. This allows different formats to be used side-by-side, as desired, without influencing one another negatively. Second, it allows a given format to be visualized in multiple notations in one editor, side-by-side. Third, the inherently deep architecture allows deep visualizations to be defined spanning as many classification levels as needed for language definition and use. Fourth, language definitions and visualizations are soft, like language applications, and can be changed at any time by modelers. The approach has been validated by means of an Eclipse EMF-based prototype implementation, called Melanee, and applied to a running example motivated by enterprise architecture modeling.