7 releases
0.2.0 | Mar 14, 2021 |
---|---|
0.1.6 | Mar 13, 2021 |
#1690 in Rust patterns
24 downloads per month
15KB
63 lines
polymorphic-constant
A macro to generate numerical constants in multiple types at once.
You can have the number Pi be available in f64 or f32
It was designed with three goals in mind:
- Catch all overflow errors on compile-time
- Minimize the code footprint
- Be readable and easy to use
This is meant as a temporary fix to the year-long debate over Rust Pre-RFC #1337
Syntax
use polymorphic_constant::polymorphic_constant;
polymorphic_constant! {
const PI: f32 | f64 = 3.141592653589793;
}
// Which can then be used as
fn pi_squared () -> f64 {
PI.f64 * PI.f64
}
A few features are supported:
use polymorphic_constant::polymorphic_constant;
polymorphic_constant! {
/// Doc comment attributes
const PI: f32 | f64 = 3.141592653589793;
// Visibility modifiers (for both constant and type)
pub (crate) const E: f32 | f64 = 2.7182818284590452;
// Nonzero numeric types (NonZeroI32, NonZeroU8, etc)
const ASCII_LINE_RETURN: u8 | nz_u8 = 10;
}
// You can handle constants like any const struct
const PI_COPY: PI = PI;
const PI_F32: f32 = PI.f32;
// Into is implemented for every variant of the constant
fn times_pi<T: std::ops::Mul<T>> (value: T) -> <T as std::ops::Mul>::Output
where
PI: Into<T>,
{
value * PI.into()
}
assert_eq!(times_pi(2.0), 6.283185307179586f64);
Safety
This system ensures that you keep all the safeties and warnings given by rust, but no more
Any incompatible type will prevent compilation:
- Float literals cannot be stored if it would convert them to infinity
const FAILS: f32 | f64 = 3141592653589793238462643383279502884197.0;
- Literals cannot be stored in a type too small to hold them
const FAILS: u64 | nz_i8 = 128;
- Negative numbers cannot be stored in unsigned types
const FAILS: i64 | u8 = -1;
- 0 cannot be stored in non-zero types
const FAILS: nz_u8 | nz_u16 | nz_u32 = 0;
- However, floats may lose precision, and a lot of it
const SUCCEEDS: f32 = 3.141592653589793238462643383279;
Warnings
Currently, the same constant cannot hold both int and float variants
const FAIL: i32 = 0.1;
const FAIL: f32 = 0;
The constant also has to be initialized with an untyped literal
const FAIL: i32 = 0u32;
It is still unclear if accepting the examples above could be dangerous, thus the conservative choice.
Example
use polymorphic_constant::polymorphic_constant;
polymorphic_constant! {
const HEIGHT: i8 | u8 | i16 | u16 | i32 | u32 = 16;
const WIDTH: i8 | u8 | i16 | u16 | i32 | u32 = 32;
}
fn main() {
let size = HEIGHT.i16 * WIDTH.i16;
assert_eq!(size, 16 * 32);
let height_copy:i32 = HEIGHT.into();
assert_eq!(HEIGHT.i32, height_copy);
}
Support
I would love any feedback on usage, for future ameliorations and features.
License: MIT