| 1 | use core::iter::Rev; |
| 2 | |
| 3 | use crate::arch::generic::memchr as generic; |
| 4 | |
| 5 | /// Search for the first occurrence of a byte in a slice. |
| 6 | /// |
| 7 | /// This returns the index corresponding to the first occurrence of `needle` in |
| 8 | /// `haystack`, or `None` if one is not found. If an index is returned, it is |
| 9 | /// guaranteed to be less than `haystack.len()`. |
| 10 | /// |
| 11 | /// While this is semantically the same as something like |
| 12 | /// `haystack.iter().position(|&b| b == needle)`, this routine will attempt to |
| 13 | /// use highly optimized vector operations that can be an order of magnitude |
| 14 | /// faster (or more). |
| 15 | /// |
| 16 | /// # Example |
| 17 | /// |
| 18 | /// This shows how to find the first position of a byte in a byte string. |
| 19 | /// |
| 20 | /// ``` |
| 21 | /// use memchr::memchr; |
| 22 | /// |
| 23 | /// let haystack = b"the quick brown fox" ; |
| 24 | /// assert_eq!(memchr(b'k' , haystack), Some(8)); |
| 25 | /// ``` |
| 26 | #[inline ] |
| 27 | pub fn memchr(needle: u8, haystack: &[u8]) -> Option<usize> { |
| 28 | // SAFETY: memchr_raw, when a match is found, always returns a valid |
| 29 | // pointer between start and end. |
| 30 | unsafe { |
| 31 | generic::search_slice_with_raw(haystack, |start: *const u8, end: *const u8| { |
| 32 | memchr_raw(needle, start, end) |
| 33 | }) |
| 34 | } |
| 35 | } |
| 36 | |
| 37 | /// Search for the last occurrence of a byte in a slice. |
| 38 | /// |
| 39 | /// This returns the index corresponding to the last occurrence of `needle` in |
| 40 | /// `haystack`, or `None` if one is not found. If an index is returned, it is |
| 41 | /// guaranteed to be less than `haystack.len()`. |
| 42 | /// |
| 43 | /// While this is semantically the same as something like |
| 44 | /// `haystack.iter().rposition(|&b| b == needle)`, this routine will attempt to |
| 45 | /// use highly optimized vector operations that can be an order of magnitude |
| 46 | /// faster (or more). |
| 47 | /// |
| 48 | /// # Example |
| 49 | /// |
| 50 | /// This shows how to find the last position of a byte in a byte string. |
| 51 | /// |
| 52 | /// ``` |
| 53 | /// use memchr::memrchr; |
| 54 | /// |
| 55 | /// let haystack = b"the quick brown fox" ; |
| 56 | /// assert_eq!(memrchr(b'o' , haystack), Some(17)); |
| 57 | /// ``` |
| 58 | #[inline ] |
| 59 | pub fn memrchr(needle: u8, haystack: &[u8]) -> Option<usize> { |
| 60 | // SAFETY: memrchr_raw, when a match is found, always returns a valid |
| 61 | // pointer between start and end. |
| 62 | unsafe { |
| 63 | generic::search_slice_with_raw(haystack, |start: *const u8, end: *const u8| { |
| 64 | memrchr_raw(needle, start, end) |
| 65 | }) |
| 66 | } |
| 67 | } |
| 68 | |
| 69 | /// Search for the first occurrence of two possible bytes in a haystack. |
| 70 | /// |
| 71 | /// This returns the index corresponding to the first occurrence of one of the |
| 72 | /// needle bytes in `haystack`, or `None` if one is not found. If an index is |
| 73 | /// returned, it is guaranteed to be less than `haystack.len()`. |
| 74 | /// |
| 75 | /// While this is semantically the same as something like |
| 76 | /// `haystack.iter().position(|&b| b == needle1 || b == needle2)`, this routine |
| 77 | /// will attempt to use highly optimized vector operations that can be an order |
| 78 | /// of magnitude faster (or more). |
| 79 | /// |
| 80 | /// # Example |
| 81 | /// |
| 82 | /// This shows how to find the first position of one of two possible bytes in a |
| 83 | /// haystack. |
| 84 | /// |
| 85 | /// ``` |
| 86 | /// use memchr::memchr2; |
| 87 | /// |
| 88 | /// let haystack = b"the quick brown fox" ; |
| 89 | /// assert_eq!(memchr2(b'k' , b'q' , haystack), Some(4)); |
| 90 | /// ``` |
| 91 | #[inline ] |
| 92 | pub fn memchr2(needle1: u8, needle2: u8, haystack: &[u8]) -> Option<usize> { |
| 93 | // SAFETY: memchr2_raw, when a match is found, always returns a valid |
| 94 | // pointer between start and end. |
| 95 | unsafe { |
| 96 | generic::search_slice_with_raw(haystack, |start: *const u8, end: *const u8| { |
