use super::UnknownUnit;
use approxeq::ApproxEq;
use length::Length;
use point::{TypedPoint2D, TypedPoint3D, point2, point3};
use size::{TypedSize2D, size2};
use scale_factor::ScaleFactor;
use num::*;
use num_traits::{Float, NumCast};
use std::fmt;
use std::ops::{Add, Neg, Mul, Sub, Div, AddAssign, SubAssign, MulAssign, DivAssign};
use std::marker::PhantomData;
define_matrix! {
pub struct TypedVector2D<T, U> {
pub x: T,
pub y: T,
}
}
pub type Vector2D<T> = TypedVector2D<T, UnknownUnit>;
impl<T: Copy + Zero, U> TypedVector2D<T, U> {
#[inline]
pub fn zero() -> Self {
TypedVector2D::new(Zero::zero(), Zero::zero())
}
#[inline]
pub fn to_3d(&self) -> TypedVector3D<T, U> {
vec3(self.x, self.y, Zero::zero())
}
}
impl<T: fmt::Debug, U> fmt::Debug for TypedVector2D<T, U> {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "({:?},{:?})", self.x, self.y)
}
}
impl<T: fmt::Display, U> fmt::Display for TypedVector2D<T, U> {
fn fmt(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
write!(formatter, "({},{})", self.x, self.y)
}
}
impl<T: Copy, U> TypedVector2D<T, U> {
#[inline]
pub fn new(x: T, y: T) -> Self {
TypedVector2D { x: x, y: y, _unit: PhantomData }
}
#[inline]
pub fn from_lengths(x: Length<T, U>, y: Length<T, U>) -> Self {
vec2(x.0, y.0)
}
#[inline]
pub fn to_point(&self) -> TypedPoint2D<T, U> {
point2(self.x, self.y)
}
#[inline]
pub fn to_size(&self) -> TypedSize2D<T, U> {
size2(self.x, self.y)
}
#[inline]
pub fn x_typed(&self) -> Length<T, U> { Length::new(self.x) }
#[inline]
pub fn y_typed(&self) -> Length<T, U> { Length::new(self.y) }
#[inline]
pub fn to_untyped(&self) -> Vector2D<T> {
vec2(self.x, self.y)
}
#[inline]
pub fn from_untyped(p: &Vector2D<T>) -> Self {
vec2(p.x, p.y)
}
#[inline]
pub fn to_array(&self) -> [T; 2] {
[self.x, self.y]
}
}
impl<T, U> TypedVector2D<T, U>
where T: Copy + Mul<T, Output=T> + Add<T, Output=T> + Sub<T, Output=T> {
#[inline]
pub fn dot(self, other: Self) -> T {
self.x * other.x + self.y * other.y
}
#[inline]
pub fn cross(self, other: Self) -> T {
self.x * other.y - self.y * other.x
}
#[inline]
pub fn normalize(self) -> Self where T: Float + ApproxEq<T> {
let dot = self.dot(self);
if dot.approx_eq(&T::zero()) {
self
} else {
self / dot.sqrt()
}
}
#[inline]
pub fn square_length(&self) -> T {
self.x * self.x + self.y * self.y
}
#[inline]
pub fn length(&self) -> T where T: Float + ApproxEq<T> {
self.square_length().sqrt()
}
}
impl<T: Copy + Add<T, Output=T>, U> Add for TypedVector2D<T, U> {
type Output = Self;
fn add(self, other: Self) -> Self {
TypedVector2D::new(self.x + other.x, self.y + other.y)
}
}
impl<T: Copy + Add<T, Output=T>, U> AddAssign for TypedVector2D<T, U> {
#[inline]
fn add_assign(&mut self, other: Self) {
*self = *self + other
}
}
impl<T: Copy + Sub<T, Output=T>, U> SubAssign<TypedVector2D<T, U>> for TypedVector2D<T, U> {
#[inline]
fn sub_assign(&mut self, other: Self) {
*self = *self - other
}
}
impl<T: Copy + Sub<T, Output=T>, U> Sub for TypedVector2D<T, U> {
type Output = Self;
#[inline]
fn sub(self, other: TypedVector2D<T, U>) -> Self {
vec2(self.x - other.x, self.y - other.y)
