1 | // Copyright Mozilla Foundation. See the COPYRIGHT |
2 | // file at the top-level directory of this distribution. |
3 | // |
4 | // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or |
5 | // https://www.apache.org/licenses/LICENSE-2.0> or the MIT license |
6 | // <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your |
7 | // option. This file may not be copied, modified, or distributed |
8 | // except according to those terms. |
9 | |
10 | //! Functions for converting between different in-RAM representations of text |
11 | //! and for quickly checking if the Unicode Bidirectional Algorithm can be |
12 | //! avoided. |
13 | //! |
14 | //! By using slices for output, the functions here seek to enable by-register |
15 | //! (ALU register or SIMD register as available) operations in order to |
16 | //! outperform iterator-based conversions available in the Rust standard |
17 | //! library. |
18 | //! |
19 | //! _Note:_ "Latin1" in this module refers to the Unicode range from U+0000 to |
20 | //! U+00FF, inclusive, and does not refer to the windows-1252 range. This |
21 | //! in-memory encoding is sometimes used as a storage optimization of text |
22 | //! when UTF-16 indexing and length semantics are exposed. |
23 | //! |
24 | //! The FFI binding for this module are in the |
25 | //! [encoding_c_mem crate](https://github.com/hsivonen/encoding_c_mem). |
26 | |
27 | #[cfg (feature = "alloc" )] |
28 | use alloc::borrow::Cow; |
29 | #[cfg (feature = "alloc" )] |
30 | use alloc::string::String; |
31 | #[cfg (feature = "alloc" )] |
32 | use alloc::vec::Vec; |
33 | |
34 | use super::in_inclusive_range16; |
35 | use super::in_inclusive_range32; |
36 | use super::in_inclusive_range8; |
37 | use super::in_range16; |
38 | use super::in_range32; |
39 | use super::DecoderResult; |
40 | use crate::ascii::*; |
41 | use crate::utf_8::*; |
42 | |
43 | macro_rules! non_fuzz_debug_assert { |
44 | ($($arg:tt)*) => (if !cfg!(fuzzing) { debug_assert!($($arg)*); }) |
45 | } |
46 | |
47 | cfg_if! { |
48 | if #[cfg(feature = "simd-accel" )] { |
49 | use ::core::intrinsics::likely; |
50 | use ::core::intrinsics::unlikely; |
51 | } else { |
52 | #[inline (always)] |
53 | fn likely(b: bool) -> bool { |
54 | b |
55 | } |
56 | #[inline (always)] |
57 | fn unlikely(b: bool) -> bool { |
58 | b |
59 | } |
60 | } |
61 | } |
62 | |
63 | /// Classification of text as Latin1 (all code points are below U+0100), |
64 | /// left-to-right with some non-Latin1 characters or as containing at least |
65 | /// some right-to-left characters. |
66 | #[must_use ] |
67 | #[derive (Debug, PartialEq, Eq)] |
68 | #[repr (C)] |
69 | pub enum Latin1Bidi { |
70 | /// Every character is below U+0100. |
71 | Latin1 = 0, |
72 | /// There is at least one character that's U+0100 or higher, but there |
73 | /// are no right-to-left characters. |
74 | LeftToRight = 1, |
75 | /// There is at least one right-to-left character. |
76 | Bidi = 2, |
77 | } |
78 | |
79 | // `as` truncates, so works on 32-bit, too. |
80 | #[allow (dead_code)] |
81 | const LATIN1_MASK: usize = 0xFF00_FF00_FF00_FF00u64 as usize; |
82 | |
83 | #[allow (unused_macros)] |
84 | macro_rules! by_unit_check_alu { |
85 | ($name:ident, $unit:ty, $bound:expr, $mask:ident) => { |
86 | #[cfg_attr(feature = "cargo-clippy" , allow(cast_ptr_alignment))] |
87 | #[inline(always)] |
88 | fn $name(buffer: &[$unit]) -> bool { |
89 | let mut offset = 0usize; |
90 | let mut accu = 0usize; |
91 | let unit_size = ::core::mem::size_of::<$unit>(); |
92 | let len = buffer.len(); |
93 | if len >= ALU_ALIGNMENT / unit_size { |
94 | // The most common reason to return `false` is for the first code |
95 | // unit to fail the test, so check that first. |
96 | if buffer[0] >= $bound { |
97 | return false; |
98 | } |
99 | let src = buffer.as_ptr(); |
100 | let mut until_alignment = ((ALU_ALIGNMENT - ((src as usize) & ALU_ALIGNMENT_MASK)) |
101 | & ALU_ALIGNMENT_MASK) |
102 | / unit_size; |
103 | if until_alignment + ALU_ALIGNMENT / unit_size <= len { |
104 | if until_alignment != 0 { |
105 | accu |= buffer[offset] as usize; |
106 | offset += 1; |
107 | until_alignment -= 1; |
108 | while until_alignment != 0 { |
109 | accu |= buffer[offset] as usize; |
110 | offset += 1; |
111 | until_alignment -= 1; |
112 | } |
113 | if accu >= $bound { |
114 | return false; |
115 | } |
116 | } |
117 | let len_minus_stride = len - ALU_ALIGNMENT / unit_size; |
118 | if offset + (4 * (ALU_ALIGNMENT / unit_size)) <= len { |
119 | // Safety: the above check lets us perform 4 consecutive reads of |
120 | // length ALU_ALIGNMENT / unit_size. ALU_ALIGNMENT is the size of usize, and unit_size |
121 | // is the size of the `src` pointer, so this is equal to performing four usize reads. |
122 | // |
123 | // This invariant is upheld on all loop iterations |
124 | let len_minus_unroll = len - (4 * (ALU_ALIGNMENT / unit_size)); |
125 | loop { |
126 | let unroll_accu = unsafe { *(src.add(offset) as *const usize) } |
127 | | unsafe { |
128 | *(src.add(offset + (ALU_ALIGNMENT / unit_size)) as *const usize) |
129 | } |
130 | | unsafe { |
131 | *(src.add(offset + (2 * (ALU_ALIGNMENT / unit_size))) |
132 | as *const usize) |
133 | } |
134 | | unsafe { |
135 | *(src.add(offset + (3 * (ALU_ALIGNMENT / unit_size))) |
136 | as *const usize) |
137 | }; |
138 | if unroll_accu & $mask != 0 { |
139 | return false; |
140 | } |
141 | offset += 4 * (ALU_ALIGNMENT / unit_size); |
142 | // Safety: this check lets us continue to perform the 4 reads earlier |
143 | if offset > len_minus_unroll { |
144 | break; |
145 | } |
146 | } |
147 | } |
148 | while offset <= len_minus_stride { |
149 | // Safety: the above check lets us perform one usize read. |
150 | accu |= unsafe { *(src.add(offset) as *const usize) }; |
151 | offset += ALU_ALIGNMENT / unit_size; |
152 | } |
153 | } |
154 | } |
155 | for &unit in &buffer[offset..] { |
156 | accu |= unit as usize; |
157 | } |
158 | accu & $mask == 0 |
159 | } |
160 | }; |
161 | } |
162 | |
163 | #[allow (unused_macros)] |
164 | macro_rules! by_unit_check_simd { |
165 | ($name:ident, $unit:ty, $splat:expr, $simd_ty:ty, $bound:expr, $func:ident) => { |
166 | #[inline(always)] |
167 | fn $name(buffer: &[$unit]) -> bool { |
168 | let mut offset = 0usize; |
169 | let mut accu = 0usize; |
170 | let unit_size = ::core::mem::size_of::<$unit>(); |
171 | let len = buffer.len(); |
172 | if len >= SIMD_STRIDE_SIZE / unit_size { |
173 | // The most common reason to return `false` is for the first code |
174 | // unit to fail the test, so check that first. |
175 | if buffer[0] >= $bound { |
176 | return false; |
177 | } |
178 | let src = buffer.as_ptr(); |
179 | let mut until_alignment = ((SIMD_ALIGNMENT |
180 | - ((src as usize) & SIMD_ALIGNMENT_MASK)) |
181 | & SIMD_ALIGNMENT_MASK) |
182 | / unit_size; |
183 | if until_alignment + SIMD_STRIDE_SIZE / unit_size <= len { |
184 | if until_alignment != 0 { |
185 | accu |= buffer[offset] as usize; |
186 | offset += 1; |
187 | until_alignment -= 1; |
188 | while until_alignment != 0 { |
189 | accu |= buffer[offset] as usize; |
190 | offset += 1; |
191 | until_alignment -= 1; |
192 | } |
193 | if accu >= $bound { |
194 | return false; |
195 | } |
196 | } |
197 | let len_minus_stride = len - SIMD_STRIDE_SIZE / unit_size; |
198 | if offset + (4 * (SIMD_STRIDE_SIZE / unit_size)) <= len { |
199 | // Safety: the above check lets us perform 4 consecutive reads of |
200 | // length SIMD_STRIDE_SIZE / unit_size. SIMD_STRIDE_SIZE is the size of $simd_ty, and unit_size |
201 | // is the size of the `src` pointer, so this is equal to performing four $simd_ty reads. |
202 | // |
203 | // This invariant is upheld on all loop iterations |
204 | let len_minus_unroll = len - (4 * (SIMD_STRIDE_SIZE / unit_size)); |
205 | loop { |
206 | let unroll_accu = unsafe { *(src.add(offset) as *const $simd_ty) } |
207 | | unsafe { |
208 | *(src.add(offset + (SIMD_STRIDE_SIZE / unit_size)) |
209 | as *const $simd_ty) |
210 | } |
211 | | unsafe { |
212 | *(src.add(offset + (2 * (SIMD_STRIDE_SIZE / unit_size))) |
213 | as *const $simd_ty) |
214 | } |
215 | | unsafe { |
216 | *(src.add(offset + (3 * (SIMD_STRIDE_SIZE / unit_size))) |
217 | as *const $simd_ty) |
218 | }; |
219 | if !$func(unroll_accu) { |
220 | return false; |
221 | } |
222 | offset += 4 * (SIMD_STRIDE_SIZE / unit_size); |
223 | // Safety: this check lets us continue to perform the 4 reads earlier |
224 | if offset > len_minus_unroll { |
225 | break; |
226 | } |
227 | } |
228 | } |
229 | let mut simd_accu = $splat; |
230 | while offset <= len_minus_stride { |
231 | // Safety: the above check lets us perform one $simd_ty read. |
232 | simd_accu = simd_accu | unsafe { *(src.add(offset) as *const $simd_ty) }; |
233 | offset += SIMD_STRIDE_SIZE / unit_size; |
234 | } |
235 | if !$func(simd_accu) { |
236 | return false; |
237 | } |
238 | } |
239 | } |
240 | for &unit in &buffer[offset..] { |
241 | accu |= unit as usize; |
242 | } |
243 | accu < $bound |
244 | } |
245 | }; |
246 | } |
247 | |
248 | cfg_if! { |
249 | if #[cfg(all(feature = "simd-accel" , any(target_feature = "sse2" , all(target_endian = "little" , target_arch = "aarch64" ), all(target_endian = "little" , target_feature = "neon" ))))] { |
250 | use crate::simd_funcs::*; |
251 | use core::simd::u8x16; |
252 | use core::simd::u16x8; |
253 | |
254 | const SIMD_ALIGNMENT: usize = 16; |
255 | |
256 | const SIMD_ALIGNMENT_MASK: usize = 15; |
257 | |
258 | by_unit_check_simd!(is_ascii_impl, u8, u8x16::splat(0), u8x16, 0x80, simd_is_ascii); |
259 | by_unit_check_simd!(is_basic_latin_impl, u16, u16x8::splat(0), u16x8, 0x80, simd_is_basic_latin); |
260 | by_unit_check_simd!(is_utf16_latin1_impl, u16, u16x8::splat(0), u16x8, 0x100, simd_is_latin1); |
261 | |
262 | #[inline(always)] |
263 | fn utf16_valid_up_to_impl(buffer: &[u16]) -> usize { |
264 | // This function is a mess, because it simultaneously tries to do |
265 | // only aligned SIMD (perhaps misguidedly) and needs to deal with |
266 | // the last code unit in a SIMD stride being part of a valid |
267 | // surrogate pair. |
268 | let unit_size = ::core::mem::size_of::<u16>(); |
269 | let src = buffer.as_ptr(); |
270 | let len = buffer.len(); |
271 | let mut offset = 0usize; |
272 | 'outer: loop { |
273 | let until_alignment = ((SIMD_ALIGNMENT - ((unsafe { src.add(offset) } as usize) & SIMD_ALIGNMENT_MASK)) & |
274 | SIMD_ALIGNMENT_MASK) / unit_size; |
275 | if until_alignment == 0 { |
276 | if offset + SIMD_STRIDE_SIZE / unit_size > len { |
277 | break; |
278 | } |
279 | } else { |
280 | let offset_plus_until_alignment = offset + until_alignment; |
281 | let offset_plus_until_alignment_plus_one = offset_plus_until_alignment + 1; |
282 | if offset_plus_until_alignment_plus_one + SIMD_STRIDE_SIZE / unit_size > len { |
283 | break; |
284 | } |
285 | let (up_to, last_valid_low) = utf16_valid_up_to_alu(&buffer[offset..offset_plus_until_alignment_plus_one]); |
286 | if up_to < until_alignment { |
287 | return offset + up_to; |
288 | } |
