MultiType

MultiType is Rust a crate for generalising fundamental types via traits.

MultiType provideds traits such as Uint and Float traits to abstract over a set of equivalent primitive types. These traits are intended to provide one-to-one copies of the interfaces that the primitive types define.

Arithmetic types

MultiType defines different traits for generalising arithmetic types:

Uint for u8, u16, u32, u64, u128, and usize;
Int for i8, i16, i32, i64, i128, and isize;
Float for f16, f32, f64, and f128.

Furthermore, StdFloat extends the Float trait with functionality that is typically only available in std's floating-point types.

Sized, arithmetic types

The basic, arithmetic traits guarantee a minimum size that is equivalent to the smallest member of its group, e.g. Uint is at least u8 and Float is at least f16.

Additionally, these three arithmetic traits have subtraits that guarantee wider types, for example:

i16 and up implement IntLeast16;
i32 and up implement IntLeast32;
Etc.

The complete list of arithmetic traits is thus:

Uint
- UintLeast16
- UintLeast32
- UintLeast64
- UintLeast128
Int
- IntLeast16
- IntLeast32
- IntLeast64
- IntLeast128
Float
- StdFloat
- FloatLeast32
- FloatLeast64
- FloatLeast128

Array types

MultiType also provides the Array trait for generalising array types – most often over their length.

An example of this trait's usecase is actually in this crate: Take, for instance, the definition of Uint: It has a Bytes associated type that is used by the bytewise constructors and destructors:

pub unsafe trait Uint: /* .. */ {
    type Bytes;

    fn from_ne_bytes(bytes: Self::Bytes) -> Self;

    fn to_ne_bytes(self) -> Self::Bytes;

    // ..
}

Now, anyone that would want to use the output of to_ne_bytes:to_ne_bytes) wouldn't really have that many choices with regard to what to do with it. So, MultiType defines the Array trait:

use multitype::array::Array;

pub unsafe trait Uint: /* .. */ {
    type Bytes: Array<Scalar = u8>;

    // ..
}

With it, it's possible for users to generically use Uint::to_ne_bytes as an array type through the trait methods.

Examples

A generic Fibonacci sequence:

use multitype::num::Uint;

fn f<T: Uint>(x: T) -> T {
    let mut y    = T::from_u8(0x0);
    let mut y_m1 = T::from_u8(0x0);
    let mut y_m2 = T::from_u8(0x1);

    let mut i = T::from_u8(0x0);
    while i < x {
        y = y_m1 + y_m2;

        y_m2 = y_m1;
        y_m1 = y;

        i += T::from_u8(0x1);
    }

    y
}

assert_eq!(f(0u8),   0);
assert_eq!(f(1u8),   1);

assert_eq!(f(2u16),  1);
assert_eq!(f(3u16),  2);

assert_eq!(f(4u32),  3);
assert_eq!(f(5u32),  5);

assert_eq!(f(6u64),  8);
assert_eq!(f(7u64), 13);

Generic array indexing:

use core::f32;
use multitype::array::Array;
use multitype::num::Float;

fn complicated_neg<T: Float>(value: T) -> T {
    let mut bytes = value.to_le_bytes();

    // Invert the sign bit -- which is always the most
    // significant bit of the most significant byte.
    *bytes.as_mut_slice().last_mut().unwrap() ^= 0b10000000;

    T::from_le_bytes(bytes)
}

assert_eq!(complicated_neg( 1.0f64), -1.0f64);
assert_eq!(complicated_neg(-1.0f64),  1.0f64);

assert_eq!(complicated_neg(f32::NEG_INFINITY), f32::INFINITY);

Copyright & Licence.