Rule splitting and attribute-directed parsing

Abstract

Rule splitting is a phenomenon, most clearly exhibited by attribute grammars and affix grammars, in which the syntactic structure of a phrase is constrained by its attributes. In this paper, rule splitting is illustrated by examples taken from real programming languages, and two varieties of rule splitting are identified and formalized. Implementations of rule splitting (attribute-directed parsing) are demonstrated for top-down and bottom-up parsers, both one-pass and multi-pass. Finally, the problems of exploiting rule splitting in a compiler writing system based on attribute grammars are explored.