lark

Lark is a Python library for parsing text that follows a formal grammar. A parser is a program that reads structured input, such as a programming language, a configuration format, or a custom mini-language, and turns it into a structured tree your code can work with. Lark handles a wide class of grammars called context-free grammars, which covers virtually every programming language and many other structured text formats.

You define the rules of your language using a notation called EBNF, a standard way to describe grammars with patterns, repetitions, and alternatives. Once you provide a grammar, Lark reads it and builds a parser automatically. You do not need to write the tree-building code yourself: Lark produces an annotated parse tree from your input, ready to traverse or transform.

Lark offers two main parsing strategies. Earley can handle any valid grammar, including ambiguous ones where a sentence can be interpreted multiple ways, and Lark will return all possible interpretations. LALR(1) is faster and more memory-efficient, suited for grammars that have no ambiguity; it can also generate a stand-alone parser file that does not depend on Lark at runtime. You pick the strategy that fits your speed and flexibility needs.

The library is pure Python with no external dependencies, so installation is a single pip command. It tracks line and column numbers automatically, which helps when you need to report errors to users. An interactive parser mode lets you step through parsing incrementally, useful for debugging or building editors with real-time feedback. Unicode is fully supported throughout.

Lark is used in several well-known Python projects, including Poetry for package management, Vyper for smart contract development, and the Hypothesis testing library. It supports syntax highlighting extensions for several popular editors, has a companion JavaScript port for LALR(1) grammars, and comes with a tutorial showing how to build a JSON parser from scratch.