| 97 | memchr2_raw(needle1, needle2, start, end) |
| 98 | }) |
| 99 | } |
| 100 | } |
| 101 | |
| 102 | /// Search for the last occurrence of two possible bytes in a haystack. |
| 103 | /// |
| 104 | /// This returns the index corresponding to the last occurrence of one of the |
| 105 | /// needle bytes in `haystack`, or `None` if one is not found. If an index is |
| 106 | /// returned, it is guaranteed to be less than `haystack.len()`. |
| 107 | /// |
| 108 | /// While this is semantically the same as something like |
| 109 | /// `haystack.iter().rposition(|&b| b == needle1 || b == needle2)`, this |
| 110 | /// routine will attempt to use highly optimized vector operations that can be |
| 111 | /// an order of magnitude faster (or more). |
| 112 | /// |
| 113 | /// # Example |
| 114 | /// |
| 115 | /// This shows how to find the last position of one of two possible bytes in a |
| 116 | /// haystack. |
| 117 | /// |
| 118 | /// ``` |
| 119 | /// use memchr::memrchr2; |
| 120 | /// |
| 121 | /// let haystack = b"the quick brown fox" ; |
| 122 | /// assert_eq!(memrchr2(b'k' , b'o' , haystack), Some(17)); |
| 123 | /// ``` |
| 124 | #[inline ] |
| 125 | pub fn memrchr2(needle1: u8, needle2: u8, haystack: &[u8]) -> Option<usize> { |
| 126 | // SAFETY: memrchr2_raw, when a match is found, always returns a valid |
| 127 | // pointer between start and end. |
| 128 | unsafe { |
| 129 | generic::search_slice_with_raw(haystack, |start: *const u8, end: *const u8| { |
| 130 | memrchr2_raw(needle1, needle2, start, end) |
| 131 | }) |
| 132 | } |
| 133 | } |
| 134 | |
| 135 | /// Search for the first occurrence of three possible bytes in a haystack. |
| 136 | /// |
| 137 | /// This returns the index corresponding to the first occurrence of one of the |
| 138 | /// needle bytes in `haystack`, or `None` if one is not found. If an index is |
| 139 | /// returned, it is guaranteed to be less than `haystack.len()`. |
| 140 | /// |
| 141 | /// While this is semantically the same as something like |
| 142 | /// `haystack.iter().position(|&b| b == needle1 || b == needle2 || b == needle3)`, |
| 143 | /// this routine will attempt to use highly optimized vector operations that |
| 144 | /// can be an order of magnitude faster (or more). |
| 145 | /// |
| 146 | /// # Example |
| 147 | /// |
| 148 | /// This shows how to find the first position of one of three possible bytes in |
| 149 | /// a haystack. |
| 150 | /// |
| 151 | /// ``` |
| 152 | /// use memchr::memchr3; |
| 153 | /// |
| 154 | /// let haystack = b"the quick brown fox" ; |
| 155 | /// assert_eq!(memchr3(b'k' , b'q' , b'u' , haystack), Some(4)); |
| 156 | /// ``` |
| 157 | #[inline ] |
| 158 | pub fn memchr3( |
| 159 | needle1: u8, |
| 160 | needle2: u8, |
| 161 | needle3: u8, |
| 162 | haystack: &[u8], |
| 163 | ) -> Option<usize> { |
| 164 | // SAFETY: memchr3_raw, when a match is found, always returns a valid |
| 165 | // pointer between start and end. |
| 166 | unsafe { |
| 167 | generic::search_slice_with_raw(haystack, |start: *const u8, end: *const u8| { |
| 168 | memchr3_raw(needle1, needle2, needle3, start, end) |
| 169 | }) |
| 170 | } |
| 171 | } |
| 172 | |
| 173 | /// Search for the last occurrence of three possible bytes in a haystack. |
| 174 | /// |
| 175 | /// This returns the index corresponding to the last occurrence of one of the |
| 176 | /// needle bytes in `haystack`, or `None` if one is not found. If an index is |
| 177 | /// returned, it is guaranteed to be less than `haystack.len()`. |
| 178 | /// |
| 179 | /// While this is semantically the same as something like |
| 180 | /// `haystack.iter().rposition(|&b| b == needle1 || b == needle2 || b == needle3)`, |
| 181 | /// this routine will attempt to use highly optimized vector operations that |
| 182 | /// can be an order of magnitude faster (or more). |
| 183 | /// |
| 184 | /// # Example |
| 185 | /// |
| 186 | /// This shows how to find the last position of one of three possible bytes in |
| 187 | /// a haystack. |
| 188 | /// |
| 189 | /// ``` |
| 190 | /// use memchr::memrchr3; |
| 191 | /// |
| 192 | /// let haystack = b"the quick brown fox" ; |
