levenjs

Fast implementation of the Levenshtein edit distance. References for the various function can be found in the docstrings [1, 2, 3].

Install

npm install levenjs

Usage

import levenjs from "levenjs";

levenjs("distance", "dist");  // 4

If you want to limit the maximum edit distance returned.

import levenjs from "levenjs";

levenjs("distance", "dist", /*maxDist=*/ 2);  // 2

Benchmark

Benchmarked against leven and fastest-levenshtein.

One word
         leven               x 646,822 ops/sec ±0.64% (98 runs sampled)
         fastest-levenshtein x 1,529,761 ops/sec ±0.76% (95 runs sampled)
         levenjs             x 1,331,218 ops/sec ±0.34% (97 runs sampled)
Short sentence, small difference
         leven               x 868,577 ops/sec ±0.63% (90 runs sampled)
         fastest-levenshtein x 6,881,822 ops/sec ±0.48% (97 runs sampled)
         levenjs             x 5,578,840 ops/sec ±0.98% (93 runs sampled)
Short sentence, similar prefix and suffix, small difference
         leven               x 2,242,494 ops/sec ±0.63% (92 runs sampled)
         fastest-levenshtein x 4,920,882 ops/sec ±0.39% (98 runs sampled)
         levenjs             x 6,336,442 ops/sec ±0.57% (92 runs sampled)
Long text, small difference
         leven               x 1,011 ops/sec ±0.67% (95 runs sampled)
         fastest-levenshtein x 10,350 ops/sec ±0.47% (97 runs sampled)
         levenjs             x 10,796 ops/sec ±0.42% (97 runs sampled)
Long text, small difference, threshold 3
         leven               x 894 ops/sec ±0.41% (97 runs sampled)
         fastest-levenshtein x 10,374 ops/sec ±1.31% (96 runs sampled)
         levenjs             x 397,756 ops/sec ±0.52% (97 runs sampled)
Long text, small difference, threshold 20
         leven               x 1,030 ops/sec ±0.43% (97 runs sampled)
         fastest-levenshtein x 10,312 ops/sec ±0.55% (97 runs sampled)
         levenjs             x 17,998 ops/sec ±0.41% (94 runs sampled)
Long text, many difference
         leven               x 622 ops/sec ±0.58% (96 runs sampled)
         fastest-levenshtein x 7,422 ops/sec ±0.43% (98 runs sampled)
         levenjs             x 6,342 ops/sec ±0.28% (99 runs sampled)
Long text, many difference, threshold 40
         leven               x 634 ops/sec ±0.25% (95 runs sampled)
         fastest-levenshtein x 7,879 ops/sec ±0.26% (99 runs sampled)
         levenjs             x 169,151 ops/sec ±0.50% (95 runs sampled)

Slightly poorer performance than fastest-levenshtein in uncapped cases but better performance when pre-/post- fix trimming shortens string lengths past certain break points. Also better performance when a maxDist threshold is set.