starlight-numera

A fast, lightweight numerical computing library for JavaScript inspired by NumPy.

Built for performance using TypedArrays and optimized loops.

Features

• Fast ndarray implementation (TypedArray-based)
• Vectorized operations (add, sub, mul, div)
• Matrix operations (matmul, transpose, dot)
• Broadcasting (NumPy-style)
• Math functions (sqrt, exp, log, trig)
• Modular and tree-shakeable

Installation

npm install starlight-numera

Quick Start

import { array, add, matmul, sqrt } from 'starlight-numera';

// Create arrays
const a = array([[1, 2], [3, 4]]);
const b = array([[5, 6], [7, 8]]);

// Element-wise operations
const c = add(a, b);

// Matrix multiplication
const d = matmul(a, b);

// Math functions
const e = sqrt(a);

console.log(c);
console.log(d);
console.log(e);

ndarray Structure

Each array is stored as:

{
  data: Float32Array,
  shape: number[],
  stride: number[],
  size: number
}

Operations

Element-wise

add(a, b)
sub(a, b)
mul(a, b)
div(a, b)

Broadcasting

const a = array([[1, 2, 3]]);
const b = array([[10], [20]]);

add(a, b);
// [
//  [11,12,13],
//  [21,22,23]
// ]

Linear Algebra

transpose(a)
dot(a, b)
matmul(a, b)

Math Functions

sqrt(a)
exp(a)
log(a)

sin(a)
cos(a)
tan(a)

Performance

• Uses Float32Array for fast memory access
• Avoids unnecessary allocations
• Optimized loops for numerical operations

Roadmap

• Advanced broadcasting optimizations
• Slicing and indexing
• GPU acceleration (WebGL / WebGPU)
• SIMD optimizations
• Higher-dimensional tensors

Contributing

Contributions are welcome.

1. Fork the repository
2. Create a feature branch
3. Submit a pull request

License

MIT License