| 193 | /// assert_eq!(memrchr3(b'k' , b'o' , b'n' , haystack), Some(17)); |
| 194 | /// ``` |
| 195 | #[inline ] |
| 196 | pub fn memrchr3( |
| 197 | needle1: u8, |
| 198 | needle2: u8, |
| 199 | needle3: u8, |
| 200 | haystack: &[u8], |
| 201 | ) -> Option<usize> { |
| 202 | // SAFETY: memrchr3_raw, when a match is found, always returns a valid |
| 203 | // pointer between start and end. |
| 204 | unsafe { |
| 205 | generic::search_slice_with_raw(haystack, |start: *const u8, end: *const u8| { |
| 206 | memrchr3_raw(needle1, needle2, needle3, start, end) |
| 207 | }) |
| 208 | } |
| 209 | } |
| 210 | |
| 211 | /// Returns an iterator over all occurrences of the needle in a haystack. |
| 212 | /// |
| 213 | /// The iterator returned implements `DoubleEndedIterator`. This means it |
| 214 | /// can also be used to find occurrences in reverse order. |
| 215 | #[inline ] |
| 216 | pub fn memchr_iter<'h>(needle: u8, haystack: &'h [u8]) -> Memchr<'h> { |
| 217 | Memchr::new(needle1:needle, haystack) |
| 218 | } |
| 219 | |
| 220 | /// Returns an iterator over all occurrences of the needle in a haystack, in |
| 221 | /// reverse. |
| 222 | #[inline ] |
| 223 | pub fn memrchr_iter(needle: u8, haystack: &[u8]) -> Rev<Memchr<'_>> { |
| 224 | Memchr::new(needle1:needle, haystack).rev() |
| 225 | } |
| 226 | |
| 227 | /// Returns an iterator over all occurrences of the needles in a haystack. |
| 228 | /// |
| 229 | /// The iterator returned implements `DoubleEndedIterator`. This means it |
| 230 | /// can also be used to find occurrences in reverse order. |
| 231 | #[inline ] |
| 232 | pub fn memchr2_iter<'h>( |
| 233 | needle1: u8, |
| 234 | needle2: u8, |
| 235 | haystack: &'h [u8], |
| 236 | ) -> Memchr2<'h> { |
| 237 | Memchr2::new(needle1, needle2, haystack) |
| 238 | } |
| 239 | |
| 240 | /// Returns an iterator over all occurrences of the needles in a haystack, in |
| 241 | /// reverse. |
| 242 | #[inline ] |
| 243 | pub fn memrchr2_iter( |
| 244 | needle1: u8, |
| 245 | needle2: u8, |
| 246 | haystack: &[u8], |
| 247 | ) -> Rev<Memchr2<'_>> { |
| 248 | Memchr2::new(needle1, needle2, haystack).rev() |
| 249 | } |
| 250 | |
| 251 | /// Returns an iterator over all occurrences of the needles in a haystack. |
| 252 | /// |
| 253 | /// The iterator returned implements `DoubleEndedIterator`. This means it |
| 254 | /// can also be used to find occurrences in reverse order. |
| 255 | #[inline ] |
| 256 | pub fn memchr3_iter<'h>( |
| 257 | needle1: u8, |
| 258 | needle2: u8, |
| 259 | needle3: u8, |
| 260 | haystack: &'h [u8], |
| 261 | ) -> Memchr3<'h> { |
| 262 | Memchr3::new(needle1, needle2, needle3, haystack) |
| 263 | } |
| 264 | |
| 265 | /// Returns an iterator over all occurrences of the needles in a haystack, in |
| 266 | /// reverse. |
| 267 | #[inline ] |
| 268 | pub fn memrchr3_iter( |
| 269 | needle1: u8, |
| 270 | needle2: u8, |
| 271 | needle3: u8, |
| 272 | haystack: &[u8], |
| 273 | ) -> Rev<Memchr3<'_>> { |
| 274 | Memchr3::new(needle1, needle2, needle3, haystack).rev() |
| 275 | } |
| 276 | |
| 277 | /// An iterator over all occurrences of a single byte in a haystack. |
| 278 | /// |
| 279 | /// This iterator implements `DoubleEndedIterator`, which means it can also be |
| 280 | /// used to find occurrences in reverse order. |
| 281 | /// |
| 282 | /// This iterator is created by the [`memchr_iter`] or `[memrchr_iter`] |
| 283 | /// functions. It can also be created with the [`Memchr::new`] method. |
| 284 | /// |
| 285 | /// The lifetime parameter `'h` refers to the lifetime of the haystack being |
| 286 | /// searched. |
| 287 | #[derive (Clone, Debug)] |
| 288 | pub struct Memchr<'h> { |
| 289 | needle1: u8, |
| 290 | it: crate::arch::generic::memchr::Iter<'h>, |
| 291 | } |
| 292 | |
| 293 | impl<'h> Memchr<'h> { |
| 294 | /// Returns an iterator over all occurrences of the needle byte in the |
| 295 | /// given haystack. |
| 296 | /// |
| 297 | /// The iterator returned implements `DoubleEndedIterator`. This means it |
| 298 | /// can also be used to find occurrences in reverse order. |
| 299 | #[inline ] |
| 300 | pub fn new(needle1: u8, haystack: &'h [u8]) -> Memchr<'h> { |
| 301 | Memchr { |
