Tool-Supported Multi-Level Language Evolution

Through their high degree of specialization, domain specific languages (DSLs) promise higher productivity and thus shorter development time and lower costs than general purpose programming languages. Since many domains are subject to continuous evolution, the associated DSLs inevitably have to evolve too, to retain their value. However, the continuous evolution of a DSL itself can be very expensive, since its compiler as well as existing words (i. e. programs) have to be adapted according to the changes to a DSL’s specification. These maintenance costs compromise the expected reduction of development costs and thus limit the success of domain specific languages in practice. This paper proposes a concept and a tool for the evolutionary development of domain specific languages. It provides language evolution operations that automate the adaptation of the compiler and existing DSL programs according to changes to the DSL specification. This significantly reduces the cost of DSL maintenance and paves the ground for bottom-up development of domain specific languages. 1 Domain Specific Chances and Limitations Albeit three decades of intense research and significant progress in research and practice, the development and maintenance of software systems still constitutes a time-consuming, costly and risky endeavor. The reduction of software development and maintenance costs thus remains a research topic of paramount importance to software engineering. A basic idea behind many approaches that attempt to reduce these costs is to increase the productivity of software developers. One approach to raise productivity that has received increased attention in recent years, are domain specific languages. 1.1 Productivity Through Specialization A domain-specific language (DSL) is a programming language or executable specification language that offers, through appropriate notations and abstractions, expressive power focused on, and usually restricted to, a particular problem domain [1]. Hence, the key characteristic of domain specific languages is their specialization to a problem domain. This specialization allows them to offer language constructs and abstractions tailored to the class of problems from this domain. DSLs typically allow these problems to be described very directly and concisely, requiring less developer effort than general purpose programming languages. DSLs thus have the potential to increase productivity and decrease costs of software development. Particularly prominent examples for the benefits of DSLs are found in the compiler construction field with specialized languages and generators (amongst others) for hashing (e. g. gperf), scanners, and parsers. Tools for the interactive design of graphical user interfaces, the query language SQL, interface definition languages like WSDL or modeling languages like MATLAB/Simulink add further examples for the benefits of DSLs, i. e. purpose-built languages with powerful generators. Besides reduced development time and cost, DSLs are a promising mean to decrease the maintenance cost of software because of the reduced code size and increased comprehensibility, both due to the higher expressiveness of DSLs compared to general-purpose languages. The decrease of maintenance costs is particularly important since 60-80% of the costs of software are usually not devoted to initial development but to maintenance [2–4]. However, DSLs only help to reduce overall maintenance costs, as long as the costs for development and maintenance of the DSL itself can be amortized. If unanticipated changes to a domain require changes to the DSL definition, maintenance costs can be high, as noted in [5].