Revision control
Copy as Markdown
Other Tools
// Copyright 2013 The Servo Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// option. This file may not be copied, modified, or distributed
// except according to those terms.
use crate::approxeq::ApproxEq;
use crate::trig::Trig;
use core::cmp::{Eq, PartialEq};
use core::hash::Hash;
use core::iter::Sum;
use core::ops::{Add, AddAssign, Div, DivAssign, Mul, MulAssign, Neg, Rem, Sub, SubAssign};
#[cfg(feature = "bytemuck")]
use bytemuck::{Pod, Zeroable};
use num_traits::real::Real;
use num_traits::{Float, FloatConst, NumCast, One, Zero};
#[cfg(feature = "serde")]
use serde::{Deserialize, Serialize};
/// An angle in radians
#[derive(Copy, Clone, Default, Debug, PartialEq, Eq, PartialOrd, Hash)]
#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
pub struct Angle<T> {
pub radians: T,
}
#[cfg(feature = "bytemuck")]
unsafe impl<T: Zeroable> Zeroable for Angle<T> {}
#[cfg(feature = "bytemuck")]
unsafe impl<T: Pod> Pod for Angle<T> {}
#[cfg(feature = "arbitrary")]
impl<'a, T> arbitrary::Arbitrary<'a> for Angle<T>
where
T: arbitrary::Arbitrary<'a>,
{
// This implementation could be derived, but the derive would require an `extern crate std`.
fn arbitrary(u: &mut arbitrary::Unstructured<'a>) -> arbitrary::Result<Self> {
Ok(Angle {
radians: arbitrary::Arbitrary::arbitrary(u)?,
})
}
fn size_hint(depth: usize) -> (usize, Option<usize>) {
<T as arbitrary::Arbitrary>::size_hint(depth)
}
}
impl<T> Angle<T> {
#[inline]
pub fn radians(radians: T) -> Self {
Angle { radians }
}
#[inline]
pub fn get(self) -> T {
self.radians
}
}
impl<T> Angle<T>
where
T: Trig,
{
#[inline]
pub fn degrees(deg: T) -> Self {
Angle {
radians: T::degrees_to_radians(deg),
}
}
#[inline]
pub fn to_degrees(self) -> T {
T::radians_to_degrees(self.radians)
}
}
impl<T> Angle<T>
where
T: Rem<Output = T> + Sub<Output = T> + Add<Output = T> + Zero + FloatConst + PartialOrd + Copy,
{
/// Returns this angle in the [0..2*PI[ range.
pub fn positive(&self) -> Self {
let two_pi = T::PI() + T::PI();
let mut a = self.radians % two_pi;
if a < T::zero() {
a = a + two_pi;
}
Angle::radians(a)
}
/// Returns this angle in the ]-PI..PI] range.
pub fn signed(&self) -> Self {
Angle::pi() - (Angle::pi() - *self).positive()
}
}
impl<T> Angle<T>
where
T: Rem<Output = T>
+ Mul<Output = T>
+ Sub<Output = T>
+ Add<Output = T>
+ One
+ FloatConst
+ Copy,
{
/// Returns the shortest signed angle between two angles.
///
/// Takes wrapping and signs into account.
pub fn angle_to(&self, to: Self) -> Self {
let two = T::one() + T::one();
let max = T::PI() * two;
let d = (to.radians - self.radians) % max;
Angle::radians(two * d % max - d)
}
/// Linear interpolation between two angles, using the shortest path.
pub fn lerp(&self, other: Self, t: T) -> Self {
*self + self.angle_to(other) * t
}
}
impl<T> Angle<T>
where
T: Float,
{
/// Returns true if the angle is a finite number.
#[inline]
pub fn is_finite(self) -> bool {
self.radians.is_finite()
}
}
impl<T> Angle<T>
where
T: Real,
{
/// Returns (sin(self), cos(self)).
pub fn sin_cos(self) -> (T, T) {
self.radians.sin_cos()
}
}
impl<T> Angle<T>
where
T: Zero,
{
pub fn zero() -> Self {
Angle::radians(T::zero())
}
}
impl<T> Angle<T>
where
T: FloatConst + Add<Output = T>,
{
pub fn pi() -> Self {
Angle::radians(T::PI())
}
pub fn two_pi() -> Self {
Angle::radians(T::PI() + T::PI())
}
pub fn frac_pi_2() -> Self {
Angle::radians(T::FRAC_PI_2())
}
pub fn frac_pi_3() -> Self {
Angle::radians(T::FRAC_PI_3())
}
pub fn frac_pi_4() -> Self {
Angle::radians(T::FRAC_PI_4())
}
}
impl<T> Angle<T>
where
T: NumCast + Copy,
{
/// Cast from one numeric representation to another.
#[inline]
pub fn cast<NewT: NumCast>(&self) -> Angle<NewT> {
self.try_cast().unwrap()
}
/// Fallible cast from one numeric representation to another.
pub fn try_cast<NewT: NumCast>(&self) -> Option<Angle<NewT>> {
NumCast::from(self.radians).map(|radians| Angle { radians })
}
// Convenience functions for common casts.
/// Cast angle to `f32`.
#[inline]
pub fn to_f32(&self) -> Angle<f32> {
self.cast()
}
/// Cast angle `f64`.
