cgmath/src/mat.rs

939 lines
32 KiB
Rust
Raw Normal View History

// Copyright 2013 The Lmath Developers. For a full listing of the authors,
// refer to the AUTHORS file at the top-level directory of this distribution.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
pub use dim::Dimensional;
2012-12-13 13:01:42 +00:00
2013-06-16 08:03:14 +00:00
use quat::{Quat, ToQuat};
use vec::{Vec2, Vec3, Vec4};
2013-06-10 23:02:25 +00:00
2013-06-16 08:03:14 +00:00
mod num_macros;
mod dim_macros;
mod mat_macros;
#[deriving(Eq)]
pub struct Mat2<T> {
x: Vec2<T>,
y: Vec2<T>,
}
2013-06-12 00:02:39 +00:00
2013-06-16 08:03:14 +00:00
// GLSL-style type aliases
2013-06-10 23:02:25 +00:00
pub type mat2 = Mat2<f32>;
pub type dmat2 = Mat2<f64>;
// Rust-style type aliases
pub type Mat2f = Mat2<float>;
pub type Mat2f32 = Mat2<f32>;
pub type Mat2f64 = Mat2<f64>;
2013-06-16 08:03:14 +00:00
impl_dimensional!(Mat2, Vec2<T>, 2)
impl_dimensional_fns!(Mat2, Vec2<T>, 2)
impl_approx!(Mat2)
impl_mat!(Mat2, Vec2)
impl_mat_copyable!(Mat2, Vec2)
impl_mat_numeric!(Mat2, Vec2)
impl_mat_approx_numeric!(Mat2)
impl_mat_neg!(Mat2)
pub trait ToMat2<T> {
pub fn to_mat2(&self) -> Mat2<T>;
}
impl<T> Mat2<T> {
#[inline]
pub fn new(c0r0: T, c0r1: T,
c1r0: T, c1r1: T) -> Mat2<T> {
Mat2::from_cols(Vec2::new(c0r0, c0r1),
Vec2::new(c1r0, c1r1))
}
#[inline]
pub fn from_cols(c0: Vec2<T>,
c1: Vec2<T>) -> Mat2<T> {
Mat2 { x: c0, y: c1 }
}
}
impl<T:Copy + Num> ToMat3<T> for Mat2<T> {
#[inline]
pub fn to_mat3(&self) -> Mat3<T> {
Mat3::new(*self.elem(0, 0), *self.elem(0, 1), zero!(T),
*self.elem(1, 0), *self.elem(1, 1), zero!(T),
zero!(T), zero!(T), one!(T))
}
}
impl<T:Copy + Num> ToMat4<T> for Mat2<T> {
#[inline]
pub fn to_mat4(&self) -> Mat4<T> {
Mat4::new(*self.elem(0, 0), *self.elem(0, 1), zero!(T), zero!(T),
*self.elem(1, 0), *self.elem(1, 1), zero!(T), zero!(T),
zero!(T), zero!(T), one!(T), zero!(T),
zero!(T), zero!(T), zero!(T), one!(T))
}
}
impl<T:Copy + Real> Mat2<T> {
#[inline]
pub fn from_angle(radians: T) -> Mat2<T> {
let cos_theta = radians.cos();
let sin_theta = radians.sin();
Mat2::new(cos_theta, -sin_theta,
sin_theta, cos_theta)
}
}
#[cfg(test)]
mod mat2_tests{
use mat::*;
use vec::*;
#[test]
fn test_mat2() {
let a = Mat2 { x: Vec2 { x: 1.0, y: 3.0 },
y: Vec2 { x: 2.0, y: 4.0 } };
let b = Mat2 { x: Vec2 { x: 2.0, y: 4.0 },
y: Vec2 { x: 3.0, y: 5.0 } };
let v1 = Vec2::new::<float>(1.0, 2.0);
let f1 = 0.5;
assert_eq!(a, Mat2::new::<float>(1.0, 3.0,
2.0, 4.0));
assert_eq!(a, Mat2::from_cols::<float>(Vec2::new::<float>(1.0, 3.0),
Vec2::new::<float>(2.0, 4.0)));
assert_eq!(Mat2::from_value::<float>(4.0),
Mat2::new::<float>(4.0, 0.0,
0.0, 4.0));
assert_eq!(*a.col(0), Vec2::new::<float>(1.0, 3.0));
assert_eq!(*a.col(1), Vec2::new::<float>(2.0, 4.0));
assert_eq!(a.row(0), Vec2::new::<float>(1.0, 2.0));
assert_eq!(a.row(1), Vec2::new::<float>(3.0, 4.0));
assert_eq!(*a.col(0), Vec2::new::<float>(1.0, 3.0));
