| 1 | // Copyright Mozilla Foundation |
| 2 | // |
| 3 | // Licensed under the Apache License (Version 2.0), or the MIT license, |
| 4 | // (the "Licenses") at your option. You may not use this file except in |
| 5 | // compliance with one of the Licenses. You may obtain copies of the |
| 6 | // Licenses at: |
| 7 | // |
| 8 | // https://www.apache.org/licenses/LICENSE-2.0 |
| 9 | // https://opensource.org/licenses/MIT |
| 10 | // |
| 11 | // Unless required by applicable law or agreed to in writing, software |
| 12 | // distributed under the Licenses is distributed on an "AS IS" BASIS, |
| 13 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 14 | // See the Licenses for the specific language governing permissions and |
| 15 | // limitations under the Licenses. |
| 16 | |
| 17 | #![no_std ] |
| 18 | |
| 19 | //! Provides iteration by `char` over `&[u16]` containing potentially-invalid |
| 20 | //! UTF-16 such that errors are replaced with the REPLACEMENT CHARACTER. |
| 21 | //! |
| 22 | //! The trait `Utf16CharsEx` provides the convenience method `chars()` on |
| 23 | //! byte slices themselves instead of having to use the more verbose |
| 24 | //! `Utf16Chars::new(slice)`. |
| 25 | |
| 26 | mod indices; |
| 27 | mod report; |
| 28 | |
| 29 | pub use crate::indices::Utf16CharIndices; |
| 30 | pub use crate::report::ErrorReportingUtf16Chars; |
| 31 | pub use crate::report::Utf16CharsError; |
| 32 | use core::iter::FusedIterator; |
| 33 | |
| 34 | #[inline (always)] |
| 35 | fn in_inclusive_range16(i: u16, start: u16, end: u16) -> bool { |
| 36 | i.wrapping_sub(start) <= (end - start) |
| 37 | } |
| 38 | |
| 39 | /// Iterator by `char` over `&[u16]` that contains |
| 40 | /// potentially-invalid UTF-16. See the crate documentation. |
| 41 | #[derive (Debug, Clone)] |
| 42 | pub struct Utf16Chars<'a> { |
| 43 | remaining: &'a [u16], |
| 44 | } |
| 45 | |
| 46 | impl<'a> Utf16Chars<'a> { |
| 47 | #[inline (always)] |
| 48 | /// Creates the iterator from a `u16` slice. |
| 49 | pub fn new(code_units: &'a [u16]) -> Self { |
| 50 | Utf16Chars::<'a> { |
| 51 | remaining: code_units, |
| 52 | } |
| 53 | } |
| 54 | |
| 55 | /// Views the current remaining data in the iterator as a subslice |
| 56 | /// of the original slice. |
| 57 | #[inline (always)] |
| 58 | pub fn as_slice(&self) -> &'a [u16] { |
| 59 | self.remaining |
| 60 | } |
| 61 | |
| 62 | #[inline (never)] |
| 63 | fn surrogate_next(&mut self, surrogate_base: u16, first: u16) -> char { |
| 64 | if surrogate_base <= (0xDBFF - 0xD800) { |
| 65 | if let Some((&low, tail_tail)) = self.remaining.split_first() { |
| 66 | if in_inclusive_range16(low, 0xDC00, 0xDFFF) { |
| 67 | self.remaining = tail_tail; |
| 68 | return unsafe { |
| 69 | char::from_u32_unchecked( |
| 70 | (u32::from(first) << 10) + u32::from(low) |
| 71 | - (((0xD800u32 << 10) - 0x10000u32) + 0xDC00u32), |
| 72 | ) |
| 73 | }; |
| 74 | } |
| 75 | } |
| 76 | } |
| 77 | ' \u{FFFD}' |
| 78 | } |
| 79 | |
| 80 | #[inline (never)] |
| 81 | fn surrogate_next_back(&mut self, last: u16) -> char { |
| 82 | if in_inclusive_range16(last, 0xDC00, 0xDFFF) { |
| 83 | if let Some((&high, head_head)) = self.remaining.split_last() { |
| 84 | if in_inclusive_range16(high, 0xD800, 0xDBFF) { |
| 85 | self.remaining = head_head; |
| 86 | return unsafe { |
| 87 | char::from_u32_unchecked( |
| 88 | (u32::from(high) << 10) + u32::from(last) |
| 89 | - (((0xD800u32 << 10) - 0x10000u32) + 0xDC00u32), |
| 90 | ) |
| 91 | }; |
| 92 | } |
| 93 | } |
| 94 | } |
| 95 | ' \u{FFFD}' |
| 96 | } |
| 97 | } |
| 98 | |
| 99 | impl<'a> Iterator for Utf16Chars<'a> { |
| 100 | type Item = char; |
| 101 | |
| 102 | #[inline (always)] |
| 103 | fn next(&mut self) -> Option<char> { |
| 104 | // It might be OK to delegate to `ErrorReportingUtf16Chars`, but since |
| 105 | // the methods are rather small, copypaste is probably clearer. Also, |
| 106 | // copypaste would _not_ be equivalent if any part of this was delegated |
| 107 | // to an `inline(never)` helper. However, previous experimentation indicated |
| 108 | // that such a helper didn't help performance here. |
| 109 | let (&first: u16, tail: &[u16]) = self.remaining.split_first()?; |
