murmur-random

Deterministic random generator powered by MurmurHash3

Usage

const murmurRandom = require('murmur-random')
 
const r = murmurRandom('whatever seed you prefer')
 
console.log(r.value('static string key', 3, 12)) // behaves like Math.random(), but it's repeatable
console.log(r.value('followed by numbers', 42))
console.log(r.value('as many as you prefer', 1, 1, 2, 3, 5, 8, 13, 21))
console.log(r.value('yes, zero is an option'))
 
const sub = r.subgen('same key and numbers thing', 420, 69)
 
console.log(sub.value('static string key', 3, 12)) // now it's a different one because the generator is different
 
// advanced stuff
 
console.log(r.util.int(r.value('foo', 2), 5, 7)) // int between 5 and 7
console.log(r.util.from(r.value('bar'), ['you', 'can', 'pick', 'random', 'stuff', 'from', 'arrays']))
 
const seed = r.seed // array of numbers, this is the actual seed the generator uses
 
const cloned = murmurRandom.raw(seed) // you can save and restore generators with this technique

The random generator of murmurRandom can be constructed with the exported function in two ways:

• murmurRandom(string, length = 4): uses a string seed, and derives the numeric seed from it, where length lets you configure the numeric seed length
• murmurRandom.raw(number[]): uses the numeric seed directly

Both functions return the same kind of random generator, which has the following properties:

• r.value(key, ...params): returns a value computed from a string key and an arbitrary amount of numbers
• r.subgen(key, ...params): returns another random generator, computing its seed from the same kind of inputs
• r.util: various utility functions, see below
• r.seed: the numeric seed array (read-only, write is unsupported but technically possible)

The string key is meant to be a static identifier of the value you expect from the generator. Do not use dynamic "key" arguments, they can lead to a memory leak (because keys are cached indefinitely). String literals or constants are recommended. Any dynamic change you need can be included in the numeric arguments following the key.

Why? It's because murmur-random is optimized for speed. Previously, a very similar deterministic random generator was used in the artgen project, which showed that long strings like encoded JSON aren't sufficiently performant with a hash like murmur3

Utils

The random generator has a util object, which is meant to help expand a single random value from 0 to 1 into something more usable. It is meant for murmurRandom, but it can also be combined with any other random generator (and imported as murmur-random/util if the rest of the library is not needed).

The provided utility functions are the following:

• r.util.float(value, min, max): returns a float between the specified bounds
• r.util.int(value, min, max): returns an int between the specified bounds
• r.util.from(value, array): returns a random element from an array
• r.util.repeat(value, func, min, max): executes the second argument a random amount of times, between min and max. An index is passed as an argument.
• r.util.chance(value, probability): returns a boolean that's true with the given probability
• r.util.point(value, min, max, ymin, ymax): returns a point in the {x, y} format between the specified bounds
• r.util.pick(value, array, amount): picks amount random elements from the array

In all of the utility functions, the first value parameter is a value between 0 and 1, generated by r.value() (or any other generator you prefer).

If a util specifies a min, max pair, you can provide just one value, which will set the bounds between 0 and that. Furthermore, if ymin is not passed, r.util.point() uses the same bounds for y as it did on x.

Contributing

Pull requests are welcome. As always, be respectful towards each other and maybe run or create tests, as appropriate. It's on npm test, as usual. (that's a todo)

Special thanks to Gary Court, whose MurmurHash3 implementation powers the project.

murmur-random is available under the MIT license.