289 | if last_valid_low { |
290 | offset = offset_plus_until_alignment_plus_one; |
291 | continue; |
292 | } |
293 | offset = offset_plus_until_alignment; |
294 | } |
295 | let len_minus_stride = len - SIMD_STRIDE_SIZE / unit_size; |
296 | loop { |
297 | let offset_plus_stride = offset + SIMD_STRIDE_SIZE / unit_size; |
298 | if contains_surrogates(unsafe { *(src.add(offset) as *const u16x8) }) { |
299 | if offset_plus_stride == len { |
300 | break 'outer; |
301 | } |
302 | let offset_plus_stride_plus_one = offset_plus_stride + 1; |
303 | let (up_to, last_valid_low) = utf16_valid_up_to_alu(&buffer[offset..offset_plus_stride_plus_one]); |
304 | if up_to < SIMD_STRIDE_SIZE / unit_size { |
305 | return offset + up_to; |
306 | } |
307 | if last_valid_low { |
308 | offset = offset_plus_stride_plus_one; |
309 | continue 'outer; |
310 | } |
311 | } |
312 | offset = offset_plus_stride; |
313 | if offset > len_minus_stride { |
314 | break 'outer; |
315 | } |
316 | } |
317 | } |
318 | let (up_to, _) = utf16_valid_up_to_alu(&buffer[offset..]); |
319 | offset + up_to |
320 | } |
321 | } else { |
322 | by_unit_check_alu!(is_ascii_impl, u8, 0x80, ASCII_MASK); |
323 | by_unit_check_alu!(is_basic_latin_impl, u16, 0x80, BASIC_LATIN_MASK); |
324 | by_unit_check_alu!(is_utf16_latin1_impl, u16, 0x100, LATIN1_MASK); |
325 | |
326 | #[inline (always)] |
327 | fn utf16_valid_up_to_impl(buffer: &[u16]) -> usize { |
328 | let (up_to, _) = utf16_valid_up_to_alu(buffer); |
329 | up_to |
330 | } |
331 | } |
332 | } |
333 | |
334 | /// The second return value is true iff the last code unit of the slice was |
335 | /// reached and turned out to be a low surrogate that is part of a valid pair. |
336 | #[cfg_attr (feature = "cargo-clippy" , allow(collapsible_if))] |
337 | #[inline (always)] |
338 | fn utf16_valid_up_to_alu(buffer: &[u16]) -> (usize, bool) { |
339 | let len = buffer.len(); |
340 | if len == 0 { |
341 | return (0, false); |
342 | } |
343 | let mut offset = 0usize; |
344 | loop { |
345 | let unit = buffer[offset]; |
346 | let next = offset + 1; |
347 | let unit_minus_surrogate_start = unit.wrapping_sub(0xD800); |
348 | if unit_minus_surrogate_start > (0xDFFF - 0xD800) { |
349 | // Not a surrogate |
350 | offset = next; |
351 | if offset == len { |
352 | return (offset, false); |
353 | } |
354 | continue; |
355 | } |
356 | if unit_minus_surrogate_start <= (0xDBFF - 0xD800) { |
357 | // high surrogate |
358 | if next < len { |
359 | let second = buffer[next]; |
360 | let second_minus_low_surrogate_start = second.wrapping_sub(0xDC00); |
361 | if second_minus_low_surrogate_start <= (0xDFFF - 0xDC00) { |
362 | // The next code unit is a low surrogate. Advance position. |
363 | offset = next + 1; |
364 | if offset == len { |
365 | return (offset, true); |
366 | } |
367 | continue; |
368 | } |
369 | // The next code unit is not a low surrogate. Don't advance |
370 | // position and treat the high surrogate as unpaired. |
371 | // fall through |
372 | } |
373 | // Unpaired, fall through |
374 | } |
375 | // Unpaired surrogate |
376 | return (offset, false); |
377 | } |
378 | } |
379 | |
380 | cfg_if! { |
381 | if #[cfg(all(feature = "simd-accel" , any(target_feature = "sse2" , all(target_endian = "little" , target_arch = "aarch64" ), all(target_endian = "little" , target_feature = "neon" ))))] { |
382 | #[inline(always)] |
383 | fn is_str_latin1_impl(buffer: &str) -> Option<usize> { |
384 | let mut offset = 0usize; |
385 | let bytes = buffer.as_bytes(); |
386 | let len = bytes.len(); |
387 | if len >= SIMD_STRIDE_SIZE { |
388 | let src = bytes.as_ptr(); |
389 | let mut until_alignment = (SIMD_ALIGNMENT - ((src as usize) & SIMD_ALIGNMENT_MASK)) & |
390 | SIMD_ALIGNMENT_MASK; |
391 | if until_alignment + SIMD_STRIDE_SIZE <= len { |
392 | while until_alignment != 0 { |
393 | if bytes[offset] > 0xC3 { |
394 | return Some(offset); |
395 | } |
396 | offset += 1; |
397 | until_alignment -= 1; |
398 | } |
399 | let len_minus_stride = len - SIMD_STRIDE_SIZE; |
400 | loop { |
401 | if !simd_is_str_latin1(unsafe { *(src.add(offset) as *const u8x16) }) { |
402 | // TODO: Ensure this compiles away when inlined into `is_str_latin1()`. |
403 | while bytes[offset] & 0xC0 == 0x80 { |
404 | offset += 1; |
405 | } |
406 | return Some(offset); |
407 | } |
408 | offset += SIMD_STRIDE_SIZE; |
409 | if offset > len_minus_stride { |
410 | break; |
411 | } |
412 | } |
413 | } |
414 | } |
415 | for i in offset..len { |
416 | if bytes[i] > 0xC3 { |
417 | return Some(i); |
418 | } |
419 | } |
420 | None |
421 | } |
422 | } else { |
423 | #[inline (always)] |
424 | fn is_str_latin1_impl(buffer: &str) -> Option<usize> { |
425 | let mut bytes = buffer.as_bytes(); |
426 | let mut total = 0; |
427 | loop { |
428 | if let Some((byte, offset)) = validate_ascii(bytes) { |
429 | total += offset; |
430 | if byte > 0xC3 { |
431 | return Some(total); |
432 | } |
433 | bytes = &bytes[offset + 2..]; |
434 | total += 2; |
435 | } else { |
436 | return None; |
437 | } |
438 | } |
439 | } |
440 | } |
441 | } |
442 | |
443 | #[inline (always)] |
444 | fn is_utf8_latin1_impl(buffer: &[u8]) -> Option<usize> { |
445 | let mut bytes: &[u8] = buffer; |
446 | let mut total: usize = 0; |
447 | loop { |
448 | if let Some((byte: u8, offset: usize)) = validate_ascii(slice:bytes) { |
449 | total += offset; |
450 | if in_inclusive_range8(i:byte, start:0xC2, end:0xC3) { |
451 | let next: usize = offset + 1; |
452 | if next == bytes.len() { |
453 | return Some(total); |
454 | } |
455 | if bytes[next] & 0xC0 != 0x80 { |
456 | return Some(total); |
457 | } |
458 | bytes = &bytes[offset + 2..]; |
459 | total += 2; |
460 | } else { |
461 | return Some(total); |
462 | } |
463 | } else { |
464 | return None; |
465 | } |
466 | } |
467 | } |
468 | |
469 | cfg_if! { |
470 | if #[cfg(all(feature = "simd-accel" , any(target_feature = "sse2" , all(target_endian = "little" , target_arch = "aarch64" ), all(target_endian = "little" , target_feature = "neon" ))))] { |
471 | #[inline(always)] |
472 | fn is_utf16_bidi_impl(buffer: &[u16]) -> bool { |
473 | let mut offset = 0usize; |
474 | let len = buffer.len(); |
475 | if len >= SIMD_STRIDE_SIZE / 2 { |
476 | let src = buffer.as_ptr(); |
477 | let mut until_alignment = ((SIMD_ALIGNMENT - ((src as usize) & SIMD_ALIGNMENT_MASK)) & |
478 | SIMD_ALIGNMENT_MASK) / 2; |
479 | if until_alignment + (SIMD_STRIDE_SIZE / 2) <= len { |
480 | while until_alignment != 0 { |
481 | if is_utf16_code_unit_bidi(buffer[offset]) { |
482 | return true; |
483 | } |
484 | offset += 1; |
485 | until_alignment -= 1; |
486 | } |
487 | let len_minus_stride = len - (SIMD_STRIDE_SIZE / 2); |
488 | loop { |
489 | if is_u16x8_bidi(unsafe { *(src.add(offset) as *const u16x8) }) { |
490 | return true; |
491 | } |
492 | offset += SIMD_STRIDE_SIZE / 2; |
493 | if offset > len_minus_stride { |
494 | break; |
495 | } |
496 | } |
497 | } |
498 | } |
499 | for &u in &buffer[offset..] { |
500 | if is_utf16_code_unit_bidi(u) { |
501 | return true; |
502 | } |
503 | } |
504 | false |
505 | } |
506 | } else { |
507 | #[inline (always)] |
508 | fn is_utf16_bidi_impl(buffer: &[u16]) -> bool { |
509 | for &u in buffer { |
510 | if is_utf16_code_unit_bidi(u) { |
511 | return true; |
512 | } |
513 | } |
514 | false |
515 | } |
516 | } |
517 | } |
518 | |
519 | cfg_if! { |
520 | if #[cfg(all(feature = "simd-accel" , any(target_feature = "sse2" , all(target_endian = "little" , target_arch = "aarch64" ), all(target_endian = "little" , target_feature = "neon" ))))] { |
521 | #[inline(always)] |
522 | fn check_utf16_for_latin1_and_bidi_impl(buffer: &[u16]) -> Latin1Bidi { |
523 | let mut offset = 0usize; |
524 | let len = buffer.len(); |
525 | if len >= SIMD_STRIDE_SIZE / 2 { |
526 | let src = buffer.as_ptr(); |
527 | let mut until_alignment = ((SIMD_ALIGNMENT - ((src as usize) & SIMD_ALIGNMENT_MASK)) & |
528 | SIMD_ALIGNMENT_MASK) / 2; |
529 | if until_alignment + (SIMD_STRIDE_SIZE / 2) <= len { |
530 | while until_alignment != 0 { |
531 | if buffer[offset] > 0xFF { |
532 | // This transition isn't optimal, since the aligment is recomputing |
533 | // but not tweaking further today. |
534 | if is_utf16_bidi_impl(&buffer[offset..]) { |
535 | return Latin1Bidi::Bidi; |
536 | } |
537 | return Latin1Bidi::LeftToRight; |
538 | } |
539 | offset += 1; |
540 | until_alignment -= 1; |
541 | } |
542 | let len_minus_stride = len - (SIMD_STRIDE_SIZE / 2); |
543 | loop { |
544 | let mut s = unsafe { *(src.add(offset) as *const u16x8) }; |
545 | if !simd_is_latin1(s) { |
546 | loop { |
547 | if is_u16x8_bidi(s) { |
548 | return Latin1Bidi::Bidi; |
549 | } |
550 | offset += SIMD_STRIDE_SIZE / 2; |
551 | if offset > len_minus_stride { |
552 | for &u in &buffer[offset..] { |
553 | if is_utf16_code_unit_bidi(u) { |
554 | return Latin1Bidi::Bidi; |
555 | } |
556 | } |
557 | return Latin1Bidi::LeftToRight; |
558 | } |
559 | s = unsafe { *(src.add(offset) as *const u16x8) }; |
560 | } |
561 | } |
562 | offset += SIMD_STRIDE_SIZE / 2; |
563 | if offset > len_minus_stride { |
564 | break; |
565 | } |
566 | } |
567 | } |
568 | } |
569 | let mut iter = (&buffer[offset..]).iter(); |
570 | loop { |
571 | if let Some(&u) = iter.next() { |
572 | if u > 0xFF { |
573 | let mut inner_u = u; |
574 | loop { |
575 | if is_utf16_code_unit_bidi(inner_u) { |
576 | return Latin1Bidi::Bidi; |
577 | } |
578 | if let Some(&code_unit) = iter.next() { |
579 | inner_u = code_unit; |
580 | } else { |
581 | return Latin1Bidi::LeftToRight; |
582 | } |
583 | } |
584 | } |
585 | } else { |
586 | return Latin1Bidi::Latin1; |
587 | } |
588 | } |
589 | } |
590 | } else { |
591 | #[cfg_attr (feature = "cargo-clippy" , allow(cast_ptr_alignment))] |
592 | #[inline (always)] |
593 | fn check_utf16_for_latin1_and_bidi_impl(buffer: &[u16]) -> Latin1Bidi { |
594 | let mut offset = 0usize; |
595 | let len = buffer.len(); |
596 | if len >= ALU_ALIGNMENT / 2 { |
597 | let src = buffer.as_ptr(); |
598 | let mut until_alignment = ((ALU_ALIGNMENT - ((src as usize) & ALU_ALIGNMENT_MASK)) & |
599 | ALU_ALIGNMENT_MASK) / 2; |
600 | if until_alignment + ALU_ALIGNMENT / 2 <= len { |
601 | while until_alignment != 0 { |
602 | if buffer[offset] > 0xFF { |
603 | if is_utf16_bidi_impl(&buffer[offset..]) { |
604 | return Latin1Bidi::Bidi; |
605 | } |
606 | return Latin1Bidi::LeftToRight; |
607 | } |
608 | offset += 1; |
609 | until_alignment -= 1; |
610 | } |
611 | let len_minus_stride = len - ALU_ALIGNMENT / 2; |
612 | loop { |
613 | if unsafe { *(src.add(offset) as *const usize) } & LATIN1_MASK != 0 { |
614 | if is_utf16_bidi_impl(&buffer[offset..]) { |
615 | return Latin1Bidi::Bidi; |
616 | } |
617 | return Latin1Bidi::LeftToRight; |
618 | } |
619 | offset += ALU_ALIGNMENT / 2; |
620 | if offset > len_minus_stride { |
621 | break; |
622 | } |
623 | } |
624 | } |
625 | } |
626 | let mut iter = (&buffer[offset..]).iter(); |
627 | loop { |
628 | if let Some(&u) = iter.next() { |
629 | if u > 0xFF { |
630 | let mut inner_u = u; |
631 | loop { |
632 | if is_utf16_code_unit_bidi(inner_u) { |