assert_eq!(*a.col(1), Vec2::new::<float>(2.0, 4.0));
assert_eq!(Mat2::identity::<float>(),
Mat2::new::<float>(1.0, 0.0,
0.0, 1.0));
assert_eq!(Mat2::zero::<float>(),
Mat2::new::<float>(0.0, 0.0,
0.0, 0.0));
assert_eq!(a.determinant(), -2.0);
assert_eq!(a.trace(), 5.0);
assert_eq!(a.neg(),
Mat2::new::<float>(-1.0, -3.0,
-2.0, -4.0));
assert_eq!(-a, a.neg());
assert_eq!(a.mul_t(f1),
Mat2::new::<float>(0.5, 1.5,
1.0, 2.0));
assert_eq!(a.mul_v(&v1), Vec2::new::<float>(5.0, 11.0));
assert_eq!(a.add_m(&b),
Mat2::new::<float>(3.0, 7.0,
5.0, 9.0));
assert_eq!(a.sub_m(&b),
Mat2::new::<float>(-1.0, -1.0,
-1.0, -1.0));
assert_eq!(a.mul_m(&b),
Mat2::new::<float>(10.0, 22.0,
13.0, 29.0));
assert_eq!(a.dot(&b), 40.0);
assert_eq!(a.transpose(),
Mat2::new::<float>(1.0, 2.0,
3.0, 4.0));
assert_eq!(a.inverse().unwrap(),
Mat2::new::<float>(-2.0, 1.5,
1.0, -0.5));
assert!(Mat2::new::<float>(0.0, 2.0,
0.0, 5.0).inverse().is_none());
let ident = Mat2::identity::<float>();
assert!(ident.is_identity());
assert!(ident.is_symmetric());
assert!(ident.is_diagonal());
assert!(!ident.is_rotated());
assert!(ident.is_invertible());
assert!(!a.is_identity());
assert!(!a.is_symmetric());
assert!(!a.is_diagonal());
assert!(a.is_rotated());
assert!(a.is_invertible());
let c = Mat2::new::<float>(2.0, 1.0,
1.0, 2.0);
assert!(!c.is_identity());
assert!(c.is_symmetric());
assert!(!c.is_diagonal());
assert!(c.is_rotated());
assert!(c.is_invertible());
assert!(Mat2::from_value::<float>(6.0).is_diagonal());
assert_eq!(a.to_mat3(),
Mat3::new::<float>(1.0, 3.0, 0.0,
2.0, 4.0, 0.0,
0.0, 0.0, 1.0));
assert_eq!(a.to_mat4(),
Mat4::new::<float>(1.0, 3.0, 0.0, 0.0,
2.0, 4.0, 0.0, 0.0,
0.0, 0.0, 1.0, 0.0,
0.0, 0.0, 0.0, 1.0));
}
fn test_mat2_mut() {
let a = Mat2 { x: Vec2 { x: 1.0, y: 3.0 },
y: Vec2 { x: 2.0, y: 4.0 } };
let b = Mat2 { x: Vec2 { x: 2.0, y: 4.0 },
y: Vec2 { x: 3.0, y: 5.0 } };
let f1 = 0.5;
let mut mut_a = a;
mut_a.swap_cols(0, 1);
assert_eq!(mut_a.col(0), a.col(1));
assert_eq!(mut_a.col(1), a.col(0));
mut_a = a;
mut_a.swap_rows(0, 1);
assert_eq!(mut_a.row(0), a.row(1));
assert_eq!(mut_a.row(1), a.row(0));
mut_a = a;
mut_a.to_identity();
assert!(mut_a.is_identity());
mut_a = a;
mut_a.to_zero();
assert_eq!(mut_a, Mat2::zero::<float>());
mut_a = a;
mut_a.mul_self_t(f1);
assert_eq!(mut_a, a.mul_t(f1));
mut_a = a;
mut_a.add_self_m(&b);
assert_eq!(mut_a, a.add_m(&b));
mut_a = a;
mut_a.sub_self_m(&b);
assert_eq!(mut_a, a.sub_m(&b));
mut_a = a;
mut_a.invert_self();
assert_eq!(mut_a, a.inverse().unwrap());
mut_a = a;
mut_a.transpose_self();
assert_eq!(mut_a, a.transpose());
// mut_a = a;
}
#[test]
fn test_mat2_approx_eq() {
assert!(!Mat2::new::<float>(0.000001, 0.000001,
0.000001, 0.000001).approx_eq(&Mat2::zero::<float>()));
assert!(Mat2::new::<float>(0.0000001, 0.0000001,
0.0000001, 0.0000001).approx_eq(&Mat2::zero::<float>()));
}
}
#[deriving(Eq)]
