Shifted to crate fmt from num

Author: madhav-madhusoodanan

library/core/src/fmt/int_format.rs

@@ -0,0 +1,24 @@
+use crate::mem::MaybeUninit;
+
+/// 41 is chosen as the buffer length, as it is larger
+/// than that required to accommodate i128::MIN
+/// (39 decimal digits, 1 for negative sign).
+const BUF_SIZE: usize = 41;
+
+/// A minimal buffer implementation containing elements of type
+/// `MaybeUninit<u8>`.
+#[unstable(feature = "int_format_into", issue = "138215")]
+#[derive(Debug)]
+pub struct NumBuffer {
+    /// An array of elements of type `MaybeUninit<u8>`.
+    pub contents: [MaybeUninit<u8>; BUF_SIZE],
+}
+
+#[unstable(feature = "int_format_into", issue = "138215")]
+impl NumBuffer {
+    /// Initializes `contents` as an uninitialized array of `MaybeUninit<u8>`.
+    #[unstable(feature = "int_format_into", issue = "138215")]
+    pub fn new() -> Self {
+        NumBuffer { contents: [MaybeUninit::<u8>::uninit(); BUF_SIZE] }
+    }
+}

library/core/src/fmt/mod.rs

@@ -12,6 +12,7 @@ use crate::{iter, mem, result, str};
 mod builders;
 #[cfg(not(no_fp_fmt_parse))]
 mod float;
+mod int_format;
 #[cfg(no_fp_fmt_parse)]
 mod nofloat;
 mod num;
@@ -46,6 +47,8 @@ impl From<rt::Alignment> for Option<Alignment> {
     }
 }
 
+#[unstable(feature = "int_format_into", issue = "138215")]
+pub use int_format::NumBuffer;
 #[stable(feature = "debug_builders", since = "1.2.0")]
 pub use self::builders::{DebugList, DebugMap, DebugSet, DebugStruct, DebugTuple};
 #[unstable(feature = "debug_closure_helpers", issue = "117729")]

library/core/src/fmt/num.rs

@@ -1,5 +1,6 @@
 //! Integer and floating-point number formatting
 
+use crate::fmt::NumBuffer;
 use crate::mem::MaybeUninit;
 use crate::num::fmt as numfmt;
 use crate::ops::{Div, Rem, Sub};
@@ -199,6 +200,42 @@ static DEC_DIGITS_LUT: &[u8; 200] = b"\
       6061626364656667686970717273747576777879\
       8081828384858687888990919293949596979899";
 
+static NEGATIVE_SIGN: &[u8; 1] = b"-";
+
+// Unsafe because
+// 1. The contents of `buf` is not validated in the function
+// 2. offset is not bound checked
+unsafe fn _extract_str_from_buf(buf: &NumBuffer, offset: usize) -> &str {
+    let written = unsafe { buf.contents.get_unchecked(offset..) };
+    let as_str = unsafe {
+        str::from_utf8_unchecked(slice::from_raw_parts(
+            MaybeUninit::slice_as_ptr(written),
+            written.len(),
+        ))
+    };
+
+    as_str
+}
+
+// Unsafe because
+// 1. Offset is not bound checked
+unsafe fn _add_negative_sign(
+    is_nonnegative: bool,
+    buf: &mut NumBuffer,
+    start_offset: usize,
+) -> usize {
+    if is_nonnegative {
+        return start_offset;
+    }
+
+    let offset = start_offset - 1;
+
+    // Setting sign for the negative number
+    buf.contents[offset].write(NEGATIVE_SIGN[0]);
+
+    offset
+}
+
 macro_rules! impl_Display {
     ($($signed:ident, $unsigned:ident,)* ; as $u:ident via $conv_fn:ident named $gen_name:ident) => {
 
