diff options
Diffstat (limited to 'third_party/rust/euclid/src/rect.rs')
-rw-r--r-- | third_party/rust/euclid/src/rect.rs | 931 |
1 files changed, 931 insertions, 0 deletions
diff --git a/third_party/rust/euclid/src/rect.rs b/third_party/rust/euclid/src/rect.rs new file mode 100644 index 0000000000..04721db811 --- /dev/null +++ b/third_party/rust/euclid/src/rect.rs @@ -0,0 +1,931 @@ +// 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 +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +use super::UnknownUnit; +use crate::box2d::Box2D; +use crate::num::*; +use crate::point::Point2D; +use crate::scale::Scale; +use crate::side_offsets::SideOffsets2D; +use crate::size::Size2D; +use crate::vector::Vector2D; + +use num_traits::{NumCast, Float}; +#[cfg(feature = "serde")] +use serde::{Deserialize, Serialize}; +#[cfg(feature = "bytemuck")] +use bytemuck::{Zeroable, Pod}; + +use core::borrow::Borrow; +use core::cmp::PartialOrd; +use core::fmt; +use core::hash::{Hash, Hasher}; +use core::ops::{Add, Div, DivAssign, Mul, MulAssign, Range, Sub}; + +/// A 2d Rectangle optionally tagged with a unit. +/// +/// # Representation +/// +/// `Rect` is represented by an origin point and a size. +/// +/// See [`Box2D`] for a rectangle represented by two endpoints. +/// +/// # Empty rectangle +/// +/// A rectangle is considered empty (see [`is_empty`]) if any of the following is true: +/// - it's area is empty, +/// - it's area is negative (`size.x < 0` or `size.y < 0`), +/// - it contains NaNs. +/// +/// [`is_empty`]: #method.is_empty +/// [`Box2D`]: struct.Box2D.html +#[repr(C)] +#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))] +#[cfg_attr( + feature = "serde", + serde(bound(serialize = "T: Serialize", deserialize = "T: Deserialize<'de>")) +)] +pub struct Rect<T, U> { + pub origin: Point2D<T, U>, + pub size: Size2D<T, U>, +} + +#[cfg(feature = "arbitrary")] +impl<'a, T, U> arbitrary::Arbitrary<'a> for Rect<T, U> +where + T: arbitrary::Arbitrary<'a>, +{ + fn arbitrary(u: &mut arbitrary::Unstructured<'a>) -> arbitrary::Result<Self> + { + let (origin, size) = arbitrary::Arbitrary::arbitrary(u)?; + Ok(Rect { + origin, + size, + }) + } +} + +#[cfg(feature = "bytemuck")] +unsafe impl<T: Zeroable, U> Zeroable for Rect<T, U> {} + +#[cfg(feature = "bytemuck")] +unsafe impl<T: Pod, U: 'static> Pod for Rect<T, U> {} + +impl<T: Hash, U> Hash for Rect<T, U> { + fn hash<H: Hasher>(&self, h: &mut H) { + self.origin.hash(h); + self.size.hash(h); + } +} + +impl<T: Copy, U> Copy for Rect<T, U> {} + +impl<T: Clone, U> Clone for Rect<T, U> { + fn clone(&self) -> Self { + Self::new(self.origin.clone(), self.size.clone()) + } +} + +impl<T: PartialEq, U> PartialEq for Rect<T, U> { + fn eq(&self, other: &Self) -> bool { + self.origin.eq(&other.origin) && self.size.eq(&other.size) + } +} + +impl<T: Eq, U> Eq for Rect<T, U> {} + +impl<T: fmt::Debug, U> fmt::Debug for Rect<T, U> { + fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { + write!(f, "Rect(")?; + fmt::Debug::fmt(&self.size, f)?; + write!(f, " at ")?; + fmt::Debug::fmt(&self.origin, f)?; + write!(f, ")") + } +} + +impl<T: Default, U> Default for Rect<T, U> { + fn default() -> Self { + Rect::new(Default::default(), Default::default()) + } +} + +impl<T, U> Rect<T, U> { + /// Constructor. + #[inline] + pub const fn new(origin: Point2D<T, U>, size: Size2D<T, U>) -> Self { + Rect { origin, size } + } +} + +impl<T, U> Rect<T, U> +where + T: Zero, +{ + /// Constructor, setting all sides to zero. + #[inline] + pub fn zero() -> Self { + Rect::new(Point2D::origin(), Size2D::zero()) + } + + /// Creates a rect of the given size, at offset zero. + #[inline] + pub fn from_size(size: Size2D<T, U>) -> Self { + Rect { + origin: Point2D::zero(), + size, + } + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + Add<T, Output = T>, +{ + #[inline] + pub fn min(&self) -> Point2D<T, U> { + self.origin + } + + #[inline] + pub fn max(&self) -> Point2D<T, U> { + self.origin + self.size + } + + #[inline] + pub fn max_x(&self) -> T { + self.origin.x + self.size.width + } + + #[inline] + pub fn min_x(&self) -> T { + self.origin.x + } + + #[inline] + pub fn max_y(&self) -> T { + self.origin.y + self.size.height + } + + #[inline] + pub fn min_y(&self) -> T { + self.origin.y + } + + #[inline] + pub fn width(&self) -> T { + self.size.width + } + + #[inline] + pub fn height(&self) -> T { + self.size.height + } + + #[inline] + pub fn x_range(&self) -> Range<T> { + self.min_x()..self.max_x() + } + + #[inline] + pub fn y_range(&self) -> Range<T> { + self.min_y()..self.max_y() + } + + /// Returns the same rectangle, translated by a vector. + #[inline] + #[must_use] + pub fn translate(&self, by: Vector2D<T, U>) -> Self { + Self::new(self.origin + by, self.size) + } + + #[inline] + pub fn to_box2d(&self) -> Box2D<T, U> { + Box2D { + min: self.min(), + max: self.max(), + } + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + PartialOrd + Add<T, Output = T>, +{ + /// Returns true if this rectangle contains the point. Points are considered + /// in the rectangle if they are on the left or top edge, but outside if they + /// are on the right or bottom edge. + #[inline] + pub fn contains(&self, p: Point2D<T, U>) -> bool { + self.to_box2d().contains(p) + } + + #[inline] + pub fn intersects(&self, other: &Self) -> bool { + self.to_box2d().intersects(&other.to_box2d()) + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + PartialOrd + Add<T, Output = T> + Sub<T, Output = T>, +{ + #[inline] + pub fn intersection(&self, other: &Self) -> Option<Self> { + let box2d = self.to_box2d().intersection_unchecked(&other.to_box2d()); + + if box2d.is_empty() { + return None; + } + + Some(box2d.to_rect()) + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + Add<T, Output = T> + Sub<T, Output = T>, +{ + #[inline] + #[must_use] + pub fn inflate(&self, width: T, height: T) -> Self { + Rect::new( + Point2D::new(self.origin.x - width, self.origin.y - height), + Size2D::new( + self.size.width + width + width, + self.size.height + height + height, + ), + ) + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + Zero + PartialOrd + Add<T, Output = T>, +{ + /// Returns true if this rectangle contains the interior of rect. Always + /// returns true if rect is empty, and always returns false if rect is + /// nonempty but this rectangle is empty. + #[inline] + pub fn contains_rect(&self, rect: &Self) -> bool { + rect.is_empty() + || (self.min_x() <= rect.min_x() + && rect.max_x() <= self.max_x() + && self.min_y() <= rect.min_y() + && rect.max_y() <= self.max_y()) + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + Zero + PartialOrd + Add<T, Output = T> + Sub<T, Output = T>, +{ + /// Calculate the size and position of an inner rectangle. + /// + /// Subtracts the side offsets from all sides. The horizontal and vertical + /// offsets must not be larger than the original side length. + /// This method assumes y oriented downward. + pub fn inner_rect(&self, offsets: SideOffsets2D<T, U>) -> Self { + let rect = Rect::new( + Point2D::new(self.origin.x + offsets.left, self.origin.y + offsets.top), + Size2D::new( + self.size.width - offsets.horizontal(), + self.size.height - offsets.vertical(), + ), + ); + debug_assert!(rect.size.width >= Zero::zero()); + debug_assert!