#[inline]
pub fn to_f64(&self) -> Angle<f64> {
self.cast()
}
}
impl<T: Add<T, Output = T>> Add for Angle<T> {
type Output = Self;
fn add(self, other: Self) -> Self {
Self::radians(self.radians + other.radians)
}
}
impl<T: Copy + Add<T, Output = T>> Add<&Self> for Angle<T> {
type Output = Self;
fn add(self, other: &Self) -> Self {
Self::radians(self.radians + other.radians)
}
}
impl<T: Add + Zero> Sum for Angle<T> {
fn sum<I: Iterator<Item = Self>>(iter: I) -> Self {
iter.fold(Self::zero(), Add::add)
}
}
impl<'a, T: 'a + Add + Copy + Zero> Sum<&'a Self> for Angle<T> {
fn sum<I: Iterator<Item = &'a Self>>(iter: I) -> Self {
iter.fold(Self::zero(), Add::add)
}
}
impl<T: AddAssign<T>> AddAssign for Angle<T> {
fn add_assign(&mut self, other: Angle<T>) {
self.radians += other.radians;
}
}
impl<T: Sub<T, Output = T>> Sub<Angle<T>> for Angle<T> {
type Output = Angle<T>;
fn sub(self, other: Angle<T>) -> <Self as Sub>::Output {
Angle::radians(self.radians - other.radians)
}
}
impl<T: SubAssign<T>> SubAssign for Angle<T> {
fn sub_assign(&mut self, other: Angle<T>) {
self.radians -= other.radians;
}
}
impl<T: Div<T, Output = T>> Div<Angle<T>> for Angle<T> {
type Output = T;
#[inline]
fn div(self, other: Angle<T>) -> T {
self.radians / other.radians
}
}
impl<T: Div<T, Output = T>> Div<T> for Angle<T> {
type Output = Angle<T>;
#[inline]
fn div(self, factor: T) -> Angle<T> {
Angle::radians(self.radians / factor)
}
}
impl<T: DivAssign<T>> DivAssign<T> for Angle<T> {
fn div_assign(&mut self, factor: T) {
self.radians /= factor;
}
}
impl<T: Mul<T, Output = T>> Mul<T> for Angle<T> {
type Output = Angle<T>;
#[inline]
fn mul(self, factor: T) -> Angle<T> {
Angle::radians(self.radians * factor)
}
}
impl<T: MulAssign<T>> MulAssign<T> for Angle<T> {
fn mul_assign(&mut self, factor: T) {
self.radians *= factor;
}
}
impl<T: Neg<Output = T>> Neg for Angle<T> {
type Output = Self;
fn neg(self) -> Self {
Angle::radians(-self.radians)
}
}
impl<T: ApproxEq<T>> ApproxEq<T> for Angle<T> {
#[inline]
fn approx_epsilon() -> T {
T::approx_epsilon()
}
#[inline]
fn approx_eq_eps(&self, other: &Angle<T>, approx_epsilon: &T) -> bool {
self.radians.approx_eq_eps(&other.radians, approx_epsilon)
}
}
#[test]
fn wrap_angles() {
use core::f32::consts::{FRAC_PI_2, PI};
assert!(Angle::radians(0.0).positive().approx_eq(&Angle::zero()));
assert!(Angle::radians(FRAC_PI_2)
.positive()
.approx_eq(&Angle::frac_pi_2()));
assert!(Angle::radians(-FRAC_PI_2)
.positive()
.approx_eq(&Angle::radians(3.0 * FRAC_PI_2)));
assert!(Angle::radians(3.0 * FRAC_PI_2)
.positive()
.approx_eq(&Angle::radians(3.0 * FRAC_PI_2)));
assert!(Angle::radians(5.0 * FRAC_PI_2)
.positive()
.approx_eq(&Angle::frac_pi_2()));
assert!(Angle::radians(2.0 * PI)
.positive()
.approx_eq(&Angle::zero()));
assert!(Angle::radians(-2.0 * PI)
.positive()
.approx_eq(&Angle::zero()));
assert!(Angle::radians(PI).positive().approx_eq(&Angle::pi()));
assert!(Angle::radians(-PI).positive().approx_eq(&Angle::pi()));
assert!(Angle::radians(FRAC_PI_2)
.signed()
.approx_eq(&Angle::frac_pi_2()));
assert!(Angle::radians(3.0 * FRAC_PI_2)
.signed()
.approx_eq(&-Angle::frac_pi_2()));
assert!(Angle::radians(5.0 * FRAC_PI_2)
.signed()
.approx_eq(&Angle::frac_pi_2()));
assert!(Angle::radians(2.0 * PI).signed().approx_eq(&Angle::zero()));
assert!(Angle::radians(-2.0 * PI).signed().approx_eq(&Angle::zero()));
assert!(Angle::radians(-PI).signed().approx_eq(&Angle::pi()));
assert!(Angle::radians(PI).signed().approx_eq(&Angle::pi()));
}
#[test]
fn lerp() {
type A = Angle<f32>;
let a = A::radians(1.0);
let b = A::radians(2.0);
assert!(a.lerp(b, 0.25).approx_eq(&Angle::radians(1.25)));
assert!(a.lerp(b, 0.5).approx_eq(&Angle::radians(1.5)));
assert!(a.lerp(b, 0.75).approx_eq(&Angle::radians(1.75)));
assert!(a
.lerp(b + A::two_pi(), 0.75)
.approx_eq(&Angle::radians(1.75)));
assert!(a
.lerp(b - A::two_pi(), 0.75)
.approx_eq(&Angle::radians(1.75)));
assert!(a
.lerp(b + A::two_pi() * 5.0, 0.75)
.approx_eq(&Angle::radians(1.75)));
}
#[test]
fn sum() {
type A = Angle<f32>;
let angles = [A::radians(1.0), A::radians(2.0), A::radians(3.0)];
let sum = A::radians(6.0);
assert_eq!(angles.iter().sum::<A>(), sum);
}