Grass

A greedy, streaming lexer written in vanilla Javascript.

Features

• Fixed memory usage regardless of input size
• Emits both token type and matched text
• 100% statement and branch coverage

Streaming Usage

Grass expects an NFA created with finite-automata.

var grass = require('grass')
  , Fragment = require('finite-automata')
  , fs = require('fs')
  , red
  , blue
  , dog
  , rule
  , lexerStream

// Make a DFA that matches /(red|blue)dog/ and labels it 'animalToken'
red = new Fragment('red')
blue = new Fragment('blue')
dog = new Fragment('dog', 'animalToken')

rule = red.copy().union(blue).concat(dog)

fs.createReadStream('input.txt')
  .pipe(grass(rule))
  .pipe(consumer)

// consumer will recieve tuple ['animalToken', match <string>] for every match

Sync Usage

grass.sync(rulea)('reddogredcatbluedog')
// ==> '[['animalToken', 'reddog'], ['animalToken, 'bluedog']]'

Greediness

Say you have a rule like /11*2*3*4*5*6*7*8*9*/. This matches ascending series of digits that start with 1.

Example matches: 1, '1234', '11223', '199'

Grass is greedy. This means that it will try to match as much text as possible, and only backtrack when it has no way to continue.

This means that grass might not emit a match at the end of the stream's data event. Sending an ascending series like 123 to grass will not result in a match until the stream is ended, or a digit lesser than 3 occurs. This is because grass knows that there is a possibility of a longer match, and it will wait for the next data event to decide on the fate of the data in its internal buffer.

This is expected behavior, because if the series 1234 was sent in two parts, 12 and 34, it should only result in one match.