633 | return Latin1Bidi::Bidi; |
634 | } |
635 | if let Some(&code_unit) = iter.next() { |
636 | inner_u = code_unit; |
637 | } else { |
638 | return Latin1Bidi::LeftToRight; |
639 | } |
640 | } |
641 | } |
642 | } else { |
643 | return Latin1Bidi::Latin1; |
644 | } |
645 | } |
646 | } |
647 | } |
648 | } |
649 | |
650 | /// Checks whether the buffer is all-ASCII. |
651 | /// |
652 | /// May read the entire buffer even if it isn't all-ASCII. (I.e. the function |
653 | /// is not guaranteed to fail fast.) |
654 | pub fn is_ascii(buffer: &[u8]) -> bool { |
655 | is_ascii_impl(buffer) |
656 | } |
657 | |
658 | /// Checks whether the buffer is all-Basic Latin (i.e. UTF-16 representing |
659 | /// only ASCII characters). |
660 | /// |
661 | /// May read the entire buffer even if it isn't all-ASCII. (I.e. the function |
662 | /// is not guaranteed to fail fast.) |
663 | pub fn is_basic_latin(buffer: &[u16]) -> bool { |
664 | is_basic_latin_impl(buffer) |
665 | } |
666 | |
667 | /// Checks whether the buffer is valid UTF-8 representing only code points |
668 | /// less than or equal to U+00FF. |
669 | /// |
670 | /// Fails fast. (I.e. returns before having read the whole buffer if UTF-8 |
671 | /// invalidity or code points above U+00FF are discovered. |
672 | pub fn is_utf8_latin1(buffer: &[u8]) -> bool { |
673 | is_utf8_latin1_impl(buffer).is_none() |
674 | } |
675 | |
676 | /// Checks whether the buffer represents only code points less than or equal |
677 | /// to U+00FF. |
678 | /// |
679 | /// Fails fast. (I.e. returns before having read the whole buffer if code |
680 | /// points above U+00FF are discovered. |
681 | pub fn is_str_latin1(buffer: &str) -> bool { |
682 | is_str_latin1_impl(buffer).is_none() |
683 | } |
684 | |
685 | /// Checks whether the buffer represents only code point less than or equal |
686 | /// to U+00FF. |
687 | /// |
688 | /// May read the entire buffer even if it isn't all-Latin1. (I.e. the function |
689 | /// is not guaranteed to fail fast.) |
690 | pub fn is_utf16_latin1(buffer: &[u16]) -> bool { |
691 | is_utf16_latin1_impl(buffer) |
692 | } |
693 | |
694 | /// Checks whether a potentially-invalid UTF-8 buffer contains code points |
695 | /// that trigger right-to-left processing. |
696 | /// |
697 | /// The check is done on a Unicode block basis without regard to assigned |
698 | /// vs. unassigned code points in the block. Hebrew presentation forms in |
699 | /// the Alphabetic Presentation Forms block are treated as if they formed |
700 | /// a block on their own (i.e. it treated as right-to-left). Additionally, |
701 | /// the four RIGHT-TO-LEFT FOO controls in General Punctuation are checked |
702 | /// for. Control characters that are technically bidi controls but do not |
703 | /// cause right-to-left behavior without the presence of right-to-left |
704 | /// characters or right-to-left controls are not checked for. As a special |
705 | /// case, U+FEFF is excluded from Arabic Presentation Forms-B. |
706 | /// |
707 | /// Returns `true` if the input is invalid UTF-8 or the input contains an |
708 | /// RTL character. Returns `false` if the input is valid UTF-8 and contains |
709 | /// no RTL characters. |
710 | #[cfg_attr (feature = "cargo-clippy" , allow(collapsible_if, cyclomatic_complexity))] |
711 | #[inline ] |
712 | pub fn is_utf8_bidi(buffer: &[u8]) -> bool { |
713 | // As of rustc 1.25.0-nightly (73ac5d6a8 2018-01-11), this is faster |
714 | // than UTF-8 validation followed by `is_str_bidi()` for German, |
715 | // Russian and Japanese. However, this is considerably slower for Thai. |
716 | // Chances are that the compiler makes some branch predictions that are |
717 | // unfortunate for Thai. Not spending the time to manually optimize |
718 | // further at this time, since it's unclear if this variant even has |
719 | // use cases. However, this is worth revisiting once Rust gets the |
720 | // ability to annotate relative priorities of match arms. |
721 | |
722 | // U+058F: D6 8F |
723 | // U+0590: D6 90 |
724 | // U+08FF: E0 A3 BF |
725 | // U+0900: E0 A4 80 |
726 | // |
727 | // U+200F: E2 80 8F |
728 | // U+202B: E2 80 AB |
729 | // U+202E: E2 80 AE |
730 | // U+2067: E2 81 A7 |
731 | // |
732 | // U+FB1C: EF AC 9C |
733 | // U+FB1D: EF AC 9D |
734 | // U+FDFF: EF B7 BF |
735 | // U+FE00: EF B8 80 |
736 | // |
737 | // U+FE6F: EF B9 AF |
738 | // U+FE70: EF B9 B0 |
739 | // U+FEFE: EF BB BE |
740 | // U+FEFF: EF BB BF |
741 | // |
742 | // U+107FF: F0 90 9F BF |
743 | // U+10800: F0 90 A0 80 |
744 | // U+10FFF: F0 90 BF BF |
745 | // U+11000: F0 91 80 80 |
746 | // |
747 | // U+1E7FF: F0 9E 9F BF |
748 | // U+1E800: F0 9E A0 80 |
749 | // U+1EFFF: F0 9E BF BF |
750 | // U+1F000: F0 9F 80 80 |
751 | let mut src = buffer; |
752 | 'outer: loop { |
753 | if let Some((mut byte, mut read)) = validate_ascii(src) { |
754 | // Check for the longest sequence to avoid checking twice for the |
755 | // multi-byte sequences. |
756 | if read + 4 <= src.len() { |
757 | 'inner: loop { |
758 | // At this point, `byte` is not included in `read`. |
759 | match byte { |
760 | 0..=0x7F => { |
761 | // ASCII: go back to SIMD. |
762 | read += 1; |
763 | src = &src[read..]; |
764 | continue 'outer; |
765 | } |
766 | 0xC2..=0xD5 => { |
767 | // Two-byte |
768 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
769 | if !in_inclusive_range8(second, 0x80, 0xBF) { |
770 | return true; |
771 | } |
772 | read += 2; |
773 | } |
774 | 0xD6 => { |
775 | // Two-byte |
776 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
777 | if !in_inclusive_range8(second, 0x80, 0xBF) { |
778 | return true; |
779 | } |
780 | // XXX consider folding the above and below checks |
781 | if second > 0x8F { |
782 | return true; |
783 | } |
784 | read += 2; |
785 | } |
786 | // two-byte starting with 0xD7 and above is bidi |
787 | 0xE1 | 0xE3..=0xEC | 0xEE => { |
788 | // Three-byte normal |
789 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
790 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
791 | if ((UTF8_DATA.table[usize::from(second)] |
792 | & unsafe { |
793 | *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) |
794 | }) |
795 | | (third >> 6)) |
796 | != 2 |
797 | { |
798 | return true; |
799 | } |
800 | read += 3; |
801 | } |
802 | 0xE2 => { |
803 | // Three-byte normal, potentially bidi |
804 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
805 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
806 | if ((UTF8_DATA.table[usize::from(second)] |
807 | & unsafe { |
808 | *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) |
809 | }) |
810 | | (third >> 6)) |
811 | != 2 |
812 | { |
813 | return true; |
814 | } |
815 | if second == 0x80 { |
816 | if third == 0x8F || third == 0xAB || third == 0xAE { |
817 | return true; |
818 | } |
819 | } else if second == 0x81 { |
820 | if third == 0xA7 { |
821 | return true; |
822 | } |
823 | } |
824 | read += 3; |
825 | } |
826 | 0xEF => { |
827 | // Three-byte normal, potentially bidi |
828 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
829 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
830 | if ((UTF8_DATA.table[usize::from(second)] |
831 | & unsafe { |
832 | *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) |
833 | }) |
834 | | (third >> 6)) |
835 | != 2 |
836 | { |
837 | return true; |
838 | } |
839 | if in_inclusive_range8(second, 0xAC, 0xB7) { |
840 | if second == 0xAC { |
841 | if third > 0x9C { |
842 | return true; |
843 | } |
844 | } else { |
845 | return true; |
846 | } |
847 | } else if in_inclusive_range8(second, 0xB9, 0xBB) { |
848 | if second == 0xB9 { |
849 | if third > 0xAF { |
850 | return true; |
851 | } |
852 | } else if second == 0xBB { |
853 | if third != 0xBF { |
854 | return true; |
855 | } |
856 | } else { |
857 | return true; |
858 | } |
859 | } |
860 | read += 3; |
861 | } |
862 | 0xE0 => { |
863 | // Three-byte special lower bound, potentially bidi |
864 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
865 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
866 | if ((UTF8_DATA.table[usize::from(second)] |
867 | & unsafe { |
868 | *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) |
869 | }) |
870 | | (third >> 6)) |
871 | != 2 |
872 | { |
873 | return true; |
874 | } |
875 | // XXX can this be folded into the above validity check |
876 | if second < 0xA4 { |
877 | return true; |
878 | } |
879 | read += 3; |
880 | } |
881 | 0xED => { |
882 | // Three-byte special upper bound |
883 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
884 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
885 | if ((UTF8_DATA.table[usize::from(second)] |
886 | & unsafe { |
887 | *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) |
888 | }) |
889 | | (third >> 6)) |
890 | != 2 |
891 | { |
892 | return true; |
893 | } |
894 | read += 3; |
895 | } |
896 | 0xF1..=0xF4 => { |
897 | // Four-byte normal |
898 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
899 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
900 | let fourth = unsafe { *(src.get_unchecked(read + 3)) }; |
901 | if (u16::from( |
902 | UTF8_DATA.table[usize::from(second)] |
903 | & unsafe { |
904 | *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) |
905 | }, |
906 | ) | u16::from(third >> 6) |
907 | | (u16::from(fourth & 0xC0) << 2)) |
908 | != 0x202 |
909 | { |
910 | return true; |
911 | } |
912 | read += 4; |
913 | } |
914 | 0xF0 => { |
915 | // Four-byte special lower bound, potentially bidi |
916 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
917 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
918 | let fourth = unsafe { *(src.get_unchecked(read + 3)) }; |
919 | if (u16::from( |
920 | UTF8_DATA.table[usize::from(second)] |
921 | & unsafe { |
922 | *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) |
923 | }, |
924 | ) | u16::from(third >> 6) |
925 | | (u16::from(fourth & 0xC0) << 2)) |
926 | != 0x202 |
927 | { |
928 | return true; |
929 | } |
930 | if unlikely(second == 0x90 || second == 0x9E) { |
931 | let third = src[read + 2]; |
932 | if third >= 0xA0 { |
933 | return true; |
934 | } |
935 | } |
936 | read += 4; |
937 | } |
938 | _ => { |
939 | // Invalid lead or bidi-only lead |
940 | return true; |
941 | } |
942 | } |
943 | if read + 4 > src.len() { |
944 | if read == src.len() { |
945 | return false; |
946 | } |
947 | byte = src[read]; |
948 | break 'inner; |
949 | } |
950 | byte = src[read]; |
951 | continue 'inner; |
952 | } |
953 | } |
954 | // We can't have a complete 4-byte sequence, but we could still have |
955 | // a complete shorter sequence. |
956 | |
957 | // At this point, `byte` is not included in `read`. |
958 | match byte { |
959 | 0..=0x7F => { |
960 | // ASCII: go back to SIMD. |
961 | read += 1; |
962 | src = &src[read..]; |
963 | continue 'outer; |
964 | } |
965 | 0xC2..=0xD5 => { |
966 | // Two-byte |
967 | let new_read = read + 2; |
968 | if new_read > src.len() { |
969 | return true; |
970 | } |
971 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
