node package manager


Efficient implementation of Levenshtein algorithm with locale-specific collator support.

fast-levenshtein - Levenshtein algorithm in Javascript

An efficient Javascript implementation of the Levenshtein algorithm with locale-specific collator support.

  • Works in node.js and in the browser.
  • Better performance than other implementations by not needing to store the whole matrix (more info).
  • Locale-sensitive string comparisions if needed.
  • Comprehensive test suite and performance benchmark.
  • Small: <1 KB minified and gzipped

Install using npm:

$ npm install fast-levenshtein

Using bower:

$ bower install fast-levenshtein

If you are not using any module loader system then the API will then be accessible via the window.Levenshtein object.

Default usage

var levenshtein = require('fast-levenshtein');
var distance = levenshtein.get('back', 'book');   // 2 
var distance = levenshtein.get('我愛你', '我叫你');   // 1 

Locale-sensitive string comparisons

It supports using Intl.Collator for locale-sensitive string comparisons:

var levenshtein = require('fast-levenshtein');
levenshtein.get('mikailovitch', 'Mikhaïlovitch', { useCollator: true});
// 1 

To build the code and run the tests:

$ npm install -g grunt-cli
$ npm install
$ npm run build

Thanks to Titus Wormer for encouraging me to do this.

Benchmarked against other node.js levenshtein distance modules (on Macbook Air 2012, Core i7, 8GB RAM):

Running suite Implementation comparison [benchmark/speed.js]...
>> levenshtein-edit-distance x 234 ops/sec ±3.02% (73 runs sampled)
>> levenshtein-component x 422 ops/sec ±4.38% (83 runs sampled)
>> levenshtein-deltas x 283 ops/sec ±3.83% (78 runs sampled)
>> natural x 255 ops/sec ±0.76% (88 runs sampled)
>> levenshtein x 180 ops/sec ±3.55% (86 runs sampled)
>> fast-levenshtein x 1,792 ops/sec ±2.72% (95 runs sampled)
Benchmark done.
Fastest test is fast-levenshtein at 4.2x faster than levenshtein-component

You can run this benchmark yourself by doing:

$ npm install
$ npm run build
$ npm run benchmark

If you wish to submit a pull request please update and/or create new tests for any changes you make and ensure the grunt build passes.

See for details.

MIT - see