This is a simple linear algebra library for golang, intended solely for 3D graphics. It defines structs and functions for points, vectors, and small matrices.
This package is currently under construction, and it’s by no means complete. While the name of the git repository is govmath, the name of the actual go package is vmath.
vmath defines six types, each of which has a tiny two-character name:
| V3 | Vector in 3-space |
| V4 | Homogeneous vector |
| P3 | Point in 3-space; similar to a V4 with w=1 |
| M3 | Represents rotation and scale |
| T3 | 4×3 matrix for affine transformations; adds translation to M3 |
| M4 | Homogeneous transform (suitable for perpective projection) |
Matrices are stored in row-major order but they can be easily transposed (e.g., for consumption by GLSL).
When vectors are premultiplied with matrices, as in V.Mul(M), they are treated as row vectors. When postmultipled, as in M.Mul(V), they are treated as column vectors.
M4 provides construction methods for right-handed and left-handed projections.
Vectors and points are distinct concepts; a V3 added to a P3 yields another P3:
v := vmath.V3{0, 0, 1}
p := vmath.P3{1, 0, 0}
p = p.Add(v)
Vectors can be added to points, but points cannot be added vectors. Adding a V3 to another V3 yields a V3:
v1 := vmath.V3{0, 0, 1}
v2 := vmath.V3{1, 0, 0}
var v vmath.V3 = v1.Add(v2)
Unlike go’s own math package, vmath is designed for 32-bit floats rather than doubles. This is mostly for OpenGL compatibility:
v := vmath.V3{1, 1, 1}
glVertexAttrib3fv(mySlot, &v.X)
All non-method functions are prefixed with the type they return (e.g., M3Identity). Functions also sometimes have a type suffix:
p1 := P3FromV3(a) p2 := P3FromV4(b) m := M4Translate(1, 0, 0) ᴨ := float32(math.Atan(1) * 4) t := T3RotateX(ᴨ/2)
Matrix methods and functions operate on pointers, while vectors and points operate on values:
//todo
You can always modify a vector or matrix by accessing its struct members, but all methods and functions treat them as immutable:
var m M4 = M4Translate(0, 0, 1) m.Transpose() // this does nothing m2 := m.Transpose() // better!
The concept of equality is dodgy with floating-point vectors, but you can always call Equivalent with an epsilon of your choice:
var ε float32 = 1e-4
i := vmath.V3{1, 0, 0}
j := vmath.V3{0, 1, 0}
k := vmath.V3{0, 0, 1}
if !i.Cross(j).Equivalent(k, ε) {
panic
}
Installation
git clone git@github.com:prideout/govmath cd govmath export GOPATH=$GOPATH:$PWD go install ./src/vmath go test ./src/vmath
TODO:
- Create a clean filter ala http://progit.org/book/ch7-2.html
- gofmt -tabs=false -tabwidth=4 -w=true *.go