972 | if !in_inclusive_range8(second, 0x80, 0xBF) { |
973 | return true; |
974 | } |
975 | read = new_read; |
976 | // We need to deal with the case where we came here with 3 bytes |
977 | // left, so we need to take a look at the last one. |
978 | src = &src[read..]; |
979 | continue 'outer; |
980 | } |
981 | 0xD6 => { |
982 | // Two-byte, potentially bidi |
983 | let new_read = read + 2; |
984 | if new_read > src.len() { |
985 | return true; |
986 | } |
987 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
988 | if !in_inclusive_range8(second, 0x80, 0xBF) { |
989 | return true; |
990 | } |
991 | // XXX consider folding the above and below checks |
992 | if second > 0x8F { |
993 | return true; |
994 | } |
995 | read = new_read; |
996 | // We need to deal with the case where we came here with 3 bytes |
997 | // left, so we need to take a look at the last one. |
998 | src = &src[read..]; |
999 | continue 'outer; |
1000 | } |
1001 | // two-byte starting with 0xD7 and above is bidi |
1002 | 0xE1 | 0xE3..=0xEC | 0xEE => { |
1003 | // Three-byte normal |
1004 | let new_read = read + 3; |
1005 | if new_read > src.len() { |
1006 | return true; |
1007 | } |
1008 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
1009 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
1010 | if ((UTF8_DATA.table[usize::from(second)] |
1011 | & unsafe { *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) }) |
1012 | | (third >> 6)) |
1013 | != 2 |
1014 | { |
1015 | return true; |
1016 | } |
1017 | } |
1018 | 0xE2 => { |
1019 | // Three-byte normal, potentially bidi |
1020 | let new_read = read + 3; |
1021 | if new_read > src.len() { |
1022 | return true; |
1023 | } |
1024 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
1025 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
1026 | if ((UTF8_DATA.table[usize::from(second)] |
1027 | & unsafe { *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) }) |
1028 | | (third >> 6)) |
1029 | != 2 |
1030 | { |
1031 | return true; |
1032 | } |
1033 | if second == 0x80 { |
1034 | if third == 0x8F || third == 0xAB || third == 0xAE { |
1035 | return true; |
1036 | } |
1037 | } else if second == 0x81 { |
1038 | if third == 0xA7 { |
1039 | return true; |
1040 | } |
1041 | } |
1042 | } |
1043 | 0xEF => { |
1044 | // Three-byte normal, potentially bidi |
1045 | let new_read = read + 3; |
1046 | if new_read > src.len() { |
1047 | return true; |
1048 | } |
1049 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
1050 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
1051 | if ((UTF8_DATA.table[usize::from(second)] |
1052 | & unsafe { *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) }) |
1053 | | (third >> 6)) |
1054 | != 2 |
1055 | { |
1056 | return true; |
1057 | } |
1058 | if in_inclusive_range8(second, 0xAC, 0xB7) { |
1059 | if second == 0xAC { |
1060 | if third > 0x9C { |
1061 | return true; |
1062 | } |
1063 | } else { |
1064 | return true; |
1065 | } |
1066 | } else if in_inclusive_range8(second, 0xB9, 0xBB) { |
1067 | if second == 0xB9 { |
1068 | if third > 0xAF { |
1069 | return true; |
1070 | } |
1071 | } else if second == 0xBB { |
1072 | if third != 0xBF { |
1073 | return true; |
1074 | } |
1075 | } else { |
1076 | return true; |
1077 | } |
1078 | } |
1079 | } |
1080 | 0xE0 => { |
1081 | // Three-byte special lower bound, potentially bidi |
1082 | let new_read = read + 3; |
1083 | if new_read > src.len() { |
1084 | return true; |
1085 | } |
1086 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
1087 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
1088 | if ((UTF8_DATA.table[usize::from(second)] |
1089 | & unsafe { *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) }) |
1090 | | (third >> 6)) |
1091 | != 2 |
1092 | { |
1093 | return true; |
1094 | } |
1095 | // XXX can this be folded into the above validity check |
1096 | if second < 0xA4 { |
1097 | return true; |
1098 | } |
1099 | } |
1100 | 0xED => { |
1101 | // Three-byte special upper bound |
1102 | let new_read = read + 3; |
1103 | if new_read > src.len() { |
1104 | return true; |
1105 | } |
1106 | let second = unsafe { *(src.get_unchecked(read + 1)) }; |
1107 | let third = unsafe { *(src.get_unchecked(read + 2)) }; |
1108 | if ((UTF8_DATA.table[usize::from(second)] |
1109 | & unsafe { *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) }) |
1110 | | (third >> 6)) |
1111 | != 2 |
1112 | { |
1113 | return true; |
1114 | } |
1115 | } |
1116 | _ => { |
1117 | // Invalid lead, 4-byte lead or 2-byte bidi-only lead |
1118 | return true; |
1119 | } |
1120 | } |
1121 | return false; |
1122 | } else { |
1123 | return false; |
1124 | } |
1125 | } |
1126 | } |
1127 | |
1128 | /// Checks whether a valid UTF-8 buffer contains code points that trigger |
1129 | /// right-to-left processing. |
1130 | /// |
1131 | /// The check is done on a Unicode block basis without regard to assigned |
1132 | /// vs. unassigned code points in the block. Hebrew presentation forms in |
1133 | /// the Alphabetic Presentation Forms block are treated as if they formed |
1134 | /// a block on their own (i.e. it treated as right-to-left). Additionally, |
1135 | /// the four RIGHT-TO-LEFT FOO controls in General Punctuation are checked |
1136 | /// for. Control characters that are technically bidi controls but do not |
1137 | /// cause right-to-left behavior without the presence of right-to-left |
1138 | /// characters or right-to-left controls are not checked for. As a special |
1139 | /// case, U+FEFF is excluded from Arabic Presentation Forms-B. |
1140 | #[cfg_attr (feature = "cargo-clippy" , allow(collapsible_if))] |
1141 | #[inline ] |
1142 | pub fn is_str_bidi(buffer: &str) -> bool { |
1143 | // U+058F: D6 8F |
1144 | // U+0590: D6 90 |
1145 | // U+08FF: E0 A3 BF |
1146 | // U+0900: E0 A4 80 |
1147 | // |
1148 | // U+200F: E2 80 8F |
1149 | // U+202B: E2 80 AB |
1150 | // U+202E: E2 80 AE |
1151 | // U+2067: E2 81 A7 |
1152 | // |
1153 | // U+FB1C: EF AC 9C |
1154 | // U+FB1D: EF AC 9D |
1155 | // U+FDFF: EF B7 BF |
1156 | // U+FE00: EF B8 80 |
1157 | // |
1158 | // U+FE6F: EF B9 AF |
1159 | // U+FE70: EF B9 B0 |
1160 | // U+FEFE: EF BB BE |
1161 | // U+FEFF: EF BB BF |
1162 | // |
1163 | // U+107FF: F0 90 9F BF |
1164 | // U+10800: F0 90 A0 80 |
1165 | // U+10FFF: F0 90 BF BF |
1166 | // U+11000: F0 91 80 80 |
1167 | // |
1168 | // U+1E7FF: F0 9E 9F BF |
1169 | // U+1E800: F0 9E A0 80 |
1170 | // U+1EFFF: F0 9E BF BF |
1171 | // U+1F000: F0 9F 80 80 |
1172 | let mut bytes = buffer.as_bytes(); |
1173 | 'outer: loop { |
1174 | // TODO: Instead of just validating ASCII using SIMD, use SIMD |
1175 | // to check for non-ASCII lead bytes, too, to quickly conclude |
1176 | // that the vector consist entirely of CJK and below-Hebrew |
1177 | // code points. |
1178 | // Unfortunately, scripts above Arabic but below CJK share |
1179 | // lead bytes with RTL. |
1180 | if let Some((mut byte, mut read)) = validate_ascii(bytes) { |
1181 | 'inner: loop { |
1182 | // At this point, `byte` is not included in `read`. |
1183 | if byte < 0xE0 { |
1184 | if byte >= 0x80 { |
1185 | // Two-byte |
1186 | // Adding `unlikely` here improved throughput on |
1187 | // Russian plain text by 33%! |
1188 | if unlikely(byte >= 0xD6) { |
1189 | if byte == 0xD6 { |
1190 | let second = bytes[read + 1]; |
1191 | if second > 0x8F { |
1192 | return true; |
1193 | } |
1194 | } else { |
1195 | return true; |
1196 | } |
1197 | } |
1198 | read += 2; |
1199 | } else { |
1200 | // ASCII: write and go back to SIMD. |
1201 | read += 1; |
1202 | // Intuitively, we should go back to the outer loop only |
1203 | // if byte is 0x30 or above, so as to avoid trashing on |
1204 | // ASCII space, comma and period in non-Latin context. |
1205 | // However, the extra branch seems to cost more than it's |
1206 | // worth. |
1207 | bytes = &bytes[read..]; |
1208 | continue 'outer; |
1209 | } |
1210 | } else if byte < 0xF0 { |
1211 | // Three-byte |
1212 | if unlikely(!in_inclusive_range8(byte, 0xE3, 0xEE) && byte != 0xE1) { |
1213 | let second = bytes[read + 1]; |
1214 | if byte == 0xE0 { |
1215 | if second < 0xA4 { |
1216 | return true; |
1217 | } |
1218 | } else if byte == 0xE2 { |
1219 | let third = bytes[read + 2]; |
1220 | if second == 0x80 { |
1221 | if third == 0x8F || third == 0xAB || third == 0xAE { |
1222 | return true; |
1223 | } |
1224 | } else if second == 0x81 { |
1225 | if third == 0xA7 { |
1226 | return true; |
1227 | } |
1228 | } |
1229 | } else { |
1230 | debug_assert_eq!(byte, 0xEF); |
1231 | if in_inclusive_range8(second, 0xAC, 0xB7) { |
1232 | if second == 0xAC { |
1233 | let third = bytes[read + 2]; |
1234 | if third > 0x9C { |
1235 | return true; |
1236 | } |
1237 | } else { |
1238 | return true; |
1239 | } |
1240 | } else if in_inclusive_range8(second, 0xB9, 0xBB) { |
1241 | if second == 0xB9 { |
1242 | let third = bytes[read + 2]; |
1243 | if third > 0xAF { |
1244 | return true; |
1245 | } |
1246 | } else if second == 0xBB { |
1247 | let third = bytes[read + 2]; |
1248 | if third != 0xBF { |
1249 | return true; |
1250 | } |
1251 | } else { |
1252 | return true; |
1253 | } |
1254 | } |
1255 | } |
1256 | } |
1257 | read += 3; |
1258 | } else { |
1259 | // Four-byte |
1260 | let second = bytes[read + 1]; |
1261 | if unlikely(byte == 0xF0 && (second == 0x90 || second == 0x9E)) { |
1262 | let third = bytes[read + 2]; |
1263 | if third >= 0xA0 { |
1264 | return true; |
1265 | } |
1266 | } |
1267 | read += 4; |
1268 | } |
1269 | // The comparison is always < or == and never >, but including |
1270 | // > here to let the compiler assume that < is true if this |
1271 | // comparison is false. |
1272 | if read >= bytes.len() { |
1273 | return false; |
1274 | } |
1275 | byte = bytes[read]; |
1276 | continue 'inner; |
1277 | } |
1278 | } else { |
1279 | return false; |
1280 | } |
1281 | } |
1282 | } |
1283 | |
1284 | /// Checks whether a UTF-16 buffer contains code points that trigger |
1285 | /// right-to-left processing. |
1286 | /// |
1287 | /// The check is done on a Unicode block basis without regard to assigned |
1288 | /// vs. unassigned code points in the block. Hebrew presentation forms in |
1289 | /// the Alphabetic Presentation Forms block are treated as if they formed |
1290 | /// a block on their own (i.e. it treated as right-to-left). Additionally, |
1291 | /// the four RIGHT-TO-LEFT FOO controls in General Punctuation are checked |
1292 | /// for. Control characters that are technically bidi controls but do not |
1293 | /// cause right-to-left behavior without the presence of right-to-left |
1294 | /// characters or right-to-left controls are not checked for. As a special |
1295 | /// case, U+FEFF is excluded from Arabic Presentation Forms-B. |
1296 | /// |
1297 | /// Returns `true` if the input contains an RTL character or an unpaired |
1298 | /// high surrogate that could be the high half of an RTL character. |
1299 | /// Returns `false` if the input contains neither RTL characters nor |
