Primality

A high-performance TypeScript library for primality testing using Miller-Rabin and Fermat algorithms. Designed with BigInt support for cryptographic-grade calculations.

Features

• Miller-Rabin Test: Robust probabilistic test.
• Fermat Primality Test: Fast probabilistic test based on Fermat's Little Theorem.
• BigInt Support: Test extremely large numbers without precision loss.
• Zero Dependencies: Lightweight and optimized for performance.

Installation

pnpm add @marlonwq/primality

Usage

import { isPrimeMillerRabin, isPrimeFermat } from '@marlonwq/primality';

// Using Miller-Rabin (Recommended for high accuracy)
// Use the 'n' suffix for BigInt literals
const num1 = 104729n;
console.log(isPrimeMillerRabin(num1)); // true

// Using Fermat (Faster, but watch out for Carmichael numbers)
const num2 = 561n; 
console.log(isPrimeFermat(num2)); // true (Fermat's false positive)
console.log(isPrimeMillerRabin(num2)); // false (Miller-Rabin's correct answer)

How it Works

The Miller-Rabin test is an evolution of Fermat's Little Theorem. Fermat states that if $p$ is prime, then for any integer $a$, $a^{p-1} \equiv 1 \pmod p$. However, some composite numbers (known as Carmichael numbers) also satisfy this condition, leading to false positives.

The Miller-Rabin Secret

The core of Miller-Rabin is the fact that in a prime field $\mathbb{Z}_p$, the equation $x^2 \equiv 1 \pmod p$ has only two solutions: $x = 1$ and $x = p - 1$ (or $-1$).

The Algorithm Steps:

1. Write $n - 1$ as $2^s \cdot d$ by repeatedly factoring out powers of 2.
2. Pick a random base $a$ in the range $[2, n - 2]$.
3. Compute $x = a^d \pmod n$. If $x = 1$ or $x = n - 1$, the number is a probable prime.
4. Otherwise, square $x$ repeatedly ($x = x^2 \pmod n$) up to $s - 1$ times.
5. If at any point $x = n - 1$, the number is a probable prime.
6. If the loop finishes without ever hitting $n - 1$, the number is definitely composite.

Accuracy

Miller-Rabin is a probabilistic algorithm. Each iteration ($k$) reduces the probability of a composite number being declared prime to less than $1/4$. With $k=40$, the error probability is less than $4^{-40}$ — a value so infinitesimal that it is statistically more likely for a meteor to strike your computer than for the test to provide a wrong answer.

Contributing

We use pnpm as our package manager. To get started:

pnpm install
pnpm build
pnpm test

If you find this project helpful, please consider contributing in the following ways: Submitting a pull request, opening an issue, giving the project a star or buying me a coffee!!

License

MIT