}
}
impl <T: Copy + Neg<Output=T>, U> Neg for TypedVector2D<T, U> {
type Output = Self;
#[inline]
fn neg(self) -> Self {
vec2(-self.x, -self.y)
}
}
impl<T: Float, U> TypedVector2D<T, U> {
#[inline]
pub fn min(self, other: TypedVector2D<T, U>) -> Self {
vec2(self.x.min(other.x), self.y.min(other.y))
}
#[inline]
pub fn max(self, other: TypedVector2D<T, U>) -> Self {
vec2(self.x.max(other.x), self.y.max(other.y))
}
}
impl<T: Copy + Mul<T, Output=T>, U> Mul<T> for TypedVector2D<T, U> {
type Output = Self;
#[inline]
fn mul(self, scale: T) -> Self {
vec2(self.x * scale, self.y * scale)
}
}
impl<T: Copy + Div<T, Output=T>, U> Div<T> for TypedVector2D<T, U> {
type Output = Self;
#[inline]
fn div(self, scale: T) -> Self {
vec2(self.x / scale, self.y / scale)
}
}
impl<T: Copy + Mul<T, Output=T>, U> MulAssign<T> for TypedVector2D<T, U> {
#[inline]
fn mul_assign(&mut self, scale: T) {
*self = *self * scale
}
}
impl<T: Copy + Div<T, Output=T>, U> DivAssign<T> for TypedVector2D<T, U> {
#[inline]
fn div_assign(&mut self, scale: T) {
*self = *self / scale
}
}
impl<T: Copy + Mul<T, Output=T>, U1, U2> Mul<ScaleFactor<T, U1, U2>> for TypedVector2D<T, U1> {
type Output = TypedVector2D<T, U2>;
#[inline]
fn mul(self, scale: ScaleFactor<T, U1, U2>) -> TypedVector2D<T, U2> {
vec2(self.x * scale.get(), self.y * scale.get())
}
}
impl<T: Copy + Div<T, Output=T>, U1, U2> Div<ScaleFactor<T, U1, U2>> for TypedVector2D<T, U2> {
type Output = TypedVector2D<T, U1>;
#[inline]
fn div(self, scale: ScaleFactor<T, U1, U2>) -> TypedVector2D<T, U1> {
vec2(self.x / scale.get(), self.y / scale.get())
}
}
impl<T: Round, U> TypedVector2D<T, U> {
#[inline]
#[must_use]
pub fn round(&self) -> Self {
vec2(self.x.round(), self.y.round())
}
}
impl<T: Ceil, U> TypedVector2D<T, U> {
#[inline]
#[must_use]
pub fn ceil(&self) -> Self {
vec2(self.x.ceil(), self.y.ceil())
}
}
impl<T: Floor, U> TypedVector2D<T, U> {
#[inline]
#[must_use]
pub fn floor(&self) -> Self {
vec2(self.x.floor(), self.y.floor())
}
}
impl<T: NumCast + Copy, U> TypedVector2D<T, U> {
#[inline]
pub fn cast<NewT: NumCast + Copy>(&self) -> Option<TypedVector2D<NewT, U>> {
match (NumCast::from(self.x), NumCast::from(self.y)) {
(Some(x), Some(y)) => Some(TypedVector2D::new(x, y)),
_ => None
}
}
#[inline]
pub fn to_f32(&self) -> TypedVector2D<f32, U> {
self.cast().unwrap()
}
#[inline]
pub fn to_usize(&self) -> TypedVector2D<usize, U> {
self.cast().unwrap()
}
#[inline]
pub fn to_i32(&self) -> TypedVector2D<i32, U> {
self.cast().unwrap()
}
#[inline]
pub fn to_i64(&self) -> TypedVector2D<i64, U> {
self.cast().unwrap()
}
}
impl<T: Copy+ApproxEq<T>, U> ApproxEq<TypedVector2D<T, U>> for TypedVector2D<T, U> {
#[inline]
fn approx_epsilon() -> Self {
vec2(T::approx_epsilon(), T::approx_epsilon())
}
#[inline]
fn approx_eq(&self, other: &Self) -> bool {
self.x.approx_eq(&other.x) && self.y.approx_eq(&other.y)
}
#[inline]
fn approx_eq_eps(&self, other: &Self, eps: &Self) -> bool {
self.x.approx_eq_eps(&other.x, &eps.x) && self.y.approx_eq_eps(&other.y, &eps.y)
}
}
impl<T: Copy, U> Into<[T; 2]> for TypedVector2D<T, U> {