@@ -212,7 +249,9 @@ macro_rules! impl_Display {
                 }
                 #[cfg(feature = "optimize_for_size")]
                 {
-                    $gen_name(self.$conv_fn(), true, f)
+                    let mut buf = NumBuffer::new();
+                    let as_str = $gen_name(self.$conv_fn(), true, &mut buf);
+                    f.pad_integral(true, "", as_str)
                 }
             }
         }
@@ -226,31 +265,122 @@ macro_rules! impl_Display {
                 }
                 #[cfg(feature = "optimize_for_size")]
                 {
-                    return $gen_name(self.unsigned_abs().$conv_fn(), *self >= 0, f);
+                    let mut buf = NumBuffer::new();
+
+                    // not setting the sign here, hence sending in `true`
+                    let as_str = $gen_name(self.unsigned_abs().$conv_fn(), true, &mut buf);
+                    f.pad_integral(*self >= 0, "", as_str)
+                }
+            }
+        }
+
+        #[unstable(feature = "int_format_into", issue = "138215")]
+        impl $signed {
+            /// Allows users to write an integer (in signed decimal format) into a variable `buf` of
+            /// type [`NumBuffer`] that is passed by the caller by mutable reference.
+            ///
+            /// # Examples
+            /// ```
+            /// #![feature(int_format_into)]
+            /// use core::fmt::NumBuffer;
+            ///
+            #[doc = concat!("let n = -32", stringify!($signed), ";")]
+            /// let mut buf = NumBuffer::new();
+            /// assert_eq!(n.format_into(&mut buf), "-32");
+            ///
+            #[doc = concat!("let n2 = ", stringify!($signed::MIN), ";")]
+            /// let mut buf2 = NumBuffer::new();
+            #[doc = concat!("assert_eq!(n2.format_into(&mut buf2), ", stringify!($signed::MIN), ".to_string());")]
+            ///
+            #[doc = concat!("let n3 = ", stringify!($signed::MAX), ";")]
+            /// let mut buf3 = NumBuffer::new();
+            #[doc = concat!("assert_eq!(n3.format_into(&mut buf3), ", stringify!($signed::MAX), ".to_string());")]
+            /// ```
+            ///
+            pub fn format_into(self, buf: &mut NumBuffer) -> &str {
+                #[cfg(not(feature = "optimize_for_size"))]
+                {
+                    let is_nonnegative = self > 0;
+                    let offset = self.unsigned_abs()._write_into_buf(buf);
+
+                    // SAFETY: `offset >= 2` (unchanged if `is_nonnegative` is true) and
+                    // `offset >= 1` (incase `is_nonnegative` is false) so
+                    // `buf_ptr[offset..offset + 1]` is safe to access.
+                    unsafe { _add_negative_sign(is_nonnegative, buf, offset); }
+
+                    // SAFETY: All buf content since offset is set, and
+                    // writes use ASCII from the lookup table exclusively.
+                    let as_str = unsafe { _extract_str_from_buf(buf, offset) };
+
+                    as_str
+                }
+
+                #[cfg(feature = "optimize_for_size")]
+                {
+                    let is_nonnegative = self > 0;
+                    $gen_name(self.unsigned_abs(), is_nonnegative, buf)
+                }
+
+            }
+        }
+
+        #[unstable(feature = "int_format_into", issue = "138215")]
+        impl $unsigned {
+            /// Allows users to write an integer (in signed decimal format) into a variable `buf` of
+            /// type [`NumBuffer`] that is passed by the caller by mutable reference.
+            ///
+            /// # Examples
+            /// ```
+            /// #![feature(int_format_into)]
+            /// use core::fmt::NumBuffer;
+            ///
+            #[doc = concat!("let n = 32", stringify!($unsigned), ";")]
+            /// let mut buf = NumBuffer::new();
+            /// assert_eq!(n.format_into(&mut buf), "32");
+            ///
+            #[doc = concat!("let n2 = ", stringify!($unsigned::MAX), ";")]
+            /// let mut buf2 = NumBuffer::new();
+            #[doc = concat!("assert_eq!(n2.format_into(&mut buf2), ", stringify!($unsigned::MAX), ".to_string());")]
+            /// ```
+            ///
+            pub fn format_into(self, buf: &mut NumBuffer) -> &str {
+                #[cfg(not(feature = "optimize_for_size"))]
+                {
+                    let offset = self._write_into_buf(buf);
+
+                    // SAFETY: All contents in `buf` since offset is set, and
+                    // writes use ASCII from the lookup table exclusively.
+                    let as_str = unsafe { _extract_str_from_buf(buf, offset) };
+
+                    as_str
+                }
+
+                #[cfg(feature = "optimize_for_size")]
+                {
+                    $gen_name(self, true, buf)
                 }
+
             }
         }
 
