@danielsimonjr/mathts-core

Core types, utilities, and configuration for MathTS.

Installation

npm install @danielsimonjr/mathts-core

Usage

import { DEFAULT_CONFIG, isNumber, isMatrix } from '@danielsimonjr/mathts-core';
import { DenseMatrix } from '@danielsimonjr/mathts-matrix';

// Factory default configuration (MathTSConfig)
const config = { ...DEFAULT_CONFIG, preferredBackend: 'wasm' };

// Type guards
console.log(isNumber(42)); // true
console.log(
  isMatrix([
    [1, 2],
    [3, 4],
  ])
); // false — nested arrays are not matrices
const m = DenseMatrix.fromArray([
  [1, 2],
  [3, 4],
]);
console.log(isMatrix(m)); // true

API

Types

• BackendType - Available backends: 'js' | 'wasm' | 'gpu'
• NumericType - Numeric precision types
• MathTSConfig - Configuration interface

SI dimension vectors

toSiDimensionVector returns exact Fraction exponents [L, M, T, I, Θ, N, J]; use vec.map((f) => f.toNumber()) when you need plain numbers (there is no toSiDimensionNumbers helper). fromSiDimensionVector accepts SiDimensionExponent (number | Fraction) and produces the length-10 Unit exponent array.

Functions

• createSeededRng(seed) / SeededRng — forkable Mulberry32 stream (next, nextInt, normal, fork); number seeds match mulberry32, strings are FNV-1a UTF-8 (not cryptographically secure)

  const rng = createSeededRng(42);
  rng.next(); // 0.6011037519201636
  rng.fork('a').next(); // 0.668102843221277
  createSeededRng('a').seed; // -468965076

• mulberry32(seed) — seeded Mulberry32 PRNG returning U[0, 1) samples (not cryptographically secure)

• boxMuller(u1, u2) — Box–Muller pair of N(0, 1) samples from two uniforms

• rowMajorStrides(shape) — row-major stride vector for a dense tensor shape

• DEFAULT_CONFIG - Default configuration object

• isNumber(value) - Type guard for numbers

• isComplex(value) - Type guard for complex numbers

• isMatrix(value) - Type guard for matrix-like objects (rows, cols, get, type)

License

MIT