Source code
Revision control
Copy as Markdown
Other Tools
use core::ops::{Div, Rem};
pub trait Euclid: Sized + Div<Self, Output = Self> + Rem<Self, Output = Self> {
/// Calculates Euclidean division, the matching method for `rem_euclid`.
///
/// This computes the integer `n` such that
/// `self = n * v + self.rem_euclid(v)`.
/// In other words, the result is `self / v` rounded to the integer `n`
/// such that `self >= n * v`.
///
/// # Examples
///
/// ```
/// use num_traits::Euclid;
///
/// let a: i32 = 7;
/// let b: i32 = 4;
/// assert_eq!(Euclid::div_euclid(&a, &b), 1); // 7 > 4 * 1
/// assert_eq!(Euclid::div_euclid(&-a, &b), -2); // -7 >= 4 * -2
/// assert_eq!(Euclid::div_euclid(&a, &-b), -1); // 7 >= -4 * -1
/// assert_eq!(Euclid::div_euclid(&-a, &-b), 2); // -7 >= -4 * 2
/// ```
fn div_euclid(&self, v: &Self) -> Self;
/// Calculates the least nonnegative remainder of `self (mod v)`.
///
/// In particular, the return value `r` satisfies `0.0 <= r < v.abs()` in
/// most cases. However, due to a floating point round-off error it can
/// result in `r == v.abs()`, violating the mathematical definition, if
/// `self` is much smaller than `v.abs()` in magnitude and `self < 0.0`.
/// This result is not an element of the function's codomain, but it is the
/// closest floating point number in the real numbers and thus fulfills the
/// property `self == self.div_euclid(v) * v + self.rem_euclid(v)`
/// approximatively.
///
/// # Examples
///
/// ```
/// use num_traits::Euclid;
///
/// let a: i32 = 7;
/// let b: i32 = 4;
/// assert_eq!(Euclid::rem_euclid(&a, &b), 3);
/// assert_eq!(Euclid::rem_euclid(&-a, &b), 1);
/// assert_eq!(Euclid::rem_euclid(&a, &-b), 3);
/// assert_eq!(Euclid::rem_euclid(&-a, &-b), 1);
/// ```
fn rem_euclid(&self, v: &Self) -> Self;
/// Returns both the quotient and remainder from Euclidean division.
///
/// By default, it internally calls both `Euclid::div_euclid` and `Euclid::rem_euclid`,
/// but it can be overridden in order to implement some optimization.
///
/// # Examples
///
/// ```
/// # use num_traits::Euclid;
/// let x = 5u8;
/// let y = 3u8;
///
/// let div = Euclid::div_euclid(&x, &y);
/// let rem = Euclid::rem_euclid(&x, &y);
///
/// assert_eq!((div, rem), Euclid::div_rem_euclid(&x, &y));
/// ```
fn div_rem_euclid(&self, v: &Self) -> (Self, Self) {
(self.div_euclid(v), self.rem_euclid(v))
}
}
macro_rules! euclid_forward_impl {
($($t:ty)*) => {$(
impl Euclid for $t {
#[inline]
fn div_euclid(&self, v: &$t) -> Self {
<$t>::div_euclid(*self, *v)
}
#[inline]
fn rem_euclid(&self, v: &$t) -> Self {
<$t>::rem_euclid(*self, *v)
}
}
)*}
}
euclid_forward_impl!(isize i8 i16 i32 i64 i128);
euclid_forward_impl!(usize u8 u16 u32 u64 u128);
#[cfg(feature = "std")]
euclid_forward_impl!(f32 f64);
#[cfg(not(feature = "std"))]
impl Euclid for f32 {
#[inline]
fn div_euclid(&self, v: &f32) -> f32 {
let q = <f32 as crate::float::FloatCore>::trunc(self / v);
if self % v < 0.0 {
return if *v > 0.0 { q - 1.0 } else { q + 1.0 };
}
q
}
#[inline]
fn rem_euclid(&self, v: &f32) -> f32 {
let r = self % v;
if r < 0.0 {
r + <f32 as crate::float::FloatCore>::abs(*v)
} else {
r
}
}
}
#[cfg(not(feature = "std"))]
impl Euclid for f64 {
#[inline]
fn div_euclid(&self, v: &f64) -> f64 {
let q = <f64 as crate::float::FloatCore>::trunc(self / v);
if self % v < 0.0 {
return if *v > 0.0 { q - 1.0 } else { q + 1.0 };
}
q
}
#[inline]
fn rem_euclid(&self, v: &f64) -> f64 {
let r = self % v;
if r < 0.0 {
r + <f64 as crate::float::FloatCore>::abs(*v)
} else {
r
}
}
}
pub trait CheckedEuclid: Euclid {
/// Performs euclid division that returns `None` instead of panicking on division by zero
/// and instead of wrapping around on underflow and overflow.
fn checked_div_euclid(&self, v: &Self) -> Option<Self>;
/// Finds the euclid remainder of dividing two numbers, checking for underflow, overflow and
/// division by zero. If any of that happens, `None` is returned.
fn checked_rem_euclid(&self, v: &Self) -> Option<Self>;
/// Returns both the quotient and remainder from checked Euclidean division.
