Skip to content

Latest commit

 

History

History

vectors

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

@thi.ng/vectors

npm version npm downloads Twitter Follow

This project is part of the @thi.ng/umbrella monorepo.

About

This package provides vector and matrix operations as plain functions and class wrappers with fluid interface. All functions support any array / typed array storage, incl. mapped views of larger buffers (e.g. for WebGL / WASM interop, pixel buffers). Additionally, vectors support flexible data layouts, incl. AOS / SOA, striped, interleaved, aligned etc.

This package will soon be replaced by the currently still unreleased @thi.ng/vectors3 and @thi.ng/matrices packages

Vectors

In addition to arbitrary sized vectors, the library provides these optimized fixed-sized versions:

Vector classes

All of the vector operations listed below are also available via class wrappers of strided buffer views. These vector classes (Vec2/3/4) are array-like themselves and provide array index and .x, .y, .z, .w property accessors (including .length). The GVec class wrapper only provides .length read access and element access via getAt() and setAt(). All classes are iterable and provide toString() and toJSON() implementations.

buf = [0, 1, 0, 2, 0, 3];

// create Vec3 view from index 1, w/ stride 2
a = new v.Vec3(buf, 1, 2);

a[0] *= 10;
a[1] *= 100;
a[2] *= 1000;

a.x // 10
a.y // 200
a.z // 3000
a.length // 3

// iterable
[...a]
// [ 10, 200, 3000 ]

buf
// [ 0, 10, 0, 200, 0, 3000 ]

Supported operations

Note: Most functions are provided in different (optimized) versions, depending on vector size. E.g. add performs vector addition for arbitrary sizes, add2 for 2D vectors, add3 for 3D, add4 for 4D... Class wrapper methods use the non-suffixed naming.

All vector operations (regardless of size) operate on any array-like buffer and accept optional start indices and component strides (number of elements (+1) between individual vector components). This allows for zero-copy vector operations on sections of larger buffers. The default start index is 0, default stride 1. See examples below and tests.

Naming conventions (suffixes):

