Tracking Issue for `#![feature(mixed_integer_ops)]`

[a1phyr]

Feature gate: #![feature(mixed_integer_ops)]

Public API

// `uX` is `u8`, `u16`, `u32`, `u64`,`u128`, `usize`
impl uX {
    pub const fn checked_add_signed(self, iX) -> Option<Self>;
    pub const fn overflowing_add_signed(self, iX) -> (Self, bool);
    pub const fn saturating_add_signed(self, iX) -> Self;
    pub const fn wrapping_add_signed(self, iX) -> Self;
}

impl iX {
    pub const fn checked_add_unsigned(self, uX) -> Option<Self>;
    pub const fn overflowing_add_unsigned(self, uX) -> (Self, bool);
    pub const fn saturating_add_unsigned(self, uX) -> Self;
    pub const fn wrapping_add_unsigned(self, uX) -> Self;

    pub const fn checked_sub_unsigned(self, uX) -> Option<Self>;
    pub const fn overflowing_sub_unsigned(self, uX) -> (Self, bool);
    pub const fn saturating_sub_unsigned(self, uX) -> Self;
    pub const fn wrapping_sub_unsigned(self, uX) -> Self;
}

Steps / History

• Implementation: Add functions to add unsigned and signed integers #87601
• Final comment period (FCP)
• Stabilization PR: Stabilize #![feature(mixed_integer_ops)] #101555

[leonardo-m]

Somewhere you should write about use cases & purposes of these ops :)

[brendanzab]

Somewhere you should write about use cases & purposes of these ops :)

I've needed checked_add_signed for working with io::SeekFrom, for example! Had to roll my own (and I'm not sure how correct it is), inspired by this implementation:

fn checked_add_signed(pos: u64, offset: i64) -> Option<u64> {
    if offset >= 0 {
        u64::checked_add(pos, offset as u64)
    } else {
        u64::checked_sub(pos, offset.unsigned_abs())
    }
}

It would be great to have this in the stdlib!

[a1phyr]

This is exactly what prompted me to write these !

Also it is what std uses for Cursor now.

[tjallingt]

Would love to see this get stabilized. Not sure what the right process is to start an FCP

[avl]

I love these methods!

I would love to see *_sub_signed and friends too. I just had a use for saturating_sub_signed (subtracting a signed residual from an unsigned value to create a corrected value). In my case, I can just negate the residual. However, that might cause overflow, in which case the final calculated value would not be correct. This won't really matter in practice, but saturating_sub_signed still would have been nice!

[joshtriplett]

I just ran into a use case for these as well, in Cargo, for working with CPU counts (e.g. "use all but 1 CPU").

[joshtriplett]

These have been around for a long time, and it seems like several people are interested in them.

Shall we stabilize these?

@rfcbot merge