Number Generator

Generate repeatable pseudo random numbers and non-cryptographic hash numbers for usage in Node.js and browser environments.

Update to version 4
Usage
Development
Disclaimer

Update to version 4

Change murmurHash to murmurhash2_x86_32 (same function since v2, alias now removed).
Non ASCII characters (e.g. chinese or emoji) are now handled properly. This can produce different results if these characters where used before inside an input hash for all murmur hash functions.

Usage

This library (7.45 KB, gzipped size: 2.94 KB) contains the following methods: one PRNG (pseudo random number generator) called Alea and four number hash generators, MurmurHash2 and MurmurHash3 for 32 and 128 bit (x86 and x64) hash strings.

More about the hash function MurmurHash can be found here on wikipedia.

Install

Based on your package manager you can install it via:

# NPM 
npm install number-generator
 
# Yarn 
yarn add number-generator
 
# pnpm 
pnpm install number-generator

After that you can import it either as a library, e.g.:

// ESM
import * as numberGenerator from "number-generator";
 
// CJS
const numberGenerator = require("number-generator");

Or as single functions:

// ESM
import aleaRNGFactory from "number-generator/lib/aleaRNGFactory";
import murmurhash2_x86_32 from "number-generator/lib/murmurhash2_x86_32";
import murmurhash3_x86_32 from "number-generator/lib/murmurhash3_x86_32";
import murmurhash3_x86_128 from "number-generator/lib/murmurhash3_x86_128";
import murmurhash3_x64_128 from "number-generator/lib/murmurhash3_x64_128";
 
// CJS
const aleaRNGFactory = require("number-generator/lib/aleaRNGFactory");
const murmurhash2_x86_32 = require("number-generator/lib/murmurhash2_x86_32");
const murmurhash3_x86_32 = require("number-generator/lib/murmurhash3_x86_32");
const murmurhash3_x86_128 = require("number-generator/lib/murmurhash3_x86_128");
const murmurhash3_x64_128 = require("number-generator/lib/murmurhash3_x64_128");

Also the library can safely be tree shaked. If tree shaking is used in e.g. Rollup or Webpack this will only put the used function with helpers in your bundle:

import { aleaRNGFactory } from "number-generator";

For use with TypeScript take a look at the usage with typescript section of this document.

Remark: For direct browser usage you can use the exposed numberGenerator global, e.g.:

// Direct browser usage e.g.:
numberGenerator.murmurhash2_x86_32("something");

All supported environments are listed under the support section.

Random numbers

You can use the aleaRNGFactory method to generate (pseudo) random numbers based an a seed (default seed is 1). Every seed produces the same result for the created getter method.

Attention: The default seed 1 should not be used! It produces one duplicate at 4370 calls. This can be avoided by using a seed larger or equal to 2.

Create a new random number generator

First step is to include the library functions you want to use in your application. If you only want to use the Alea implementation you can import it directly by:

const aleaRNGFactory = require("number-generator/lib/aleaRNGFactory");

Now you can create a new generator by calling the function with a seed equal or larger to 2. The number 0, float or negative numbers are not valid and will throw a TypeError.

// Valid:
const generator1 = aleaRNGFactory(2);
const generator2 = aleaRNGFactory(4836325);
 
// Invalid:
const notValidGen1 = aleaRNGFactory(0);
const notValidGen2 = aleaRNGFactory(0.47);
const notValidGen3 = aleaRNGFactory(-1);

Create an unsigned integer

If you have a valid generator object you can use the uInt32 method to get a random unsigned integer. Call it multiple times to get new numbers.

const { uInt32 } = aleaRNGFactory(10);
uInt32(); // 20916391
uInt32(); // 1567221093

This will create the exact same result on your machine! You get always the same values for the same seed used.

This means if you create multiple generators with the same seed, you get the same result for the n-th call:

const generator1 = aleaRNGFactory(2);
const generator2 = aleaRNGFactory(2);
 
const value1 = generator1.uInt32();
const value2 = generator2.uInt32();
 
value1 === value2; // true

Create an unsigned float

The same that works for the uInt32 method applies to the uFloat32 method. But this time you get an unsigned float.

const { uFloat32 } = aleaRNGFactory(5);
uFloat32(); // 0.0024349885061383247
uFloat32(); // 0.1826920467428863

Again, this should create the exact same result on your machine!

If you create multiple generators with the same seed, you get the same result for the n-th call:

const generator1 = aleaRNGFactory(4);
const generator2 = aleaRNGFactory(4);
 
const value1 = generator1.uFloat32();
const value2 = generator2.uFloat32();
 
value1 === value2; // true

Change the seed

You can change the seed used by the generator object with the setSeed method.

const generator = aleaRNGFactory(2);
 
// Get some random numbers
generator.uInt32();
generator.uInt32();
 
// Change seed
generator.setSeed(3);
 
// Get some more random numbers
generator.uInt32();
generator.uInt32();

Get the state

You can get and restore the internal state with getState and setState.

const generator = aleaRNGFactory(42);
generator.uInt32();
generator.uInt32();
const state = generator.getState(); // Get the generator state
const value1 = generator.uInt32();
generator.uInt32();
generator.uInt32();
generator.setState(state); // Restore the previous state
const value2 = generator.uInt32();
 
value1 === value2; // true

For TypeScript the state interface is NumberGeneratorState.

import { aleaRNGFactory, NumberGeneratorState } from "number-generator";
 
const generator = aleaRNGFactory(2);
generator.uInt32();
 
const state: NumberGeneratorState = generator.getState();

Set the state

You can set the state with setState on two ways. Either you don't pass any parameter to the state function, where it will reset the state to the initial state. Or you can pass a state object to restore a previous state:

const generator = aleaRNGFactory(2);
generator.uInt32();
 
const state = generator.getState();
generator.setState(); // Reset the state
generator.uInt32(); // Get a new value
 
generator.setState(state); // Restore saved state

Something like Math.random?