  • N = uniform: 2nd or 3rd arg is a scalar which will be used for all vector components, e.g. setN3([], 0) => [0, 0, 0]
  • S = scalar: args are individual scalar values, e.g. setS3([], 10, 20, 30) => [10, 20, 30]
  • o = output: operation writes to separate output vector (1st arg), e.g. add2o([], [1,2], [10, 20])
Operation Generic 2D 3D 4D
Get vector (dense copy) get get2 get3 get4
Set vector components (vector) set set2 set3 set4
Set vector components (uniform) setN setN2 setN3 setN4
Set vector components (scalars) setS2 setS3 setS4
Swizzle vector components swizzle2 swizzle3 swizzle4
Swap vectors swap2 swap3 swap4
Equality (w/ epsilon) eqDelta eqDelta2 eqDelta3 eqDelta4
Vector addition add add2 add3 add4
add2o add3o add4o
Vector subtraction sub sub2 sub3 sub4
sub2o sub3o sub4o
Vector multiplication mul mul2 mul3 mul4
mul2o mul3o mul4o
Vector division div div2 div3 div4
div2o div3o div4o
Uniform scalar addition addN addN2 addN3 addN4
addN2o addN3o addN4o
Uniform scalar subtraction subN subN2 subN3 subN4
subN2o subN3o subN4o
Uniform scalar multiply mulN mulN2 mulN3 mulN4
mulN2o mulN3o mulN4o
Uniform scalar multiply divN divN2 divN3 divN4
divN2o divN3o divN4o
Vector negation neg neg2 neg3 neg4
Multiply-add vectors madd madd2 madd3 madd4
Multiply-add scalar maddN maddN2 maddN3 maddN4
Linear interpolation (vector) mix mix2 mix3 mix4
mix2o mix3o mix4o
Linear interpolation (uniform) mixN mixN2 mixN3 mixN4
mixN2o mixN3o mixN4o
Bilinear interpolation* mixBilinear2 mixBilinear3 mixBilinear4
Dot product dot dot2 dot3 dot4
Cross product cross2 cross3
Magnitude mag mag2 mag3 mag4
Magnitude (squared) magSq magSq2 magSq3 magSq4
Normalize (w/ opt length) normalize normalize2 normalize3 normalize4
Limit to length limit2 limit3 limit4
Distance dist dist2 dist3 dist4
Distance (squared) distSq distSq2 distSq3 distSq4
Manhattan distance distManhattan2 distManhattan3 distManhattan4
Chebyshev distance distChebyshev2 distChebyshev3 distChebyshev4
Reflection reflect2 reflect3 reflect4
Refraction refract2 refract3 refract4
Perpendicular perpendicularLeft2
perpendicularRight2
RotationX rotateX3
RotationY rotateY3
RotationZ rotate2 rotateZ3
Rotation around point rotateAroundPoint2
Rotation around axis rotateAroundAxis3
Heading XY heading2 headingXY3
Heading XZ headingXZ3
Heading YZ headingYZ3
Angle between vectors angleBetween2 angleBetween3
Bisector angle bisect2
Cartesian -> Polar toPolar2 toSpherical3
Polar -> Cartesian toCartesian2 toCartesian3
Cartesian -> Cylindrical toCylindrical3
Cylindrical -> Cartesian fromCylindrical3
Minor axis minorAxis2 minorAxis3 minorAxis4
Major axis majorAxis2 majorAxis3 majorAxis4
Minimum min min2 min3 min4
Maximum max max2 max3 max4
Range clamping clamp clamp2 clamp3 clamp4
Step (like GLSL) step step2 step3 step4
SmoothStep (like GLSL) smoothStep smoothStep2 smoothStep3 smoothStep4
Absolute value abs abs2 abs3 abs4
Sign (w/ opt epsilon) sign sign2 sign3 sign4
Round down floor floor2 floor3 floor4
Round up ceil ceil2 ceil3 ceil4
Square root sqrt sqrt2 sqrt3 sqrt4
Power (vector) pow pow2 pow3 pow4
Power (uniform) powN powN2 powN3 powN4

(*) Static method in class wrapper

Matrices

All matrix types are in WebGL layout (column major) and densely packed (stride always 1). As with vectors, class wrapper methods use the non-suffixed naming.

Operation 2x3 3x3 4x4
Set identity* identity23 identity33 identity44
Get matrix components (dense copy) get23 get33 get44
Set matrix components (matrix) set23 set33 set44
Set matrix components (scalars) setS23 setS33 setS44
Create rotation matrix* rotationX33 rotationX44
rotationY33 rotationY44
rotation23 rotationZ33 rotationZ44
rotationAroundPoint23
Create scale matrix* (vector) scaleV23 scaleV33 scaleV44
Create scale matrix* (uniform) scaleN23 scaleN33 scaleN44
Create scale matrix* (scalars) scaleS23 scaleS33 scaleS44
scaleWithCenter23 scaleWithCenter44
Create translation matrix* (vector) translationV23 translationV44
Create translation matrix* (scalars) translationS23 translationS44
Create skew matrix* skewX23 / shearX23
skewY23 / shearY23
Create projection matrix* projection
ortho
frustum
Create camera matrix* lookAt
Matrix multiply mul23 mul33 mul44
Matrix concatenation* (multiple) concat23 concat33 concat44
Matrix vector multiply mulV23 mulV33 mulV44 (Vec4)
mulV344 (Vec3)
Determinant det23 det33 det44
Matrix inversion invert23 invert33 invert44
Matrix transpose transpose33 transpose44

(*) Static method in class wrapper

Installation

yarn add @thi.ng/vectors

Dependencies

Usage examples

Basics

import * as v from "@thi.ng/vectors";

