The EDSL’s struggle for their sources

Abstract

Embedded Domain Specific Languages make language design and implementation easier, because lexical and syntactical analysis and part of the semantic checks can be completed by the compiler of the host language. On the other hand, by the nature of embedding, EDSL compilers have to work with a syntax tree that stores no information about the source file processed and the location of the program entities within the source file. This makes it hard to produce user-friendly error messages and connect the generated target code with the source code for debugging and profiling purposes. This lecture note presents this problem in detail and shows possible solutions. The first, lightweight solution uses macro preprocessing. The second one is based on syntax tree transformations to add missing source-related information. This is more powerful, but also more heavyweight. The last technique avoids the problem by turning the embedded language implementation to a standalone one (with own parser) after the experimental phase of the language development process: It turns out that most of the embedded implementation can be reused in the standalone one.