1300 | /// unpaired high surrogates that could be higher halves of RTL characters. |
1301 | pub fn is_utf16_bidi(buffer: &[u16]) -> bool { |
1302 | is_utf16_bidi_impl(buffer) |
1303 | } |
1304 | |
1305 | /// Checks whether a scalar value triggers right-to-left processing. |
1306 | /// |
1307 | /// The check is done on a Unicode block basis without regard to assigned |
1308 | /// vs. unassigned code points in the block. Hebrew presentation forms in |
1309 | /// the Alphabetic Presentation Forms block are treated as if they formed |
1310 | /// a block on their own (i.e. it treated as right-to-left). Additionally, |
1311 | /// the four RIGHT-TO-LEFT FOO controls in General Punctuation are checked |
1312 | /// for. Control characters that are technically bidi controls but do not |
1313 | /// cause right-to-left behavior without the presence of right-to-left |
1314 | /// characters or right-to-left controls are not checked for. As a special |
1315 | /// case, U+FEFF is excluded from Arabic Presentation Forms-B. |
1316 | #[inline (always)] |
1317 | pub fn is_char_bidi(c: char) -> bool { |
1318 | // Controls: |
1319 | // Every control with RIGHT-TO-LEFT in its name in |
1320 | // https://www.unicode.org/charts/PDF/U2000.pdf |
1321 | // U+200F RLM |
1322 | // U+202B RLE |
1323 | // U+202E RLO |
1324 | // U+2067 RLI |
1325 | // |
1326 | // BMP RTL: |
1327 | // https://www.unicode.org/roadmaps/bmp/ |
1328 | // U+0590...U+08FF |
1329 | // U+FB1D...U+FDFF Hebrew presentation forms and |
1330 | // Arabic Presentation Forms A |
1331 | // U+FE70...U+FEFE Arabic Presentation Forms B (excl. BOM) |
1332 | // |
1333 | // Supplementary RTL: |
1334 | // https://www.unicode.org/roadmaps/smp/ |
1335 | // U+10800...U+10FFF (Lead surrogate U+D802 or U+D803) |
1336 | // U+1E800...U+1EFFF (Lead surrogate U+D83A or U+D83B) |
1337 | let code_point = u32::from(c); |
1338 | if code_point < 0x0590 { |
1339 | // Below Hebrew |
1340 | return false; |
1341 | } |
1342 | if in_range32(code_point, 0x0900, 0xFB1D) { |
1343 | // Above Arabic Extended-A and below Hebrew presentation forms |
1344 | if in_inclusive_range32(code_point, 0x200F, 0x2067) { |
1345 | // In the range that contains the RTL controls |
1346 | return code_point == 0x200F |
1347 | || code_point == 0x202B |
1348 | || code_point == 0x202E |
1349 | || code_point == 0x2067; |
1350 | } |
1351 | return false; |
1352 | } |
1353 | if code_point > 0x1EFFF { |
1354 | // Above second astral RTL. (Emoji is here.) |
1355 | return false; |
1356 | } |
1357 | if in_range32(code_point, 0x11000, 0x1E800) { |
1358 | // Between astral RTL blocks |
1359 | return false; |
1360 | } |
1361 | if in_range32(code_point, 0xFEFF, 0x10800) { |
1362 | // Above Arabic Presentations Forms B (excl. BOM) and below first |
1363 | // astral RTL |
1364 | return false; |
1365 | } |
1366 | if in_range32(code_point, 0xFE00, 0xFE70) { |
1367 | // Between Arabic Presentations Forms |
1368 | return false; |
1369 | } |
1370 | true |
1371 | } |
1372 | |
1373 | /// Checks whether a UTF-16 code unit triggers right-to-left processing. |
1374 | /// |
1375 | /// The check is done on a Unicode block basis without regard to assigned |
1376 | /// vs. unassigned code points in the block. Hebrew presentation forms in |
1377 | /// the Alphabetic Presentation Forms block are treated as if they formed |
1378 | /// a block on their own (i.e. it treated as right-to-left). Additionally, |
1379 | /// the four RIGHT-TO-LEFT FOO controls in General Punctuation are checked |
1380 | /// for. Control characters that are technically bidi controls but do not |
1381 | /// cause right-to-left behavior without the presence of right-to-left |
1382 | /// characters or right-to-left controls are not checked for. As a special |
1383 | /// case, U+FEFF is excluded from Arabic Presentation Forms-B. |
1384 | /// |
1385 | /// Since supplementary-plane right-to-left blocks are identifiable from the |
1386 | /// high surrogate without examining the low surrogate, this function returns |
1387 | /// `true` for such high surrogates making the function suitable for handling |
1388 | /// supplementary-plane text without decoding surrogate pairs to scalar |
1389 | /// values. Obviously, such high surrogates are then reported as right-to-left |
1390 | /// even if actually unpaired. |
1391 | #[inline (always)] |
1392 | pub fn is_utf16_code_unit_bidi(u: u16) -> bool { |
1393 | if u < 0x0590 { |
1394 | // Below Hebrew |
1395 | return false; |
1396 | } |
1397 | if in_range16(u, 0x0900, 0xD802) { |
1398 | // Above Arabic Extended-A and below first RTL surrogate |
1399 | if in_inclusive_range16(u, 0x200F, 0x2067) { |
1400 | // In the range that contains the RTL controls |
1401 | return u == 0x200F || u == 0x202B || u == 0x202E || u == 0x2067; |
1402 | } |
1403 | return false; |
1404 | } |
1405 | if in_range16(u, 0xD83C, 0xFB1D) { |
1406 | // Between astral RTL high surrogates and Hebrew presentation forms |
1407 | // (Emoji is here) |
1408 | return false; |
1409 | } |
1410 | if in_range16(u, 0xD804, 0xD83A) { |
1411 | // Between RTL high surragates |
1412 | return false; |
1413 | } |
1414 | if u > 0xFEFE { |
1415 | // Above Arabic Presentation Forms (excl. BOM) |
1416 | return false; |
1417 | } |
1418 | if in_range16(u, 0xFE00, 0xFE70) { |
1419 | // Between Arabic Presentations Forms |
1420 | return false; |
1421 | } |
1422 | true |
1423 | } |
1424 | |
1425 | /// Checks whether a potentially invalid UTF-8 buffer contains code points |
1426 | /// that trigger right-to-left processing or is all-Latin1. |
1427 | /// |
1428 | /// Possibly more efficient than performing the checks separately. |
1429 | /// |
1430 | /// Returns `Latin1Bidi::Latin1` if `is_utf8_latin1()` would return `true`. |
1431 | /// Otherwise, returns `Latin1Bidi::Bidi` if `is_utf8_bidi()` would return |
1432 | /// `true`. Otherwise, returns `Latin1Bidi::LeftToRight`. |
1433 | pub fn check_utf8_for_latin1_and_bidi(buffer: &[u8]) -> Latin1Bidi { |
1434 | if let Some(offset: usize) = is_utf8_latin1_impl(buffer) { |
1435 | if is_utf8_bidi(&buffer[offset..]) { |
1436 | Latin1Bidi::Bidi |
1437 | } else { |
1438 | Latin1Bidi::LeftToRight |
1439 | } |
1440 | } else { |
1441 | Latin1Bidi::Latin1 |
1442 | } |
1443 | } |
1444 | |
1445 | /// Checks whether a valid UTF-8 buffer contains code points |
1446 | /// that trigger right-to-left processing or is all-Latin1. |
1447 | /// |
1448 | /// Possibly more efficient than performing the checks separately. |
1449 | /// |
1450 | /// Returns `Latin1Bidi::Latin1` if `is_str_latin1()` would return `true`. |
1451 | /// Otherwise, returns `Latin1Bidi::Bidi` if `is_str_bidi()` would return |
1452 | /// `true`. Otherwise, returns `Latin1Bidi::LeftToRight`. |
1453 | pub fn check_str_for_latin1_and_bidi(buffer: &str) -> Latin1Bidi { |
1454 | // The transition from the latin1 check to the bidi check isn't |
1455 | // optimal but not tweaking it to perfection today. |
1456 | if let Some(offset: usize) = is_str_latin1_impl(buffer) { |
1457 | if is_str_bidi(&buffer[offset..]) { |
1458 | Latin1Bidi::Bidi |
1459 | } else { |
1460 | Latin1Bidi::LeftToRight |
1461 | } |
1462 | } else { |
1463 | Latin1Bidi::Latin1 |
1464 | } |
1465 | } |
1466 | |
1467 | /// Checks whether a potentially invalid UTF-16 buffer contains code points |
1468 | /// that trigger right-to-left processing or is all-Latin1. |
1469 | /// |
1470 | /// Possibly more efficient than performing the checks separately. |
1471 | /// |
1472 | /// Returns `Latin1Bidi::Latin1` if `is_utf16_latin1()` would return `true`. |
1473 | /// Otherwise, returns `Latin1Bidi::Bidi` if `is_utf16_bidi()` would return |
1474 | /// `true`. Otherwise, returns `Latin1Bidi::LeftToRight`. |
1475 | pub fn check_utf16_for_latin1_and_bidi(buffer: &[u16]) -> Latin1Bidi { |
1476 | check_utf16_for_latin1_and_bidi_impl(buffer) |
1477 | } |
1478 | |
1479 | /// Converts potentially-invalid UTF-8 to valid UTF-16 with errors replaced |
1480 | /// with the REPLACEMENT CHARACTER. |
1481 | /// |
1482 | /// The length of the destination buffer must be at least the length of the |
1483 | /// source buffer _plus one_. |
1484 | /// |
1485 | /// Returns the number of `u16`s written. |
1486 | /// |
1487 | /// # Panics |
1488 | /// |
1489 | /// Panics if the destination buffer is shorter than stated above. |
1490 | pub fn convert_utf8_to_utf16(src: &[u8], dst: &mut [u16]) -> usize { |
1491 | // TODO: Can the requirement for dst to be at least one unit longer |
1492 | // be eliminated? |
1493 | assert!(dst.len() > src.len()); |
1494 | let mut decoder = Utf8Decoder::new_inner(); |
1495 | let mut total_read = 0usize; |
1496 | let mut total_written = 0usize; |
1497 | loop { |
1498 | let (result, read, written) = |
1499 | decoder.decode_to_utf16_raw(&src[total_read..], &mut dst[total_written..], true); |
1500 | total_read += read; |
1501 | total_written += written; |
1502 | match result { |
1503 | DecoderResult::InputEmpty => { |
1504 | return total_written; |
1505 | } |
1506 | DecoderResult::OutputFull => { |
1507 | unreachable!("The assert at the top of the function should have caught this." ); |
1508 | } |
1509 | DecoderResult::Malformed(_, _) => { |
1510 | // There should always be space for the U+FFFD, because |
1511 | // otherwise we'd have gotten OutputFull already. |
1512 | dst[total_written] = 0xFFFD; |
1513 | total_written += 1; |
1514 | } |
1515 | } |
1516 | } |
1517 | } |
1518 | |
1519 | /// Converts valid UTF-8 to valid UTF-16. |
1520 | /// |
1521 | /// The length of the destination buffer must be at least the length of the |
1522 | /// source buffer. |
1523 | /// |
1524 | /// Returns the number of `u16`s written. |
1525 | /// |
1526 | /// # Panics |
1527 | /// |
1528 | /// Panics if the destination buffer is shorter than stated above. |
1529 | pub fn convert_str_to_utf16(src: &str, dst: &mut [u16]) -> usize { |
1530 | assert!( |
1531 | dst.len() >= src.len(), |
1532 | "Destination must not be shorter than the source." |
1533 | ); |
1534 | let bytes = src.as_bytes(); |
1535 | let mut read = 0; |
1536 | let mut written = 0; |
1537 | 'outer: loop { |
1538 | let mut byte = { |
1539 | let src_remaining = &bytes[read..]; |
1540 | let dst_remaining = &mut dst[written..]; |
1541 | let length = src_remaining.len(); |
1542 | match unsafe { |
1543 | ascii_to_basic_latin(src_remaining.as_ptr(), dst_remaining.as_mut_ptr(), length) |
1544 | } { |
1545 | None => { |
1546 | written += length; |
1547 | return written; |
1548 | } |
1549 | Some((non_ascii, consumed)) => { |
1550 | read += consumed; |
1551 | written += consumed; |
1552 | non_ascii |
1553 | } |
1554 | } |
1555 | }; |
1556 | 'inner: loop { |
1557 | // At this point, `byte` is not included in `read`. |
1558 | if byte < 0xE0 { |
1559 | if byte >= 0x80 { |
1560 | // Two-byte |
1561 | let second = unsafe { *(bytes.get_unchecked(read + 1)) }; |
1562 | let point = ((u16::from(byte) & 0x1F) << 6) | (u16::from(second) & 0x3F); |
1563 | unsafe { *(dst.get_unchecked_mut(written)) = point }; |
1564 | read += 2; |