// raw vector addition
v.add4([1, 2, 3, 4], [10, 20, 30, 40]);
// [ 11, 22, 33, 44 ]

// with custom layout
// here 3x 3D vectors in SOA layout:
//       [x, x, x, y, y, y, z, z, z]
points = [1, 4, 7, 2, 5, 8, 3, 6, 9];

// specify start indices and stride lengths
// update 1st vector
v.add3(points, [100, 200, 300], 0, 0, 3, 1);
// [ 101, 4, 7, 202, 5, 8, 303, 6, 9 ]

// update 2nd vector
v.add3(points, [100, 200, 300], 1, 0, 3, 1);
// [ 101, 104, 7, 202, 205, 8, 303, 306, 9 ]

// update 3rd vector
v.add3(points, [100, 200, 300], 2, 0, 3, 1);
// [ 101, 104, 107, 202, 205, 208, 303, 306, 309 ]

// add 1st and 3rd vector and extract result
v.get3(v.add3(points, points, 0, 2, 3, 3), 0, 3);
// [ 208, 410, 612 ]

// re-arrange vector components into new vector
// the last 4 args define component order:

// YXWZ
v.swizzle4([], [10, 20, 30, 40], 1, 0, 3, 2);
// [ 20, 10, 40, 30 ]

// XXZZ
v.swizzle4([], [10, 20, 30, 40], 0, 0, 2, 2);
// [ 10, 10, 30, 30 ]

// arbitrary length vectors
norm = v.normalize([1, 2, 3, 4, 5, 6, 7, 8, 6, 4]);
// [ 0.0625, 0.125, 0.1875, 0.25, 0.3125, 0.375, 0.4375, 0.5, 0.375, 0.25 ]

v.mag(norm);
// 1

Vector classes & interleaved vectors in large buffer

// element stride 3 + 2 + 4 = 9
buf = [
    // pos     uv   color (rgba)
    0,0,0,     0,0, 1,0,0,1,
    100,0,0,   1,0, 1,1,0,1,
    100,100,0, 1,1, 1,0,1,1,
    0,100,0,   0,1, 0,1,1,1,
];

// create memory mapped vector instances (using classes)
pos = v.Vec3.mapBuffer(buf, 4, 0, 1, 9); // offset = 0
uv  = v.Vec2.mapBuffer(buf, 4, 3, 1, 9); // offset = 3
col = v.Vec4.mapBuffer(buf, 4, 5, 1, 9); // offset = 5

console.log(`pos: ${pos[1]}, uv: ${uv[1]}, color: ${col[1]}`);
// pos: [100, 0, 0], uv: [1, 0], color: [1, 1, 0, 1]

// compute centroid
centroid = pos.reduce((c, p) => c.add(p), v.vec3()).divN(pos.length);
// Vec3 { buf: [ 50, 50, 0 ], i: 0, s: 1 }

// build matrix to transform geometry
tx = v.Mat44.concat(
    v.Mat44.scale(0.01),
    v.Mat44.translation(centroid.copy().neg()),
    v.Mat44.rotationZ(v.rad(90)),
);

// apply transform to all positions
pos.forEach((p) => tx.mulV3(p));

Image RGB grayscale conversion

canvas = document.getElementById("main");
img = canvas.getContext("2d").getImageData(0,0, canvas.width, canvas.height);

v.transformVectors1(
    // multiply each RGB vector w/ weights
    // then use result for all 3 color channels
    (a, b, ia, ib, sa, sb) =>
        v.setN3(a, v.dot3(a, b, ia, ib, sa, sb), ia, sa),
    // pixel buffer
    img,
    // RGB weight coefficients
    [0.29, 0.6, 0.11],
    // num pixels (RGBA vectors)
    canvas.width * canvas.height,
    // start indices
    0, 0,
    // component strides
    1, 1,
    // pixel stride
    4
);

Authors

  • Karsten Schmidt

License

© 2016 - 2018 Karsten Schmidt // Apache Software License 2.0