Purely functional GLL parsing

Abstract

Generalised parsing has become increasingly important in the context of software language design and several compiler generators and language workbenches have adopted generalised parsing algorithms such as GLR and GLL. The original GLL parsing algorithms are described in low-level pseudo-code as the output of a parser generator. This paper explains GLL parsing differently, defining the FUN-GLL algorithm as a collection of pure, mathematical functions and focussing on the logic of the algorithm by omitting implementation details. In particular, the data structures are modelled by abstract sets and relations rather than specialised implementations. The description is further simplified by omitting lookahead and adopting the binary subtree representation of derivations to avoid the clerical overhead of graph construction. Conventional parser combinators inherit the drawbacks from the recursive descent algorithms they implement. Based on FUN-GLL, this paper defines generalised parser combinators that overcome these problems. The algorithm is described in the same notation and style as FUN-GLL and uses the same data structures. Both algorithms are explained as a generalisation of basic recursive descent algorithms. The generalised parser combinators of this paper have several advantages over combinator libraries that generate internal grammars. For example, with the generalised parser combinators it is possible to parse larger permutation phrases and to write parsers for languages that are not context-free. The 'BNF combinator library' is built around the generalised parser combinators. With the library, embedded and executable syntax specifications are written. The specifications contain semantic actions for interpreting programs and constructing syntax trees. The library takes advantage of Haskell's type-system to type-check semantic actions and Haskell's abstraction mechanism enables 'reuse through abstraction'. The practicality of the library is demonstrated by running parsers obtained from the syntax descriptions of several software languages.

RNGSGLR: Generalization of the Context-Aware Scanning Architecture for All Character-Level Context-Free Languages

“…We will attempt to create an incremental version of our architecture, which will support insertions anywhere in the input string. Our architecture could be adapted to other parsing methods that have the same power as GLR, such as GLL [44]. Since our architecture can parse any specification, it can be useful for applications where the specifications are generated procedurally, such as grammar synthesis [45].…”

A new view on parser combinators

“…Hence, the different parse tree of eRight is wrong. The left-corner transformation is an algorithm to transform possible left-recursive grammars to right-recursive grammars [2,11,21,23,28]. It solves the non-termination problem for left-recursive grammars, but there are two major disadvantages.…”