(rect.size.height >= Zero::zero()); + rect + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + Add<T, Output = T> + Sub<T, Output = T>, +{ + /// Calculate the size and position of an outer rectangle. + /// + /// Add the offsets to all sides. The expanded rectangle is returned. + /// This method assumes y oriented downward. + pub fn outer_rect(&self, offsets: SideOffsets2D<T, U>) -> Self { + Rect::new( + Point2D::new(self.origin.x - offsets.left, self.origin.y - offsets.top), + Size2D::new( + self.size.width + offsets.horizontal(), + self.size.height + offsets.vertical(), + ), + ) + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + Zero + PartialOrd + Sub<T, Output = T>, +{ + /// Returns the smallest rectangle defined by the top/bottom/left/right-most + /// points provided as parameter. + /// + /// Note: This function has a behavior that can be surprising because + /// the right-most and bottom-most points are exactly on the edge + /// of the rectangle while the `contains` function is has exclusive + /// semantic on these edges. This means that the right-most and bottom-most + /// points provided to `from_points` will count as not contained by the rect. + /// This behavior may change in the future. + pub fn from_points<I>(points: I) -> Self + where + I: IntoIterator, + I::Item: Borrow<Point2D<T, U>>, + { + Box2D::from_points(points).to_rect() + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + One + Add<Output = T> + Sub<Output = T> + Mul<Output = T>, +{ + /// Linearly interpolate between this rectangle and another rectangle. + #[inline] + pub fn lerp(&self, other: Self, t: T) -> Self { + Self::new( + self.origin.lerp(other.origin, t), + self.size.lerp(other.size, t), + ) + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + One + Add<Output = T> + Div<Output = T>, +{ + pub fn center(&self) -> Point2D<T, U> { + let two = T::one() + T::one(); + self.origin + self.size.to_vector() / two + } +} + +impl<T, U> Rect<T, U> +where + T: Copy + PartialOrd + Add<T, Output = T> + Sub<T, Output = T> + Zero, +{ + #[inline] + pub fn union(&self, other: &Self) -> Self { + self.to_box2d().union(&other.to_box2d()).to_rect() + } +} + +impl<T, U> Rect<T, U> { + #[inline] + pub fn scale<S: Copy>(&self, x: S, y: S) -> Self + where + T: Copy + Mul<S, Output = T>, + { + Rect::new( + Point2D::new(self.origin.x * x, self.origin.y * y), + Size2D::new(self.size.width * x, self.size.height * y), + ) + } +} + +impl<T: Copy + Mul<T, Output = T>, U> Rect<T, U> { + #[inline] + pub fn area(&self) -> T { + self.size.area() + } +} + +impl<T: Copy + Zero + PartialOrd, U> Rect<T, U> { + #[inline] + pub fn is_empty(&self) -> bool { + self.size.is_empty() + } +} + +impl<T: Copy + Zero + PartialOrd, U> Rect<T, U> { + #[inline] + pub fn to_non_empty(&self) -> Option<Self> { + if self.is_empty() { + return None; + } + + Some(*self) + } +} + +impl<T: Copy + Mul, U> Mul<T> for Rect<T, U> { + type Output = Rect<T::Output, U>; + + #[inline] + fn mul(self, scale: T) -> Self::Output { + Rect::new(self.origin * scale, self.size * scale) + } +} + +impl<T: Copy + MulAssign, U> MulAssign<T> for Rect<T, U> { + #[inline] + fn mul_assign(&mut self, scale: T) { + *self *= Scale::new(scale); + } +} + +impl<T: Copy + Div, U> Div<T> for Rect<T, U> { + type Output = Rect<T::Output, U>; + + #[inline] + fn div(self, scale: T) -> Self::Output { + Rect::new(self.origin / scale.clone(), self.size / scale) + } +} + +impl<T: Copy + DivAssign, U> DivAssign<T> for Rect<T, U> { + #[inline] + fn div_assign(&mut self, scale: T) { + *self /= Scale::new(scale); + } +} + +impl<T: Copy + Mul, U1, U2> Mul<Scale<T, U1, U2>> for Rect<T, U1> { + type Output = Rect<T::Output, U2>; + + #[inline] + fn mul(self, scale: Scale<T, U1, U2>) -> Self::Output { + Rect::new(self.origin * scale.clone(), self.size * scale) + } +} + +impl<T: Copy + MulAssign, U> MulAssign<Scale<T, U, U>> for Rect<T, U> { + #[inline] + fn mul_assign(&mut self, scale: Scale<T, U, U>) { + self.origin *= scale.clone(); + self.size *= scale; + } +} + +impl<T: Copy + Div, U1, U2> Div<Scale<T, U1, U2>> for Rect<T, U2> { + type Output = Rect<T::Output, U1>; + + #[inline] + fn div(self, scale: Scale<T, U1, U2>) -> Self::Output { + Rect::new(self.origin / scale.clone(), self.size / scale) + } +} + +impl<T: Copy + DivAssign, U> DivAssign<Scale<T, U, U>> for Rect<T, U> { + #[inline] + fn div_assign(&mut self, scale: Scale<T, U, U>) { + self.origin /= scale.clone(); + self.size /= scale; + } +} + +impl<T: Copy, U> Rect<T, U> { + /// Drop the units, preserving only the numeric value. + #[inline] + pub fn to_untyped(&self) -> Rect<T, UnknownUnit> { + Rect::new(self.origin.to_untyped(), self.size.to_untyped()) + } + + /// Tag a unitless value with units. + #[inline] + pub fn from_untyped(r: &Rect<T, UnknownUnit>) -> Rect<T, U> { + Rect::new( + Point2D::from_untyped(r.origin), + Size2D::from_untyped(r.size), + ) + } + + /// Cast the unit + #[inline] + pub fn cast_unit<V>(&self) -> Rect<T, V> { + Rect::new(self.origin.cast_unit(), self.size.cast_unit()) + } +} + +impl<T: NumCast + Copy, U> Rect<T, U> { + /// Cast from one numeric representation to another, preserving the units. + /// + /// When casting from floating point to integer coordinates, the decimals are truncated + /// as one would expect from a simple cast, but this behavior does not always make sense + /// geometrically. Consider using round(), round_in or round_out() before casting. + #[inline] + pub fn cast<NewT: NumCast>(&self) -> Rect<NewT, U> { + Rect::new(self.origin.cast(), self.size.cast()) + } + + /// Fallible cast from one numeric representation to another, preserving the units. + /// + /// When casting from floating point to integer coordinates, the decimals are truncated + /// as one would expect from a simple cast, but this behavior does not always make sense + /// geometrically. Consider using round(), round_in or round_out() before casting. + pub fn try_cast<NewT: NumCast>(&self) -> Option<Rect<NewT, U>> { + match (self.origin.try_cast(), self.size.try_cast()) { + (Some(origin), Some(size)) => Some(Rect::new(origin, size)), + _ => None, + } + } + + // Convenience functions for common casts + + /// Cast into an `f32` rectangle. + #[inline] + pub fn to_f32(&self) -> Rect<f32, U> { + self.cast() + } + + /// Cast into an `f64` rectangle. + #[inline] + pub fn to_f64(&self) -> Rect<f64, U> { + self.cast() + } + + /// Cast into an `usize` rectangle, truncating decimals if any. + /// + /// When casting from floating point rectangles, it is worth considering whether + /// to `round()`, `round_in()` or `round_out()` before the cast in order to + /// obtain the desired conversion behavior. + #[inline] + pub fn to_usize(&self) -> Rect<usize, U> { + self.cast() + } + + /// Cast into an `u32` rectangle, truncating decimals if any. + /// + /// When casting from floating point rectangles, it is worth considering whether + /// to `round()`, `round_in()` or `round_out()` before the cast in order to + /// obtain the desired conversion behavior. + #[inline] + pub fn to_u32(&self) -> Rect<u32, U> { + self.cast() + } + + /// Cast into an `u64` rectangle, truncating decimals if any. + /// + /// When casting from floating point rectangles, it is worth considering whether + /// to `round()`, `round_in()` or `round_out()` before the cast in order to + /// obtain the desired conversion behavior. + #[inline] + pub fn to_u64(&self) -> Rect<u64, U> { + self.cast() + } + + /// Cast into an `i32` rectangle, truncating decimals if any. + /// + /// When