8 比例缩放（scale.rs）

scale.rs代码是几何变换库euclid中典型的数据结构和方法的例子，用于处理二维和三维空间中的缩放变换。

一、scale.rs文件源码

//! A type-checked scaling factor between units.use crate::num::One;use crate::approxord::{max, min};
use crate::{Box2D, Box3D, Point2D, Point3D, Rect, Size2D, Vector2D};use core::cmp::Ordering;
use core::fmt;
use core::hash::{Hash, Hasher};
use core::marker::PhantomData;
use core::ops::{Add, Div, Mul, Sub};#[cfg(feature = "bytemuck")]
use bytemuck::{Pod, Zeroable};
use num_traits::NumCast;
#[cfg(feature = "serde")]
use serde::{Deserialize, Serialize};/// A scaling factor between two different units of measurement.
///
/// This is effectively a type-safe float, intended to be used in combination with other types like
/// `length::Length` to enforce conversion between systems of measurement at compile time.
///
/// `Src` and `Dst` represent the units before and after multiplying a value by a `Scale`. They
/// may be types without values, such as empty enums.  For example:
///
/// ```rust
/// use euclid::Scale;
/// use euclid::Length;
/// enum Mm {};
/// enum Inch {};
///
/// let mm_per_inch: Scale<f32, Inch, Mm> = Scale::new(25.4);
///
/// let one_foot: Length<f32, Inch> = Length::new(12.0);
/// let one_foot_in_mm: Length<f32, Mm> = one_foot * mm_per_inch;
/// ```
#[repr(C)]
#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
#[cfg_attr(feature = "serde",serde(bound(serialize = "T: serde::Serialize",deserialize = "T: serde::Deserialize<'de>"))
)]
pub struct Scale<T, Src, Dst>(pub T, #[doc(hidden)] pub PhantomData<(Src, Dst)>);impl<T, Src, Dst> Scale<T, Src, Dst> {#[inline]pub const fn new(x: T) -> Self {Scale(x, PhantomData)}/// 创建标识比例（1.0）#[inline]pub fn identity() -> Selfwhere T: One,{Scale::new(T::one())}/// Returns the given point transformed by this scale.////// # Example////// ```rust/// use euclid::{Scale, point2};/// enum Mm {};/// enum Cm {};////// let to_mm: Scale<i32, Cm, Mm> = Scale::new(10);////// assert_eq!(to_mm.transform_point(point2(42, -42)), point2(420, -420));/// ```#[inline]pub fn transform_point(self, point: Point2D<T, Src>) -> Point2D<T::Output, Dst>whereT: Copy + Mul,{Point2D::new(point.x * self.0, point.y * self.0)}/// Returns the given point transformed by this scale.#[inline]pub fn transform_point3d(self, point: Point3D<T, Src>) -> Point3D<T::Output, Dst>whereT: Copy + Mul,{Point3D::new(point.x * self.0, point.y * self.0, point.z * self.0)}/// Returns the given vector transformed by this scale.////// # Example////// ```rust/// use euclid::{Scale, vec2};/// enum Mm {};/// enum Cm {};////// let to_mm: Scale<i32, Cm, Mm> = Scale::new(10);////// assert_eq!(to_mm.transform_vector(vec2(42, -42)), vec2(420, -420));/// ```#[inline]pub fn transform_vector(self, vec: Vector2D<T, Src>) -> Vector2D<T::Output, Dst>whereT: Copy + Mul,{Vector2D::new(vec.x * self.0, vec.y * self.0)}/// Returns the given size transformed by this scale.////// # Example////// ```rust/// use euclid::{Scale, size2};/// enum Mm {};/// enum Cm {};////// let to_mm: Scale<i32, Cm, Mm> = Scale::new(10);////// assert_eq!(to_mm.transform_size(size2(42, -42)), size2(420, -420));/// ```#[inline]pub fn transform_size(self, size: Size2D<T, Src>) -> Size2D<T::Output, Dst>whereT: Copy + Mul,{Size2D::new(size.width * self.0, size.height * self.0)}/// Returns the given rect transformed by this scale.////// # Example////// ```rust/// use euclid::{Scale, rect};/// enum Mm {};/// enum Cm {};////// let to_mm: Scale<i32, Cm, Mm> = Scale::new(10);////// assert_eq!