When implementing a primitive fixed-size vector type (float2
for example), I want to support the Add
and Sub
traits. Later, I will want to support Mul
and *Assign
.
Looking up the documentation and other examples, I came up with this:
use std::ops::{Add, Sub};
#[derive(Copy, Clone)]
struct float2(f64, f64);
impl Add for float2 {
type Output = float2;
fn add(self, _rhs: float2) -> float2 {
float2(self.0 + _rhs.0, self.1 + _rhs.1)
}
}
impl Sub for float2 {
type Output = float2;
fn sub(self, _rhs: float2) -> float2 {
float2(self.0 - _rhs.0, self.1 - _rhs.1)
}
}
This works for basic examples, however I found in practice I would often end up with references passed in as arguments as well as local float2
's on the stack.
To mix these I needed to either:
Example:
impl<'a, 'b> Add<&'b float2> for &'a float2 {
type Output = float2;
fn add(self, _rhs: &'b float2) -> float2 {
float2(self.0 + _rhs.0, self.1 + _rhs.1)
}
}
impl<'a> Add<float2> for &'a float2 {
type Output = float2;
fn add(self, _rhs: float2) -> float2 {
float2(self.0 + _rhs.0, self.1 + _rhs.1)
}
}
impl<'b> Add<&'b float2> for float2 {
type Output = float2;
fn add(self, _rhs: &'b float2) -> float2 {
float2(self.0 + _rhs.0, self.1 + _rhs.1)
}
}
/*... and again for Sub */
While this allows to write expressions without de-referencing. it becomes quite tedious to enumerate each combinations, especially when adding more operations & types (float3
, float4
...).
Is there a generally accepted way to...
Or is it expected that developers either:
Note, I'm currently a beginner, I've checked some quite advanced math libraries in Rust, they're way over my head, while I could use them - I would like to understand how to write operator overloading for my own types.
The great thing about Rust is that it's open source. This means you can see how the authors of the language have solved a problem. The closest analogue is primitive integer types:
macro_rules! add_impl {
($($t:ty)*) => ($(
#[stable(feature = "rust1", since = "1.0.0")]
impl Add for $t {
type Output = $t;
#[inline]
fn add(self, other: $t) -> $t { self + other }
}
forward_ref_binop! { impl Add, add for $t, $t }
)*)
}
forward_ref_binop
is defined as:
macro_rules! forward_ref_binop {
(impl $imp:ident, $method:ident for $t:ty, $u:ty) => {
#[stable(feature = "rust1", since = "1.0.0")]
impl<'a> $imp<$u> for &'a $t {
type Output = <$t as $imp<$u>>::Output;
#[inline]
fn $method(self, other: $u) -> <$t as $imp<$u>>::Output {
$imp::$method(*self, other)
}
}
#[stable(feature = "rust1", since = "1.0.0")]
impl<'a> $imp<&'a $u> for $t {
type Output = <$t as $imp<$u>>::Output;
#[inline]
fn $method(self, other: &'a $u) -> <$t as $imp<$u>>::Output {
$imp::$method(self, *other)
}
}
#[stable(feature = "rust1", since = "1.0.0")]
impl<'a, 'b> $imp<&'a $u> for &'b $t {
type Output = <$t as $imp<$u>>::Output;
#[inline]
fn $method(self, other: &'a $u) -> <$t as $imp<$u>>::Output {
$imp::$method(*self, *other)
}
}
}
}
It's certainly valid to write wrapper implementations of the traits for references that simply dereference and call the value-oriented version.
I advise you to use the impl_os crate for that purpose, see that other answer that I wrote.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With