Dynamic semantics for language-based programming environments revisited

Abstract

Montages are a semi-visual language specification formalism recently proposed by some of the authors. This framework has been successfully used to give complete executable specifications of languages like C and Java. This paper compares Montages with action equations, both being paradigms for incorporating a means for specifying dynamic semantics in an environment generator using an attribute grammar framework. The major differences between the two lie in the way they specify the semantic processing at the nodes of the abstract syntax tree and how they specify control flow. Action equations use equations, similar in spirit to the semantic equations in attribute grammars, while Montages make use of Abstract State Machine rules to specify the semantic processing. Control flow in Montages is represented as state-transitions in a finite state machine which is specified using a graphical notation, while action equations specify control flow by propagating events from one node of the syntax tree to the other. Finally, the specifications of common control structures found in any imperative or object-oriented language are compared using the two formalisms and their relative merits are discussed.