Skip to main content

Matrix4

A 4x4 matrix. Any arguments to Matrix4 methods can be plain JavaScript arrays or other math.gl objects.

Usage​

import {Matrix4} from `math.gl`;

Copy a matrix to a Matrix4 so that it can be manipulated (and mutated) with Matrix4 methods:

const IDENTITY = [1, 0, ..., 1];
const m = new Matrix4(IDENTITY).translate([1, 0, 0]);

Create a perspective projection matrix

const projectionMatrix = new Matrix4().perspective({fov, aspect, near, far});

Create an orthograhic projection matrix


Invert a matrix

const inverse = matrix.invert();

Transform a vector as a point (including translations)

const transform = new Matrix4();
const vector2 = transform.transformPoint([0, 0]);
const vector3 = transform.transformPoint([0, 1, 2]);
const vector4 = transform.transformPoint([0, 1, 2, 1]);

Transform a vector as a direction (NOT including translations)

const transform = new Matrix4();
const vector2 = transform.transformDirection([0, 0]);
const vector3 = transform.transformDirection([0, 1, 2]);
const vector4 = transform.transformDirection([0, 1, 2, 1]);

Inheritance​

class Matrix4 extends Matrix extends MathArray extends Array`

Many basic methods are inherited:

  • matrix4.clone()
  • matrix4.copy(array)
  • matrix4.set(...args)
  • matrix4.fromArray(array, offset = 0)
  • matrix4.toString()
  • matrix4.toArray(array = [], offset = 0)
  • matrix4.equals(array)
  • matrix4.exactEquals(array)
  • matrix4.validate(array = this)
  • matrix4.check(array = this)
  • matrix4.normalize()

Since Matrix4 is a subclass of the built in JavaScript Array it can be used directly as a parameter to any function expecting an Array.

Methods​

constructor()​

Creates an empty Matrix4

new Matrix4()

identity(): this​

Sets the matrix to the multiplicative identity matrix.

matrix4.identity()

set(...number): this​

Sets the elements of the matrix.

matrix4.set(m00, m01, m02, m03, m10, m11, m12, m13, m20, m21, m22, m23, m30, m31, m32, m33)

fromQuaternion(quaternion: Quaternion): this​

Sets the matrix to a transformation corresponding to the rotations represented by the given quaternion.

matrix4.fromQuaternion(quaternion)

  • quaternion (Quaternion) - the quaternion to create matrix from
frustum(options: {left: number, right: number, bottom: number, top: number, near: number, far: number}): this​

Generates a frustum matrix with the given bounds. The frustum far plane can be infinite.

matrix4.frustum({left, right, bottom, top, near, far})

  • left (number) - Left bound of the frustum
  • right (number) - Right bound of the frustum
  • bottom (number) - Bottom bound of the frustum
  • top (number) - Top bound of the frustum
  • near (number) - Near bound of the frustum
  • far (number|Infinity) - Far bound of the frustum
lookAt(options?: {eye: number, center: number, up: number}): this​

Generates a look-at matrix with the given eye position, focal point, and up axis

matrix4.lookAt({eye, center, up})

  • eye (Vector3|number[3]) - Position of the viewer
  • center=[0, 0, 0] (Vector3|number[3]) vec3 Point the viewer is looking at
  • up=[0, 1, 0] (Vector3|number[3]) vec3 vec3 pointing up
ortho(options: {left: number, right: number, bottom: number, top: number, near?: number, far: number}): this​

Generates a orthogonal projection matrix with the given bounds

matrix4.ortho({left, right, bottom, top, near = 0.1, far = 500})

  • left (number) - Left bound of the frustum
  • right (number) - Right bound of the frustum
  • bottom (number) - Bottom bound of the frustum
  • top (number) - Top bound of the frustum
  • near (number) - Near bound of the frustum
  • far (number) - Far bound of the frustum
orthographic​

Generates an orthogonal projection matrix with the same parameters as a perspective matrix (plus focalDistance).

  • Matrix4.orthographic({fovy, aspect, focalDistance, near, far})

  • fovy (number) - Vertical field of view in radians

  • aspect (number) - Aspect ratio. typically viewport width/height

  • focalDistance (number) - selects which plane in the perspective view frustum should be used to calculate the size of the orthographic view box.

  • near=0.1 (number) - Near bound of the frustum

  • far=500 (number) - Far bound of the frustum

In applications it is not unusual to want to offer both perspective and orthographic views and this method is supplied to make this as simple as possible.

perspective​

Generates a perspective projection matrix with the given bounds. The frustum far plane can be infinite.

matrix4.perspective({ fovy = 45 * Math.PI - / 180, aspect = 1, near = 0.1, far = 500 })

  • fovy=45 (number) - Vertical field of view in radians (default is 45 degrees specified in radians)
  • aspect=1 (number) - Aspect ratio. typically viewport width/height
  • near=0.1 (number) - Near bound of the frustum
  • far=500 (number|Infinity) - Far bound of the frustum
determinant(): number​

Returns the determinant of the matrix (does not modify the matrix).

