Automatic Generation of Multiple Exit Parsing Subroutines

Abstract

The parser construction method presented here might be characterized as the offspring of the successful marriage of LR(k) methodology [4] with the transition diagram systems of Conway [2]. Like transition diagram systems, the parsers constructed by the method presented here, called the LLP(k) method, consist of small, finite state automata linked by “subroutine calls” and provide a mixed top-down/bottom-up parse. Transition diagram systems with one exit state per diagram correspond to top-down parsers and have been extensively studied [6,9,10]. Like Conway’s transition diagrams, however, LLP(k) subroutines can parse multiple non-terminals simultaneously and return an indication of what they have discovered. It has been shown [7] that transition diagram systems composed of such multiple exit diagrams can parse all deterministic context free languages. Further, LLP(k) parsers can be realized as directly executing (non-interpretive) subroutines. The method has been implemented [8]. The results demonstrate the feasibility of the approach.