Taming Context-Sensitive Languages with Principled Stateful Parsing

TL;DR

This paper introduces a novel approach called principled stateful parsing, enabling context-sensitive features in programming languages through mutable parse state, implemented in the Autumn library, improving grammar expressiveness and maintainability.

Contribution

It presents a formal, transactional discipline for stateful parsing that handles context-sensitive features more transparently and modularly than previous ad-hoc methods.

Findings

Successfully implemented context-sensitive features like whitespace and namespace handling.

Enables more maintainable and expressive grammars for complex language features.

Provides a practical library, Autumn, demonstrating the approach.

Abstract

Historically, true context-sensitive parsing has seldom been applied to programming languages, due to its inherent complexity. However, many mainstream programming and markup languages (C, Haskell, Python, XML, and more) possess context-sensitive features. These features are traditionally handled with ad-hoc code (e.g., custom lexers), outside of the scope of parsing theory. Current grammar formalisms struggle to express context-sensitive features. Most solutions lack context transparency: they make grammars hard to write, maintain and compose by hardwiring context through the entire grammar. Instead, we approach context-sensitive parsing through the idea that parsers may recall previously matched input (or data derived therefrom) in order to make parsing decisions. We make use of mutable parse state to enable this form of recall. We introduce principled stateful parsing as a new transactional discipline that makes state changes transparent to parsing mechanisms such as backtracking and memoization. To enforce this discipline, users specify parsers using formally specified primitive state manipulation operations. Our solution is available as a parsing library named Autumn. We illustrate our solution by implementing some practical context-sensitive grammar features such as significant whitespace handling and namespace classification.