(to_mm.transform_rect(&rect(1, 2, 42, -42)), rect(10, 20, 420, -420));/// ```#[inline]pub fn transform_rect(self, rect: &Rect<T, Src>) -> Rect<T::Output, Dst>whereT: Copy + Mul,{Rect::new(self.transform_point(rect.origin),self.transform_size(rect.size),)}/// Returns the given box transformed by this scale.#[inline]pub fn transform_box2d(self, b: &Box2D<T, Src>) -> Box2D<T::Output, Dst>whereT: Copy + Mul,{Box2D {min: self.transform_point(b.min),max: self.transform_point(b.max),}}/// Returns the given box transformed by this scale.#[inline]pub fn transform_box3d(self, b: &Box3D<T, Src>) -> Box3D<T::Output, Dst>where T: Copy + Mul,{Box3D {min: self.transform_point3d(b.min),max: self.transform_point3d(b.max),}}/// Returns `true` if this scale has no effect.////// # Example////// ```rust/// use euclid::Scale;/// use euclid::num::One;/// enum Mm {};/// enum Cm {};////// let cm_per_mm: Scale<f32, Mm, Cm> = Scale::new(0.1);/// let mm_per_mm: Scale<f32, Mm, Mm> = Scale::new(1.0);////// assert_eq!(cm_per_mm.is_identity(), false);/// assert_eq!(mm_per_mm.is_identity(), true);/// assert_eq!(mm_per_mm, Scale::one());/// ```#[inline]pub fn is_identity(self) -> boolwhere T: PartialEq + One,{self.0 == T::one()}/// Returns the underlying scalar scale factor.#[inline]pub fn get(self) -> T {self.0}/// The inverse Scale (1.0 / self).////// # Example////// ```rust/// use euclid::Scale;/// enum Mm {};/// enum Cm {};////// let cm_per_mm: Scale<f32, Cm, Mm> = Scale::new(0.1);////// assert_eq!(cm_per_mm.inverse(), Scale::new(10.0));/// ```pub fn inverse(self) -> Scale<T::Output, Dst, Src>whereT: One + Div,{let one: T = One::one();Scale::new(one / self.0)}
}impl<T: PartialOrd, Src, Dst> Scale<T, Src, Dst> {#[inline]pub fn min(self, other: Self) -> Self {Self::new(min(self.0, other.0))}#[inline]pub fn max(self, other: Self) -> Self {Self::new(max(self.0, other.0))}/// Returns the point each component of which clamped by corresponding/// components of `start` and `end`.////// Shortcut for `self.max(start).min(end)`.#[inline]pub fn clamp(self, start: Self, end: Self) -> SelfwhereT: Copy,{self.max(start).min(end)}
}impl<T: NumCast, Src, Dst> Scale<T, Src, Dst> {/// Cast from one numeric representation to another, preserving the units.////// # Panics////// If the source value cannot be represented by the target type `NewT`, then/// method panics. Use `try_cast` if that must be case.////// # Example////// ```rust/// use euclid::Scale;/// enum Mm {};/// enum Cm {};////// let to_mm: Scale<i32, Cm, Mm> = Scale::new(10);////// assert_eq!(to_mm.cast::<f32>(), Scale::new(10.0));/// ```/// That conversion will panic, because `i32` not enough to store such big numbers:/// ```rust,should_panic/// use euclid::Scale;/// enum Mm {};// millimeter = 10^-2 meters/// enum Em {};// exameter   = 10^18 meters////// // Panics/// let to_em: Scale<i32, Mm, Em> = Scale::new(10e20).cast();/// ```#[inline]pub fn cast<NewT: NumCast>(self) -> Scale<NewT, Src, Dst> {self.try_cast().unwrap()}/// Fallible cast from one numeric representation to another, preserving the units./// If the source value cannot be represented by the target type `NewT`, then `None`/// is returned.////// # Example////// ```rust/// use euclid::Scale;/// enum Mm {};/// enum Cm {};/// enum Em {};// Exameter = 10^18 meters////// let to_mm: Scale<i32, Cm, Mm> = Scale::new(10);/// let to_em: Scale<f32, Mm, Em> = Scale::new(10e20);////// assert_eq!(to_mm.try_cast::<f32>(), Some(Scale::new(10.0)));/// // Integer to small to store that number/// assert_eq!(to_em.try_cast::<i32>(), None);/// ```pub fn try_cast<NewT: NumCast>(self) -> Option<Scale<NewT, Src, Dst>> {NumCast::from(self.0).map(Scale::new)}