const determinant = matrix4.determinant()

Returns (number) - the determinant

  • If the determinant is zero, the matrix is not invertible.
  • Determinant calculation is somewhat expensive.
transpose(): this​

Sets this matrix to its transpose matrix.

matrix4.transpose()

  • The transpose matrix mirrors the original matrix elements in the diagonal.
invert(): this​

Sets this matrix to its inverse matrix.

matrix4.invert()

  • The inverse matrix mirrors the original matrix elements in the diagonal.
multiplyLeft(matrix: number[16]): this​

Multiplies in another matrix from the left

matrix4.multiplyLeft(matrix4)

  • When using Matrix4 to transform vectors, the vectors are multiplied in from the right. This means that the multiplying in a matrix from the left will cause it to be applied last during transformation (unless additional matrices are multiplied in from the left of course).
multiplyRight(matrix: number[16]): this​

matrix4.multiplyRight(matrix4)

  • When using Matrix4 to transform vectors, the vectors are multiplied in from the right. This means that the multiplying in a matrix from the left will cause it to be applied last during transformation (unless additional matrices are multiplied in from the left of course).
rotateX(radians: number): this​

Adds a rotation by the given angle around the X axis. Equivalent to right multiplying the new transform into the matrix but more performant.

matrix4.rotateX(radians)

rotateY(radians: number): this​

Adds a rotation by the given angle around the Y axis.

rotateY(radians)

  • Equivalent to right multiplying the new transform into the matrix but more performant.
rotateZ(radians: number): this​

Adds a rotation by the given angle around the Z axis.

matrix4.rotateZ(radians)

  • Equivalent to right multiplying the new transform into the matrix but more performant.
rotateXYZ(angles: [rx: number, ry: number, rz: number]): this​

Adds successive rotations by the given angles around the X, Y and Z axis.

rotateXYZ([rx, ry, rz])

  • Equivalent to right multiplying the new transform into the matrix but more performant.
rotateAxis(radians: number, axis: number[3]): this​

Adds successive rotations by the given angles around the X, Y and Z axis.

rotateAxis(radians, axis)

Equivalent to right multiplying the new transform into the matrix but more performant.

scale(factor: number | number[3]): this​

Adds a scaling transform, each axis can be scaled independently.

matrix4.scale(factor)

  • factor (number) - scale factor to be applied to each axis.

matrix4.scale([x, y, z])

  • x (number) - scale factor to be multiplied into x component
  • y (number) - scale factor to be multiplied into y component
  • z (number) - scale factor to be multiplied into z component

Equivalent to right multiplying the new transform into the matrix but more performant.

  • During vector transformation all coordinates will be multiplied with the given factors.
  • Scale with -1 will flip the coordinate system in that axis.
  • Scale with 0 will drop that component.
translate(scale: number[3]): this​

Adds a translation to the matrix.

matrix4.translate([x, y, z])

  • x (number) - translation to be added to the x component
  • y (number) - translation to be added to the y component
  • z (number) - translation to be added to the z component

Equivalent to right multiplying the new transform into the matrix but more performant.

During vector transformation the given translation values are added to each component of the vector being transformed.

Decomposition​

getRotation(result?: number[16]) : number[16]​

Returns a 4x4 rotation matrix.

getRotationMatrix3(result?: number[9]) : number[9]​

Returns a 3x3 rotation matrix.

getTranslation(result?: number[3]) : number[3]​

Returns the 3-element translation vector component of the affine transform described by the matrix.

For performance, an existing vector can be provided, if not a new vector will be returned.

getScale(result?: number[3]) : number[3]​

Returns the 3-element scale vector component of the affine transform described by the matrix.

For performance, an existing vector can be provided, if not a new vector will be returned.

Point Transformations​

transformAsPoint(vector : number[4]) : number[4]​

Transforms any 2, 3 or 4 element vector as a "point" by multiplying it (from the right) with this matrix. Point here means that the returned vector will include any translations in this matrix.

const vector = matrix4.transformPoint(vector, out=)

  • vector (Array|Vector2|Vector3|Vector4)

  • out - unless supplied, will be a Vector2, Vector3 or Vector4, matching the length of input vector. Returns out, or a newly minted Vector2, Vector3 or Vector4

  • If vector is specified in homogeneous coordinates, w coordinate must NOT be 0.

  • If vector is specified in homogeneous coordinates the returned vector will be w adjusted, (i.e. w coordinate will be 1, even if the supplied vector was not normalized).

transformAsVector(vector : number[4]) : number[4]​

Transforms any 2, 3 or 4 element vector interpreted as a direction (i.e. all vectors are based in the origin so the transformation not pick up any translations from the matrix).

const vector = matrix4.transformDirection(vector, out)

  • If vector is specified in homogeneous coordinates, w coordinate must be 0.

Remarks​

  • All transforms are effectively "right multiplied" onto the matrix (meaning that during transform they will be applied in opposite order).
  • Matrix4 is stored internally in column major format (per WebGL conventions). This only matters when you read out the matrix to use it with other software.