pub struct Mat3<T> {
x: Vec3<T>,
y: Vec3<T>,
z: Vec3<T>,
}
// GLSL-style type aliases
pub type mat3 = Mat3<f32>;
pub type dmat3 = Mat3<f64>;
// Rust-style type aliases
2013-06-10 23:02:25 +00:00
pub type Mat3f = Mat3<float>;
pub type Mat3f32 = Mat3<f32>;
pub type Mat3f64 = Mat3<f64>;
2013-06-16 08:03:14 +00:00
impl_dimensional!(Mat3, Vec3<T>, 3)
impl_dimensional_fns!(Mat3, Vec3<T>, 3)
impl_approx!(Mat3)
impl_mat!(Mat3, Vec3)
impl_mat_copyable!(Mat3, Vec3)
impl_mat_numeric!(Mat3, Vec3)
impl_mat_approx_numeric!(Mat3)
impl_mat_neg!(Mat3)
pub trait ToMat3<T> {
pub fn to_mat3(&self) -> Mat3<T>;
}
impl<T> Mat3<T> {
#[inline]
pub fn new(c0r0:T, c0r1:T, c0r2:T,
c1r0:T, c1r1:T, c1r2:T,
c2r0:T, c2r1:T, c2r2:T) -> Mat3<T> {
Mat3::from_cols(Vec3::new(c0r0, c0r1, c0r2),
Vec3::new(c1r0, c1r1, c1r2),
Vec3::new(c2r0, c2r1, c2r2))
}
#[inline]
pub fn from_cols(c0: Vec3<T>,
c1: Vec3<T>,
c2: Vec3<T>) -> Mat3<T> {
Mat3 { x: c0, y: c1, z: c2 }
}
}
impl<T:Copy + Num> ToMat4<T> for Mat3<T> {
#[inline]
pub fn to_mat4(&self) -> Mat4<T> {
Mat4::new(*self.elem(0, 0), *self.elem(0, 1), *self.elem(0, 2), zero!(T),
*self.elem(1, 0), *self.elem(1, 1), *self.elem(1, 2), zero!(T),
*self.elem(2, 0), *self.elem(2, 1), *self.elem(2, 2), zero!(T),
zero!(T), zero!(T), zero!(T), one!(T))
}
}
impl<T:Copy + Real> Mat3<T> {
/// Construct a matrix from an angular rotation around the `x` axis
pub fn from_angle_x(radians: T) -> Mat3<T> {
// http://en.wikipedia.org/wiki/Rotation_matrix#Basic_rotations
let cos_theta = radians.cos();
let sin_theta = radians.sin();
Mat3::new(one!(T), zero!(T), zero!(T),
zero!(T), cos_theta, sin_theta,
zero!(T), -sin_theta, cos_theta)
}
/// Construct a matrix from an angular rotation around the `y` axis
pub fn from_angle_y(radians: T) -> Mat3<T> {
// http://en.wikipedia.org/wiki/Rotation_matrix#Basic_rotations
let cos_theta = radians.cos();
let sin_theta = radians.sin();
Mat3::new(cos_theta, zero!(T), -sin_theta,
zero!(T), one!(T), zero!(T),
sin_theta, zero!(T), cos_theta)
}
/// Construct a matrix from an angular rotation around the `z` axis
pub fn from_angle_z(radians: T) -> Mat3<T> {
// http://en.wikipedia.org/wiki/Rotation_matrix#Basic_rotations
let cos_theta = radians.cos();
let sin_theta = radians.sin();
Mat3::new(cos_theta, sin_theta, zero!(T),
-sin_theta, cos_theta, zero!(T),
zero!(T), zero!(T), one!(T))
}
/// Construct a matrix from Euler angles
///
/// # Arguments
///
/// - `theta_x`: the angular rotation around the `x` axis (pitch)
/// - `theta_y`: the angular rotation around the `y` axis (yaw)
/// - `theta_z`: the angular rotation around the `z` axis (roll)
pub fn from_angle_xyz(radians_x: T, radians_y: T, radians_z: T) -> Mat3<T> {
// http://en.wikipedia.org/wiki/Rotation_matrix#General_rotations
let cx = radians_x.cos();
let sx = radians_x.sin();
let cy = radians_y.cos();
let sy = radians_y.sin();
let cz = radians_z.cos();