}#[cfg(feature = "arbitrary")]
impl<'a, T, Src, Dst> arbitrary::Arbitrary<'a> for Scale<T, Src, Dst>
whereT: arbitrary::Arbitrary<'a>,
{fn arbitrary(u: &mut arbitrary::Unstructured<'a>) -> arbitrary::Result<Self> {Ok(Scale::new(arbitrary::Arbitrary::arbitrary(u)?))}
}#[cfg(feature = "bytemuck")]
unsafe impl<T: Zeroable, Src, Dst> Zeroable for Scale<T, Src, Dst> {}#[cfg(feature = "bytemuck")]
unsafe impl<T: Pod, Src: 'static, Dst: 'static> Pod for Scale<T, Src, Dst> {}// scale0 * scale1
// (A,B) * (B,C) = (A,C)
impl<T: Mul, A, B, C> Mul<Scale<T, B, C>> for Scale<T, A, B> {type Output = Scale<T::Output, A, C>;#[inline]fn mul(self, other: Scale<T, B, C>) -> Self::Output {Scale::new(self.0 * other.0)}
}// scale0 + scale1
impl<T: Add, Src, Dst> Add for Scale<T, Src, Dst> {type Output = Scale<T::Output, Src, Dst>;#[inline]fn add(self, other: Scale<T, Src, Dst>) -> Self::Output {Scale::new(self.0 + other.0)}
}// scale0 - scale1
impl<T: Sub, Src, Dst> Sub for Scale<T, Src, Dst> {type Output = Scale<T::Output, Src, Dst>;#[inline]fn sub(self, other: Scale<T, Src, Dst>) -> Self::Output {Scale::new(self.0 - other.0)}
}// FIXME: Switch to `derive(PartialEq, Clone)` after this Rust issue is fixed:
// https://github.com/rust-lang/rust/issues/26925impl<T: PartialEq, Src, Dst> PartialEq for Scale<T, Src, Dst> {fn eq(&self, other: &Scale<T, Src, Dst>) -> bool {self.0 == other.0}
}impl<T: Eq, Src, Dst> Eq for Scale<T, Src, Dst> {}impl<T: PartialOrd, Src, Dst> PartialOrd for Scale<T, Src, Dst> {fn partial_cmp(&self, other: &Self) -> Option<Ordering> {self.0.partial_cmp(&other.0)}
}impl<T: Ord, Src, Dst> Ord for Scale<T, Src, Dst> {fn cmp(&self, other: &Self) -> Ordering {self.0.cmp(&other.0)}
}impl<T: Clone, Src, Dst> Clone for Scale<T, Src, Dst> {fn clone(&self) -> Scale<T, Src, Dst> {Scale::new(self.0.clone())}
}impl<T: Copy, Src, Dst> Copy for Scale<T, Src, Dst> {}impl<T: fmt::Debug, Src, Dst> fmt::Debug for Scale<T, Src, Dst> {fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {self.0.fmt(f)}
}impl<T: Default, Src, Dst> Default for Scale<T, Src, Dst> {fn default() -> Self {Self::new(T::default())}
}impl<T: Hash, Src, Dst> Hash for Scale<T, Src, Dst> {fn hash<H: Hasher>(&self, state: &mut H) {self.0.hash(state);}
}impl<T: One, Src, Dst> One for Scale<T, Src, Dst> {#[inline]fn one() -> Self {Scale::new(T::one())}
}#[cfg(test)]
mod tests {use super::Scale;enum Inch {}enum Cm {}enum Mm {}#[test]fn test_scale() {let mm_per_inch: Scale<f32, Inch, Mm> = Scale::new(25.4);let cm_per_mm: Scale<f32, Mm, Cm> = Scale::new(0.1);let mm_per_cm: Scale<f32, Cm, Mm> = cm_per_mm.inverse();assert_eq!(mm_per_cm.get(), 10.0);let one: Scale<f32, Mm, Mm> = cm_per_mm * mm_per_cm;assert_eq!(one.get(), 1.0);let one: Scale<f32, Cm, Cm> = mm_per_cm * cm_per_mm;assert_eq!(one.get(), 1.0);let cm_per_inch: Scale<f32, Inch, Cm> = mm_per_inch * cm_per_mm;//  mm     cm     cm// ---- x ---- = ----// inch    mm    inchassert_eq!(cm_per_inch, Scale::new(2.54));let a: Scale<isize, Inch, Inch> = Scale::new(2);let b: Scale<isize, Inch, Inch> = Scale::new(3);assert_ne!(a, b);assert_eq!(a, a.clone());assert_eq!(a.clone() + b.clone(), Scale::new(5));assert_eq!(a - b, Scale::new(-1));// Clampassert_eq!(Scale::identity().clamp(a, b), a);assert_eq!(Scale::new(5).clamp(a, b), b);let a = Scale::<f32, Inch, Inch>::new(2.0);let b = Scale::<f32, Inch, Inch>::new(3.0);let c = Scale::<f32, Inch, Inch>::new(2.5);assert_eq!(c.clamp(a, b), c);}
}

