diff options
Diffstat (limited to 'vendor/num-traits/src/ops/saturating.rs')
-rw-r--r-- | vendor/num-traits/src/ops/saturating.rs | 130 |
1 files changed, 130 insertions, 0 deletions
diff --git a/vendor/num-traits/src/ops/saturating.rs b/vendor/num-traits/src/ops/saturating.rs new file mode 100644 index 0000000..16a0045 --- /dev/null +++ b/vendor/num-traits/src/ops/saturating.rs @@ -0,0 +1,130 @@ +use core::ops::{Add, Mul, Sub}; + +/// Saturating math operations. Deprecated, use `SaturatingAdd`, `SaturatingSub` and +/// `SaturatingMul` instead. +pub trait Saturating { + /// Saturating addition operator. + /// Returns a+b, saturating at the numeric bounds instead of overflowing. + fn saturating_add(self, v: Self) -> Self; + + /// Saturating subtraction operator. + /// Returns a-b, saturating at the numeric bounds instead of overflowing. + fn saturating_sub(self, v: Self) -> Self; +} + +macro_rules! deprecated_saturating_impl { + ($trait_name:ident for $($t:ty)*) => {$( + impl $trait_name for $t { + #[inline] + fn saturating_add(self, v: Self) -> Self { + Self::saturating_add(self, v) + } + + #[inline] + fn saturating_sub(self, v: Self) -> Self { + Self::saturating_sub(self, v) + } + } + )*} +} + +deprecated_saturating_impl!(Saturating for isize i8 i16 i32 i64 i128); +deprecated_saturating_impl!(Saturating for usize u8 u16 u32 u64 u128); + +macro_rules! saturating_impl { + ($trait_name:ident, $method:ident, $t:ty) => { + impl $trait_name for $t { + #[inline] + fn $method(&self, v: &Self) -> Self { + <$t>::$method(*self, *v) + } + } + }; +} + +/// Performs addition that saturates at the numeric bounds instead of overflowing. +pub trait SaturatingAdd: Sized + Add<Self, Output = Self> { + /// Saturating addition. Computes `self + other`, saturating at the relevant high or low boundary of + /// the type. + fn saturating_add(&self, v: &Self) -> Self; +} + +saturating_impl!(SaturatingAdd, saturating_add, u8); +saturating_impl!(SaturatingAdd, saturating_add, u16); +saturating_impl!(SaturatingAdd, saturating_add, u32); +saturating_impl!(SaturatingAdd, saturating_add, u64); +saturating_impl!(SaturatingAdd, saturating_add, usize); +saturating_impl!(SaturatingAdd, saturating_add, u128); + +saturating_impl!(SaturatingAdd, saturating_add, i8); +saturating_impl!(SaturatingAdd, saturating_add, i16); +saturating_impl!(SaturatingAdd, saturating_add, i32); +saturating_impl!(SaturatingAdd, saturating_add, i64); +saturating_impl!(SaturatingAdd, saturating_add, isize); +saturating_impl!(SaturatingAdd, saturating_add, i128); + +/// Performs subtraction that saturates at the numeric bounds instead of overflowing. +pub trait SaturatingSub: Sized + Sub<Self, Output = Self> { + /// Saturating subtraction. Computes `self - other`, saturating at the relevant high or low boundary of + /// the type. + fn saturating_sub(&self, v: &Self) -> Self; +} + +saturating_impl!(SaturatingSub, saturating_sub, u8); +saturating_impl!(SaturatingSub, saturating_sub, u16); +saturating_impl!(SaturatingSub, saturating_sub, u32); +saturating_impl!(SaturatingSub, saturating_sub, u64); +saturating_impl!(SaturatingSub, saturating_sub, usize); +saturating_impl!(SaturatingSub, saturating_sub, u128); + +saturating_impl!(SaturatingSub, saturating_sub, i8); +saturating_impl!(SaturatingSub, saturating_sub, i16); +saturating_impl!(SaturatingSub, saturating_sub, i32); +saturating_impl!(SaturatingSub, saturating_sub, i64); +saturating_impl!(SaturatingSub, saturating_sub, isize); +saturating_impl!(SaturatingSub, saturating_sub, i128); + +/// Performs multiplication that saturates at the numeric bounds instead of overflowing. +pub trait SaturatingMul: Sized + Mul<Self, Output = Self> { + /// Saturating multiplication. Computes `self * other`, saturating at the relevant high or low boundary of + /// the type. + fn saturating_mul(&self, v: &Self) -> Self; +} + +saturating_impl!(SaturatingMul, saturating_mul, u8); +saturating_impl!(SaturatingMul, saturating_mul, u16); +saturating_impl!(SaturatingMul, saturating_mul, u32); +saturating_impl!(SaturatingMul, saturating_mul, u64); +saturating_impl!(SaturatingMul, saturating_mul, usize); +saturating_impl!(SaturatingMul, saturating_mul, u128); + +saturating_impl!(SaturatingMul, saturating_mul, i8); +saturating_impl!(SaturatingMul, saturating_mul, i16); +saturating_impl!(SaturatingMul, saturating_mul, i32); +saturating_impl!(SaturatingMul, saturating_mul, i64); +saturating_impl!(SaturatingMul, saturating_mul, isize); +saturating_impl!(SaturatingMul, saturating_mul, i128); + +// TODO: add SaturatingNeg for signed integer primitives once the saturating_neg() API is stable. + +#[test] +fn test_saturating_traits() { + fn saturating_add<T: SaturatingAdd>(a: T, b: T) -> T { + a.saturating_add(&b) + } + fn saturating_sub<T: SaturatingSub>(a: T, b: T) -> T { + a.saturating_sub(&b) + } + fn saturating_mul<T: SaturatingMul>(a: T, b: T) -> T { + a.saturating_mul(&b) + } + assert_eq!(saturating_add(255, 1), 255u8); + assert_eq!(saturating_add(127, 1), 127i8); + assert_eq!(saturating_add(-128, -1), -128i8); + assert_eq!(saturating_sub(0, 1), 0u8); + assert_eq!(saturating_sub(-128, 1), -128i8); + assert_eq!(saturating_sub(127, -1), 127i8); + assert_eq!(saturating_mul(255, 2), 255u8); + assert_eq!(saturating_mul(127, 2), 127i8); + assert_eq!(saturating_mul(-128, 2), -128i8); +} |