node-levenshtein :rocket:

Fastest JavaScript implementation of the Levenshtein distance algorithm - (up to 10x+ faster than closest competitor). This algorithm can be used to measure the similarity of two strings. The performance has been achieved by using Myers' bit-parallel algorithm.

Note: if you need to take the edit distance of one string against many strings (maybe billions), I have created a GPU-implementation that will be atleast 1000x faster than this. Contact me, if you're interested.

$ npm install node-levenshtein --save

Usage

1const levenshtein = require('node-levenshtein')
2
3// Print edit-distance between 'fast' and 'faster' 
4console.log(levenshtein.compare('fast', 'faster'))
5//=> 2
6
7// Print string from array with lowest edit-distance to 'fast'
8console.log(levenshtein.find('fast', ['slow', 'faster', 'fastest']))
9//=> 'faster'

Benchmark

I generated 500 pairs of strings with length N. I measured the ops/sec each library achieves to process all the given pairs. Higher is better. node-levenshtein is a lot faster in all cases.

Test Target	N=4	N=8	N=16	N=32	N=64	N=128	N=256	N=512	N=1024
node-levenshtein	42423	17237	5577	3428	2013	596.1	182.6	49.88	12.99
js-levenshtein	24012	12684	4100	1311	371	97.25	25.40	6.404	1.632
leven	30291	11002	3378	907	227	59.94	15.59	3.958	0.998
talisman	29859	9782	3319	926	234	63.76	16.12	3.989	0.986
fast-levenshtein	23517	7686	2291	576	154	40.13	10.16	2.558	0.643
levenshtein-edit-distance	31423	10783	3209	851	214	56.13	14.24	3.694	0.931

Relative Performance

This image shows the relative performance between node-levenshtein and js-levenshtein (the 2nd fastest). As you can see, node-levenshtein is a lot faster for small strings, but especially for long strings (8x faster when N = 1024).

License

This project is licensed under the MIT License - see the LICENSE.md file for details