二、结构体定义

#[repr(C)]
#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
#[cfg_attr(feature = "serde",serde(bound(serialize = "T: serde::Serialize",deserialize = "T: serde::Deserialize<'de>"))
)]
pub struct Scale<T, Src, Dst>(pub T, #[doc(hidden)] pub PhantomData<(Src, Dst)>);

上面代码定义了一个使用Rust语言特性的泛型结构体Scale，这个结构体不仅使用了Rust的条件编译指令，还考虑了序列化/反序列化的能力，如果项目中启用了serde特性。下面是对这段代码的详细解释：

#[repr( C )]：这个属性指定了结构体应该使用C语言的布局规则。这意味着结构体的内存布局将与C语言中的等效结构相匹配，这对于与C语言代码交互或与需要特定内存布局的系统API交互非常有用。
#[cfg_attr(feature = “serde”, derive(Serialize, Deserialize))]：这是条件编译属性的一个示例。如果项目配置了serde特性（通常在Cargo.toml文件中通过features部分指定），那么将为这个结构体派生Serialize和Deserialize特性。这两个特性来自serde库，允许结构体实例被序列化为JSON、RON等格式或从这些格式反序列化。
#[cfg_attr(… serde(bound(…)) )]：这部分进一步细化了serde特性的条件编译。它指定了当序列化或反序列化时，泛型参数T必须满足的条件。具体来说，如果正在序列化，T必须实现serde::Serialize；如果正在反序列化，T必须实现serde::Deserialize<'de>。这里，'de是一个生命周期参数，它是反序列化过程中使用的，表明反序列化操作与特定数据的生命周期相关联。
pub struct Scale<T, Src, Dst>(pub T, #[doc(hidden)] pub PhantomData<(Src, Dst)>);：这是Scale结构体的定义。它是一个泛型结构体，有三个类型参数：T、Src和Dst。结构体内部包含两个字段：

