An attribute language definition for adaptable parsing expression grammars

Abstract

Adaptable Parsing Expression Grammars (APEG) are a formal model whose main purpose is to formally describe the syntax of extensible languages and their extension mechanisms. APEG extends Parsing Expression Grammar model with the notion of syntactic attributes, which are values passed through parse tree nodes and used during the parsing process. A grammar is a first-class value passed to every nonterminal, and the rules used during the parsing are fetched from this grammar. The ability to change and pass different grammars is the key to dynamically extend the original language grammar. The reported implementation of APEG attributes uses strings and ad hoc Java code to manipulate and build grammars during parsing time. This approach has at least three disadvantages: a grammar specification becomes dependent on the language in which the functions to manipulate and build new grammars were implemented; we may not assure that the grammars built are always syntactically correct; and it is virtually impossible to prove that the generated parser does not lead to a infinite loop. In this work, we formally define an attribute language for APEG containing operators to manipulate grammars. As a result, new rules and grammars built during parsing time are syntactically correct. In addition, we define a restriction on APEG rules that assures that any generated parser will terminate on all inputs.