let sz = radians_z.sin();
Mat3::new(cy*cz, cy*sz, -sy,
-cx*sz + sx*sy*cz, cx*cz + sx*sy*sz, sx*cy,
sx*sz + cx*sy*cz, -sx*cz + cx*sy*sz, cx*cy)
}
/// Construct a matrix from an axis and an angular rotation
pub fn from_angle_axis(radians: T, axis: &Vec3<T>) -> Mat3<T> {
let c = radians.cos();
let s = radians.sin();
let _1_c = one!(T) - c;
let x = axis.x;
let y = axis.y;
let z = axis.z;
Mat3::new(_1_c*x*x + c, _1_c*x*y + s*z, _1_c*x*z - s*y,
_1_c*x*y - s*z, _1_c*y*y + c, _1_c*y*z + s*x,
_1_c*x*z + s*y, _1_c*y*z - s*x, _1_c*z*z + c)
}
#[inline]
pub fn from_axes(x: Vec3<T>, y: Vec3<T>, z: Vec3<T>) -> Mat3<T> {
Mat3::from_cols(x, y, z)
}
pub fn look_at(dir: &Vec3<T>, up: &Vec3<T>) -> Mat3<T> {
let dir_ = dir.normalize();
let side = dir_.cross(&up.normalize());
let up_ = side.cross(&dir_).normalize();
Mat3::from_axes(up_, side, dir_)
}
}
impl<T:Copy + Real> ToQuat<T> for Mat3<T> {
/// Convert the matrix to a quaternion
pub fn to_quat(&self) -> Quat<T> {
// Implemented using a mix of ideas from jMonkeyEngine and Ken Shoemake's
// paper on Quaternions: http://www.cs.ucr.edu/~vbz/resources/Quatut.pdf
let mut s;
let w; let x; let y; let z;
let trace = self.trace();
// FIXME: We don't have any numeric conversions in std yet :P
let half = one!(T) / two!(T);
cond! (
(trace >= zero!(T)) {
s = (one!(T) + trace).sqrt();
w = half * s;
s = half / s;
x = (*self.elem(1, 2) - *self.elem(2, 1)) * s;
y = (*self.elem(2, 0) - *self.elem(0, 2)) * s;
z = (*self.elem(0, 1) - *self.elem(1, 0)) * s;
}
((*self.elem(0, 0) > *self.elem(1, 1))
&& (*self.elem(0, 0) > *self.elem(2, 2))) {
s = (half + (*self.elem(0, 0) - *self.elem(1, 1) - *self.elem(2, 2))).sqrt();
w = half * s;
s = half / s;
x = (*self.elem(0, 1) - *self.elem(1, 0)) * s;
y = (*self.elem(2, 0) - *self.elem(0, 2)) * s;
z = (*self.elem(1, 2) - *self.elem(2, 1)) * s;
}
(*self.elem(1, 1) > *self.elem(2, 2)) {
s = (half + (*self.elem(1, 1) - *self.elem(0, 0) - *self.elem(2, 2))).sqrt();
w = half * s;
s = half / s;
x = (*self.elem(0, 1) - *self.elem(1, 0)) * s;
y = (*self.elem(1, 2) - *self.elem(2, 1)) * s;
z = (*self.elem(2, 0) - *self.elem(0, 2)) * s;
}
_ {
s = (half + (*self.elem(2, 2) - *self.elem(0, 0) - *self.elem(1, 1))).sqrt();
w = half * s;
s = half / s;
x = (*self.elem(2, 0) - *self.elem(0, 2)) * s;
y = (*self.elem(1, 2) - *self.elem(2, 1)) * s;
z = (*self.elem(0, 1) - *self.elem(1, 0)) * s;
}
)
Quat::new(w, x, y, z)
}
}
#[cfg(test)]
mod mat3_tests{
use mat::*;
use vec::*;
#[test]
fn test_mat3() {
let a = Mat3 { x: Vec3 { x: 1.0, y: 4.0, z: 7.0 },
y: Vec3 { x: 2.0, y: 5.0, z: 8.0 },
z: Vec3 { x: 3.0, y: 6.0, z: 9.0 } };
let b = Mat3 { x: Vec3 { x: 2.0, y: 5.0, z: 8.0 },
y: Vec3 { x: 3.0, y: 6.0, z: 9.0 },
z: Vec3 { x: 4.0, y: 7.0, z: 10.0 } };
let v1 = Vec3::new::<float>(1.0, 2.0, 3.0);
let f1 = 0.5;
assert_eq!(a, Mat3::new::<float>(1.0, 4.0, 7.0,
2.0, 5.0, 8.0,
3.0, 6.0, 9.0));
