Yet another generation of LALR parsers for regular right part grammars

Abstract

In this paper we introduce two methods for building LALR parsers for regular right part grammars (RRPGs). Both methods build a parser directly from a grammar, require no extra state or data structure, and can deal with all LALR RRPGs.The first method is quite simple. For almost all LALR RRPGs, including the majority of grammars with stacking conflicts, parsing actions are similar to those of LALR parsers for usual context free grammars. No extra action is required to recognize a handle in this case. For other LALR RRPGs, the right hand side of a production is checked to recognize a handle.The second method does not require checking of the right hand side of a production to recognize a handle. Instead, it records the number of conflicts in LR items and in the stack. Unlike previous methods, our method needs no extra data structure.

Fast deterministic parsers for transition networks

“…Many authors have proposed to extend the standard LR method to EBNF grammars, each proposal purporting to improve previous attempts, but no clear-cut optimal solution had surfaced so far. The following discussion particularly draws from the later papers [14,16,22], which also include relevant surveys. An oldfashioned distinction concerns the source language specification: either an EBNF grammar, i.e., a set of rules with regular expressions as right-hand sides, or a TN.…”

“…Others, notably [22], specify the rule right parts by DFAs, i.e., they use a TN as we also do. Notice that the use of NFAs instead of DFAs has not been considered presumably because the lower state complexity thus achievable would make little difference in terms of readability, since the typical practical languages considered have a very small state complexity.…”

“…In [5] a systematic transformation from EBNF to BNF is used to obtain an ELR(1) parser that simulates the standard Knuth's one. It is generally agreed that approach (B) is superior, because transforming to BNF may add inefficiency and may obscure the semantic structure [22] of the language. Furthermore, we have seen in Sect.…”

Parsing methods streamlined

“…Most papers usually start with critical reviews of related proposals, from which we can grasp the difficulties and motivations then perceived. The following discussion particularly draws from the later papers [Morimoto and Sassa 2001;Kannapinn 2001;Hemerik 2009].…”

“…Assuming that the grammar is specified by a net of F A's, two approaches for building a parser have been followed: (A) transform the grammar into a BN F grammar and apply Knuth's LR (1) construction, or (B) directly construct an ELR (1) parser from the given machine net. It is generally agreed that "approach (B) is better than approach (A) because the transformation adds inefficiency and makes it harder to determine the semantic structure due to the additional structure added by the transformation" [Morimoto and Sassa 2001]. But since approach (A) leverages on existing parser generators such as Bison, it is quite common for language reference manuals featuring syntax chart notations to include also an equivalent BN F LR (1) (or even LALR (1)) grammar.…”