Phasor.js is a wasm (via Rust) implementation of complex numbers that strives to yield correct results.

import {i, rect} from 'phasor.js';

i(Math.PI).exp().add(rect(1)); // approximately rect(0);

Phasor.js was originally developed as part of Steady to provide the core complex algebra primitives required in the implementation of algorithms that solve electrical circuits.

While it wasn't Steady's goal to implement complex numbers from scratch, circumstances proved it necessary given the fact that, at the time (Feb 2018), no alternative library available on NPM was able to correctly handle edge cases such as complex numbers with infinite magnitude.

For example, all four alternatives tested^[a][b][c][d] compute ∞i × -∞i to NaN instead of ∞ and three of them also generally fail to prevent overflow/underflow even in trivial expressions such as 1E200i / 1E200i and 1E-200i / 1E-200i.

Phasor.js on the other hand successfully passes all of the following assertions.

i(Infinity).mul(i(-Infinity)).ulpsEq(rect(Infinity));
i(1E200).div(i(1E200)).ulpsEq(rect(1));
i(1E-200).div(i(1E-200)).ulpsEq(rect(1));

In 2020, Phasor.js was re-implemented from scratch in Rust.

Constructs a complex number given it's real and imaginary parts.

Constructs a complex number given it's magnitude and angle.

Constructs a purely imaginary number.

Example:

i(42).ulpsEq(rect(0, 42));

Extracts the real part of a complex number.

Example:

rect(3).real() === 3;
rect(3, 4).real() === 3;

Extracts the imaginary part of a complex number.

Example:

rect(3).real() === 0;
rect(3, 4).imag() === 4;

Extracts the magnitude of a complex number.

Example:

rect(3, 4).norm() === 5;

Extracts the angle of a complex number.

Example:

rect(3, 4).angle() === Math.atan2(4, 3);

Compares two complex numbers for approximate equality, optionally taking a positive residue.

Compares two complex numbers for approximate equality, optionally taking a positive residue and a maximum relative distance.

Compares two complex numbers for approximate equality, optionally taking a positive residue and the maximum distance of Units in the Last Place.

Computes the addition of two complex numbers.

Example:

rect(3).add(rect(0, 4)).ulpsEq(rect(3, 4));

Computes the subtraction of two complex numbers.

Example:

rect(3).sub(rect(0, 4)).ulpsEq(rect(3, -4));

Computes the multiplication of two complex numbers.

Example:

rect(3).mul(rect(0, 4)).ulpsEq(i(12));

Computes the division of two complex numbers.

Example:

rect(3).div(rect(0, 4)).ulpsEq(i(-0.75));

Computes the opposite of a complex number.

Example:

rect(3, 4).neg().ulpsEq(rect(-3, -4));

Computes the conjugate of a complex number.

Example:

rect(3, 4).conj().ulpsEq(rect(3, -4));

Computes the reciprocal of a complex number.

Example:

rect(3, 4).recip().ulpsEq(rect(3 / 25, -4 / 25));

Computes the exponential of a complex number.

Example:

rect(3, 4).exp().ulpsEq(polar(Math.exp(3), 4));

Computes the principal natural logarithm of a complex number.

Example:

rect(3, 4).ln().ulpsEq(rect(Math.log(5), Math.atan(4 / 3)));

Computes the principal logarithm of a complex number to an arbitrary base.

Example:

rect(3, 4).log(10).ulpsEq(rect(Math.log10(5), Math.atan(4 / 3) / Math.log(10)));

Computes the hyperbolic sine of a complex number.

Example:

const p = rect(3, 4);
const q = polar(
    -Math.hypot(Math.sinh(3), Math.sin(4)),
    Math.atan2(Math.tan(4), Math.tanh(3)),
);

p.sinh().ulpsEq(q);

Computes the hyperbolic cosine of a complex number.

Example:

const p = rect(3, 4);
const q = polar(
    -Math.hypot(Math.sinh(3), Math.cos(4)),
    Math.atan2(Math.tan(4), 1 / Math.tanh(3)),
);

p.cosh().ulpsEq(q);

Returns true if either the imaginary or real part (or both) of a complex number is NaN.

Example:

console.assert(!i(0).isNaN());
console.assert(!i(1).isNaN());
console.assert(!i(1E-315).isNaN())
console.assert(!i(Infinity).isNaN());
console.assert(i(NaN).isNaN());

Returns true if the magnitude of a complex number is Infinity.

Example:

console.assert(!i(0).isInfinite());
console.assert(!i(1).isInfinite());
console.assert(!i(1E-315).isInfinite())
console.assert(i(Infinity).isInfinite());
console.assert(!i(NaN).isInfinite());

Returns true if a complex number is neither NaN nor infinite.

Example:

console.assert(i(0).isFinite());
console.assert(i(1).isFinite());
console.assert(i(1E-315).isFinite())
console.assert(!i(Infinity).isFinite());
console.assert(!i(NaN).isFinite());

Returns true if the magnitude of a complex number is zero.

Example:

console.assert(i(0).isZero());
console.assert(!i(1).isZero());
console.assert(!i(1E-315).isZero())
console.assert(!i(Infinity).isZero());
console.assert(!i(NaN).isZero());

Returns true if the magnitude of a complex number is subnormal.

Example:

console.assert(!i(0).isSubnormal());
console.assert(!i(1).isSubnormal());
console.assert(i(1E-315).isSubnormal())
console.assert(!i(Infinity).isSubnormal());
console.assert(!i(NaN).isSubnormal());

Returns true if a complex number is not NaN, infinite, zero, or subnormal.

Example:

console.assert(!i(0).isNormal());
console.assert(i(1).isNormal());
console.assert(!i(1E-315).isNormal())
console.assert(!i(Infinity).isNormal());
console.assert(!i(NaN).isNormal());

Returns true if a complex number is purely real.

Example:

console.assert(rect(3).isReal());
console.assert(!rect(0, 4).isReal());
console.assert(!rect(3, 4).isReal());

Returns true if a complex number is purely imaginary.

Example:

console.assert(!rect(3).isImaginary());
console.assert(rect(0, 4).isImaginary());
console.assert(!rect(3, 4).isImaginary());

Complex numbers are represented under the hood by their magnitude and the tangent of its angle, which might seem unusual, but it's one of the main reasons why Phasor.js is able to yield numerically correct results.

The tangent of an angle has many advantages from the point of view of numerical algorithms than angles expressed in radians, the most important one being the fact that it makes it possible to implement most algorithms, in particular the four fundamental operations (add, sub, mul and div), without relying on any trigonometric function, which are highly non-linear and a major source of numerical error.

Phasor.js essentially takes advantage of the following two trigonometric identities that rely on the less non-linear square root function:

cos(tan(α)) ≡ 1 / sqrt(1 + α^2)
sin(tan(α)) ≡ sign(α) / sqrt(1 + 1 / α^2)

These identities are particularly convenient, because those square roots in the denominators can take advantage of the hypot algorithm, which efficiently avoids unnecessary over/underflows.