Syntactic error recovery in LR parsers

Abstract

In this thesis, we describe an error recovery for syntactic errors in LR parsers. For the first level recovery, the pattern mapping approach described in [2] was slightly modified and applied to incorporate the cost function concept. A detailed description on the pattern generation process is presented. A new restricted set of pattern RP (c,k,l) is defined. It is simple but the efficiency and quality of recovery using RP (c,k,l) were as good as those using MCL(k) defined in [2]. Some heuristics to improve the efficiency and correcting power are introduced. Suggestions for further augmentations are contained also in this thesis. The second level recovery for the case of failure from the pattern mapping recovery is described for the sake of self-containment. We have applied our error recovery scheme to some simple grammars which have almost all the important constructs of common programming languages.