         #[cfg(not(feature = "optimize_for_size"))]
         impl $unsigned {
-            fn _fmt(self, is_nonnegative: bool, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-                const MAX_DEC_N: usize = $unsigned::MAX.ilog(10) as usize + 1;
-                // Buffer decimals for $unsigned with right alignment.
-                let mut buf = [MaybeUninit::<u8>::uninit(); MAX_DEC_N];
+            fn _write_into_buf(self, buf: &mut NumBuffer) -> usize {
                 // Count the number of bytes in buf that are not initialized.
-                let mut offset = buf.len();
+                let mut offset = buf.contents.len();
                 // Consume the least-significant decimals from a working copy.
                 let mut remain = self;
 
                 // Format per four digits from the lookup table.
                 // Four digits need a 16-bit $unsigned or wider.
                 while size_of::<Self>() > 1 && remain > 999.try_into().expect("branch is not hit for types that cannot fit 999 (u8)") {
-                    // SAFETY: All of the decimals fit in buf due to MAX_DEC_N
+                    // SAFETY: All of the decimals fit in buf since buf is large enough to
+                    // accommodate the largest representation of a number possible (that of i128::MIN)
                     // and the while condition ensures at least 4 more decimals.
                     unsafe { core::hint::assert_unchecked(offset >= 4) }
-                    // SAFETY: The offset counts down from its initial buf.len()
+                    // SAFETY: The offset counts down from its initial size
                     // without underflow due to the previous precondition.
-                    unsafe { core::hint::assert_unchecked(offset <= buf.len()) }
+                    unsafe { core::hint::assert_unchecked(offset <= buf.contents.len()) }
                     offset -= 4;
 
                     // pull two pairs
@@ -259,26 +389,27 @@ macro_rules! impl_Display {
                     remain /= scale;
                     let pair1 = (quad / 100) as usize;
                     let pair2 = (quad % 100) as usize;
-                    buf[offset + 0].write(DEC_DIGITS_LUT[pair1 * 2 + 0]);
-                    buf[offset + 1].write(DEC_DIGITS_LUT[pair1 * 2 + 1]);
-                    buf[offset + 2].write(DEC_DIGITS_LUT[pair2 * 2 + 0]);
-                    buf[offset + 3].write(DEC_DIGITS_LUT[pair2 * 2 + 1]);
+                    buf.contents[offset + 0].write(DEC_DIGITS_LUT[pair1 * 2 + 0]);
+                    buf.contents[offset + 1].write(DEC_DIGITS_LUT[pair1 * 2 + 1]);
+                    buf.contents[offset + 2].write(DEC_DIGITS_LUT[pair2 * 2 + 0]);
+                    buf.contents[offset + 3].write(DEC_DIGITS_LUT[pair2 * 2 + 1]);
                 }
 
                 // Format per two digits from the lookup table.
                 if remain > 9 {
-                    // SAFETY: All of the decimals fit in buf due to MAX_DEC_N
+                    // SAFETY: All of the decimals fit in buf since buf is large enough to
+                    // accommodate the largest representation of a number possilble (that of i128::MIN)
                     // and the while condition ensures at least 2 more decimals.
                     unsafe { core::hint::assert_unchecked(offset >= 2) }
-                    // SAFETY: The offset counts down from its initial buf.len()
+                    // SAFETY: The offset counts down from its initial size
                     // without underflow due to the previous precondition.
-                    unsafe { core::hint::assert_unchecked(offset <= buf.len()) }
+                    unsafe { core::hint::assert_unchecked(offset <= buf.contents.len()) }
                     offset -= 2;
 
                     let pair = (remain % 100) as usize;
                     remain /= 100;
-                    buf[offset + 0].write(DEC_DIGITS_LUT[pair * 2 + 0]);
-                    buf[offset + 1].write(DEC_DIGITS_LUT[pair * 2 + 1]);
+                    buf.contents[offset + 0].write(DEC_DIGITS_LUT[pair * 2 + 0]);
+                    buf.contents[offset + 1].write(DEC_DIGITS_LUT[pair * 2 + 1]);
                 }
 
                 // Format the last remaining digit, if any.
@@ -288,40 +419,43 @@ macro_rules! impl_Display {
                     unsafe { core::hint::assert_unchecked(offset >= 1) }
                     // SAFETY: The offset counts down from its initial buf.len()
                     // without underflow due to the previous precondition.
-                    unsafe { core::hint::assert_unchecked(offset <= buf.len()) }
+                    unsafe { core::hint::assert_unchecked(offset <= buf.contents.len()) }
                     offset -= 1;
 