| 110 | self.remaining = tail; |
| 111 | let surrogate_base: u16 = first.wrapping_sub(0xD800); |
| 112 | if surrogate_base > (0xDFFF - 0xD800) { |
| 113 | return Some(unsafe { char::from_u32_unchecked(u32::from(first)) }); |
| 114 | } |
| 115 | Some(self.surrogate_next(surrogate_base, first)) |
| 116 | } |
| 117 | } |
| 118 | |
| 119 | impl<'a> DoubleEndedIterator for Utf16Chars<'a> { |
| 120 | #[inline (always)] |
| 121 | fn next_back(&mut self) -> Option<char> { |
| 122 | let (&last: u16, head: &[u16]) = self.remaining.split_last()?; |
| 123 | self.remaining = head; |
| 124 | if !in_inclusive_range16(i:last, start:0xD800, end:0xDFFF) { |
| 125 | return Some(unsafe { char::from_u32_unchecked(u32::from(last)) }); |
| 126 | } |
| 127 | Some(self.surrogate_next_back(last)) |
| 128 | } |
| 129 | } |
| 130 | |
| 131 | impl FusedIterator for Utf16Chars<'_> {} |
| 132 | |
| 133 | /// Convenience trait that adds `chars()` and `char_indices()` methods |
| 134 | /// similar to the ones on string slices to `u16` slices. |
| 135 | pub trait Utf16CharsEx { |
| 136 | fn chars(&self) -> Utf16Chars<'_>; |
| 137 | fn char_indices(&self) -> Utf16CharIndices<'_>; |
| 138 | } |
| 139 | |
| 140 | impl Utf16CharsEx for [u16] { |
| 141 | /// Convenience method for creating an UTF-16 iterator |
| 142 | /// for the slice. |
| 143 | #[inline ] |
| 144 | fn chars(&self) -> Utf16Chars<'_> { |
| 145 | Utf16Chars::new(self) |
| 146 | } |
| 147 | /// Convenience method for creating a code unit index and |
| 148 | /// UTF-16 iterator for the slice. |
| 149 | #[inline ] |
| 150 | fn char_indices(&self) -> Utf16CharIndices<'_> { |
| 151 | Utf16CharIndices::new(self) |
| 152 | } |
| 153 | } |
| 154 | |
| 155 | #[cfg (test)] |
| 156 | mod tests { |
| 157 | use crate::Utf16CharsEx; |
| 158 | |
| 159 | #[test ] |
| 160 | fn test_boundaries() { |
| 161 | assert!([0xD7FFu16] |
| 162 | .as_slice() |
| 163 | .chars() |
| 164 | .eq(core::iter::once(' \u{D7FF}' ))); |
| 165 | assert!([0xE000u16] |
| 166 | .as_slice() |
| 167 | .chars() |
| 168 | .eq(core::iter::once(' \u{E000}' ))); |
| 169 | assert!([0xD800u16] |
| 170 | .as_slice() |
| 171 | .chars() |
| 172 | .eq(core::iter::once(' \u{FFFD}' ))); |
| 173 | assert!([0xDFFFu16] |
| 174 | .as_slice() |
| 175 | .chars() |
| 176 | .eq(core::iter::once(' \u{FFFD}' ))); |
| 177 | } |
| 178 | |
| 179 | #[test ] |
| 180 | fn test_unpaired() { |
| 181 | assert!([0xD800u16, 0x0061u16] |
| 182 | .as_slice() |
| 183 | .chars() |
| 184 | .eq([0xFFFDu16, 0x0061u16].as_slice().chars())); |
| 185 | assert!([0xDFFFu16, 0x0061u16] |
| 186 | .as_slice() |
| 187 | .chars() |
| 188 | .eq([0xFFFDu16, 0x0061u16].as_slice().chars())); |
| 189 | } |
| 190 | |
| 191 | #[test ] |
| 192 | fn test_unpaired_rev() { |
| 193 | assert!([0xD800u16, 0x0061u16] |
| 194 | .as_slice() |
| 195 | .chars() |
| 196 | .rev() |
| 197 | .eq([0xFFFDu16, 0x0061u16].as_slice().chars().rev())); |
| 198 | assert!([0xDFFFu16, 0x0061u16] |
| 199 | .as_slice() |
| 200 | .chars() |
| 201 | .rev() |
| 202 | .eq([0xFFFDu16, 0x0061u16].as_slice().chars().rev())); |
| 203 | } |
| 204 | |
| 205 | #[test ] |
| 206 | fn test_paired() { |
| 207 | assert!([0xD83Eu16, 0xDD73u16] |
| 208 | .as_slice() |
| 209 | .chars() |
| 210 | .eq(core::iter::once('🥳' ))); |
| 211 | } |
| 212 | |
| 213 | #[test ] |
| 214 | fn test_paired_rev() { |
| 215 | assert!([0xD83Eu16, 0xDD73u16] |
| 216 | .as_slice() |
| 217 | .chars() |
| 218 | .rev() |
| 219 | .eq(core::iter::once('🥳' ))); |
| 220 | } |
| 221 | |
| 222 | #[test ] |
| 223 | fn test_as_slice() { |
| 224 | let mut iter = [0x0061u16, 0x0062u16].as_slice().chars(); |
| 225 | let at_start = iter.as_slice(); |
| 226 | assert_eq!(iter.next(), Some('a' )); |
| 227 | let in_middle = iter.as_slice(); |
| 228 | assert_eq!(iter.next(), Some('b' )); |
| 229 | let at_end = iter.as_slice(); |
| 230 | assert_eq!(at_start.len(), 2); |
| 231 | assert_eq!(in_middle.len(), 1); |
| 232 | assert_eq!(at_end.len(), 0); |
| 233 | assert_eq!(at_start[0], 0x0061u16); |
| 234 | assert_eq!(at_start[1], 0x0062u16); |
| 235 | assert_eq!(in_middle[0], 0x0062u16); |
| 236 | } |
| 237 | } |
| 238 | |