If you want something similar to Math.random() (without duplicate values) you can use the JavaScript Date API with a timestamp and combine it with the uFloat32 method from the aleaRNGFactory e.g.:

const { uFloat32: random } = aleaRNGFactory(Date.now());
 
// Get a random float number
random();

Murmur hash

To generate a hash there are four functions, murmurhash2_x86_32, murmurhash3_x86_32, murmurhash3_x86_128 and murmurhash3_x64_128. The murmurhash functions implement the MurmurHash algorithm for 32 and 128 bit in JavaScript (murmurhash2 and 3) for x86 and x64. They take a string and generate a non-cryptographic hash number as unsigned integer with 32 bit or a string hash with 128 bit.

You can import the functions directly:

const murmurhash2_x86_32 = require("number-generator/lib/murmurhash2_x86_32");
const murmurhash3_x86_32 = require("number-generator/lib/murmurhash3_x86_32");
const murmurhash3_x86_128 = require("number-generator/lib/murmurhash3_x86_128");
const murmurhash3_x64_128 = require("number-generator/lib/murmurhash3_x64_128");

Both murmurhash2_x86_32 and murmurhash3_x86_32 will generate a unsigned 32 bit number. The murmurhash3_x86_128 and murmurhash3_x64_128 functions will generate a 128 bit string. To showcase the difference:

murmurhash2_x86_32("Hello"); // 1826530862
murmurhash3_x86_32("Hello"); // 316307400
murmurhash3_x86_128("Hello"); // "2360ae465e6336c6ad45b3f4ad45b3f4"
murmurhash3_x64_128("Hello"); // "35b974ff55d4c41ca000eacf29125544"

Basic hash generation

All murmur hash functions work the same. So the following examples will take the murmur hash 2 function to demonstrate the usage. To use it pass a string to generate the hash number. The default seed used is 0.

const hash1 = murmurhash2_x86_32("My string.");
const hash2 = murmurhash2_x86_32("My string.", 0);
 
hash1; // 1836966117
hash1 === hash2; // true

This will create the exact same result on your machine!

Hash based on different seeds

Different seeds generate different results for the same input string. Only whole numbers are valid seeds for any murmur hash function!

const hash1 = murmurhash2_x86_32("My string.", 1);
const hash2 = murmurhash2_x86_32("My string.", 2);
 
hash1 === hash2; // false

A float number as a seed throws a TypeError:

const hash = murmurhash2_x86_32("My string.", 0.7); // TypeError!

TypeScript

This package contains all the type definitions for TypeScript. Every murmur hash function implements the NumberHashGenerator interface.

import {
  NumberGenerator,
  NumberHashGenerator,
  aleaRNGFactory,
  murmurhash2_x86_32,
  murmurhash3_x86_32
} from "number-generator";
 
const generator: NumberGenerator = aleaRNGFactory(2);
// const factory: () => NumberGenerator = aleaRNGFactory;
 
const hashFn1: NumberHashGenerator = murmurhash2_x86_32;
const hashFn2: NumberHashGenerator = murmurhash3_x86_32;
 
generator.uInt32();
hashFn1("What?", 42);
hashFn2("something", 14);

Support

This library was tested on the following environments:

Node >= 8
All major browsers and IE >= 10

Exceptions for Node are versions that reached the End-of-Life as defined under https://github.com/nodejs/Release#end-of-life-releases.

Benchmarks

Disclaimer: The following benchmarks were created on a MacBook Pro, Processor 2,7 GHz Intel Core i5 with 8 GB 1867 MHz DDR3 memory and run under Node v10.16.3.

aleaRNGFactory

Comparison between uInt32 and uFloat32 methods:

// v4.0.0
aleaRNGFactory#uInt32()   x 8,037,159 ops/sec
aleaRNGFactory#uFloat32() x 7,435,745 ops/sec

murmurhash

Comparison between murmurhash2_x86_32, murmurhash3_x86_32, murmurhash3_x86_128 and murmurhash3_x64_128 function:

// v4.0.0
murmurhash2_x86_32  x 506,674 ops/sec
murmurhash3_x86_32  x 510,627 ops/sec
murmurhash3_x86_128 x 245,012 ops/sec
murmurhash3_x64_128 x 135,522 ops/sec

To run them on your machine execute pnpm run test:benchmark.

Development

If you want to contribute see the CONTRIBUTING.md

Disclaimer

"Why pseudo random number generators and number hash functions" you may ask? Read more in this fantastic blog post about "A primer on repeatable random numbers" from Rune Skovbo Johansen.

Thanks to Johannes Baagøe for the Alea port and Ray Morgan for the MurmurHash2 algorithm implementation in JavaScript. Also thanks to Karan Lyons for the MurmurHash3 implementation.

Resources used to tests against implementations in other languages are:

JSON for modern C++ by Niels Lohmann: https://github.com/nlohmann/json
MurmurHash2 and 3 in C++ by Austin Appleby: https://github.com/aappleby/smhasher

Big thanks as well to Alex Ciminian for raising the issue with non ASCII characters: https://cimi.io/murmurhash3js-revisited/

npm