fn into(self) -> [T; 2] {
self.to_array()
}
}
impl<T: Copy, U> From<[T; 2]> for TypedVector2D<T, U> {
fn from(array: [T; 2]) -> Self {
vec2(array[0], array[1])
}
}
define_matrix! {
pub struct TypedVector3D<T, U> {
pub x: T,
pub y: T,
pub z: T,
}
}
pub type Vector3D<T> = TypedVector3D<T, UnknownUnit>;
impl<T: Copy + Zero, U> TypedVector3D<T, U> {
#[inline]
pub fn zero() -> Self {
vec3(Zero::zero(), Zero::zero(), Zero::zero())
}
#[inline]
pub fn to_array_4d(&self) -> [T; 4] {
[self.x, self.y, self.z, Zero::zero()]
}
}
impl<T: fmt::Debug, U> fmt::Debug for TypedVector3D<T, U> {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "({:?},{:?},{:?})", self.x, self.y, self.z)
}
}
impl<T: fmt::Display, U> fmt::Display for TypedVector3D<T, U> {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "({},{},{})", self.x, self.y, self.z)
}
}
impl<T: Copy, U> TypedVector3D<T, U> {
#[inline]
pub fn new(x: T, y: T, z: T) -> Self {
TypedVector3D { x: x, y: y, z: z, _unit: PhantomData }
}
#[inline]
pub fn from_lengths(x: Length<T, U>, y: Length<T, U>, z: Length<T, U>) -> TypedVector3D<T, U> {
vec3(x.0, y.0, z.0)
}
#[inline]
pub fn to_point(&self) -> TypedPoint3D<T, U> {
point3(self.x, self.y, self.z)
}
#[inline]
pub fn x_typed(&self) -> Length<T, U> { Length::new(self.x) }
#[inline]
pub fn y_typed(&self) -> Length<T, U> { Length::new(self.y) }
#[inline]
pub fn z_typed(&self) -> Length<T, U> { Length::new(self.z) }
#[inline]
pub fn to_array(&self) -> [T; 3] { [self.x, self.y, self.z] }
#[inline]
pub fn to_untyped(&self) -> Vector3D<T> {
vec3(self.x, self.y, self.z)
}
#[inline]
pub fn from_untyped(p: &Vector3D<T>) -> Self {
vec3(p.x, p.y, p.z)
}
#[inline]
pub fn to_2d(&self) -> TypedVector2D<T, U> {
vec2(self.x, self.y)
}
}
impl<T: Mul<T, Output=T> +
Add<T, Output=T> +
Sub<T, Output=T> +
Copy, U> TypedVector3D<T, U> {
#[inline]
pub fn dot(self, other: Self) -> T {
self.x * other.x +
self.y * other.y +
self.z * other.z
}
#[inline]
pub fn cross(self, other: TypedVector3D<T, U>) -> Self {
vec3(
self.y * other.z - self.z * other.y,
self.z * other.x - self.x * other.z,
self.x * other.y - self.y * other.x
)
}
#[inline]
pub fn normalize(self) -> Self where T: Float + ApproxEq<T> {
let dot = self.dot(self);
if dot.approx_eq(&T::zero()) {
self
} else {
self / dot.sqrt()
}
}
#[inline]
pub fn square_length(&self) -> T {
self.x * self.x + self.y * self.y + self.z * self.z
}
#[inline]
pub fn length(&self) -> T where T: Float + ApproxEq<T> {
self.square_length().sqrt()
}
}
impl<T: Copy + Add<T, Output=T>, U> Add for TypedVector3D<T, U> {
type Output = TypedVector3D<T, U>;
#[inline]
fn add(self, other: TypedVector3D<T, U>) -> Self {
vec3(self.x + other.x, self.y + other.y, self.z + other.z)
}
}
impl<T: Copy + Sub<T, Output=T>, U> Sub for TypedVector3D<T, U> {
type Output = TypedVector3D<T, U>;
#[inline]
fn sub(self, other: TypedVector3D<T, U>) -> Self {
vec3(self.x - other.x, self.y - other.y, self.z - other.z)
}
}
impl<T: Copy + Add<T, Output=T>, U> AddAssign for TypedVector3D<T, U> {
#[inline]
fn add_assign(&mut self, other: Self) {
*self = *self + other
}
}
