The Converge programming language

class ML1_Element { name : String; inv nonempty_name: name != null and name.len() > 0 } Note that the DSL fragment is written in an entirely different syntax than Converge itself. Currently DSL blocks are automatically tokenized by the Converge compiler using its default tokenization rules — this is not a fundamental requirement of the technique, but a peculiarity of the current implementation. More sophisticated implementations might choose to defer tokenization to the DSL implementation function. However using the Converge tokenizer has the advantage that normal Converge code can be embedded inside the DSL itself assuming an appropriate link from the DSL’s grammar to the Converge grammar. 5.1 DSL implementation functions DSL implementation functions follow a largely similar sequence of steps in order to translate the input tokens into an ITree: 1. Alter the input tokens as necessary. Since DSL’s often use keywords that are not part of the main Converge grammar, such alterations mostly take the form of replacing ID tokens with specific keyword tokens. 2. Parse the input tokens according to the DSL’s grammar. 3. Traverse the parse tree, translating it into an ITree. Section 6 explores these steps in greater detail via a concrete example.