Efficient construction of LR(k) states and tables

• Published in: ACM transactions on programming languages and systems 1991-01, Vol.13 (1), p.150-178
• Main Authors: Anicona, M., Dodero, G., Gianuzzi, V., Morgavi, M.
• Format: Article
• Language: English
• Subjects: Applied sciences; Computer science; control theory; systems; Exact sciences and technology; Formal languages and automata theory; Grammars and context-free languages; Language processing and microprogramming; Parsing; Program reasoning; Semantics and reasoning; Software; Theory of computation
• ISSN: 0164-0925; 1558-4593
• DOI: 10.1145/114005.102809

A new method for building LR(k) states and parsing tables is presented. The method aims at giving a feasible construction of a collection of LR(k) parsing tables, especially when k > 1. for nontrivial grammars. To this purpose, the algorithm first attempts to build a set of normal states for the given grammar, each one associated to a single parsing action in {accept, reduce, shift}. When such an action cannot be uniquely determined, that is, when up to k input symbols have to be examined (inadequacy), further states, belonging to a new type, called look-ahead states, are computed. The action associated with inadequate states is a new parsing action, look. States are built without actual computation of the FIRSTk and EFFk functions; that is, nonterminals are kept in the context string of items composing each state, and their expansion to terminals is deferred until indispensable to solve inadequacy. The aforementioned method is illustrated; then the canonical collection of states and the canonical tables are compared with those obtained from the proposed method. A sufficient condition is stated, by which the size of parsing tables, obtained by applying this new method, is smaller than that of canonical tables. Experimental results show that such a condition is verified by the grammars of several programming languagues and that significant speed is gained by avoiding the computation of the FIRSTk function.