casting from floating point rectangles, it is worth considering whether + /// to `round()`, `round_in()` or `round_out()` before the cast in order to + /// obtain the desired conversion behavior. + #[inline] + pub fn to_i32(&self) -> Rect<i32, U> { + self.cast() + } + + /// Cast into an `i64` rectangle, truncating decimals if any. + /// + /// When casting from floating point rectangles, it is worth considering whether + /// to `round()`, `round_in()` or `round_out()` before the cast in order to + /// obtain the desired conversion behavior. + #[inline] + pub fn to_i64(&self) -> Rect<i64, U> { + self.cast() + } +} + +impl<T: Float, U> Rect<T, U> { + /// Returns true if all members are finite. + #[inline] + pub fn is_finite(self) -> bool { + self.origin.is_finite() && self.size.is_finite() + } +} + +impl<T: Floor + Ceil + Round + Add<T, Output = T> + Sub<T, Output = T>, U> Rect<T, U> { + /// Return a rectangle with edges rounded to integer coordinates, such that + /// the returned rectangle has the same set of pixel centers as the original + /// one. + /// Edges at offset 0.5 round up. + /// Suitable for most places where integral device coordinates + /// are needed, but note that any translation should be applied first to + /// avoid pixel rounding errors. + /// Note that this is *not* rounding to nearest integer if the values are negative. + /// They are always rounding as floor(n + 0.5). + /// + /// # Usage notes + /// Note, that when using with floating-point `T` types that method can significantly + /// loose precision for large values, so if you need to call this method very often it + /// is better to use [`Box2D`]. + /// + /// [`Box2D`]: struct.Box2D.html + #[must_use] + pub fn round(&self) -> Self { + self.to_box2d().round().to_rect() + } + + /// Return a rectangle with edges rounded to integer coordinates, such that + /// the original rectangle contains the resulting rectangle. + /// + /// # Usage notes + /// Note, that when using with floating-point `T` types that method can significantly + /// loose precision for large values, so if you need to call this method very often it + /// is better to use [`Box2D`]. + /// + /// [`Box2D`]: struct.Box2D.html + #[must_use] + pub fn round_in(&self) -> Self { + self.to_box2d().round_in().to_rect() + } + + /// Return a rectangle with edges rounded to integer coordinates, such that + /// the original rectangle is contained in the resulting rectangle. + /// + /// # Usage notes + /// Note, that when using with floating-point `T` types that method can significantly + /// loose precision for large values, so if you need to call this method very often it + /// is better to use [`Box2D`]. + /// + /// [`Box2D`]: struct.Box2D.html + #[must_use] + pub fn round_out(&self) -> Self { + self.to_box2d().round_out().to_rect() + } +} + +impl<T, U> From<Size2D<T, U>> for Rect<T, U> +where + T: Zero, +{ + fn from(size: Size2D<T, U>) -> Self { + Self::from_size(size) + } +} + +/// Shorthand for `Rect::new(Point2D::new(x, y), Size2D::new(w, h))`. +pub const fn rect<T, U>(x: T, y: T, w: T, h: T) -> Rect<T, U> { + Rect::new(Point2D::new(x, y), Size2D::new(w, h)) +} + +#[cfg(test)] +mod tests { + use crate::default::{Point2D, Rect, Size2D}; + use crate::side_offsets::SideOffsets2D; + use crate::{point2, rect, size2, vec2}; + + #[test] + fn test_translate() { + let p = Rect::new(Point2D::new(0u32, 0u32), Size2D::new(50u32, 40u32)); + let pp = p.translate(vec2(10, 15)); + + assert!(pp.size.width == 50); + assert!(pp.size.height == 40); + assert!(pp.origin.x == 10); + assert!(pp.origin.y == 15); + + let r = Rect::new(Point2D::new(-10, -5), Size2D::new(50, 40)); + let rr = r.translate(vec2(0, -10)); + + assert!