1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
// This file is part of ICU4X. For terms of use, please see the file
// called LICENSE at the top level of the ICU4X source tree
// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).

#![allow(clippy::upper_case_acronyms)]
//! ULE implementation for Plain Old Data types, including all sized integers.

use super::*;
use crate::ZeroSlice;
use core::mem;

/// A u8 array of little-endian data with infallible conversions to and from &[u8].
#[repr(transparent)]
#[derive(Debug, PartialEq, Eq, Clone, Copy, PartialOrd, Ord, Hash)]
pub struct RawBytesULE<const N: usize>(pub [u8; N]);

macro_rules! impl_byte_slice_size {
    ($unsigned:ty, $size:literal) => {
        impl From<[u8; $size]> for RawBytesULE<$size> {
            #[inline]
            fn from(le_bytes: [u8; $size]) -> Self {
                Self(le_bytes)
            }
        }
        impl RawBytesULE<$size> {
            #[inline]
            pub fn as_bytes(&self) -> &[u8] {
                &self.0
            }
        }
        // Safety (based on the safety checklist on the ULE trait):
        //  1. RawBytesULE does not include any uninitialized or padding bytes.
        //     (achieved by `#[repr(transparent)]` on a type that satisfies this invariant)
        //  2. RawBytesULE is aligned to 1 byte.
        //     (achieved by `#[repr(transparent)]` on a type that satisfies this invariant)
        //  3. The impl of validate_byte_slice() returns an error if any byte is not valid (never).
        //  4. The impl of validate_byte_slice() returns an error if there are leftover bytes.
        //  5. The other ULE methods use the default impl.
        //  6. RawBytesULE byte equality is semantic equality
        unsafe impl ULE for RawBytesULE<$size> {
            #[inline]
            fn validate_byte_slice(bytes: &[u8]) -> Result<(), ZeroVecError> {
                if bytes.len() % $size == 0 {
                    // Safe because Self is transparent over [u8; $size]
                    Ok(())
                } else {
                    Err(ZeroVecError::length::<Self>(bytes.len()))
                }
            }
        }

        impl RawBytesULE<$size> {
            #[inline]
            pub fn from_byte_slice_unchecked_mut(bytes: &mut [u8]) -> &mut [Self] {
                let data = bytes.as_mut_ptr();
                let len = bytes.len() / $size;
                // Safe because Self is transparent over [u8; $size]
                unsafe { core::slice::from_raw_parts_mut(data as *mut Self, len) }
            }

            /// Gets this RawBytesULE as an unsigned int. This is equivalent to calling
            /// [AsULE::from_unaligned()] on the appropriately sized type.
            #[inline]
            pub fn as_unsigned_int(&self) -> $unsigned {
                <$unsigned as $crate::ule::AsULE>::from_unaligned(*self)
            }

            /// Convert an array of native-endian aligned integers to an array of RawBytesULE.
            pub const fn from_array<const N: usize>(arr: [$unsigned; N]) -> [Self; N] {
                let mut result = [RawBytesULE([0; $size]); N];
                let mut i = 0;
                // Won't panic because i < N and arr has length N
                #[allow(clippy::indexing_slicing)]
                while i < N {
                    result[i].0 = arr[i].to_le_bytes();
                    i += 1;
                }
                result
            }
        }
    };
}

macro_rules! impl_const_constructors {
    ($base:ty, $size:literal) => {
        impl ZeroSlice<$base> {
            /// This function can be used for constructing ZeroVecs in a const context, avoiding
            /// parsing checks.
            ///
            /// This cannot be generic over T because of current limitations in `const`, but if
            /// this method is needed in a non-const context, check out [`ZeroSlice::parse_byte_slice()`]
            /// instead.
            ///
            /// See [`ZeroSlice::cast()`] for an example.
            pub const fn try_from_bytes(bytes: &[u8]) -> Result<&Self, ZeroVecError> {
                let len = bytes.len();
                #[allow(clippy::modulo_one)]
                if len % $size == 0 {
                    unsafe {
                        // Most of the slice manipulation functions are not yet const-stable,
                        // so we construct a slice with the right metadata and cast its type
                        // https://rust-lang.github.io/unsafe-code-guidelines/layout/pointers.html#notes
                        //
                        // Safety:
                        // * [u8] and [RawBytesULE<N>] have different lengths but the same alignment
                        // * ZeroSlice<$base> is repr(transparent) with [RawBytesULE<N>]
                        let [ptr, _]: [usize; 2] = mem::transmute(bytes);
                        let new_len = len / $size;
                        let raw = [ptr, new_len];
                        Ok(mem::transmute(raw))
                    }
                } else {
                    Err(ZeroVecError::InvalidLength {
                        ty: concat!("<const construct: ", $size, ">"),
                        len,
                    })
                }
            }
        }
    };
}

macro_rules! impl_byte_slice_type {
    ($type:ty, $size:literal) => {
        impl From<$type> for RawBytesULE<$size> {
            #[inline]
            fn from(value: $type) -> Self {
                Self(value.to_le_bytes())
            }
        }
        impl AsULE for $type {
            type ULE = RawBytesULE<$size>;
            #[inline]
            fn to_unaligned(self) -> Self::ULE {
                RawBytesULE(self.to_le_bytes())
            }
            #[inline]
            fn from_unaligned(unaligned: Self::ULE) -> Self {
                <$type>::from_le_bytes(unaligned.0)
            }
        }
        // EqULE is true because $type and RawBytesULE<$size>
        // have the same byte sequence on little-endian
        unsafe impl EqULE for $type {}
    };
}

impl_byte_slice_size!(u16, 2);
impl_byte_slice_size!(u32, 4);
impl_byte_slice_size!(u64, 8);
impl_byte_slice_size!(u128, 16);

impl_byte_slice_type!(u16, 2);
impl_byte_slice_type!(u32, 4);
impl_byte_slice_type!(u64, 8);
impl_byte_slice_type!(u128, 16);

impl_byte_slice_type!(i16, 2);
impl_byte_slice_type!(i32, 4);
impl_byte_slice_type!(i64, 8);
impl_byte_slice_type!(i128, 16);

impl_const_constructors!(u8, 1);
impl_const_constructors!(u16, 2);
impl_const_constructors!(u32, 4);
impl_const_constructors!(u64, 8);
impl_const_constructors!(u128, 16);

