Unifying view of recent LALR(1) lookahead set algorithms

Abstract

Since the introduction of LALR parsing, several algorithms have been presented for the computation of the lookahead sets needed to produce an LALR parser. The algorithm in Aho and Ullman[1] has perhaps received the widest exposure. The recent algorithms by DeRemer and Pennello[2] and Park, Choe, and Chang[4] are the most efficient. A new algorithm has been developed from an algorithm originally based on the Aho and Ullman algorithm and subsequently modified to take advantage of the efficiencies introduced by the DeRemer and Pennello algorithm. The new algorithm performs better than the Park, Choe, and Chang algorithm, and both perform better than the DeRemer and Pennello algorithm. The reasons for the relative performances are easily understood when the algorithms are presented in a common light.