assert_eq!(a, Mat3::from_cols::<float>(Vec3::new::<float>(1.0, 4.0, 7.0),
Vec3::new::<float>(2.0, 5.0, 8.0),
Vec3::new::<float>(3.0, 6.0, 9.0)));
assert_eq!(*a.col(0), Vec3::new::<float>(1.0, 4.0, 7.0));
assert_eq!(*a.col(1), Vec3::new::<float>(2.0, 5.0, 8.0));
assert_eq!(*a.col(2), Vec3::new::<float>(3.0, 6.0, 9.0));
assert_eq!(a.row(0), Vec3::new::<float>(1.0, 2.0, 3.0));
assert_eq!(a.row(1), Vec3::new::<float>(4.0, 5.0, 6.0));
assert_eq!(a.row(2), Vec3::new::<float>(7.0, 8.0, 9.0));
assert_eq!(*a.col(0), Vec3::new::<float>(1.0, 4.0, 7.0));
assert_eq!(*a.col(1), Vec3::new::<float>(2.0, 5.0, 8.0));
assert_eq!(*a.col(2), Vec3::new::<float>(3.0, 6.0, 9.0));
assert_eq!(Mat3::identity::<float>(),
Mat3::new::<float>(1.0, 0.0, 0.0,
0.0, 1.0, 0.0,
0.0, 0.0, 1.0));
assert_eq!(Mat3::zero::<float>(),
Mat3::new::<float>(0.0, 0.0, 0.0,
0.0, 0.0, 0.0,
0.0, 0.0, 0.0));
assert_eq!(a.determinant(), 0.0);
assert_eq!(a.trace(), 15.0);
assert_eq!(a.neg(),
Mat3::new::<float>(-1.0, -4.0, -7.0,
-2.0, -5.0, -8.0,
-3.0, -6.0, -9.0));
assert_eq!(-a, a.neg());
assert_eq!(a.mul_t(f1),
Mat3::new::<float>(0.5, 2.0, 3.5,
1.0, 2.5, 4.0,
1.5, 3.0, 4.5));
assert_eq!(a.mul_v(&v1), Vec3::new::<float>(14.0, 32.0, 50.0));
assert_eq!(a.add_m(&b),
Mat3::new::<float>(3.0, 9.0, 15.0,
5.0, 11.0, 17.0,
7.0, 13.0, 19.0));
assert_eq!(a.sub_m(&b),
Mat3::new::<float>(-1.0, -1.0, -1.0,
-1.0, -1.0, -1.0,
-1.0, -1.0, -1.0));
assert_eq!(a.mul_m(&b),
Mat3::new::<float>(36.0, 81.0, 126.0,
42.0, 96.0, 150.0,
48.0, 111.0, 174.0));
assert_eq!(a.dot(&b), 330.0);
assert_eq!(a.transpose(),
Mat3::new::<float>(1.0, 2.0, 3.0,
4.0, 5.0, 6.0,
7.0, 8.0, 9.0));
assert!(a.inverse().is_none());
assert_eq!(Mat3::new::<float>(2.0, 4.0, 6.0,
0.0, 2.0, 4.0,
0.0, 0.0, 1.0).inverse().unwrap(),
Mat3::new::<float>(0.5, -1.0, 1.0,
0.0, 0.5, -2.0,
0.0, 0.0, 1.0));
let ident = Mat3::identity::<float>();
assert_eq!(ident.inverse().unwrap(), ident);
assert!(ident.is_identity());
assert!(ident.is_symmetric());
assert!(ident.is_diagonal());
assert!(!ident.is_rotated());
assert!(ident.is_invertible());
assert!(!a.is_identity());
assert!(!a.is_symmetric());
assert!(!a.is_diagonal());
assert!(a.is_rotated());
assert!(!a.is_invertible());
let c = Mat3::new::<float>(3.0, 2.0, 1.0,
2.0, 3.0, 2.0,
1.0, 2.0, 3.0);
assert!(!c.is_identity());
assert!(c.is_symmetric());
assert!(!c.is_diagonal());
assert!(c.is_rotated());
assert!(c.is_invertible());
assert!(Mat3::from_value::<float>(6.0).is_diagonal());
assert_eq!(a.to_mat4(),
Mat4::new::<float>(1.0, 4.0, 7.0, 0.0,
2.0, 5.0, 8.0, 0.0,
3.0, 6.0, 9.0, 0.0,
0.0, 0.0, 0.0, 1.0));
// to_Quaternion
}
fn test_mat3_mut() {
let a = Mat3 { x: Vec3 { x: 1.0, y: 4.0, z: 7.0 },
y: Vec3 { x: 2.0, y: 5.0, z: 8.0 },
z: Vec3 { x: 3.0, y: 6.0, z: 9.0 } };
let b = Mat3 { x: Vec3 { x: 2.0, y: 5.0, z: 8.0 },
y: Vec3 { x: 3.0, y: 6.0, z: 9.0 },
z: Vec3 { x: 4.0, y: 7.0, z: 10.0 } };
let c = Mat3 { x: Vec3 { x: 2.0, y: 4.0, z: 6.0 },