// Note: The f32 and f64 const constructors currently have limited use because
// `f32::to_le_bytes` is not yet const.

impl_const_constructors!(bool, 1);

// Safety (based on the safety checklist on the ULE trait):
//  1. u8 does not include any uninitialized or padding bytes.
//  2. u8 is aligned to 1 byte.
//  3. The impl of validate_byte_slice() returns an error if any byte is not valid (never).
//  4. The impl of validate_byte_slice() returns an error if there are leftover bytes (never).
//  5. The other ULE methods use the default impl.
//  6. u8 byte equality is semantic equality
unsafe impl ULE for u8 {
    #[inline]
    fn validate_byte_slice(_bytes: &[u8]) -> Result<(), ZeroVecError> {
        Ok(())
    }
}

impl AsULE for u8 {
    type ULE = Self;
    #[inline]
    fn to_unaligned(self) -> Self::ULE {
        self
    }
    #[inline]
    fn from_unaligned(unaligned: Self::ULE) -> Self {
        unaligned
    }
}

// EqULE is true because u8 is its own ULE.
unsafe impl EqULE for u8 {}

// Safety (based on the safety checklist on the ULE trait):
//  1. i8 does not include any uninitialized or padding bytes.
//  2. i8 is aligned to 1 byte.
//  3. The impl of validate_byte_slice() returns an error if any byte is not valid (never).
//  4. The impl of validate_byte_slice() returns an error if there are leftover bytes (never).
//  5. The other ULE methods use the default impl.
//  6. i8 byte equality is semantic equality
unsafe impl ULE for i8 {
    #[inline]
    fn validate_byte_slice(_bytes: &[u8]) -> Result<(), ZeroVecError> {
        Ok(())
    }
}

impl AsULE for i8 {
    type ULE = Self;
    #[inline]
    fn to_unaligned(self) -> Self::ULE {
        self
    }
    #[inline]
    fn from_unaligned(unaligned: Self::ULE) -> Self {
        unaligned
    }
}

// EqULE is true because i8 is its own ULE.
unsafe impl EqULE for i8 {}

// These impls are actually safe and portable due to Rust always using IEEE 754, see the documentation
// on f32::from_bits: https://doc.rust-lang.org/stable/std/primitive.f32.html#method.from_bits
//
// The only potential problem is that some older platforms treat signaling NaNs differently. This is
// still quite portable, signalingness is not typically super important.

impl AsULE for f32 {
    type ULE = RawBytesULE<4>;
    #[inline]
    fn to_unaligned(self) -> Self::ULE {
        self.to_bits().to_unaligned()
    }
    #[inline]
    fn from_unaligned(unaligned: Self::ULE) -> Self {
        Self::from_bits(u32::from_unaligned(unaligned))
    }
}

impl AsULE for f64 {
    type ULE = RawBytesULE<8>;
    #[inline]
    fn to_unaligned(self) -> Self::ULE {
        self.to_bits().to_unaligned()
    }
    #[inline]
    fn from_unaligned(unaligned: Self::ULE) -> Self {
        Self::from_bits(u64::from_unaligned(unaligned))
    }
}

// The from_bits documentation mentions that they have identical byte representations to integers
// and EqULE only cares about LE systems
unsafe impl EqULE for f32 {}
unsafe impl EqULE for f64 {}

// The bool impl is not as efficient as it could be
// We can, in the future, have https://github.com/unicode-org/icu4x/blob/main/utils/zerovec/design_doc.md#bitpacking
// for better bitpacking

// Safety (based on the safety checklist on the ULE trait):
//  1. bool does not include any uninitialized or padding bytes (the remaining 7 bytes in bool are by definition zero)
//  2. bool is aligned to 1 byte.
//  3. The impl of validate_byte_slice() returns an error if any byte is not valid (bytes that are not 0 or 1).
//  4. The impl of validate_byte_slice() returns an error if there are leftover bytes (never).
//  5. The other ULE methods use the default impl.
//  6. bool byte equality is semantic equality
unsafe impl ULE for bool {
    #[inline]
    fn validate_byte_slice(bytes: &[u8]) -> Result<(), ZeroVecError> {
        for byte in bytes {
            // https://doc.rust-lang.org/reference/types/boolean.html
            // Rust booleans are always size 1, align 1 values with valid bit patterns 0x0 or 0x1
            if *byte > 1 {
                return Err(ZeroVecError::parse::<Self>());
            }
        }
        Ok(())
    }
}

impl AsULE for bool {
    type ULE = Self;
    #[inline]
    fn to_unaligned(self) -> Self::ULE {
        self
    }
    #[inline]
    fn from_unaligned(unaligned: Self::ULE) -> Self {
        unaligned
    }
}

// EqULE is true because bool is its own ULE.
unsafe impl EqULE for bool {}