fastest-levenshtein

fastest-levenshtein is a zero-dependency JavaScript/TypeScript library that computes the Levenshtein edit distance between two strings faster than any other published JS implementation, according to its own benchmark suite against js-levenshtein, leven, fast-levenshtein, and others. It exposes two functions: distance() for the raw edit-distance number, and closest() for finding the string in an array with the smallest edit distance to a target.

The entire implementation lives in a single TypeScript file (mod.ts) with no runtime dependencies, and ships prebuilt CommonJS and ESM outputs alongside .d.ts type declarations. It works identically under Node.js and Deno, making it a drop-in choice anywhere fuzzy string matching, spell-check suggestion, or typo-tolerant search needs a fast core distance metric.

What You Get

• A distance(a, b) function returning the Levenshtein edit distance between two strings
• A closest(target, candidates) function returning the candidate string with the smallest edit distance to the target
• Zero runtime dependencies and a small single-file implementation, keeping install and parse cost minimal
• Prebuilt CommonJS and ESM builds plus TypeScript type declarations out of the box
• Published benchmarks showing consistently higher ops/sec than js-levenshtein, leven, fast-levenshtein, and levenshtein-edit-distance across string lengths from N=4 to N=1024

Common Use Cases

• Typo-tolerant search or autocomplete: ranking candidate strings by similarity to a user’s query
• Spell-check suggestion engines that need to find the closest dictionary word to a misspelled input
• Fuzzy matching of user input against a fixed set of commands, tags, or identifiers
• Deduplication or record-linkage pipelines that score near-duplicate strings by edit distance

Under The Hood

Architecture - The library is a single TypeScript source file, mod.ts (142 lines), exporting exactly two functions: distance and closest. There is no class hierarchy, no configuration object, and no plugin surface — the entire public API is these two pure functions operating on strings and arrays of strings. closest is implemented as a thin wrapper that iterates candidates and calls distance for each, tracking the minimum. Tech Stack - Written in TypeScript, compiled to both CommonJS (mod.js) and ESM (esm/mod.js) via tsc, with a separate tsconfig.esm.json for the ESM build target. It has zero runtime dependencies; devDependencies are limited to TypeScript, Jest for testing, ESLint/Prettier for linting, and several competing Levenshtein implementations (fast-levenshtein, js-levenshtein, leven, natural, talisman) pulled in purely to drive the comparative benchmark suite (bench.ts). Code Quality - Tests exist under test.ts and run via Jest with coverage reporting wired to Coveralls in CI. Given the algorithm’s narrow, well-defined scope (a single numeric function plus a thin wrapper), the surface area for bugs is small; the codebase favors dense, optimized loop logic over abstraction, which is typical and appropriate for a performance-focused micro-library. Repository activity has been low in recent years (last push 2024-03), consistent with a small utility that reached feature-complete maturity rather than one that is unmaintained by neglect. API Design - Getting started requires a single import and a single function call — distance('fast', 'faster') — with no configuration, options object, or setup step. This minimalism is the library’s main developer-experience strength: it does exactly one thing and exposes exactly the two functions needed to do it, in both Node and Deno without any adapter code.