nonzero_ext

Traits to represent generic nonzero integer types

Rust ships with non-zero integer types now, which let programmers promise (memory-savingly!) that a number can never be zero. That's great, but sadly the standard library has not got a whole lot of tools to help you use them ergonomically.

A macro for non-zero constant literals

Creating and handling constant literals is neat, but the standard library (and the rust parser at the moment) have no affordances to easily create values of num::NonZeroU* types from constant literals. This crate ships a nonzero! macro that lets you write nonzero!(20u32), which checks at compile time that the constant being converted is non-zero, instead of the cumbersome (and runtime-checked!) NonZeroU32::new(20).unwrap().

Traits for generic non-zeroness

The stdlib num::NonZeroU* types do not implement any common traits (and neither do their zeroable equivalents). Where this lack of traits in the standard library becomes problematic is if you want to write a function that takes a vector of integers, and that returns a vector of the corresponding non-zero integer types, minus any elements that were zero in the original. You can write that with the standard library quite easily for concrete types:

fn only_nonzeros(v: Vec<u8>) -> Vec<NonZeroU8>
{
    v.into_iter()
        .filter_map(|n| NonZeroU8::new(n))
        .collect::<Vec<NonZeroU8>>()
}
let expected: Vec<NonZeroU8> = vec![nonzero!(20u8), nonzero!(5u8)];
assert_eq!(expected, only_nonzeros(vec![0, 20, 5]));

But what if you want to allow this function to work with any integer type that has a corresponding non-zero type? This crate can help:

fn only_nonzeros<I>(v: Vec<I>) -> Vec<I::NonZero>
where
    I: Sized + NonZeroAble,
{
    v.into_iter()
        .filter_map(|n| n.as_nonzero())
        .collect::<Vec<I::NonZero>>()
}

// It works for `u8`:
let input_u8: Vec<u8> = vec![0, 20, 5];
let expected_u8: Vec<NonZeroU8> = vec![nonzero!(20u8), nonzero!(5u8)];
assert_eq!(expected_u8, only_nonzeros(input_u8));

// And it works for `u32`:
let input_u32: Vec<u32> = vec![0, 20, 5];
let expected_u32: Vec<NonZeroU32> = vec![nonzero!(20u32), nonzero!(5u32)];
assert_eq!(expected_u32, only_nonzeros(input_u32));

License: Apache-2.0