Functionality for int_format_into for integer types

[madhav-madhusoodanan]

ACP approval: rust-lang/libs-team#546 (comment)

Context

1. Implemented integer formatting into a fixed-size buffer without relying on fmt.
2. Added a NumBuffer type to help with this process.

Associated Issue

• Tracking Issue for integer formatting into a fixed-size buffer #138215

r? @hanna-kruppe

[rustbot]

Failed to set assignee to hanna-kruppe: invalid assignee

Note: Only org members with at least the repository "read" role, users with write permissions, or people who have commented on the PR may be assigned.

[madhav-madhusoodanan]

@rustbot author

[madhav-madhusoodanan]

I was able to gain more clarity on what was needed, and I've moved the updates to the fmt module.

Primarily I've broken down impl_Display into impl_NumBuffer (the primitives) and impl_Display so that I can reuse functionality into the new macro impl_FormatInto.

Sorry for the long history in this PR. Please let me know if this looks good.

[hanna-kruppe]

Thank you for working on this! The bird's eye view of the diff is now basically what I'd expect. I left some smaller comments I've noticed while scrolling through. I don't expect I'll have the time to review the implementation details as carefully as unsafe code deserves, so please work out those details with a reviewer who actually has permissions on this repository.

[madhav-madhusoodanan]

Thank you for working on this! The bird's eye view of the diff is now basically what I'd expect. I left some smaller comments I've noticed while scrolling through. I don't expect I'll have the time to review the implementation details as carefully as unsafe code deserves, so please work out those details with a reviewer who actually has permissions on this repository.

I must thank you for patiently going through my updates multiple times. There was a lot for me to learn.

[scottmcm]

I'm going to re-roll to get someone hopefully more familiar with the formatting stuff.
r? libs

[bors]

☔ The latest upstream changes (presumably #136974) made this pull request unmergeable. Please resolve the merge conflicts.

[programmerjake]

as mentioned on the ACP and listed as an unresolved issue, I think NumBuffer should have a generic parameter for the type of integer being formatted (we could extend this to support floats later), so when we add future types like u8192 you don't have to have a huge buffer for u32 even though u8192 needs a huge buffer.

so like:

#[unstable(...)] // impl details
/// Safety: `Buffer` must be `[MaybeUninit<u8>; N]` where `N` is big enough to fit any value of `Self` when formatted.
pub unsafe trait NumBufferTrait {
    type Buffer: AsMut<[MaybeUninit<u8>]> + AsRef<[MaybeUninit<u8>]> + 'static + Sync + Send;
    const UNINIT_BUFFER: Self::Buffer;
}

// impl NumBufferTrait for u8, i8, u16, i16, ...

pub struct NumBuffer<T: NumBufferTrait> {
    contents: T::Buffer,
}

impl<T: NumBufferTrait> NumBuffer<T> {
    pub const fn new() -> Self {
        Self { contents: T::UNINIT_BUFFER }
    }

    pub fn len(&self) -> usize {
        self.contents.as_ref().len()
    }

    pub(crate) unsafe fn extract<I>(&self, index: I) -> &I::Output
    where
        I: SliceIndex<[MaybeUninit<u8>]>,
    {
        // SAFETY: `index` is bound-checked by the caller.
        unsafe { self.contents.as_ref().get_unchecked(index) }
    }

    #[cfg(feature = "optimize_for_size")]
    pub(crate) fn extract_start_mut_ptr(buf: &mut Self) -> *mut u8 {
        MaybeUninit::slice_as_mut_ptr(buf.contents.as_mut())
    }

    pub(crate) unsafe fn write(&mut self, offset: usize, data: u8) {
        self.contents.as_mut()[offset].write(data);
    }
}

[hanna-kruppe]

I think this can be done in a follow up PR.