| 302 | needle1, |
| 303 | it: crate::arch::generic::memchr::Iter::new(haystack), |
| 304 | } |
| 305 | } |
| 306 | } |
| 307 | |
| 308 | impl<'h> Iterator for Memchr<'h> { |
| 309 | type Item = usize; |
| 310 | |
| 311 | #[inline ] |
| 312 | fn next(&mut self) -> Option<usize> { |
| 313 | // SAFETY: All of our implementations of memchr ensure that any |
| 314 | // pointers returns will fall within the start and end bounds, and this |
| 315 | // upholds the safety contract of `self.it.next`. |
| 316 | unsafe { |
| 317 | // NOTE: I attempted to define an enum of previously created |
| 318 | // searchers and then switch on those here instead of just |
| 319 | // calling `memchr_raw` (or `One::new(..).find_raw(..)`). But |
| 320 | // that turned out to have a fair bit of extra overhead when |
| 321 | // searching very small haystacks. |
| 322 | self.it.next(|s, e| memchr_raw(self.needle1, s, e)) |
| 323 | } |
| 324 | } |
| 325 | |
| 326 | #[inline ] |
| 327 | fn count(self) -> usize { |
| 328 | self.it.count(|s, e| { |
| 329 | // SAFETY: We rely on our generic iterator to return valid start |
| 330 | // and end pointers. |
| 331 | unsafe { count_raw(self.needle1, s, e) } |
| 332 | }) |
| 333 | } |
| 334 | |
| 335 | #[inline ] |
| 336 | fn size_hint(&self) -> (usize, Option<usize>) { |
| 337 | self.it.size_hint() |
| 338 | } |
| 339 | } |
| 340 | |
| 341 | impl<'h> DoubleEndedIterator for Memchr<'h> { |
| 342 | #[inline ] |
| 343 | fn next_back(&mut self) -> Option<usize> { |
| 344 | // SAFETY: All of our implementations of memchr ensure that any |
| 345 | // pointers returns will fall within the start and end bounds, and this |
| 346 | // upholds the safety contract of `self.it.next_back`. |
| 347 | unsafe { self.it.next_back(|s: *const u8, e: *const u8| memrchr_raw(self.needle1, start:s, end:e)) } |
| 348 | } |
| 349 | } |
| 350 | |
| 351 | impl<'h> core::iter::FusedIterator for Memchr<'h> {} |
| 352 | |
| 353 | /// An iterator over all occurrences of two possible bytes in a haystack. |
| 354 | /// |
| 355 | /// This iterator implements `DoubleEndedIterator`, which means it can also be |
| 356 | /// used to find occurrences in reverse order. |
| 357 | /// |
| 358 | /// This iterator is created by the [`memchr2_iter`] or `[memrchr2_iter`] |
| 359 | /// functions. It can also be created with the [`Memchr2::new`] method. |
| 360 | /// |
| 361 | /// The lifetime parameter `'h` refers to the lifetime of the haystack being |
| 362 | /// searched. |
| 363 | #[derive (Clone, Debug)] |
| 364 | pub struct Memchr2<'h> { |
| 365 | needle1: u8, |
| 366 | needle2: u8, |
| 367 | it: crate::arch::generic::memchr::Iter<'h>, |
| 368 | } |
| 369 | |
| 370 | impl<'h> Memchr2<'h> { |
| 371 | /// Returns an iterator over all occurrences of the needle bytes in the |
| 372 | /// given haystack. |
| 373 | /// |
| 374 | /// The iterator returned implements `DoubleEndedIterator`. This means it |
| 375 | /// can also be used to find occurrences in reverse order. |
| 376 | #[inline ] |
| 377 | pub fn new(needle1: u8, needle2: u8, haystack: &'h [u8]) -> Memchr2<'h> { |
| 378 | Memchr2 { |
| 379 | needle1, |
| 380 | needle2, |
| 381 | it: crate::arch::generic::memchr::Iter::new(haystack), |
| 382 | } |
| 383 | } |
| 384 | } |
| 385 | |
| 386 | impl<'h> Iterator for Memchr2<'h> { |
| 387 | type Item = usize; |
| 388 | |
| 389 | #[inline ] |
| 390 | fn next(&mut self) -> Option<usize> { |
| 391 | // SAFETY: All of our implementations of memchr ensure that any |
| 392 | // pointers returns will fall within the start and end bounds, and this |
| 393 | // upholds the safety contract of `self.it.next`. |
| 394 | unsafe { |
| 395 | self.it.next(|s: *const u8, e: *const u8| memchr2_raw(self.needle1, self.needle2, start:s, end:e)) |
| 396 | } |
| 397 | } |
| 398 | |
| 399 | #[inline ] |
| 400 | fn size_hint(&self) -> (usize, Option<usize>) { |
| 401 | self.it.size_hint() |
| 402 | } |
| 403 | } |
| 404 | |
| 405 | impl<'h> DoubleEndedIterator for Memchr2<'h> { |
| 406 | #[inline ] |
| 407 | fn next_back(&mut self) -> Option<usize> { |
| 408 | // SAFETY: All of our implementations of memchr ensure that any |