1565 | written += 1; |
1566 | } else { |
1567 | // ASCII: write and go back to SIMD. |
1568 | unsafe { *(dst.get_unchecked_mut(written)) = u16::from(byte) }; |
1569 | read += 1; |
1570 | written += 1; |
1571 | // Intuitively, we should go back to the outer loop only |
1572 | // if byte is 0x30 or above, so as to avoid trashing on |
1573 | // ASCII space, comma and period in non-Latin context. |
1574 | // However, the extra branch seems to cost more than it's |
1575 | // worth. |
1576 | continue 'outer; |
1577 | } |
1578 | } else if byte < 0xF0 { |
1579 | // Three-byte |
1580 | let second = unsafe { *(bytes.get_unchecked(read + 1)) }; |
1581 | let third = unsafe { *(bytes.get_unchecked(read + 2)) }; |
1582 | let point = ((u16::from(byte) & 0xF) << 12) |
1583 | | ((u16::from(second) & 0x3F) << 6) |
1584 | | (u16::from(third) & 0x3F); |
1585 | unsafe { *(dst.get_unchecked_mut(written)) = point }; |
1586 | read += 3; |
1587 | written += 1; |
1588 | } else { |
1589 | // Four-byte |
1590 | let second = unsafe { *(bytes.get_unchecked(read + 1)) }; |
1591 | let third = unsafe { *(bytes.get_unchecked(read + 2)) }; |
1592 | let fourth = unsafe { *(bytes.get_unchecked(read + 3)) }; |
1593 | let point = ((u32::from(byte) & 0x7) << 18) |
1594 | | ((u32::from(second) & 0x3F) << 12) |
1595 | | ((u32::from(third) & 0x3F) << 6) |
1596 | | (u32::from(fourth) & 0x3F); |
1597 | unsafe { *(dst.get_unchecked_mut(written)) = (0xD7C0 + (point >> 10)) as u16 }; |
1598 | unsafe { |
1599 | *(dst.get_unchecked_mut(written + 1)) = (0xDC00 + (point & 0x3FF)) as u16 |
1600 | }; |
1601 | read += 4; |
1602 | written += 2; |
1603 | } |
1604 | // The comparison is always < or == and never >, but including |
1605 | // > here to let the compiler assume that < is true if this |
1606 | // comparison is false. |
1607 | if read >= src.len() { |
1608 | return written; |
1609 | } |
1610 | byte = bytes[read]; |
1611 | continue 'inner; |
1612 | } |
1613 | } |
1614 | } |
1615 | |
1616 | /// Converts potentially-invalid UTF-8 to valid UTF-16 signaling on error. |
1617 | /// |
1618 | /// The length of the destination buffer must be at least the length of the |
1619 | /// source buffer. |
1620 | /// |
1621 | /// Returns the number of `u16`s written or `None` if the input was invalid. |
1622 | /// |
1623 | /// When the input was invalid, some output may have been written. |
1624 | /// |
1625 | /// # Panics |
1626 | /// |
1627 | /// Panics if the destination buffer is shorter than stated above. |
1628 | pub fn convert_utf8_to_utf16_without_replacement(src: &[u8], dst: &mut [u16]) -> Option<usize> { |
1629 | assert!( |
1630 | dst.len() >= src.len(), |
1631 | "Destination must not be shorter than the source." |
1632 | ); |
1633 | let (read: usize, written: usize) = convert_utf8_to_utf16_up_to_invalid(src, dst); |
1634 | if read == src.len() { |
1635 | return Some(written); |
1636 | } |
1637 | None |
1638 | } |
1639 | |
1640 | /// Converts potentially-invalid UTF-16 to valid UTF-8 with errors replaced |
1641 | /// with the REPLACEMENT CHARACTER with potentially insufficient output |
1642 | /// space. |
1643 | /// |
1644 | /// Returns the number of code units read and the number of bytes written. |
1645 | /// |
1646 | /// Guarantees that the bytes in the destination beyond the number of |
1647 | /// bytes claimed as written by the second item of the return tuple |
1648 | /// are left unmodified. |
1649 | /// |
1650 | /// Not all code units are read if there isn't enough output space. |
1651 | /// |
1652 | /// Note that this method isn't designed for general streamability but for |
1653 | /// not allocating memory for the worst case up front. Specifically, |
1654 | /// if the input starts with or ends with an unpaired surrogate, those are |
1655 | /// replaced with the REPLACEMENT CHARACTER. |
1656 | /// |
1657 | /// Matches the semantics of `TextEncoder.encodeInto()` from the |
1658 | /// Encoding Standard. |
1659 | /// |
1660 | /// # Safety |
1661 | /// |
1662 | /// If you want to convert into a `&mut str`, use |
1663 | /// `convert_utf16_to_str_partial()` instead of using this function |
1664 | /// together with the `unsafe` method `as_bytes_mut()` on `&mut str`. |
1665 | #[inline (always)] |
1666 | pub fn convert_utf16_to_utf8_partial(src: &[u16], dst: &mut [u8]) -> (usize, usize) { |
1667 | // The two functions called below are marked `inline(never)` to make |
1668 | // transitions from the hot part (first function) into the cold part |
1669 | // (second function) go through a return and another call to discouge |
1670 | // the CPU from speculating from the hot code into the cold code. |
1671 | // Letting the transitions be mere intra-function jumps, even to |
1672 | // basic blocks out-of-lined to the end of the function would wipe |
1673 | // away a quarter of Arabic encode performance on Haswell! |
1674 | let (read: usize, written: usize) = convert_utf16_to_utf8_partial_inner(src, dst); |
1675 | if likely(read == src.len()) { |
1676 | return (read, written); |
1677 | } |
1678 | let (tail_read: usize, tail_written: usize) = |
1679 | convert_utf16_to_utf8_partial_tail(&src[read..], &mut dst[written..]); |
1680 | (read + tail_read, written + tail_written) |
1681 | } |
1682 | |
1683 | /// Converts potentially-invalid UTF-16 to valid UTF-8 with errors replaced |
1684 | /// with the REPLACEMENT CHARACTER. |
1685 | /// |
1686 | /// The length of the destination buffer must be at least the length of the |
1687 | /// source buffer times three. |
1688 | /// |
1689 | /// Returns the number of bytes written. |
1690 | /// |
1691 | /// # Panics |
1692 | /// |
1693 | /// Panics if the destination buffer is shorter than stated above. |
1694 | /// |
1695 | /// # Safety |
1696 | /// |
1697 | /// If you want to convert into a `&mut str`, use `convert_utf16_to_str()` |
1698 | /// instead of using this function together with the `unsafe` method |
1699 | /// `as_bytes_mut()` on `&mut str`. |
1700 | #[inline (always)] |
1701 | pub fn convert_utf16_to_utf8(src: &[u16], dst: &mut [u8]) -> usize { |
1702 | assert!(dst.len() >= src.len() * 3); |
1703 | let (read: usize, written: usize) = convert_utf16_to_utf8_partial(src, dst); |
1704 | debug_assert_eq!(read, src.len()); |
1705 | written |
1706 | } |
1707 | |
1708 | /// Converts potentially-invalid UTF-16 to valid UTF-8 with errors replaced |
1709 | /// with the REPLACEMENT CHARACTER such that the validity of the output is |
1710 | /// signaled using the Rust type system with potentially insufficient output |
1711 | /// space. |
1712 | /// |
1713 | /// Returns the number of code units read and the number of bytes written. |
1714 | /// |
1715 | /// Not all code units are read if there isn't enough output space. |
1716 | /// |
1717 | /// Note that this method isn't designed for general streamability but for |
1718 | /// not allocating memory for the worst case up front. Specifically, |
1719 | /// if the input starts with or ends with an unpaired surrogate, those are |
1720 | /// replaced with the REPLACEMENT CHARACTER. |
1721 | pub fn convert_utf16_to_str_partial(src: &[u16], dst: &mut str) -> (usize, usize) { |
1722 | let bytes: &mut [u8] = unsafe { dst.as_bytes_mut() }; |
1723 | let (read: usize, written: usize) = convert_utf16_to_utf8_partial(src, dst:bytes); |
1724 | let len: usize = bytes.len(); |
1725 | let mut trail: usize = written; |
1726 | while trail < len && ((bytes[trail] & 0xC0) == 0x80) { |
1727 | bytes[trail] = 0; |
1728 | trail += 1; |
1729 | } |
1730 | (read, written) |
1731 | } |
1732 | |
1733 | /// Converts potentially-invalid UTF-16 to valid UTF-8 with errors replaced |
1734 | /// with the REPLACEMENT CHARACTER such that the validity of the output is |
1735 | /// signaled using the Rust type system. |
1736 | /// |
1737 | /// The length of the destination buffer must be at least the length of the |
1738 | /// source buffer times three. |
1739 | /// |
1740 | /// Returns the number of bytes written. |
1741 | /// |
1742 | /// # Panics |
1743 | /// |
1744 | /// Panics if the destination buffer is shorter than stated above. |
1745 | #[inline (always)] |
1746 | pub fn convert_utf16_to_str(src: &[u16], dst: &mut str) -> usize { |
1747 | assert!(dst.len() >= src.len() * 3); |
1748 | let (read: usize, written: usize) = convert_utf16_to_str_partial(src, dst); |
1749 | debug_assert_eq!(read, src.len()); |
1750 | written |
1751 | } |
1752 | |
1753 | /// Converts bytes whose unsigned value is interpreted as Unicode code point |
1754 | /// (i.e. U+0000 to U+00FF, inclusive) to UTF-16. |
1755 | /// |
1756 | /// The length of the destination buffer must be at least the length of the |
1757 | /// source buffer. |
1758 | /// |
1759 | /// The number of `u16`s written equals the length of the source buffer. |
1760 | /// |
1761 | /// # Panics |
1762 | /// |
1763 | /// Panics if the destination buffer is shorter than stated above. |
1764 | pub fn convert_latin1_to_utf16(src: &[u8], dst: &mut [u16]) { |
1765 | assert!( |
1766 | dst.len() >= src.len(), |
1767 | "Destination must not be shorter than the source." |
1768 | ); |
1769 | // TODO: On aarch64, the safe version autovectorizes to the same unpacking |
1770 | // instructions and this code, but, yet, the autovectorized version is |
1771 | // faster. |
1772 | unsafe { |
1773 | unpack_latin1(src.as_ptr(), dst.as_mut_ptr(), src.len()); |
1774 | } |
1775 | } |
1776 | |
1777 | /// Converts bytes whose unsigned value is interpreted as Unicode code point |
1778 | /// (i.e. U+0000 to U+00FF, inclusive) to UTF-8 with potentially insufficient |
1779 | /// output space. |
1780 | /// |
1781 | /// Returns the number of bytes read and the number of bytes written. |
1782 | /// |
1783 | /// If the output isn't large enough, not all input is consumed. |
1784 | /// |
1785 | /// # Safety |
1786 | /// |
1787 | /// If you want to convert into a `&mut str`, use |
1788 | /// `convert_utf16_to_str_partial()` instead of using this function |
1789 | /// together with the `unsafe` method `as_bytes_mut()` on `&mut str`. |
1790 | pub fn convert_latin1_to_utf8_partial(src: &[u8], dst: &mut [u8]) -> (usize, usize) { |
1791 | let src_len = src.len(); |
1792 | let src_ptr = src.as_ptr(); |
1793 | let dst_ptr = dst.as_mut_ptr(); |
1794 | let dst_len = dst.len(); |
1795 | let mut total_read = 0usize; |
1796 | let mut total_written = 0usize; |
1797 | loop { |
1798 | // src can't advance more than dst |
1799 | let src_left = src_len - total_read; |
1800 | let dst_left = dst_len - total_written; |
1801 | let min_left = ::core::cmp::min(src_left, dst_left); |
1802 | if let Some((non_ascii, consumed)) = unsafe { |
1803 | ascii_to_ascii( |
1804 | src_ptr.add(total_read), |
1805 | dst_ptr.add(total_written), |
1806 | min_left, |
1807 | ) |
1808 | } { |
1809 | total_read += consumed; |
1810 | total_written += consumed; |
1811 | if total_written.checked_add(2).unwrap() > dst_len { |
1812 | return (total_read, total_written); |
1813 | } |
1814 | |
1815 | total_read += 1; // consume `non_ascii` |
1816 | |