impl<T: Copy + Sub<T, Output=T>, U> SubAssign<TypedVector3D<T, U>> for TypedVector3D<T, U> {
#[inline]
fn sub_assign(&mut self, other: Self) {
*self = *self - other
}
}
impl <T: Copy + Neg<Output=T>, U> Neg for TypedVector3D<T, U> {
type Output = Self;
#[inline]
fn neg(self) -> Self {
vec3(-self.x, -self.y, -self.z)
}
}
impl<T: Copy + Mul<T, Output=T>, U> Mul<T> for TypedVector3D<T, U> {
type Output = Self;
#[inline]
fn mul(self, scale: T) -> Self {
Self::new(self.x * scale, self.y * scale, self.z * scale)
}
}
impl<T: Copy + Div<T, Output=T>, U> Div<T> for TypedVector3D<T, U> {
type Output = Self;
#[inline]
fn div(self, scale: T) -> Self {
Self::new(self.x / scale, self.y / scale, self.z / scale)
}
}
impl<T: Copy + Mul<T, Output=T>, U> MulAssign<T> for TypedVector3D<T, U> {
#[inline]
fn mul_assign(&mut self, scale: T) {
*self = *self * scale
}
}
impl<T: Copy + Div<T, Output=T>, U> DivAssign<T> for TypedVector3D<T, U> {
#[inline]
fn div_assign(&mut self, scale: T) {
*self = *self / scale
}
}
impl<T: Float, U> TypedVector3D<T, U> {
#[inline]
pub fn min(self, other: TypedVector3D<T, U>) -> TypedVector3D<T, U> {
vec3(self.x.min(other.x), self.y.min(other.y), self.z.min(other.z))
}
#[inline]
pub fn max(self, other: TypedVector3D<T, U>) -> TypedVector3D<T, U> {
vec3(self.x.max(other.x), self.y.max(other.y), self.z.max(other.z))
}
}
impl<T: Round, U> TypedVector3D<T, U> {
#[inline]
#[must_use]
pub fn round(&self) -> Self {
vec3(self.x.round(), self.y.round(), self.z.round())
}
}
impl<T: Ceil, U> TypedVector3D<T, U> {
#[inline]
#[must_use]
pub fn ceil(&self) -> Self {
vec3(self.x.ceil(), self.y.ceil(), self.z.ceil())
}
}
impl<T: Floor, U> TypedVector3D<T, U> {
#[inline]
#[must_use]
pub fn floor(&self) -> Self {
vec3(self.x.floor(), self.y.floor(), self.z.floor())
}
}
impl<T: NumCast + Copy, U> TypedVector3D<T, U> {
#[inline]
pub fn cast<NewT: NumCast + Copy>(&self) -> Option<TypedVector3D<NewT, U>> {
match (NumCast::from(self.x),
NumCast::from(self.y),
NumCast::from(self.z)) {
(Some(x), Some(y), Some(z)) => Some(vec3(x, y, z)),
_ => None
}
}
#[inline]
pub fn to_f32(&self) -> TypedVector3D<f32, U> {
self.cast().unwrap()
}
#[inline]
pub fn to_usize(&self) -> TypedVector3D<usize, U> {
self.cast().unwrap()
}
#[inline]
pub fn to_i32(&self) -> TypedVector3D<i32, U> {
self.cast().unwrap()
}
#[inline]
pub fn to_i64(&self) -> TypedVector3D<i64, U> {
self.cast().unwrap()
}
}
impl<T: Copy+ApproxEq<T>, U> ApproxEq<TypedVector3D<T, U>> for TypedVector3D<T, U> {
#[inline]
fn approx_epsilon() -> Self {
vec3(T::approx_epsilon(), T::approx_epsilon(), T::approx_epsilon())
}
#[inline]
fn approx_eq(&self, other: &Self) -> bool {
self.x.approx_eq(&other.x)
&& self.y.approx_eq(&other.y)
&& self.z.approx_eq(&other.z)
}
#[inline]
fn approx_eq_eps(&self, other: &Self, eps: &Self) -> bool {
self.x.approx_eq_eps(&other.x, &eps.x)
&& self.y.approx_eq_eps(&other.y, &eps.y)
&& self.z.approx_eq_eps(&other.z, &eps.z)
}
}
impl<T: Copy, U> Into<[T; 3]> for TypedVector3D<T, U> {
fn into(self) -> [T; 3] {
self.to_array()
}
}
impl<T: Copy, U> From<[T; 3]> for TypedVector3D<T, U> {