| 409 | // pointers returns will fall within the start and end bounds, and this |
| 410 | // upholds the safety contract of `self.it.next_back`. |
| 411 | unsafe { |
| 412 | self.it.next_back(|s: *const u8, e: *const u8| { |
| 413 | memrchr2_raw(self.needle1, self.needle2, start:s, end:e) |
| 414 | }) |
| 415 | } |
| 416 | } |
| 417 | } |
| 418 | |
| 419 | impl<'h> core::iter::FusedIterator for Memchr2<'h> {} |
| 420 | |
| 421 | /// An iterator over all occurrences of three possible bytes in a haystack. |
| 422 | /// |
| 423 | /// This iterator implements `DoubleEndedIterator`, which means it can also be |
| 424 | /// used to find occurrences in reverse order. |
| 425 | /// |
| 426 | /// This iterator is created by the [`memchr2_iter`] or `[memrchr2_iter`] |
| 427 | /// functions. It can also be created with the [`Memchr3::new`] method. |
| 428 | /// |
| 429 | /// The lifetime parameter `'h` refers to the lifetime of the haystack being |
| 430 | /// searched. |
| 431 | #[derive (Clone, Debug)] |
| 432 | pub struct Memchr3<'h> { |
| 433 | needle1: u8, |
| 434 | needle2: u8, |
| 435 | needle3: u8, |
| 436 | it: crate::arch::generic::memchr::Iter<'h>, |
| 437 | } |
| 438 | |
| 439 | impl<'h> Memchr3<'h> { |
| 440 | /// Returns an iterator over all occurrences of the needle bytes in the |
| 441 | /// given haystack. |
| 442 | /// |
| 443 | /// The iterator returned implements `DoubleEndedIterator`. This means it |
| 444 | /// can also be used to find occurrences in reverse order. |
| 445 | #[inline ] |
| 446 | pub fn new( |
| 447 | needle1: u8, |
| 448 | needle2: u8, |
| 449 | needle3: u8, |
| 450 | haystack: &'h [u8], |
| 451 | ) -> Memchr3<'h> { |
| 452 | Memchr3 { |
| 453 | needle1, |
| 454 | needle2, |
| 455 | needle3, |
| 456 | it: crate::arch::generic::memchr::Iter::new(haystack), |
| 457 | } |
| 458 | } |
| 459 | } |
| 460 | |
| 461 | impl<'h> Iterator for Memchr3<'h> { |
| 462 | type Item = usize; |
| 463 | |
| 464 | #[inline ] |
| 465 | fn next(&mut self) -> Option<usize> { |
| 466 | // SAFETY: All of our implementations of memchr ensure that any |
| 467 | // pointers returns will fall within the start and end bounds, and this |
| 468 | // upholds the safety contract of `self.it.next`. |
| 469 | unsafe { |
| 470 | self.it.next(|s: *const u8, e: *const u8| { |
| 471 | memchr3_raw(self.needle1, self.needle2, self.needle3, start:s, end:e) |
| 472 | }) |
| 473 | } |
| 474 | } |
| 475 | |
| 476 | #[inline ] |
| 477 | fn size_hint(&self) -> (usize, Option<usize>) { |
| 478 | self.it.size_hint() |
| 479 | } |
| 480 | } |
| 481 | |
| 482 | impl<'h> DoubleEndedIterator for Memchr3<'h> { |
| 483 | #[inline ] |
| 484 | fn next_back(&mut self) -> Option<usize> { |
| 485 | // SAFETY: All of our implementations of memchr ensure that any |
| 486 | // pointers returns will fall within the start and end bounds, and this |
| 487 | // upholds the safety contract of `self.it.next_back`. |
| 488 | unsafe { |
| 489 | self.it.next_back(|s: *const u8, e: *const u8| { |
| 490 | memrchr3_raw(self.needle1, self.needle2, self.needle3, start:s, end:e) |
| 491 | }) |
| 492 | } |
| 493 | } |
| 494 | } |
| 495 | |
| 496 | impl<'h> core::iter::FusedIterator for Memchr3<'h> {} |
| 497 | |
| 498 | /// memchr, but using raw pointers to represent the haystack. |
| 499 | /// |
| 500 | /// # Safety |
| 501 | /// |
| 502 | /// Pointers must be valid. See `One::find_raw`. |
| 503 | #[inline ] |
| 504 | unsafe fn memchr_raw( |
| 505 | needle: u8, |
| 506 | start: *const u8, |
| 507 | end: *const u8, |
| 508 | ) -> Option<*const u8> { |
| 509 | #[cfg (target_arch = "x86_64" )] |
| 510 | { |
| 511 | // x86_64 does CPU feature detection at runtime in order to use AVX2 |
| 512 | // instructions even when the `avx2` feature isn't enabled at compile |
| 513 | // time. This function also handles using a fallback if neither AVX2 |
| 514 | // nor SSE2 (unusual) are available. |
| 515 | crate::arch::x86_64::memchr::memchr_raw(needle, start, end) |
| 516 | } |
| 517 | #[cfg (all(target_arch = "wasm32" , target_feature = "simd128" ))] |
| 518 | { |
| 519 | crate::arch::wasm32::memchr::memchr_raw(needle, start, end) |