y: Vec3 { x: 0.0, y: 2.0, z: 4.0 },
z: Vec3 { x: 0.0, y: 0.0, z: 1.0 } };
let f1 = 0.5;
let mut mut_a = a;
let mut mut_c = c;
mut_a.swap_cols(0, 2);
assert_eq!(mut_a.col(0), a.col(2));
assert_eq!(mut_a.col(2), a.col(0));
mut_a = a;
mut_a.swap_cols(1, 2);
assert_eq!(mut_a.col(1), a.col(2));
assert_eq!(mut_a.col(2), a.col(1));
mut_a = a;
mut_a.swap_rows(0, 2);
assert_eq!(mut_a.row(0), a.row(2));
assert_eq!(mut_a.row(2), a.row(0));
mut_a = a;
mut_a.swap_rows(1, 2);
assert_eq!(mut_a.row(1), a.row(2));
assert_eq!(mut_a.row(2), a.row(1));
mut_a = a;
mut_a.to_identity();
assert!(mut_a.is_identity());
mut_a = a;
mut_a.to_zero();
assert_eq!(mut_a, Mat3::zero::<float>());
mut_a = a;
mut_a.mul_self_t(f1);
assert_eq!(mut_a, a.mul_t(f1));
mut_a = a;
mut_a.add_self_m(&b);
assert_eq!(mut_a, a.add_m(&b));
mut_a = a;
mut_a.sub_self_m(&b);
assert_eq!(mut_a, a.sub_m(&b));
mut_a = a;
mut_c.invert_self();
assert_eq!(mut_c, c.inverse().unwrap());
// mut_c = c;
mut_a.transpose_self();
assert_eq!(mut_a, a.transpose());
// mut_a = a;
}
#[test]
fn test_mat3_approx_eq() {
assert!(!Mat3::new::<float>(0.000001, 0.000001, 0.000001,
0.000001, 0.000001, 0.000001,
0.000001, 0.000001, 0.000001)
.approx_eq(&Mat3::zero::<float>()));
assert!(Mat3::new::<float>(0.0000001, 0.0000001, 0.0000001,
0.0000001, 0.0000001, 0.0000001,
0.0000001, 0.0000001, 0.0000001)
.approx_eq(&Mat3::zero::<float>()));
}
}
#[deriving(Eq)]
pub struct Mat4<T> {
x: Vec4<T>,
y: Vec4<T>,
z: Vec4<T>,
w: Vec4<T>,
}
// GLSL-style type aliases
pub type mat4 = Mat4<f32>;
pub type dmat4 = Mat4<f64>;
// Rust-style type aliases
pub type Mat4f = Mat4<float>;
pub type Mat4f32 = Mat4<f32>;
pub type Mat4f64 = Mat4<f64>;
2013-06-16 08:03:14 +00:00
impl_dimensional!(Mat4, Vec4<T>, 4)
impl_dimensional_fns!(Mat4, Vec4<T>, 4)
impl_approx!(Mat4)
impl_mat!(Mat4, Vec4)
impl_mat_copyable!(Mat4, Vec4)
impl_mat_numeric!(Mat4, Vec4)
impl_mat_approx_numeric!(Mat4)
impl_mat_neg!(Mat4)
pub trait ToMat4<T> {
pub fn to_mat4(&self) -> Mat4<T>;
}
impl<T> Mat4<T> {
#[inline]
pub fn new(c0r0: T, c0r1: T, c0r2: T, c0r3: T,
c1r0: T, c1r1: T, c1r2: T, c1r3: T,
c2r0: T, c2r1: T, c2r2: T, c2r3: T,
c3r0: T, c3r1: T, c3r2: T, c3r3: T) -> Mat4<T> {
Mat4::from_cols(Vec4::new(c0r0, c0r1, c0r2, c0r3),
Vec4::new(c1r0, c1r1, c1r2, c1r3),
Vec4::new(c2r0, c2r1, c2r2, c2r3),
Vec4::new(c3r0, c3r1, c3r2, c3r3))
}
#[inline]
pub fn from_cols(c0: Vec4<T>,
c1: Vec4<T>,
c2: Vec4<T>,
c3: Vec4<T>) -> Mat4<T> {
Mat4 { x: c0, y: c1, z: c2, w: c3 }
}
}
#[cfg(test)]
mod mat4_tests {
use mat::*;
use vec::*;
#[test]
fn test_mat4() {
let a = Mat4 { x: Vec4 { x: 1.0, y: 5.0, z: 9.0, w: 13.0 },
y: Vec4 { x: 2.0, y: 6.0, z: 10.0, w: 14.0 },
z: Vec4 { x: 3.0, y: 7.0, z: 11.0, w: 15.0 },
w: Vec4 { x: 4.0, y: 8.0, z: 12.0, w: 16.0 } };
let b = Mat4 { x: Vec4 { x: 2.0, y: 6.0, z: 10.0, w: 14.0 },
y: Vec4 { x: 3.0, y: 7.0, z: 11.0, w: 15.0 },
