Practical optimizations of LR(k) parsing tables

A method for optimizing LR(k) tables is presented. This method extends the algorithms of phi -inaccessible tables and error postponement for LR(k) tables defined by A. V. Aho and J. D. Ullman (1972, 1973) to a new kind of parsing tables, called reduced tables. Reduced tables are more convenient than canonical tables for all classes of LR grammars. For example, reduced tables can be effectively used to create full LR parsers for programming languages like Pascal, C, Oberon, and Eiffel. Experiments with some important programming languages show rates of optimization of the order of 70% of the eliminated tables and of 50% of nonerror deleted entries. Such results have been obtained with the grammars of programming languages like Ada, Pascal, C, Oberon, and Modula-2.<>