Random-gen

This is an experiment. I do not pretend it is well secure. This is a work in progress

Description

Many methods are used to generate pseudo-random values. Random-gen gets the current time eagerly on occasions (bootstrap, instanciating the basic class with new etc) and then generates a random number on demand again with a second current time.

Why and how

This is based only on JavaScript and does normally behaves the same on all JavaScript engines. Contary to Math.random() and crypto.getRandomValues(). It builds an array of random numbers like [124, 122, 10] basically from the current time new Date() and randomly (same way, not really recursively) pick some numbers from the array to build a new value and move on.

This comes with the assumption that calling successively generate() is not that fast, obviously it does not work with a quantom compuer.

Example

const lib = new RandomGen()
const rand1 = lib.generate()
console.log(rand1) // 2696
t.true(rand1 > 0)
const rand2 = lib.generate()
console.log(rand2) // 14861
t.true(rand2 > 0)

Benchmarks

stress test (no overflows in constructor ?)
Execution time (hr): 5s 77.9351ms
-----------------------

stress test (no overflows in methods?)
Execution time (hr): 7s 996.0777ms
-----------------------

stress test versus Math.random
Execution time (hr): 0s 71.0504ms
-----------------------

stress test versus crypto.randomInt
Execution time (hr): 0s 125.3639ms

Results

For 9999999 iterations:

• Math.random is the fastest
• crypto.randomInt is faster but in the same order (1/10) apparently (9999999 iterations)
• No memory blow or overflow apparently
• It doesn't seem to converge to a value