第一个字段是公开的（pub T），意味着它可以被外部访问。
第二个字段是PhantomData<(Src, Dst)>，它被标记为#[doc(hidden)]，意味着在生成的文档中这个字段将被隐藏。PhantomData是一个特殊的类型，用于在类型系统中表达所有权、借用或生命周期，但不占用任何运行时空间。在这里，它用于表示Scale结构体与Src和Dst类型有关联，尽管这两个类型在运行时并不实际存储任何数据。

总结：这段代码定义了一个条件编译支持序列化/反序列化，同时遵循C语言内存布局规则的泛型结构体，用于在Rust代码中表示某种形式的缩放或转换操作，其中T代表缩放或转换的值类型，而Src和Dst则用于在类型系统中表达源和目标类型的信息。

三、对应方法

方法源码

impl<T, Src, Dst> Scale<T, Src, Dst> {#[inline]pub const fn new(x: T) -> Self {Scale(x, PhantomData)}/// 创建标识比例（1.0）#[inline]pub fn identity() -> Selfwhere T: One,{Scale::new(T::one())}/// Returns the given point transformed by this scale.////// # Example////// ```rust/// use euclid::{Scale, point2};/// enum Mm {};/// enum Cm {};////// let to_mm: Scale<i32, Cm, Mm> = Scale::new(10);////// assert_eq!(to_mm.transform_point(point2(42, -42)), point2(420, -420));/// ```#[inline]pub fn transform_point(self, point: Point2D<T, Src>) -> Point2D<T::Output, Dst>whereT: Copy + Mul,{Point2D::new(point.x * self.0, point.y * self.0)}/// Returns the given point transformed by this scale.#[inline]pub fn transform_point3d(self, point: Point3D<T, Src>) -> Point3D<T::Output, Dst>whereT: Copy + Mul,{Point3D::new(point.x * self.0, point.y * self.0, point.z * self.0)}/// Returns the given vector transformed by this scale.////// # Example////// ```rust/// use euclid::{Scale, vec2};/// enum Mm {};/// enum Cm {};////// let to_mm: Scale<i32, Cm, Mm> = Scale::new(10);////// assert_eq!(to_mm.transform_vector(vec2(42, -42)), vec2(420, -420));/// ```#[inline]pub fn transform_vector(self, vec: Vector2D<T, Src>) -> Vector2D<T::Output, Dst>whereT: Copy + Mul,{Vector2D::new(vec.x * self.0, vec.y * self.0)}/// Returns the given size transformed by this scale.////// # Example////// ```rust/// use euclid::{Scale, size2};/// enum Mm {};/// enum Cm {};////// let to_mm: Scale<i32, Cm, Mm> = Scale::new(10);////// assert_eq!(to_mm.transform_size(size2(42, -42)), size2(420, -420));/// ```#[inline]pub fn transform_size(self, size: Size2D<T, Src>) -> Size2D<T::Output, Dst>whereT: Copy + Mul,{Size2D::new(size.width * self.0, size.height * self.0)}/// Returns the given rect transformed by this scale.////// # Example////// ```rust/// use euclid::{Scale, rect};/// enum Mm {};/// enum Cm {};////// let to_mm: Scale<i32, Cm, Mm> = Scale::new(10);////// assert_eq!(to_mm.transform_rect(&rect(1, 2, 42, -42)), rect(10, 20, 420, -420));/// ```#[inline]pub fn transform_rect(self, rect: &Rect<T, Src>) -> Rect<T::Output, Dst>whereT: Copy + Mul,{Rect::new(self.transform_point(rect.origin),self.transform_size(rect.size),)}/// Returns the given box transformed by this scale.#[inline]pub fn transform_box2d(self, b: &Box2D<T, Src>) -> Box2D<T::Output, Dst>whereT: Copy + Mul,{Box2D {min: self.transform_point(b.min),max: self.transform_point(b.max),}}/// Returns the given box transformed by this scale.#[inline]pub fn transform_box3d(self, b: &Box3D<T, Src>) -> Box3D<T::Output, Dst>where T: Copy + Mul,{Box3D {min: self.transform_point3d(b.min),max: self.transform_point3d(b.max),}}/// Returns `true` if this scale has no effect.////// # Example////// ```rust/// use euclid::Scale;/// use euclid::num::One;/// enum Mm {};/// enum Cm {};////// let cm_per_mm: Scale<f32, Mm, Cm> = Scale::new(0.1);/// let mm_per_mm: Scale<f32, Mm, Mm> = Scale::new(1.0);////// assert_eq!(cm_per_mm.is_identity(), false);/// assert_eq!(mm_per_mm.is_identity(), true);/// assert_eq!(mm_per_mm, Scale::one());/// ```#[inline]pub fn is_identity(self) -> boolwhere T: PartialEq + One,{self.0 == T::one()}/// Returns the underlying scalar scale factor.#[inline]pub fn get(self) -> T {self.0}/// The inverse Scale (1.0 / self).////// # Example////// ```rust/// use euclid::Scale;/// enum Mm {};/// enum Cm {};////// let cm_per_mm: Scale<f32, Cm, Mm> = Scale::new(0.1);////// assert_eq!(cm_per_mm.inverse(), Scale::new(10.0));/// ```pub fn inverse(self) -> Scale<T::Output, Dst, Src>whereT: One + Div,{let one: T = One::one();Scale::new(one / self.0)}
}

