Language Implementation Patterns

Rule-Based Translation

To build a translator, we have to define an input-to-output mapping (a set of “this goes to that” rules) no matter what translator architecture we use. Instead of coding them with a general-purpose programming language, we can use a translation DSL. There are a number of excellent rule-based systems available such as ASF+SDF,^[40] Stratego/XT,^[41] and TXL.^[42] The ANTLR project also has a simple rule engine called ANTLRMorph.^[43]

To use these engines, we feed them an input language grammar and a set of translation rules, as shown in the following image. From the grammar, the engine builds a parse tree from the input and then applies the rules to get the output. As we saw in ​Decoupling Tree Traversal from Pattern Matching ...