z: Vec4 { x: 4.0, y: 8.0, z: 12.0, w: 16.0 },
w: Vec4 { x: 5.0, y: 9.0, z: 13.0, w: 17.0 } };
let c = Mat4 { x: Vec4 { x: 3.0, y: 2.0, z: 1.0, w: 1.0 },
y: Vec4 { x: 2.0, y: 3.0, z: 2.0, w: 2.0 },
z: Vec4 { x: 1.0, y: 2.0, z: 3.0, w: 3.0 },
w: Vec4 { x: 0.0, y: 1.0, z: 1.0, w: 0.0 } };
let v1 = Vec4::new::<float>(1.0, 2.0, 3.0, 4.0);
let f1 = 0.5;
assert_eq!(a, Mat4::new::<float>(1.0, 5.0, 9.0, 13.0,
2.0, 6.0, 10.0, 14.0,
3.0, 7.0, 11.0, 15.0,
4.0, 8.0, 12.0, 16.0));
assert_eq!(a, Mat4::from_cols::<float>(Vec4::new::<float>(1.0, 5.0, 9.0, 13.0),
Vec4::new::<float>(2.0, 6.0, 10.0, 14.0),
Vec4::new::<float>(3.0, 7.0, 11.0, 15.0),
Vec4::new::<float>(4.0, 8.0, 12.0, 16.0)));
assert_eq!(Mat4::from_value::<float>(4.0),
Mat4::new::<float>(4.0, 0.0, 0.0, 0.0,
0.0, 4.0, 0.0, 0.0,
0.0, 0.0, 4.0, 0.0,
0.0, 0.0, 0.0, 4.0));
assert_eq!(*a.col(0), Vec4::new::<float>(1.0, 5.0, 9.0, 13.0));
assert_eq!(*a.col(1), Vec4::new::<float>(2.0, 6.0, 10.0, 14.0));
assert_eq!(*a.col(2), Vec4::new::<float>(3.0, 7.0, 11.0, 15.0));
assert_eq!(*a.col(3), Vec4::new::<float>(4.0, 8.0, 12.0, 16.0));
assert_eq!(a.row(0), Vec4::new::<float>( 1.0, 2.0, 3.0, 4.0));
assert_eq!(a.row(1), Vec4::new::<float>( 5.0, 6.0, 7.0, 8.0));
assert_eq!(a.row(2), Vec4::new::<float>( 9.0, 10.0, 11.0, 12.0));
assert_eq!(a.row(3), Vec4::new::<float>(13.0, 14.0, 15.0, 16.0));
assert_eq!(*a.col(0), Vec4::new::<float>(1.0, 5.0, 9.0, 13.0));
assert_eq!(*a.col(1), Vec4::new::<float>(2.0, 6.0, 10.0, 14.0));
assert_eq!(*a.col(2), Vec4::new::<float>(3.0, 7.0, 11.0, 15.0));
assert_eq!(*a.col(3), Vec4::new::<float>(4.0, 8.0, 12.0, 16.0));
assert_eq!(Mat4::identity::<float>(),
Mat4::new::<float>(1.0, 0.0, 0.0, 0.0,
0.0, 1.0, 0.0, 0.0,
0.0, 0.0, 1.0, 0.0,
0.0, 0.0, 0.0, 1.0));
assert_eq!(Mat4::zero::<float>(),
Mat4::new::<float>(0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0));
assert_eq!(a.determinant(), 0.0);
assert_eq!(a.trace(), 34.0);
assert_eq!(a.neg(),
Mat4::new::<float>(-1.0, -5.0, -9.0, -13.0,
-2.0, -6.0, -10.0, -14.0,
-3.0, -7.0, -11.0, -15.0,
-4.0, -8.0, -12.0, -16.0));
assert_eq!(-a, a.neg());
assert_eq!(a.mul_t(f1),
Mat4::new::<float>(0.5, 2.5, 4.5, 6.5,
1.0, 3.0, 5.0, 7.0,
1.5, 3.5, 5.5, 7.5,
2.0, 4.0, 6.0, 8.0));
assert_eq!(a.mul_v(&v1),
Vec4::new::<float>(30.0, 70.0, 110.0, 150.0));
assert_eq!(a.add_m(&b),
Mat4::new::<float>(3.0, 11.0, 19.0, 27.0,
5.0, 13.0, 21.0, 29.0,
7.0, 15.0, 23.0, 31.0,
9.0, 17.0, 25.0, 33.0));
assert_eq!(a.sub_m(&b),
Mat4::new::<float>(-1.0, -1.0, -1.0, -1.0,
-1.0, -1.0, -1.0, -1.0,
-1.0, -1.0, -1.0, -1.0,
-1.0, -1.0, -1.0, -1.0));
assert_eq!(a.mul_m(&b),
Mat4::new::<float>(100.0, 228.0, 356.0, 484.0,
110.0, 254.0, 398.0, 542.0,
120.0, 280.0, 440.0, 600.0,
130.0, 306.0, 482.0, 658.0));
assert_eq!(a.dot(&b), 1632.0);
assert_eq!(a.transpose(),
Mat4::new::<float>( 1.0, 2.0, 3.0, 4.0,
5.0, 6.0, 7.0, 8.0,
9.0, 10.0, 11.0, 12.0,
13.0, 14.0, 15.0, 16.0));
assert_approx_eq!(c.inverse().unwrap(),