fn from(array: [T; 3]) -> Self {
vec3(array[0], array[1], array[2])
}
}
#[inline]
pub fn vec2<T: Copy, U>(x: T, y: T) -> TypedVector2D<T, U> {
TypedVector2D::new(x, y)
}
#[inline]
pub fn vec3<T: Copy, U>(x: T, y: T, z: T) -> TypedVector3D<T, U> {
TypedVector3D::new(x, y, z)
}
#[cfg(test)]
mod vector2d {
use super::{Vector2D, vec2};
type Vec2 = Vector2D<f32>;
#[test]
pub fn test_scalar_mul() {
let p1: Vec2 = vec2(3.0, 5.0);
let result = p1 * 5.0;
assert_eq!(result, Vector2D::new(15.0, 25.0));
}
#[test]
pub fn test_dot() {
let p1: Vec2 = vec2(2.0, 7.0);
let p2: Vec2 = vec2(13.0, 11.0);
assert_eq!(p1.dot(p2), 103.0);
}
#[test]
pub fn test_cross() {
let p1: Vec2 = vec2(4.0, 7.0);
let p2: Vec2 = vec2(13.0, 8.0);
let r = p1.cross(p2);
assert_eq!(r, -59.0);
}
#[test]
pub fn test_normalize() {
let p0: Vec2 = Vec2::zero();
let p1: Vec2 = vec2(4.0, 0.0);
let p2: Vec2 = vec2(3.0, -4.0);
assert_eq!(p0.normalize(), p0);
assert_eq!(p1.normalize(), vec2(1.0, 0.0));
assert_eq!(p2.normalize(), vec2(0.6, -0.8));
}
#[test]
pub fn test_min() {
let p1: Vec2 = vec2(1.0, 3.0);
let p2: Vec2 = vec2(2.0, 2.0);
let result = p1.min(p2);
assert_eq!(result, vec2(1.0, 2.0));
}
#[test]
pub fn test_max() {
let p1: Vec2 = vec2(1.0, 3.0);
let p2: Vec2 = vec2(2.0, 2.0);
let result = p1.max(p2);
assert_eq!(result, vec2(2.0, 3.0));
}
}
#[cfg(test)]
mod typedvector2d {
use super::{TypedVector2D, vec2};
use scale_factor::ScaleFactor;
pub enum Mm {}
pub enum Cm {}
pub type Vector2DMm<T> = TypedVector2D<T, Mm>;
pub type Vector2DCm<T> = TypedVector2D<T, Cm>;
#[test]
pub fn test_add() {
let p1 = Vector2DMm::new(1.0, 2.0);
let p2 = Vector2DMm::new(3.0, 4.0);
let result = p1 + p2;
assert_eq!(result, vec2(4.0, 6.0));
}
#[test]
pub fn test_add_assign() {
let mut p1 = Vector2DMm::new(1.0, 2.0);
p1 += vec2(3.0, 4.0);
assert_eq!(p1, vec2(4.0, 6.0));
}
#[test]
pub fn test_scalar_mul() {
let p1 = Vector2DMm::new(1.0, 2.0);
let cm_per_mm: ScaleFactor<f32, Mm, Cm> = ScaleFactor::new(0.1);
let result: Vector2DCm<f32> = p1 * cm_per_mm;
assert_eq!(result, vec2(0.1, 0.2));
}
}
#[cfg(test)]
mod vector3d {
use super::{Vector3D, vec3};
type Vec3 = Vector3D<f32>;
#[test]
pub fn test_dot() {
let p1: Vec3 = vec3(7.0, 21.0, 32.0);
let p2: Vec3 = vec3(43.0, 5.0, 16.0);
assert_eq!(p1.dot(p2), 918.0);
}
#[test]
pub fn test_cross() {
let p1: Vec3 = vec3(4.0, 7.0, 9.0);
let p2: Vec3 = vec3(13.0, 8.0, 3.0);
let p3 = p1.cross(p2);
assert_eq!(p3, vec3(-51.0, 105.0, -59.0));
}
#[test]
pub fn test_normalize() {
let p0: Vec3 = Vec3::zero();
let p1: Vec3 = vec3(0.0, -6.0, 0.0);
let p2: Vec3 = vec3(1.0, 2.0, -2.0);
assert_eq!(p0.normalize(), p0);
assert_eq!(p1.normalize(), vec3(0.0, -1.0, 0.0));
assert_eq!(p2.normalize(), vec3(1.0/3.0, 2.0/3.0, -2.0/3.0));
}
#[test]
pub fn test_min() {
let p1: Vec3 = vec3(1.0, 3.0, 5.0);
let p2: Vec3 = vec3(2.0, 2.0, -1.0);
let result = p1.min(p2);
assert_eq!(result, vec3(1.0, 2.0, -1.0));
}
#[test]
pub fn test_max() {
let p1: Vec3 = vec3(1.0, 3.0, 5.0);
let p2: Vec3 = vec3(2.0, 2.0, -1.0);
let result = p1.max(p2);
assert_eq!(result, vec3(2.0, 3.0, 5.0));
}
}