| 520 | } |
| 521 | #[cfg (target_arch = "aarch64" )] |
| 522 | { |
| 523 | crate::arch::aarch64::memchr::memchr_raw(needle, start, end) |
| 524 | } |
| 525 | #[cfg (not(any( |
| 526 | target_arch = "x86_64" , |
| 527 | all(target_arch = "wasm32" , target_feature = "simd128" ), |
| 528 | target_arch = "aarch64" |
| 529 | )))] |
| 530 | { |
| 531 | crate::arch::all::memchr::One::new(needle).find_raw(start, end) |
| 532 | } |
| 533 | } |
| 534 | |
| 535 | /// memrchr, but using raw pointers to represent the haystack. |
| 536 | /// |
| 537 | /// # Safety |
| 538 | /// |
| 539 | /// Pointers must be valid. See `One::rfind_raw`. |
| 540 | #[inline ] |
| 541 | unsafe fn memrchr_raw( |
| 542 | needle: u8, |
| 543 | start: *const u8, |
| 544 | end: *const u8, |
| 545 | ) -> Option<*const u8> { |
| 546 | #[cfg (target_arch = "x86_64" )] |
| 547 | { |
| 548 | crate::arch::x86_64::memchr::memrchr_raw(n1:needle, start, end) |
| 549 | } |
| 550 | #[cfg (all(target_arch = "wasm32" , target_feature = "simd128" ))] |
| 551 | { |
| 552 | crate::arch::wasm32::memchr::memrchr_raw(needle, start, end) |
| 553 | } |
| 554 | #[cfg (target_arch = "aarch64" )] |
| 555 | { |
| 556 | crate::arch::aarch64::memchr::memrchr_raw(needle, start, end) |
| 557 | } |
| 558 | #[cfg (not(any( |
| 559 | target_arch = "x86_64" , |
| 560 | all(target_arch = "wasm32" , target_feature = "simd128" ), |
| 561 | target_arch = "aarch64" |
| 562 | )))] |
| 563 | { |
| 564 | crate::arch::all::memchr::One::new(needle).rfind_raw(start, end) |
| 565 | } |
| 566 | } |
| 567 | |
| 568 | /// memchr2, but using raw pointers to represent the haystack. |
| 569 | /// |
| 570 | /// # Safety |
| 571 | /// |
| 572 | /// Pointers must be valid. See `Two::find_raw`. |
| 573 | #[inline ] |
| 574 | unsafe fn memchr2_raw( |
| 575 | needle1: u8, |
| 576 | needle2: u8, |
| 577 | start: *const u8, |
| 578 | end: *const u8, |
| 579 | ) -> Option<*const u8> { |
| 580 | #[cfg (target_arch = "x86_64" )] |
| 581 | { |
| 582 | crate::arch::x86_64::memchr::memchr2_raw(n1:needle1, n2:needle2, start, end) |
| 583 | } |
| 584 | #[cfg (all(target_arch = "wasm32" , target_feature = "simd128" ))] |
| 585 | { |
| 586 | crate::arch::wasm32::memchr::memchr2_raw(needle1, needle2, start, end) |
| 587 | } |
| 588 | #[cfg (target_arch = "aarch64" )] |
| 589 | { |
| 590 | crate::arch::aarch64::memchr::memchr2_raw(needle1, needle2, start, end) |
| 591 | } |
| 592 | #[cfg (not(any( |
| 593 | target_arch = "x86_64" , |
| 594 | all(target_arch = "wasm32" , target_feature = "simd128" ), |
| 595 | target_arch = "aarch64" |
| 596 | )))] |
| 597 | { |
| 598 | crate::arch::all::memchr::Two::new(needle1, needle2) |
| 599 | .find_raw(start, end) |
| 600 | } |
| 601 | } |
| 602 | |
| 603 | /// memrchr2, but using raw pointers to represent the haystack. |
| 604 | /// |
| 605 | /// # Safety |
| 606 | /// |
| 607 | /// Pointers must be valid. See `Two::rfind_raw`. |
| 608 | #[inline ] |
| 609 | unsafe fn memrchr2_raw( |
| 610 | needle1: u8, |
| 611 | needle2: u8, |
| 612 | start: *const u8, |
| 613 | end: *const u8, |
| 614 | ) -> Option<*const u8> { |
| 615 | #[cfg (target_arch = "x86_64" )] |
| 616 | { |
| 617 | crate::arch::x86_64::memchr::memrchr2_raw(needle1, needle2, start, end) |
| 618 | } |
| 619 | #[cfg (all(target_arch = "wasm32" , target_feature = "simd128" ))] |
| 620 | { |
| 621 | crate::arch::wasm32::memchr::memrchr2_raw(needle1, needle2, start, end) |
| 622 | } |
| 623 | #[cfg (target_arch = "aarch64" )] |
| 624 | { |
| 625 | crate::arch::aarch64::memchr::memrchr2_raw( |
| 626 | needle1, needle2, start, end, |
| 627 | ) |
| 628 | } |
| 629 | #[cfg (not(any( |
| 630 | target_arch = "x86_64" , |
| 631 | all(target_arch = "wasm32" , target_feature = "simd128" ), |
| 632 | target_arch = "aarch64" |
| 633 | )))] |
| 634 | { |
| 635 | crate::arch::all::memchr::Two::new(needle1, needle2) |
| 636 | .rfind_raw(start, end) |
| 637 | } |
| 638 | } |
| 639 | |
| 640 | /// memchr3, but using raw pointers to represent the haystack. |
| 641 | /// |
| 642 | /// # Safety |
| 643 | /// |
| 644 | /// Pointers must be valid. See `Three::find_raw`. |
| 645 | #[inline ] |
