Does it make sense to focus on Domain Specifics Languages (DSL) development following a Software Product Line approach?
Does anyone know any other approach to create and maintain several related Domain Specifics Languages at the same time? Note that to support a custom language, requires support multiple tools, from parser, compilers, interpreters, to current state of art IDE, etc.
On
I think it makes sense to focus on DSLs following a Software Product Line approach. If you define the DSL correctly, it will essentially define a framework for creating applications within in a domain and an operating environment in which they execute in. By operating environment, I mean the OS, hardware, and database, as well as the code that implements the semantics or run time environment of the DSL. The framework and operating environment will be the artifacts that get reused across product lines. You may have to create an operating environment that consists of run time environments for multiple DSLs to support multiple product lines.
Our DMS Software Reengineering Toolkit is exactly this idea. DMS provides generic parsing, tree building, analyzes (name resolution, control flow anlaysis, data flow analyis, call graphs and points-to analysis, custom analyzers, arbitrary transformations). It has a variety of legacy language front ends, as well as some DSLs (e.g., HTML, XML, temporal logic equations, industrial controller languages, ...) off-the-shelf, but has very good support for defining other DSLs.
We use DMS both to build custom analyzers and transformation tools, but also as product-line generator. For example, we provide test coverage, profilers, smart differencers, and clone detections for a wide variety of languages... because DMS makes this possible. Yes, this lowers our dev and maintenance costs because each of these tool types uses DMS directly as a foundation. Fundamentally, it allows the amortization of language parsers, analyzers and transformers not only across specific languages, but across dialects of those languages and even different languages.