///
/// By default, it internally calls both `CheckedEuclid::checked_div_euclid` and `CheckedEuclid::checked_rem_euclid`,
/// but it can be overridden in order to implement some optimization.
/// # Examples
///
/// ```
/// # use num_traits::CheckedEuclid;
/// let x = 5u8;
/// let y = 3u8;
///
/// let div = CheckedEuclid::checked_div_euclid(&x, &y);
/// let rem = CheckedEuclid::checked_rem_euclid(&x, &y);
///
/// assert_eq!(Some((div.unwrap(), rem.unwrap())), CheckedEuclid::checked_div_rem_euclid(&x, &y));
/// ```
fn checked_div_rem_euclid(&self, v: &Self) -> Option<(Self, Self)> {
Some((self.checked_div_euclid(v)?, self.checked_rem_euclid(v)?))
}
}
macro_rules! checked_euclid_forward_impl {
($($t:ty)*) => {$(
impl CheckedEuclid for $t {
#[inline]
fn checked_div_euclid(&self, v: &$t) -> Option<Self> {
<$t>::checked_div_euclid(*self, *v)
}
#[inline]
fn checked_rem_euclid(&self, v: &$t) -> Option<Self> {
<$t>::checked_rem_euclid(*self, *v)
}
}
)*}
}
checked_euclid_forward_impl!(isize i8 i16 i32 i64 i128);
checked_euclid_forward_impl!(usize u8 u16 u32 u64 u128);
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn euclid_unsigned() {
macro_rules! test_euclid {
($($t:ident)+) => {
$(
{
let x: $t = 10;
let y: $t = 3;
let div = Euclid::div_euclid(&x, &y);
let rem = Euclid::rem_euclid(&x, &y);
assert_eq!(div, 3);
assert_eq!(rem, 1);
assert_eq!((div, rem), Euclid::div_rem_euclid(&x, &y));
}
)+
};
}
test_euclid!(usize u8 u16 u32 u64);
}
#[test]
fn euclid_signed() {
macro_rules! test_euclid {
($($t:ident)+) => {
$(
{
let x: $t = 10;
let y: $t = -3;
assert_eq!(Euclid::div_euclid(&x, &y), -3);
assert_eq!(Euclid::div_euclid(&-x, &y), 4);
assert_eq!(Euclid::rem_euclid(&x, &y), 1);
assert_eq!(Euclid::rem_euclid(&-x, &y), 2);
assert_eq!((Euclid::div_euclid(&x, &y), Euclid::rem_euclid(&x, &y)), Euclid::div_rem_euclid(&x, &y));
let x: $t = $t::min_value() + 1;
let y: $t = -1;
assert_eq!(Euclid::div_euclid(&x, &y), $t::max_value());
}
)+
};
}
test_euclid!(isize i8 i16 i32 i64 i128);
}
#[test]
fn euclid_float() {
macro_rules! test_euclid {
($($t:ident)+) => {
$(
{
let x: $t = 12.1;
let y: $t = 3.2;
assert!(Euclid::div_euclid(&x, &y) * y + Euclid::rem_euclid(&x, &y) - x
<= 46.4 * <$t as crate::float::FloatCore>::epsilon());
assert!(Euclid::div_euclid(&x, &-y) * -y + Euclid::rem_euclid(&x, &-y) - x
<= 46.4 * <$t as crate::float::FloatCore>::epsilon());
assert!(Euclid::div_euclid(&-x, &y) * y + Euclid::rem_euclid(&-x, &y) + x
<= 46.4 * <$t as crate::float::FloatCore>::epsilon());
assert!(Euclid::div_euclid(&-x, &-y) * -y + Euclid::rem_euclid(&-x, &-y) + x
<= 46.4 * <$t as crate::float::FloatCore>::epsilon());
assert_eq!((Euclid::div_euclid(&x, &y), Euclid::rem_euclid(&x, &y)), Euclid::div_rem_euclid(&x, &y));
}
)+
};
}
test_euclid!(f32 f64);
}
#[test]
fn euclid_checked() {
macro_rules! test_euclid_checked {
($($t:ident)+) => {
$(
{
assert_eq!(CheckedEuclid::checked_div_euclid(&$t::min_value(), &-1), None);
assert_eq!(CheckedEuclid::checked_rem_euclid(&$t::min_value(), &-1), None);
assert_eq!(CheckedEuclid::checked_div_euclid(&1, &0), None);
assert_eq!(CheckedEuclid::checked_rem_euclid(&1, &0), None);
}
)+
};
}
test_euclid_checked!(isize i8 i16 i32 i64 i128);
}
}