1817 | dst[total_written] = (non_ascii >> 6) | 0xC0; |
1818 | total_written += 1; |
1819 | dst[total_written] = (non_ascii & 0x3F) | 0x80; |
1820 | total_written += 1; |
1821 | continue; |
1822 | } |
1823 | return (total_read + min_left, total_written + min_left); |
1824 | } |
1825 | } |
1826 | |
1827 | /// Converts bytes whose unsigned value is interpreted as Unicode code point |
1828 | /// (i.e. U+0000 to U+00FF, inclusive) to UTF-8. |
1829 | /// |
1830 | /// The length of the destination buffer must be at least the length of the |
1831 | /// source buffer times two. |
1832 | /// |
1833 | /// Returns the number of bytes written. |
1834 | /// |
1835 | /// # Panics |
1836 | /// |
1837 | /// Panics if the destination buffer is shorter than stated above. |
1838 | /// |
1839 | /// # Safety |
1840 | /// |
1841 | /// Note that this function may write garbage beyond the number of bytes |
1842 | /// indicated by the return value, so using a `&mut str` interpreted as |
1843 | /// `&mut [u8]` as the destination is not safe. If you want to convert into |
1844 | /// a `&mut str`, use `convert_utf16_to_str()` instead of this function. |
1845 | #[inline ] |
1846 | pub fn convert_latin1_to_utf8(src: &[u8], dst: &mut [u8]) -> usize { |
1847 | assert!( |
1848 | dst.len() >= src.len() * 2, |
1849 | "Destination must not be shorter than the source times two." |
1850 | ); |
1851 | let (read: usize, written: usize) = convert_latin1_to_utf8_partial(src, dst); |
1852 | debug_assert_eq!(read, src.len()); |
1853 | written |
1854 | } |
1855 | |
1856 | /// Converts bytes whose unsigned value is interpreted as Unicode code point |
1857 | /// (i.e. U+0000 to U+00FF, inclusive) to UTF-8 such that the validity of the |
1858 | /// output is signaled using the Rust type system with potentially insufficient |
1859 | /// output space. |
1860 | /// |
1861 | /// Returns the number of bytes read and the number of bytes written. |
1862 | /// |
1863 | /// If the output isn't large enough, not all input is consumed. |
1864 | #[inline ] |
1865 | pub fn convert_latin1_to_str_partial(src: &[u8], dst: &mut str) -> (usize, usize) { |
1866 | let bytes: &mut [u8] = unsafe { dst.as_bytes_mut() }; |
1867 | let (read: usize, written: usize) = convert_latin1_to_utf8_partial(src, dst:bytes); |
1868 | let len: usize = bytes.len(); |
1869 | let mut trail: usize = written; |
1870 | let max: usize = ::core::cmp::min(v1:len, v2:trail + MAX_STRIDE_SIZE); |
1871 | while trail < max { |
1872 | bytes[trail] = 0; |
1873 | trail += 1; |
1874 | } |
1875 | while trail < len && ((bytes[trail] & 0xC0) == 0x80) { |
1876 | bytes[trail] = 0; |
1877 | trail += 1; |
1878 | } |
1879 | (read, written) |
1880 | } |
1881 | |
1882 | /// Converts bytes whose unsigned value is interpreted as Unicode code point |
1883 | /// (i.e. U+0000 to U+00FF, inclusive) to UTF-8 such that the validity of the |
1884 | /// output is signaled using the Rust type system. |
1885 | /// |
1886 | /// The length of the destination buffer must be at least the length of the |
1887 | /// source buffer times two. |
1888 | /// |
1889 | /// Returns the number of bytes written. |
1890 | /// |
1891 | /// # Panics |
1892 | /// |
1893 | /// Panics if the destination buffer is shorter than stated above. |
1894 | #[inline ] |
1895 | pub fn convert_latin1_to_str(src: &[u8], dst: &mut str) -> usize { |
1896 | assert!( |
1897 | dst.len() >= src.len() * 2, |
1898 | "Destination must not be shorter than the source times two." |
1899 | ); |
1900 | let (read: usize, written: usize) = convert_latin1_to_str_partial(src, dst); |
1901 | debug_assert_eq!(read, src.len()); |
1902 | written |
1903 | } |
1904 | |
1905 | /// If the input is valid UTF-8 representing only Unicode code points from |
1906 | /// U+0000 to U+00FF, inclusive, converts the input into output that |
1907 | /// represents the value of each code point as the unsigned byte value of |
1908 | /// each output byte. |
1909 | /// |
1910 | /// If the input does not fulfill the condition stated above, this function |
1911 | /// panics if debug assertions are enabled (and fuzzing isn't) and otherwise |
1912 | /// does something that is memory-safe without any promises about any |
1913 | /// properties of the output. In particular, callers shouldn't assume the |
1914 | /// output to be the same across crate versions or CPU architectures and |
1915 | /// should not assume that non-ASCII input can't map to ASCII output. |
1916 | /// |
1917 | /// The length of the destination buffer must be at least the length of the |
1918 | /// source buffer. |
1919 | /// |
1920 | /// Returns the number of bytes written. |
1921 | /// |
1922 | /// # Panics |
1923 | /// |
1924 | /// Panics if the destination buffer is shorter than stated above. |
1925 | /// |
1926 | /// If debug assertions are enabled (and not fuzzing) and the input is |
1927 | /// not in the range U+0000 to U+00FF, inclusive. |
1928 | pub fn convert_utf8_to_latin1_lossy(src: &[u8], dst: &mut [u8]) -> usize { |
1929 | assert!( |
1930 | dst.len() >= src.len(), |
1931 | "Destination must not be shorter than the source." |
1932 | ); |
1933 | non_fuzz_debug_assert!(is_utf8_latin1(src)); |
1934 | let src_len = src.len(); |
1935 | let src_ptr = src.as_ptr(); |
1936 | let dst_ptr = dst.as_mut_ptr(); |
1937 | let mut total_read = 0usize; |
1938 | let mut total_written = 0usize; |
1939 | loop { |
1940 | // dst can't advance more than src |
1941 | let src_left = src_len - total_read; |
1942 | if let Some((non_ascii, consumed)) = unsafe { |
1943 | ascii_to_ascii( |
1944 | src_ptr.add(total_read), |
1945 | dst_ptr.add(total_written), |
1946 | src_left, |
1947 | ) |
1948 | } { |
1949 | total_read += consumed + 1; |
1950 | total_written += consumed; |
1951 | |
1952 | if total_read == src_len { |
1953 | return total_written; |
1954 | } |
1955 | |
1956 | let trail = src[total_read]; |
1957 | total_read += 1; |
1958 | |
1959 | dst[total_written] = ((non_ascii & 0x1F) << 6) | (trail & 0x3F); |
1960 | total_written += 1; |
1961 | continue; |
1962 | } |
1963 | return total_written + src_left; |
1964 | } |
1965 | } |
1966 | |
1967 | /// If the input is valid UTF-16 representing only Unicode code points from |
1968 | /// U+0000 to U+00FF, inclusive, converts the input into output that |
1969 | /// represents the value of each code point as the unsigned byte value of |
1970 | /// each output byte. |
1971 | /// |
1972 | /// If the input does not fulfill the condition stated above, does something |
1973 | /// that is memory-safe without any promises about any properties of the |
1974 | /// output and will probably assert in debug builds in future versions. |
1975 | /// In particular, callers shouldn't assume the output to be the same across |
1976 | /// crate versions or CPU architectures and should not assume that non-ASCII |
1977 | /// input can't map to ASCII output. |
1978 | /// |
1979 | /// The length of the destination buffer must be at least the length of the |
1980 | /// source buffer. |
1981 | /// |
1982 | /// The number of bytes written equals the length of the source buffer. |
1983 | /// |
1984 | /// # Panics |
1985 | /// |
1986 | /// Panics if the destination buffer is shorter than stated above. |
1987 | /// |
1988 | /// (Probably in future versions if debug assertions are enabled (and not |
1989 | /// fuzzing) and the input is not in the range U+0000 to U+00FF, inclusive.) |
1990 | pub fn convert_utf16_to_latin1_lossy(src: &[u16], dst: &mut [u8]) { |
1991 | assert!( |
1992 | dst.len() >= src.len(), |
1993 | "Destination must not be shorter than the source." |
1994 | ); |
1995 | // non_fuzz_debug_assert!(is_utf16_latin1(src)); |
1996 | unsafe { |
1997 | pack_latin1(src.as_ptr(), dst.as_mut_ptr(), src.len()); |
1998 | } |
1999 | } |
2000 | |
2001 | /// Converts bytes whose unsigned value is interpreted as Unicode code point |
2002 | /// (i.e. U+0000 to U+00FF, inclusive) to UTF-8. |
2003 | /// |
2004 | /// Borrows if input is ASCII-only. Performs a single heap allocation |
2005 | /// otherwise. |
2006 | /// |
2007 | /// Only available if the `alloc` feature is enabled (enabled by default). |
2008 | #[cfg (feature = "alloc" )] |
2009 | pub fn decode_latin1<'a>(bytes: &'a [u8]) -> Cow<'a, str> { |
2010 | let up_to: usize = ascii_valid_up_to(bytes); |
2011 | // >= makes later things optimize better than == |
2012 | if up_to >= bytes.len() { |
2013 | debug_assert_eq!(up_to, bytes.len()); |
2014 | let s: &str = unsafe { ::core::str::from_utf8_unchecked(bytes) }; |
2015 | return Cow::Borrowed(s); |
2016 | } |
2017 | let (head: &[u8], tail: &[u8]) = bytes.split_at(mid:up_to); |
2018 | let capacity: usize = head.len() + tail.len() * 2; |
2019 | let mut vec: Vec = Vec::with_capacity(capacity); |
2020 | unsafe { |
2021 | vec.set_len(new_len:capacity); |
2022 | } |
2023 | (&mut vec[..up_to]).copy_from_slice(src:head); |
2024 | let written: usize = convert_latin1_to_utf8(src:tail, &mut vec[up_to..]); |
2025 | vec.truncate(len:up_to + written); |
2026 | Cow::Owned(unsafe { String::from_utf8_unchecked(bytes:vec) }) |
2027 | } |
2028 | |
2029 | /// If the input is valid UTF-8 representing only Unicode code points from |
2030 | /// U+0000 to U+00FF, inclusive, converts the input into output that |
2031 | /// represents the value of each code point as the unsigned byte value of |
2032 | /// each output byte. |
2033 | /// |
2034 | /// If the input does not fulfill the condition stated above, this function |
2035 | /// panics if debug assertions are enabled (and fuzzing isn't) and otherwise |
2036 | /// does something that is memory-safe without any promises about any |
2037 | /// properties of the output. In particular, callers shouldn't assume the |
2038 | /// output to be the same across crate versions or CPU architectures and |
2039 | /// should not assume that non-ASCII input can't map to ASCII output. |
2040 | /// |
2041 | /// Borrows if input is ASCII-only. Performs a single heap allocation |
2042 | /// otherwise. |
2043 | /// |
2044 | /// Only available if the `alloc` feature is enabled (enabled by default). |
2045 | #[cfg (feature = "alloc" )] |
2046 | pub fn encode_latin1_lossy<'a>(string: &'a str) -> Cow<'a, [u8]> { |
2047 | let bytes: &[u8] = string.as_bytes(); |
2048 | let up_to: usize = ascii_valid_up_to(bytes); |
2049 | // >= makes later things optimize better than == |
2050 | if up_to >= bytes.len() { |
2051 | debug_assert_eq!(up_to, bytes.len()); |
2052 | return Cow::Borrowed(bytes); |
2053 | } |
2054 | let (head: &[u8], tail: &[u8]) = bytes.split_at(mid:up_to); |
2055 | let capacity: usize = bytes.len(); |
2056 | let mut vec: Vec = Vec::with_capacity(capacity); |
2057 | unsafe { |
2058 | vec.set_len(new_len:capacity); |
2059 | } |
2060 | (&mut vec[..up_to]).copy_from_slice(src:head); |
2061 | let written: usize = convert_utf8_to_latin1_lossy(src:tail, &mut vec[up_to..]); |
2062 | vec.truncate(len:up_to + written); |
2063 | Cow::Owned(vec) |
2064 | } |
2065 | |
2066 | /// Returns the index of the first unpaired surrogate or, if the input is |