(rr.size.width == 50); + assert!(rr.size.height == 40); + assert!(rr.origin.x == -10); + assert!(rr.origin.y == -15); + } + + #[test] + fn test_union() { + let p = Rect::new(Point2D::new(0, 0), Size2D::new(50, 40)); + let q = Rect::new(Point2D::new(20, 20), Size2D::new(5, 5)); + let r = Rect::new(Point2D::new(-15, -30), Size2D::new(200, 15)); + let s = Rect::new(Point2D::new(20, -15), Size2D::new(250, 200)); + + let pq = p.union(&q); + assert!(pq.origin == Point2D::new(0, 0)); + assert!(pq.size == Size2D::new(50, 40)); + + let pr = p.union(&r); + assert!(pr.origin == Point2D::new(-15, -30)); + assert!(pr.size == Size2D::new(200, 70)); + + let ps = p.union(&s); + assert!(ps.origin == Point2D::new(0, -15)); + assert!(ps.size == Size2D::new(270, 200)); + } + + #[test] + fn test_intersection() { + let p = Rect::new(Point2D::new(0, 0), Size2D::new(10, 20)); + let q = Rect::new(Point2D::new(5, 15), Size2D::new(10, 10)); + let r = Rect::new(Point2D::new(-5, -5), Size2D::new(8, 8)); + + let pq = p.intersection(&q); + assert!(pq.is_some()); + let pq = pq.unwrap(); + assert!(pq.origin == Point2D::new(5, 15)); + assert!(pq.size == Size2D::new(5, 5)); + + let pr = p.intersection(&r); + assert!(pr.is_some()); + let pr = pr.unwrap(); + assert!(pr.origin == Point2D::new(0, 0)); + assert!(pr.size == Size2D::new(3, 3)); + + let qr = q.intersection(&r); + assert!(qr.is_none()); + } + + #[test] + fn test_intersection_overflow() { + // test some scenarios where the intersection can overflow but + // the min_x() and max_x() don't. Gecko currently fails these cases + let p = Rect::new(Point2D::new(-2147483648, -2147483648), Size2D::new(0, 0)); + let q = Rect::new( + Point2D::new(2136893440, 2136893440), + Size2D::new(279552, 279552), + ); + let r = Rect::new(Point2D::new(-2147483648, -2147483648), Size2D::new(1, 1)); + + assert!(p.is_empty()); + let pq = p.intersection(&q); + assert!(pq.is_none()); + + let qr = q.intersection(&r); + assert!(qr.is_none()); + } + + #[test] + fn test_contains() { + let r = Rect::new(Point2D::new(-20, 15), Size2D::new(100, 200)); + + assert!(r.contains(Point2D::new(0, 50))); + assert!(r.contains(Point2D::new(-10, 200))); + + // The `contains` method is inclusive of the top/left edges, but not the + // bottom/right edges. + assert!(r.contains(Point2D::new(-20, 15))); + assert!(!r.contains(Point2D::new(80, 15))); + assert!(!r.contains(Point2D::new(80, 215))); + assert!(!r.contains(Point2D::new(-20, 215))); + + // Points beyond the top-left corner. + assert!(!r.contains(Point2D::new(-25, 15))); + assert!(!r.contains(Point2D::new(-15, 10))); + + // Points beyond the top-right corner. + assert!(!r.contains(Point2D::new(85, 20))); + assert!(!r.contains(Point2D::new(75, 10))); + + // Points beyond the bottom-right corner. + assert!(!r.contains(Point2D::new(85, 210))); + assert!(!r.contains(Point2D::new(75, 220))); + + // Points beyond the bottom-left corner. + assert!(!r.contains(Point2D::new(-25, 210))); + assert!(!r.contains(Point2D::new(-15, 220))); + + let r = Rect::new(Point2D::new(-20.0, 15.0), Size2D::new(100.0, 200.0)); + assert!(r.contains_rect(&r)); + assert!(!r.contains_rect(&r.translate(vec2(0.1, 0.0)))); + assert!(!r.contains_rect(&r.translate(vec2(-0.1, 0.0)))); + assert!(!r.contains_rect(&r.translate(vec2(0.0, 0.1)))); + assert!(!r.contains_rect(&r.translate(vec2(0.0, -0.1)))); + // Empty rectangles are always considered as contained in other rectangles, + // even if their origin is not. + let p = Point2D::new(1.0, 1.0); + assert!(!r.contains(p)); + assert!(r.contains_rect(&Rect::new(p, Size2D::zero()))); + } + + #[test] + fn test_scale() { + let p = Rect::new(Point2D::new(0u32, 0u32), Size2D::new(50u32, 40u32)); + let pp = p.scale(10, 15); + + assert!