                     // Either the compiler sees that remain < 10, or it prevents
                     // a boundary check up next.
                     let last = (remain & 15) as usize;
-                    buf[offset].write(DEC_DIGITS_LUT[last * 2 + 1]);
+                    buf.contents[offset].write(DEC_DIGITS_LUT[last * 2 + 1]);
                     // not used: remain = 0;
                 }
 
-                // SAFETY: All buf content since offset is set.
-                let written = unsafe { buf.get_unchecked(offset..) };
-                // SAFETY: Writes use ASCII from the lookup table exclusively.
+                offset
+            }
+
+            fn _fmt(self, is_nonnegative: bool, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+                // Buffer decimals for $unsigned with right alignment.
+                let mut buf = NumBuffer::new();
+                let offset = self._write_into_buf(&mut buf);
+
+                // SAFETY: All buf content since offset is set, and
+                // writes use ASCII from the lookup table exclusively.
                 let as_str = unsafe {
-                    str::from_utf8_unchecked(slice::from_raw_parts(
-                          MaybeUninit::slice_as_ptr(written),
-                          written.len(),
-                    ))
+                    _extract_str_from_buf(&buf, offset)
                 };
+
                 f.pad_integral(is_nonnegative, "", as_str)
             }
         })*
 
         #[cfg(feature = "optimize_for_size")]
-        fn $gen_name(mut n: $u, is_nonnegative: bool, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-            const MAX_DEC_N: usize = $u::MAX.ilog(10) as usize + 1;
-            let mut buf = [MaybeUninit::<u8>::uninit(); MAX_DEC_N];
-            let mut curr = MAX_DEC_N;
-            let buf_ptr = MaybeUninit::slice_as_mut_ptr(&mut buf);
+        fn $gen_name(mut n: $u, is_nonnegative: bool, buf: &mut NumBuffer) -> &str {
+            let mut curr = buf.contents.len();
+            let buf_ptr = MaybeUninit::slice_as_mut_ptr(&mut buf.contents);
 
             // SAFETY: To show that it's OK to copy into `buf_ptr`, notice that at the beginning
-            // `curr == buf.len() == 39 > log(n)` since `n < 2^128 < 10^39`, and at
+            // `curr == buf.len() == 41 > log(n)` since `n < 2^128 < 10^39 < 10^41`, and at
             // each step this is kept the same as `n` is divided. Since `n` is always
-            // non-negative, this means that `curr > 0` so `buf_ptr[curr..curr + 1]`
+            // non-negative, this means that `curr >= (41 - 39) == 2 > 0` so `buf_ptr[curr..curr + 1]`
             // is safe to access.
             unsafe {
                 loop {
@@ -335,12 +469,18 @@ macro_rules! impl_Display {
                 }
             }
 
-            // SAFETY: `curr` > 0 (since we made `buf` large enough), and all the chars are valid UTF-8
+            // SAFETY: `curr >= 2` (unchanged if `is_nonnegative` is true) and
+            // `curr >= 1` (incase `is_nonnegative` is false) so `buf_ptr[curr..curr + 1]`
+            // is safe to access.
+            unsafe { _add_negative_sign(is_nonnegative, buf, curr); }
+
+            // SAFETY: `curr >= 1` (since we made `buf` large enough), and all the chars are valid UTF-8
             let buf_slice = unsafe {
                 str::from_utf8_unchecked(
                     slice::from_raw_parts(buf_ptr.add(curr), buf.len() - curr))
             };
-            f.pad_integral(is_nonnegative, "", buf_slice)
+
+            buf_slice
         }
     };
 }
@@ -529,7 +669,7 @@ mod imp {
         i32, u32,
         i64, u64,
         isize, usize,
-        ; as u64 via to_u64 named fmt_u64
+        ; as u64 via to_u64 named stringify_u64
     );
     impl_Exp!(
         i8, u8, i16, u16, i32, u32, i64, u64, usize, isize
@@ -545,10 +685,10 @@ mod imp {
         i16, u16,
         i32, u32,
         isize, usize,
-        ; as u32 via to_u32 named fmt_u32);
+        ; as u32 via to_u32 named stringify_u32);
     impl_Display!(
         i64, u64,
-        ; as u64 via to_u64 named fmt_u64);
+        ; as u64 via to_u64 named stringify_u64);
 
     impl_Exp!(i8, u8, i16, u16, i32, u32, isize, usize as u32 via to_u32 named exp_u32);
     impl_Exp!(i64, u64 as u64 via to_u64 named exp_u64);