Section: Research Program

Domain-specific Languages

Our final goal is centered around Domain-specific languages (DSLs), which are software languages tailored to a specific problem domain. DSLs can provide orders of magnitude improvement in terms of software quality and productivity. However, the implementation of DSLs is challenging and requires not only thorough knowledge of the problem domain (e.g., finance, digital forensics, insurance, auditing etc.), but also knowledge of language implementation (e.g., parsing, compilation, type checking etc.). Tools for language implementation have been around since the archetypical parser generator YACC. However, many of such tools are characterized by high learning curves, lack of integration of language implementation facets, and lead to implementations that are hard to maintain. This line of research focuses on two topics: improve the practice and experience of DSL implementation, and evaluate the success of DSLs in industrial practice.

Language workbenches [4] are integrated environments to facilitate the development of all aspects of DSLs. This includes IDE support (e.g., syntax coloring, outlining, reference resolving etc.) for the defined languages. Rascal can be seen as a language workbench that focuses on flexibility, programmability and modularity. DSL implementation is, in essence, an instance of source code analysis and transformation. As a result, Rascal's features for fact extraction, analysis, tree traversal and synthesis are an excellent fit for this area. An important aspect in this line of research is bringing the IDE closer to the source code. This will involve investigation of heterogeneous representations of source code, by integrating graphical, tabular or forms-based user interface elements. As a result, we propose Rascal as a feature-rich workbench for model-driven software development.

The second component of this research is concerned with evaluating DSLs in industrial contexts. This means that DSLs constructed using Rascal will be applied in real-life environments so that expected improvements in quality, performance, or productivity can be observed. We already have experience with this in the domain of digital forensics, computational auditing and games.

The goal of this research topic is to improve the practice of DSL-based software development through language design and tool support. A primary focus is to extend the IDE support provided by Rascal, and to facilitate incremental, and iterative design of DSLs. The latter is supported by new (meta-)language constructs for extending existing language implementations. This will require research into extensible programming and composition of compilers, interpreters and type checkers. Finally, a DSL is never an island: it will have to integrate with (third-party) source code, such as host language, libraries, runtime systems etc. This leads to the vision of multi-lingual programming environments [15] .