(pp.size.width == 500); + assert!(pp.size.height == 600); + assert!(pp.origin.x == 0); + assert!(pp.origin.y == 0); + + let r = Rect::new(Point2D::new(-10, -5), Size2D::new(50, 40)); + let rr = r.scale(1, 20); + + assert!(rr.size.width == 50); + assert!(rr.size.height == 800); + assert!(rr.origin.x == -10); + assert!(rr.origin.y == -100); + } + + #[test] + fn test_inflate() { + let p = Rect::new(Point2D::new(0, 0), Size2D::new(10, 10)); + let pp = p.inflate(10, 20); + + assert!(pp.size.width == 30); + assert!(pp.size.height == 50); + assert!(pp.origin.x == -10); + assert!(pp.origin.y == -20); + + let r = Rect::new(Point2D::new(0, 0), Size2D::new(10, 20)); + let rr = r.inflate(-2, -5); + + assert!(rr.size.width == 6); + assert!(rr.size.height == 10); + assert!(rr.origin.x == 2); + assert!(rr.origin.y == 5); + } + + #[test] + fn test_inner_outer_rect() { + let inner_rect = Rect::new(point2(20, 40), size2(80, 100)); + let offsets = SideOffsets2D::new(20, 10, 10, 10); + let outer_rect = inner_rect.outer_rect(offsets); + assert_eq!(outer_rect.origin.x, 10); + assert_eq!(outer_rect.origin.y, 20); + assert_eq!(outer_rect.size.width, 100); + assert_eq!(outer_rect.size.height, 130); + assert_eq!(outer_rect.inner_rect(offsets), inner_rect); + } + + #[test] + fn test_min_max_x_y() { + let p = Rect::new(Point2D::new(0u32, 0u32), Size2D::new(50u32, 40u32)); + assert!(p.max_y() == 40); + assert!(p.min_y() == 0); + assert!(p.max_x() == 50); + assert!(p.min_x() == 0); + + let r = Rect::new(Point2D::new(-10, -5), Size2D::new(50, 40)); + assert!(r.max_y() == 35); + assert!(r.min_y() == -5); + assert!(r.max_x() == 40); + assert!(r.min_x() == -10); + } + + #[test] + fn test_width_height() { + let r = Rect::new(Point2D::new(-10, -5), Size2D::new(50, 40)); + assert!(r.width() == 50); + assert!(r.height() == 40); + } + + #[test] + fn test_is_empty() { + assert!(Rect::new(Point2D::new(0u32, 0u32), Size2D::new(0u32, 0u32)).is_empty()); + assert!(Rect::new(Point2D::new(0u32, 0u32), Size2D::new(10u32, 0u32)).is_empty()); + assert!(Rect::new(Point2D::new(0u32, 0u32), Size2D::new(0u32, 10u32)).is_empty()); + assert!(!Rect::new(Point2D::new(0u32, 0u32), Size2D::new(1u32, 1u32)).is_empty()); + assert!(Rect::new(Point2D::new(10u32, 10u32), Size2D::new(0u32, 0u32)).is_empty()); + assert!(Rect::new(Point2D::new(10u32, 10u32), Size2D::new(10u32, 0u32)).is_empty()); + assert!(Rect::new(Point2D::new(10u32, 10u32), Size2D::new(0u32, 10u32)).is_empty()); + assert!(!Rect::new(Point2D::new(10u32, 10u32), Size2D::new(1u32, 1u32)).is_empty()); + } + + #[test] + fn test_round() { + let mut x = -2.0; + let mut y = -2.0; + let mut w = -2.0; + let mut h = -2.0; + while x < 2.0 { + while y < 2.0 { + while w < 2.0 { + while h < 2.0 { + let rect = Rect::new(Point2D::new(x, y), Size2D::new(w, h)); + + assert!(rect.contains_rect(&rect.round_in())); + assert!(rect.round_in().inflate(1.0, 1.0).contains_rect(&rect)); + + assert!(rect.round_out().contains_rect(&rect)); + assert!(rect.inflate(1.0, 1.0).contains_rect(&rect.round_out())); + + assert!(rect.inflate(1.0, 1.0).contains_rect(&rect.round())); + assert!(rect.round().inflate(1.0, 1.0).contains_rect(&rect)); + + h += 0.1; + } + w += 0.1; + } + y += 0.1; + } + x += 0.1 + } + } + + #[test] + fn test_center() { + let r: Rect<i32> = rect(-2, 5, 4, 10); + assert_eq!(r.center(), point2(0, 10)); + + let r: Rect<f32> = rect(1.0, 2.0, 3.0, 4.0); + assert_eq!(r.center(), point2(2.5, 4.0)); + } + + #[test] + fn test_nan() { + let r1: Rect<f32> = rect(-2.0, 5.0, 4.0, std::f32::NAN); + let r2: Rect<f32> = rect(std::f32::NAN, -1.0, 3.0, 10.0); + + assert_eq!(r1.intersection(&r2), None); + } +} |