Parsey

Parsey is a parser for context-free grammars. It utilizes earley parsing, a top-down chart parsing algorithm that can parse a large set of grammars.

Parsey takes a list of rules (which can be defined using the Rule constructor) and some input sentence, then constructs a parse tree for the given language. The many applications of this include abstract syntax tree creation, NLP, and arithmetic, to name a few.

Installation

Install from npm:

npm install parsey --save

Usage

const Rule    = require('parsey').Rule;
const Sym     = require('parsey').Sym;
const parse   = require('parsey').parse;

const sum     = new Sym('sum');
const prod    = new Sym('prod');
const factor  = new Sym('factor');

// Rule( [lhs : Sym], [rhs : Array], [valuator : Function] )

let grammar = [
  // sum
  new Rule(sum    , [sum, '+', prod]    , (x, _, y) => x + y),
  new Rule(sum    , [prod]              , (x) => x),

  // product
  new Rule(prod   , [prod, '*', factor] , (x, _, y) => x * y),
  new Rule(prod   , [factor]            , (x) => x),

  // factor
  new Rule(factor , ['(', sum, ')']     , (_, x) => x),
  new Rule(factor , [/\d+/]             , (n) => parseFloat(n))
];

let expr = '2 + 3 * (1 + 4)';
let parseTree = parse(exp, grammar);

console.log(evaluate(parseTree));
// => 17

Rules and Sym(bol)s

A Sym is nothing more than an object that references a specific symbol for use in grammar rules, with a name attribute that describes the symbol (names need not be unique).

const Sym = require('parsey').Sym;
let sum = new Sym('sum');
let prod = new Sym('prod');

Rules describe the structure of a language. They contain a left-hand side consisting of a sole symbol, a right-hand side consisting of a list of symbols, and an optional evaluation function.

const Rule = require('parsey').Rule;
//                     lhs        rhs               valuator
let sumRule = new Rule(sum, [sum, '+', prod], (x, _, y) => x + y);

This rule states that a sum can be any sum, followed by a +, followed by any prod. The rhs can be populated with Sym objects, strings, or RegExp objects.

This rule's valuator is a function that takes 3 arguments (one for each symbol on the rhs), and adds the first and last (we don't care about the '+').

The Rule class is extended from Array, so the rhs symbols correspond to a Rule's indices and all of the methods on Array.prototype will apply to the rhs.

The CFG

The CFG is a glorified array, a container for your rules. It has some methods like rule() and getSymbols(), which can be handy for creating rules from strings.

const CFG = require('parsey').CFG;
let grammar = new CFG();

grammar.rule('S -> NP VP');
grammar.rule('NP -> Det N');
grammar.rule('VP -> V NP');

grammar.rule('Det -> "the");
grammar.rule('Det -> "a");

grammar.rule('V -> /runs?/');
grammar.rule('V -> "ran"');

rule() will try to find existing symbols in the grammar that have the same names as the ones that appear in a production string, and will create new Sym objects if a symbol could not be found.

It will also treat string-looking symbols like "the" as terminal symbols and leave them as strings, and likewise will turn things like /regex/ into RegExp objects, which are also terminals.

Parsing

Given a list of Rules, parsing becomes as simple as

const parse = require('parsey').parse;
parse('1 + 2 * 3', grammar);

The parse() function returns a parse tree, which is a node of the following structure:

{
  item: [Rule],
  children: [Array]
}

Elements in children can be either nodes of the above structure, or strings.

Example:

// 1 * 2
{
  item: <Rule 'prod' [prod, '*', factor]>,
  children: [
    {
      item: <Rule 'factor' [/\d+/]>,
      children: [ '1' ]
    },
    '*',
    {
      item: <Rule 'factor' [/\d+/]>,
      children: [ '2' ]
    }
  ]
}

Examples

Check out examples to see parsey in action for various use cases!

Testing

Testing parsey requres jasmine (2.0+). All specs can be found in spec/. To run all tests, simply run npm test in the project's root.

License

MIT

patgrasso/parsey