2、new 方法：

这是一个常量函数，用于创建一个新的 Scale 实例。它接受一个类型为 T 的参数 x 作为缩放因子。

3、identity 方法：

用于创建一个表示没有缩放的 Scale 实例（即缩放因子为 1.0）。这需要 T 类型实现 One trait，该 trait 提供了一个获取类型中“1”的方法。

4、transform_point 方法：

用于将二维点（Point2D<T, Src>）根据缩放因子变换到新的单位系统（Dst）。这个方法要求 T 类型实现 Copy 和 Mul traits，以便可以复制和乘法操作。

5、transform_point3d 方法：

类似于 transform_point，但是用于三维点（Point3D<T, Src>）。

6、is_identity 方法：

检查当前的缩放因子是否为 1.0（即是否没有缩放）。这要求 T 类型实现 PartialEq 和 One traits。

7、get 方法：

返回缩放因子 x。

8、inverse 方法：

返回当前缩放的逆缩放，即将 1.0 / self.0 作为新的缩放因子。这要求 T 类型实现 One 和 Div traits。

9、min和max方法的实现：

min和max方法分别返回两个Scale实例之间较小或较大的那个。它们使用内部存储的值（self.0）来调用标准库中的min和max函数。

10、clamp方法的实现：

clamp方法将当前Scale实例的每个分量限制在start和end指定的范围内。这是通过先对start使用max方法，然后对end使用min方法来实现的，从而确保结果位于start和end之间。

11、NumCast trait相关的实现：

这部分代码被注释掉了，似乎原本打算为Scale实现一个与数值类型转换相关的功能，允许从一种数值表示转换到另一种，但具体的实现被省略了。

12、partial_cmp、cmp方法的实现：

partial_cmp方法提供了一个可选的比较结果，这在类型T只能部分排序时非常有用（例如，浮点数与NaN的比较）。
cmp方法提供了一个确定性的比较结果，要求T必须完全可排序（实现Ord trait）。

13、Clone、Copy trait的实现：

为Scale实现了Clone和Copy trait，这取决于T是否也实现了这些trait。如果T可以克隆或复制，那么Scale<T, Src, Dst>也可以。

14、Debug trait的实现：

为Scale实现了Debug trait，允许其实例使用{:?}格式化时打印调试信息。这依赖于T也实现了Debug trait。

15、Default trait的实现：

为Scale实现了Default trait，提供了一个默认构造函数，它依赖于T也实现了Default trait。

16、Hash trait的实现：

为Scale实现了Hash trait，允许其实例被哈希。这依赖于T也实现了Hash trait。

17、One trait的实现：

为Scale实现了One trait，提供了一个单位元素的构造函数（即值为1的Scale实例）。这依赖于T也实现了One trait。