fp::prime

Trait Prime

pub trait Prime:
    Debug
    + Clone
    + Copy
    + Display
    + Hash
    + PartialEq
    + Eq
    + PartialEq<u32>
    + PartialOrd<u32>
    + Add<u32, Output = u32>
    + Sub<u32, Output = u32>
    + Mul<u32, Output = u32>
    + Div<u32, Output = u32>
    + Rem<u32, Output = u32>
    + Shl<u32, Output = u32>
    + Shr<u32, Output = u32>
    + Serialize
    + for<'de> Deserialize<'de>
    + MaybeArbitrary<Option<NonZeroU32>>
    + 'static {
    // Required methods
    fn as_i32(self) -> i32;
    fn to_dyn(self) -> ValidPrime;

    // Provided methods
    fn as_u32(self) -> u32 { ... }
    fn as_usize(self) -> usize { ... }
    fn sum(self, n1: u32, n2: u32) -> u32 { ... }
    fn product(self, n1: u32, n2: u32) -> u32 { ... }
    fn inverse(self, k: u32) -> u32 { ... }
    fn pow(self, exp: u32) -> u32 { ... }
    fn pow_mod(self, b: u32, e: u32) -> u32 { ... }
}

Expand description

A trait that represents a prime number. There are currently two kinds of structs that implement this trait: static primes and ValidPrime, the dynamic prime.

The methods in this trait take a self receiver so that the dynamic prime ValidPrime can implement it. We could also have a &self receiver, but since Prime is a supertrait of Copy, the two are equivalent. Using self might also be useful in the future if we ever want to play with autoref specialization.

The fact that e.g. P2::to_u32 is hardcoded to return 2 means that a condition like p.to_u32() == 2 (or even better, just p == 2) will reduce to true at compile time, and allow the compiler to eliminate an entire branch, while also leaving that check in for when the prime is chosen at runtime.