Typesafe Code Reuse Across ASTs via Code Generation

Writing data structures for abstract syntax trees (ASTs) in a conventional OO programming language is tedious and error-prone. Hence, programmers often use AST generators to generate OO code from a higher-level description. This article argues that the existing AST generators do not provide good support for programs that manipulate several similar structural variations of an AST. Using a conventional code generator for such family of ASTs either leads to code duplication or a loss of static type-safety. This article introduces extended ASTs (XASTs), a small language for capturing ASTs, and a code-generation algorithm for XASTs. Since XASTs are a variation of regular tree types they are ordered by a similar structural subtyping relation which is preserved by the generative algorithm. In contrast to existing AST code generators, our solution facilitates both code reuse and preserves static type-safety. We demonstrate the feasibility of our approach by providing an implementation of a code generator from XASTs to Java.