Mat4::new::<float>( 5.0, -4.0, 1.0, 0.0,
-4.0, 8.0, -4.0, 0.0,
4.0, -8.0, 4.0, 8.0,
-3.0, 4.0, 1.0, -8.0).mul_t(0.125));
let ident = Mat4::identity::<float>();
assert_eq!(ident.inverse().unwrap(), ident);
assert!(ident.is_identity());
assert!(ident.is_symmetric());
assert!(ident.is_diagonal());
assert!(!ident.is_rotated());
assert!(ident.is_invertible());
assert!(!a.is_identity());
assert!(!a.is_symmetric());
assert!(!a.is_diagonal());
assert!(a.is_rotated());
assert!(!a.is_invertible());
let c = Mat4::new::<float>(4.0, 3.0, 2.0, 1.0,
3.0, 4.0, 3.0, 2.0,
2.0, 3.0, 4.0, 3.0,
1.0, 2.0, 3.0, 4.0);
assert!(!c.is_identity());
assert!(c.is_symmetric());
assert!(!c.is_diagonal());
assert!(c.is_rotated());
assert!(c.is_invertible());
assert!(Mat4::from_value::<float>(6.0).is_diagonal());
}
fn test_mat4_mut() {
let a = Mat4 { x: Vec4 { x: 1.0, y: 5.0, z: 9.0, w: 13.0 },
y: Vec4 { x: 2.0, y: 6.0, z: 10.0, w: 14.0 },
z: Vec4 { x: 3.0, y: 7.0, z: 11.0, w: 15.0 },
w: Vec4 { x: 4.0, y: 8.0, z: 12.0, w: 16.0 } };
let b = Mat4 { x: Vec4 { x: 2.0, y: 6.0, z: 10.0, w: 14.0 },
y: Vec4 { x: 3.0, y: 7.0, z: 11.0, w: 15.0 },
z: Vec4 { x: 4.0, y: 8.0, z: 12.0, w: 16.0 },
w: Vec4 { x: 5.0, y: 9.0, z: 13.0, w: 17.0 } };
let c = Mat4 { x: Vec4 { x: 3.0, y: 2.0, z: 1.0, w: 1.0 },
y: Vec4 { x: 2.0, y: 3.0, z: 2.0, w: 2.0 },
z: Vec4 { x: 1.0, y: 2.0, z: 3.0, w: 3.0 },
w: Vec4 { x: 0.0, y: 1.0, z: 1.0, w: 0.0 } };
let f1 = 0.5;
let mut mut_a = a;
let mut mut_c = c;
mut_a.swap_cols(0, 3);
assert_eq!(mut_a.col(0), a.col(3));
assert_eq!(mut_a.col(3), a.col(0));
mut_a = a;
mut_a.swap_cols(1, 2);
assert_eq!(mut_a.col(1), a.col(2));
assert_eq!(mut_a.col(2), a.col(1));
mut_a = a;
mut_a.swap_rows(0, 3);
assert_eq!(mut_a.row(0), a.row(3));
assert_eq!(mut_a.row(3), a.row(0));
mut_a = a;
mut_a.swap_rows(1, 2);
assert_eq!(mut_a.row(1), a.row(2));
assert_eq!(mut_a.row(2), a.row(1));
mut_a = a;
mut_a.to_identity();
assert!(mut_a.is_identity());
mut_a = a;
mut_a.to_zero();
assert_eq!(mut_a, Mat4::zero::<float>());
mut_a = a;
mut_a.mul_self_t(f1);
assert_eq!(mut_a, a.mul_t(f1));
mut_a = a;
mut_a.add_self_m(&b);
assert_eq!(mut_a, a.add_m(&b));
mut_a = a;
mut_a.sub_self_m(&b);
assert_eq!(mut_a, a.sub_m(&b));
mut_a = a;
mut_c.invert_self();
assert_eq!(mut_c, c.inverse().unwrap());
// mut_c = c;
mut_a.transpose_self();
assert_eq!(mut_a, a.transpose());
// mut_a = a;
}
#[test]
fn test_mat4_approx_eq() {
assert!(!Mat4::new::<float>(0.000001, 0.000001, 0.000001, 0.000001,
0.000001, 0.000001, 0.000001, 0.000001,
0.000001, 0.000001, 0.000001, 0.000001,
0.000001, 0.000001, 0.000001, 0.000001)
.approx_eq(&Mat4::zero::<float>()));
assert!(Mat4::new::<float>(0.0000001, 0.0000001, 0.0000001, 0.0000001,
0.0000001, 0.0000001, 0.0000001, 0.0000001,
0.0000001, 0.0000001, 0.0000001, 0.0000001,
0.0000001, 0.0000001, 0.0000001, 0.0000001)
.approx_eq(&Mat4::zero::<float>()));
}
}