| 646 | unsafe fn memchr3_raw( |
| 647 | needle1: u8, |
| 648 | needle2: u8, |
| 649 | needle3: u8, |
| 650 | start: *const u8, |
| 651 | end: *const u8, |
| 652 | ) -> Option<*const u8> { |
| 653 | #[cfg (target_arch = "x86_64" )] |
| 654 | { |
| 655 | crate::arch::x86_64::memchr::memchr3_raw( |
| 656 | needle1, needle2, needle3, start, end, |
| 657 | ) |
| 658 | } |
| 659 | #[cfg (all(target_arch = "wasm32" , target_feature = "simd128" ))] |
| 660 | { |
| 661 | crate::arch::wasm32::memchr::memchr3_raw( |
| 662 | needle1, needle2, needle3, start, end, |
| 663 | ) |
| 664 | } |
| 665 | #[cfg (target_arch = "aarch64" )] |
| 666 | { |
| 667 | crate::arch::aarch64::memchr::memchr3_raw( |
| 668 | needle1, needle2, needle3, start, end, |
| 669 | ) |
| 670 | } |
| 671 | #[cfg (not(any( |
| 672 | target_arch = "x86_64" , |
| 673 | all(target_arch = "wasm32" , target_feature = "simd128" ), |
| 674 | target_arch = "aarch64" |
| 675 | )))] |
| 676 | { |
| 677 | crate::arch::all::memchr::Three::new(needle1, needle2, needle3) |
| 678 | .find_raw(start, end) |
| 679 | } |
| 680 | } |
| 681 | |
| 682 | /// memrchr3, but using raw pointers to represent the haystack. |
| 683 | /// |
| 684 | /// # Safety |
| 685 | /// |
| 686 | /// Pointers must be valid. See `Three::rfind_raw`. |
| 687 | #[inline ] |
| 688 | unsafe fn memrchr3_raw( |
| 689 | needle1: u8, |
| 690 | needle2: u8, |
| 691 | needle3: u8, |
| 692 | start: *const u8, |
| 693 | end: *const u8, |
| 694 | ) -> Option<*const u8> { |
| 695 | #[cfg (target_arch = "x86_64" )] |
| 696 | { |
| 697 | crate::arch::x86_64::memchr::memrchr3_raw( |
| 698 | needle1, needle2, needle3, start, end, |
| 699 | ) |
| 700 | } |
| 701 | #[cfg (all(target_arch = "wasm32" , target_feature = "simd128" ))] |
| 702 | { |
| 703 | crate::arch::wasm32::memchr::memrchr3_raw( |
| 704 | needle1, needle2, needle3, start, end, |
| 705 | ) |
| 706 | } |
| 707 | #[cfg (target_arch = "aarch64" )] |
| 708 | { |
| 709 | crate::arch::aarch64::memchr::memrchr3_raw( |
| 710 | needle1, needle2, needle3, start, end, |
| 711 | ) |
| 712 | } |
| 713 | #[cfg (not(any( |
| 714 | target_arch = "x86_64" , |
| 715 | all(target_arch = "wasm32" , target_feature = "simd128" ), |
| 716 | target_arch = "aarch64" |
| 717 | )))] |
| 718 | { |
| 719 | crate::arch::all::memchr::Three::new(needle1, needle2, needle3) |
| 720 | .rfind_raw(start, end) |
| 721 | } |
| 722 | } |
| 723 | |
| 724 | /// Count all matching bytes, but using raw pointers to represent the haystack. |
| 725 | /// |
| 726 | /// # Safety |
| 727 | /// |
| 728 | /// Pointers must be valid. See `One::count_raw`. |
| 729 | #[inline ] |
| 730 | unsafe fn count_raw(needle: u8, start: *const u8, end: *const u8) -> usize { |
| 731 | #[cfg (target_arch = "x86_64" )] |
| 732 | { |
| 733 | crate::arch::x86_64::memchr::count_raw(n1:needle, start, end) |
| 734 | } |
| 735 | #[cfg (all(target_arch = "wasm32" , target_feature = "simd128" ))] |
| 736 | { |
| 737 | crate::arch::wasm32::memchr::count_raw(needle, start, end) |
| 738 | } |
| 739 | #[cfg (target_arch = "aarch64" )] |
| 740 | { |
| 741 | crate::arch::aarch64::memchr::count_raw(needle, start, end) |
| 742 | } |
| 743 | #[cfg (not(any( |
| 744 | target_arch = "x86_64" , |
| 745 | all(target_arch = "wasm32" , target_feature = "simd128" ), |
| 746 | target_arch = "aarch64" |
| 747 | )))] |
| 748 | { |
| 749 | crate::arch::all::memchr::One::new(needle).count_raw(start, end) |
| 750 | } |
| 751 | } |
| 752 | |
| 753 | #[cfg (test)] |
| 754 | mod tests { |
| 755 | use super::*; |
| 756 | |
| 757 | #[test ] |
| 758 | fn forward1_iter() { |
| 759 | crate::tests::memchr::Runner::new(1).forward_iter( |
| 760 | |haystack, needles| { |
| 761 | Some(memchr_iter(needles[0], haystack).collect()) |
| 762 | }, |
| 763 | ) |
| 764 | } |
| 765 | |
| 766 | #[test ] |
| 767 | fn forward1_oneshot() { |
| 768 | crate::tests::memchr::Runner::new(1).forward_oneshot( |
| 769 | |haystack, needles| Some(memchr(needles[0], haystack)), |
| 770 | ) |
| 771 | } |
| 772 | |
| 773 | #[test ] |
| 774 | fn reverse1_iter() { |
| 775 | crate::tests::memchr::Runner::new(1).reverse_iter( |
| 776 | |haystack, needles| { |
