@lukeed/uuid

> A tiny (~230B) and fast UUID (v4) generator for Node and the browser.

This module offers two modes for your needs:

• @lukeed/uuid
  The default is "non-secure", which uses Math.random to produce UUIDs.
• @lukeed/uuid/secure
  The "secure" mode produces cryptographically secure (CSPRNG) UUIDs using the current environment's crypto module.

> Important:
Version 1.0.0 only offered a "secure" implementation.
In v2.0.0, this is now exported as the "@lukeed/uuid/secure" entry.

Additionally, this module is preconfigured for native ESM support in Node.js with fallback to CommonJS. It will also work with any Rollup and webpack configuration.

Install

$ npm install --save @lukeed/uuid

Modes

There are two "versions" of @lukeed/uuid available:

@lukeed/uuid

> Size (gzip): 231 bytes
> Availability: CommonJS, ES Module, UMD

Relies on Math.random, which means that, while faster, this mode is not cryptographically secure.
Works in Node.js and all browsers.

@lukeed/uuid/secure

> Size (gzip): 235 bytes
> Availability: CommonJS, ES Module, UMD

Relies on the environment's crypto module in order to produce cryptographically secure (CSPRNG) values.
Works in all versions of Node.js. Works in all browsers with crypto.getRandomValues() support.

Usage

import { v4 as uuid } from '@lukeed/uuid';
import { v4 as secure } from '@lukeed/uuid/secure';

uuid(); //=> '400fa120-5e9f-411e-94bd-2a23f6695704'
uuid(); //=> 'cd6ffb4d-2eda-4c84-aef5-71eb360ac8c5'

secure(); //=> '8641f70e-8112-4168-9d81-d38170bfa612'
secure(); //=> 'd175fabc-2a4d-475f-be56-29ba8104c2f2'

API

uuid.v4()

Returns: string

Creates a new Version 4 (random) RFC4122 UUID.

Benchmarks

> Running on Node.js v12.18.4

Validation:
  ✔ String.replace(Math.random)
  ✔ String.replace(crypto)
  ✔ uuid/v4
  ✔ @lukeed/uuid
  ✔ @lukeed/uuid/secure

Benchmark:
  String.replace(Math.random)  x    381,358 ops/sec ±0.31% (93 runs sampled)
  String.replace(crypto)       x     15,842 ops/sec ±1.16% (86 runs sampled)
  uuid/v4                      x  1,259,600 ops/sec ±0.45% (91 runs sampled)
  @lukeed/uuid                 x  6,384,840 ops/sec ±0.22% (95 runs sampled)
  @lukeed/uuid/secure          x  5,439,096 ops/sec ±0.23% (98 runs sampled)

> Running on Chrome v85.0.4183.121

Validation:
  ✔ String.replace(Math.random)
  ✔ uuid/v4
  ✔ @lukeed/uuid
  ✔ @lukeed/uuid/secure

Benchmark:
  String.replace(Math.random)  x    313,213 ops/sec ±0.58% (65 runs sampled)
  uuid/v4                      x    302,914 ops/sec ±0.94% (64 runs sampled)
  @lukeed/uuid                 x  5,881,761 ops/sec ±1.29% (62 runs sampled)
  @lukeed/uuid/secure          x    852,939 ops/sec ±0.88% (65 runs sampled)

Performance

The reason why this UUID.V4 implementation is so much faster is two-fold:

1. It composes an output with hexadecimal pairs (from a cached dictionary) instead of single characters.
2. It allocates a larger Buffer/ArrayBuffer up front (expensive) and slices off chunks as needed (cheap).

The @lukeed/uuid/secure module maintains an internal ArrayBuffer of 4096 bytes, which supplies 256 uuid.v4() invocations. However, the default module preallocates 256 invocations using less memory upfront. Both implementations will regenerate its internal allocation as needed.

A larger buffer would result in higher performance over time, but I found this to be a good balance of performance and memory space.

License

MIT © Luke Edwards