| 777 | Some(memrchr_iter(needles[0], haystack).collect()) |
| 778 | }, |
| 779 | ) |
| 780 | } |
| 781 | |
| 782 | #[test ] |
| 783 | fn reverse1_oneshot() { |
| 784 | crate::tests::memchr::Runner::new(1).reverse_oneshot( |
| 785 | |haystack, needles| Some(memrchr(needles[0], haystack)), |
| 786 | ) |
| 787 | } |
| 788 | |
| 789 | #[test ] |
| 790 | fn count1_iter() { |
| 791 | crate::tests::memchr::Runner::new(1).count_iter(|haystack, needles| { |
| 792 | Some(memchr_iter(needles[0], haystack).count()) |
| 793 | }) |
| 794 | } |
| 795 | |
| 796 | #[test ] |
| 797 | fn forward2_iter() { |
| 798 | crate::tests::memchr::Runner::new(2).forward_iter( |
| 799 | |haystack, needles| { |
| 800 | let n1 = needles.get(0).copied()?; |
| 801 | let n2 = needles.get(1).copied()?; |
| 802 | Some(memchr2_iter(n1, n2, haystack).collect()) |
| 803 | }, |
| 804 | ) |
| 805 | } |
| 806 | |
| 807 | #[test ] |
| 808 | fn forward2_oneshot() { |
| 809 | crate::tests::memchr::Runner::new(2).forward_oneshot( |
| 810 | |haystack, needles| { |
| 811 | let n1 = needles.get(0).copied()?; |
| 812 | let n2 = needles.get(1).copied()?; |
| 813 | Some(memchr2(n1, n2, haystack)) |
| 814 | }, |
| 815 | ) |
| 816 | } |
| 817 | |
| 818 | #[test ] |
| 819 | fn reverse2_iter() { |
| 820 | crate::tests::memchr::Runner::new(2).reverse_iter( |
| 821 | |haystack, needles| { |
| 822 | let n1 = needles.get(0).copied()?; |
| 823 | let n2 = needles.get(1).copied()?; |
| 824 | Some(memrchr2_iter(n1, n2, haystack).collect()) |
| 825 | }, |
| 826 | ) |
| 827 | } |
| 828 | |
| 829 | #[test ] |
| 830 | fn reverse2_oneshot() { |
| 831 | crate::tests::memchr::Runner::new(2).reverse_oneshot( |
| 832 | |haystack, needles| { |
| 833 | let n1 = needles.get(0).copied()?; |
| 834 | let n2 = needles.get(1).copied()?; |
| 835 | Some(memrchr2(n1, n2, haystack)) |
| 836 | }, |
| 837 | ) |
| 838 | } |
| 839 | |
| 840 | #[test ] |
| 841 | fn forward3_iter() { |
| 842 | crate::tests::memchr::Runner::new(3).forward_iter( |
| 843 | |haystack, needles| { |
| 844 | let n1 = needles.get(0).copied()?; |
| 845 | let n2 = needles.get(1).copied()?; |
| 846 | let n3 = needles.get(2).copied()?; |
| 847 | Some(memchr3_iter(n1, n2, n3, haystack).collect()) |
| 848 | }, |
| 849 | ) |
| 850 | } |
| 851 | |
| 852 | #[test ] |
| 853 | fn forward3_oneshot() { |
| 854 | crate::tests::memchr::Runner::new(3).forward_oneshot( |
| 855 | |haystack, needles| { |
| 856 | let n1 = needles.get(0).copied()?; |
| 857 | let n2 = needles.get(1).copied()?; |
| 858 | let n3 = needles.get(2).copied()?; |
| 859 | Some(memchr3(n1, n2, n3, haystack)) |
| 860 | }, |
| 861 | ) |
| 862 | } |
| 863 | |
| 864 | #[test ] |
| 865 | fn reverse3_iter() { |
| 866 | crate::tests::memchr::Runner::new(3).reverse_iter( |
| 867 | |haystack, needles| { |
| 868 | let n1 = needles.get(0).copied()?; |
| 869 | let n2 = needles.get(1).copied()?; |
| 870 | let n3 = needles.get(2).copied()?; |
| 871 | Some(memrchr3_iter(n1, n2, n3, haystack).collect()) |
| 872 | }, |
| 873 | ) |
| 874 | } |
| 875 | |
| 876 | #[test ] |
| 877 | fn reverse3_oneshot() { |
| 878 | crate::tests::memchr::Runner::new(3).reverse_oneshot( |
| 879 | |haystack, needles| { |
| 880 | let n1 = needles.get(0).copied()?; |
| 881 | let n2 = needles.get(1).copied()?; |
| 882 | let n3 = needles.get(2).copied()?; |
| 883 | Some(memrchr3(n1, n2, n3, haystack)) |
| 884 | }, |
| 885 | ) |
| 886 | } |
| 887 | |
| 888 | // Prior to memchr 2.6, the memchr iterators both implemented Send and |
| 889 | // Sync. But in memchr 2.6, the iterator changed to use raw pointers |
| 890 | // internally and I didn't add explicit Send/Sync impls. This ended up |
| 891 | // regressing the API. This test ensures we don't do that again. |
| 892 | // |
| 893 | // See: https://github.com/BurntSushi/memchr/issues/133 |
| 894 | #[test ] |
| 895 | fn sync_regression() { |
| 896 | use core::panic::{RefUnwindSafe, UnwindSafe}; |
| 897 | |
| 898 | fn assert_send_sync<T: Send + Sync + UnwindSafe + RefUnwindSafe>() {} |
| 899 | assert_send_sync::<Memchr>(); |
| 900 | assert_send_sync::<Memchr2>(); |
| 901 | assert_send_sync::<Memchr3>() |
| 902 | } |
| 903 | } |
| 904 | |