2067 | /// valid UTF-16 in its entirety, the length of the input. |
2068 | pub fn utf16_valid_up_to(buffer: &[u16]) -> usize { |
2069 | utf16_valid_up_to_impl(buffer) |
2070 | } |
2071 | |
2072 | /// Returns the index of first byte that starts an invalid byte |
2073 | /// sequence or a non-Latin1 byte sequence, or the length of the |
2074 | /// string if there are neither. |
2075 | pub fn utf8_latin1_up_to(buffer: &[u8]) -> usize { |
2076 | is_utf8_latin1_impl(buffer).unwrap_or(default:buffer.len()) |
2077 | } |
2078 | |
2079 | /// Returns the index of first byte that starts a non-Latin1 byte |
2080 | /// sequence, or the length of the string if there are none. |
2081 | pub fn str_latin1_up_to(buffer: &str) -> usize { |
2082 | is_str_latin1_impl(buffer).unwrap_or_else(|| buffer.len()) |
2083 | } |
2084 | |
2085 | /// Replaces unpaired surrogates in the input with the REPLACEMENT CHARACTER. |
2086 | #[inline ] |
2087 | pub fn ensure_utf16_validity(buffer: &mut [u16]) { |
2088 | let mut offset: usize = 0; |
2089 | loop { |
2090 | offset += utf16_valid_up_to(&buffer[offset..]); |
2091 | if offset == buffer.len() { |
2092 | return; |
2093 | } |
2094 | buffer[offset] = 0xFFFD; |
2095 | offset += 1; |
2096 | } |
2097 | } |
2098 | |
2099 | /// Copies ASCII from source to destination up to the first non-ASCII byte |
2100 | /// (or the end of the input if it is ASCII in its entirety). |
2101 | /// |
2102 | /// The length of the destination buffer must be at least the length of the |
2103 | /// source buffer. |
2104 | /// |
2105 | /// Returns the number of bytes written. |
2106 | /// |
2107 | /// # Panics |
2108 | /// |
2109 | /// Panics if the destination buffer is shorter than stated above. |
2110 | pub fn copy_ascii_to_ascii(src: &[u8], dst: &mut [u8]) -> usize { |
2111 | assert!( |
2112 | dst.len() >= src.len(), |
2113 | "Destination must not be shorter than the source." |
2114 | ); |
2115 | if let Some((_, consumed: usize)) = |
2116 | unsafe { ascii_to_ascii(src.as_ptr(), dst.as_mut_ptr(), src.len()) } |
2117 | { |
2118 | consumed |
2119 | } else { |
2120 | src.len() |
2121 | } |
2122 | } |
2123 | |
2124 | /// Copies ASCII from source to destination zero-extending it to UTF-16 up to |
2125 | /// the first non-ASCII byte (or the end of the input if it is ASCII in its |
2126 | /// entirety). |
2127 | /// |
2128 | /// The length of the destination buffer must be at least the length of the |
2129 | /// source buffer. |
2130 | /// |
2131 | /// Returns the number of `u16`s written. |
2132 | /// |
2133 | /// # Panics |
2134 | /// |
2135 | /// Panics if the destination buffer is shorter than stated above. |
2136 | pub fn copy_ascii_to_basic_latin(src: &[u8], dst: &mut [u16]) -> usize { |
2137 | assert!( |
2138 | dst.len() >= src.len(), |
2139 | "Destination must not be shorter than the source." |
2140 | ); |
2141 | if let Some((_, consumed: usize)) = |
2142 | unsafe { ascii_to_basic_latin(src.as_ptr(), dst.as_mut_ptr(), src.len()) } |
2143 | { |
2144 | consumed |
2145 | } else { |
2146 | src.len() |
2147 | } |
2148 | } |
2149 | |
2150 | /// Copies Basic Latin from source to destination narrowing it to ASCII up to |
2151 | /// the first non-Basic Latin code unit (or the end of the input if it is |
2152 | /// Basic Latin in its entirety). |
2153 | /// |
2154 | /// The length of the destination buffer must be at least the length of the |
2155 | /// source buffer. |
2156 | /// |
2157 | /// Returns the number of bytes written. |
2158 | /// |
2159 | /// # Panics |
2160 | /// |
2161 | /// Panics if the destination buffer is shorter than stated above. |
2162 | pub fn copy_basic_latin_to_ascii(src: &[u16], dst: &mut [u8]) -> usize { |
2163 | assert!( |
2164 | dst.len() >= src.len(), |
2165 | "Destination must not be shorter than the source." |
2166 | ); |
2167 | if let Some((_, consumed: usize)) = |
2168 | unsafe { basic_latin_to_ascii(src.as_ptr(), dst.as_mut_ptr(), src.len()) } |
2169 | { |
2170 | consumed |
2171 | } else { |
2172 | src.len() |
2173 | } |
2174 | } |
2175 | |
2176 | // Any copyright to the test code below this comment is dedicated to the |
2177 | // Public Domain. http://creativecommons.org/publicdomain/zero/1.0/ |
2178 | |
2179 | #[cfg (all(test, feature = "alloc" ))] |
2180 | mod tests { |
2181 | use super::*; |
2182 | |
2183 | #[test ] |
2184 | fn test_is_ascii_success() { |
2185 | let mut src: Vec<u8> = Vec::with_capacity(128); |
2186 | src.resize(128, 0); |
2187 | for i in 0..src.len() { |
2188 | src[i] = i as u8; |
2189 | } |
2190 | for i in 0..src.len() { |
2191 | assert!(is_ascii(&src[i..])); |
2192 | } |
2193 | } |
2194 | |
2195 | #[test ] |
2196 | fn test_is_ascii_fail() { |
2197 | let mut src: Vec<u8> = Vec::with_capacity(128); |
2198 | src.resize(128, 0); |
2199 | for i in 0..src.len() { |
2200 | src[i] = i as u8; |
2201 | } |
2202 | for i in 0..src.len() { |
2203 | let tail = &mut src[i..]; |
2204 | for j in 0..tail.len() { |
2205 | tail[j] = 0xA0; |
2206 | assert!(!is_ascii(tail)); |
2207 | } |
2208 | } |
2209 | } |
2210 | |
2211 | #[test ] |
2212 | fn test_is_basic_latin_success() { |
2213 | let mut src: Vec<u16> = Vec::with_capacity(128); |
2214 | src.resize(128, 0); |
2215 | for i in 0..src.len() { |
2216 | src[i] = i as u16; |
2217 | } |
2218 | for i in 0..src.len() { |
2219 | assert!(is_basic_latin(&src[i..])); |
2220 | } |
2221 | } |
2222 | |
2223 | #[test ] |
2224 | fn test_is_basic_latin_fail() { |
2225 | let mut src: Vec<u16> = Vec::with_capacity(128); |
2226 | src.resize(128, 0); |
2227 | for i in 0..src.len() { |
2228 | src[i] = i as u16; |
2229 | } |
2230 | for i in 0..src.len() { |
2231 | let tail = &mut src[i..]; |
2232 | for j in 0..tail.len() { |
2233 | tail[j] = 0xA0; |
2234 | assert!(!is_basic_latin(tail)); |
2235 | } |
2236 | } |
2237 | } |
2238 | |
2239 | #[test ] |
2240 | fn test_is_utf16_latin1_success() { |
2241 | let mut src: Vec<u16> = Vec::with_capacity(256); |
2242 | src.resize(256, 0); |
2243 | for i in 0..src.len() { |
2244 | src[i] = i as u16; |
2245 | } |
2246 | for i in 0..src.len() { |
2247 | assert!(is_utf16_latin1(&src[i..])); |
2248 | assert_eq!( |
2249 | check_utf16_for_latin1_and_bidi(&src[i..]), |
2250 | Latin1Bidi::Latin1 |
2251 | ); |
2252 | } |
2253 | } |
2254 | |
2255 | #[test ] |
2256 | fn test_is_utf16_latin1_fail() { |
2257 | let len = if cfg!(miri) { 64 } else { 256 }; // Miri is too slow |
2258 | let mut src: Vec<u16> = Vec::with_capacity(len); |
2259 | src.resize(len, 0); |
2260 | for i in 0..src.len() { |
2261 | src[i] = i as u16; |
2262 | } |
2263 | for i in 0..src.len() { |
2264 | let tail = &mut src[i..]; |
2265 | for j in 0..tail.len() { |
2266 | tail[j] = 0x100 + j as u16; |
2267 | assert!(!is_utf16_latin1(tail)); |
2268 | assert_ne!(check_utf16_for_latin1_and_bidi(tail), Latin1Bidi::Latin1); |
2269 | } |
2270 | } |
2271 | } |
2272 | |
2273 | #[test ] |
2274 | fn test_is_str_latin1_success() { |
2275 | let len = if cfg!(miri) { 64 } else { 256 }; // Miri is too slow |
2276 | let mut src: Vec<u16> = Vec::with_capacity(len); |
2277 | src.resize(len, 0); |
2278 | for i in 0..src.len() { |
2279 | src[i] = i as u16; |
2280 | } |
2281 | for i in 0..src.len() { |
2282 | let s = String::from_utf16(&src[i..]).unwrap(); |
2283 | assert!(is_str_latin1(&s[..])); |
2284 | assert_eq!(check_str_for_latin1_and_bidi(&s[..]), Latin1Bidi::Latin1); |
2285 | } |
2286 | } |
2287 | |
2288 | #[test ] |
2289 | fn test_is_str_latin1_fail() { |
2290 | let len = if cfg!(miri) { 32 } else { 256 }; // Miri is too slow |
2291 | let mut src: Vec<u16> = Vec::with_capacity(len); |
2292 | src.resize(len, 0); |
2293 | for i in 0..src.len() { |
2294 | src[i] = i as u16; |
2295 | } |
2296 | for i in 0..src.len() { |
2297 | let tail = &mut src[i..]; |
2298 | for j in 0..tail.len() { |
2299 | tail[j] = 0x100 + j as u16; |
2300 | let s = String::from_utf16(tail).unwrap(); |
2301 | assert!(!is_str_latin1(&s[..])); |
2302 | assert_ne!(check_str_for_latin1_and_bidi(&s[..]), Latin1Bidi::Latin1); |
2303 | } |
2304 | } |
2305 | } |
2306 | |
2307 | #[test ] |
2308 | fn test_is_utf8_latin1_success() { |
2309 | let len = if cfg!(miri) { 64 } else { 256 }; // Miri is too slow |
2310 | let mut src: Vec<u16> = Vec::with_capacity(len); |
2311 | src.resize(len, 0); |
2312 | for i in 0..src.len() { |
2313 | src[i] = i as u16; |
2314 | } |
2315 | for i in 0..src.len() { |
2316 | let s = String::from_utf16(&src[i..]).unwrap(); |
2317 | assert!(is_utf8_latin1(s.as_bytes())); |
2318 | assert_eq!( |
2319 | check_utf8_for_latin1_and_bidi(s.as_bytes()), |
2320 | Latin1Bidi::Latin1 |
2321 | ); |
2322 | } |
2323 | } |
2324 | |
2325 | #[test ] |
2326 | fn test_is_utf8_latin1_fail() { |
2327 | let len = if cfg!(miri) { 32 } else { 256 }; // Miri is too slow |
2328 | let mut src: Vec<u16> = Vec::with_capacity(len); |
2329 | src.resize(len, 0); |
2330 | for i in 0..src.len() { |
2331 | src[i] = i as u16; |
2332 | } |
2333 | for i in 0..src.len() { |
2334 | let tail = &mut src[i..]; |
2335 | for j in 0..tail.len() { |
2336 | tail[j] = 0x100 + j as u16; |
2337 | let s = String::from_utf16(tail).unwrap(); |
2338 | assert!(!is_utf8_latin1(s.as_bytes())); |
2339 | assert_ne!( |
2340 | check_utf8_for_latin1_and_bidi(s.as_bytes()), |
2341 | Latin1Bidi::Latin1 |
2342 | ); |
2343 | } |
2344 | } |
2345 | } |
2346 | |
2347 | #[test ] |
2348 | fn test_is_utf8_latin1_invalid() { |
2349 | assert!(!is_utf8_latin1(b" \xC3" )); |
2350 | assert!(!is_utf8_latin1(b"a \xC3" )); |
2351 | assert!(!is_utf8_latin1(b" \xFF" )); |
2352 | assert!(!is_utf8_latin1(b"a \xFF" )); |
2353 | assert!(!is_utf8_latin1(b" \xC3\xFF" )); |
2354 | assert!(!is_utf8_latin1(b"a \xC3\xFF" )); |
2355 | } |
2356 | |
2357 | #[test ] |
2358 | fn test_convert_utf8_to_utf16() { |
2359 | let src = "abcdefghijklmnopqrstu \u{1F4A9}v \u{2603}w \u{00B6}xyzz" ; |
2360 | let mut dst: Vec<u16> = Vec::with_capacity(src.len() + 1); |
2361 | dst.resize(src.len() + 1, 0); |
2362 | let len = convert_utf8_to_utf16(src.as_bytes(), &mut dst[..]); |
2363 | dst.truncate(len); |
2364 | let reference: Vec<u16> = src.encode_utf16().collect(); |
2365 | assert_eq!(dst, reference); |
2366 | } |
2367 | |
2368 | #[test ] |
2369 | fn test_convert_str_to_utf16() { |
2370 | let src = "abcdefghijklmnopqrstu \u{1F4A9}v \u{2603}w \u{00B6}xyzz" ; |
2371 | let mut dst: Vec<u16> = Vec::with_capacity(src.len()); |
2372 | dst.resize(src.len(), 0); |
2373 | let len = convert_str_to_utf16(src, &mut dst[..]); |
2374 | dst.truncate(len); |
2375 | let reference: Vec<u16> = src.encode_utf16().collect(); |
2376 | assert_eq!(dst, reference); |
2377 | } |
2378 | |
2379 | #[test ] |
2380 | fn test_convert_utf16_to_utf8_partial() { |
2381 | let reference = "abcdefghijklmnopqrstu \u{1F4A9}v \u{2603}w \u{00B6}xyzz" ; |
2382 | let src: Vec<u16> = reference.encode_utf16().collect(); |
2383 | let mut dst: Vec<u8> = Vec::with_capacity(src.len() * 3 + 1); |
2384 | dst.resize(src.len() * 3 + 1, 0); |
2385 | let (read, written) = convert_utf16_to_utf8_partial(&src[..], &mut dst[..24]); |
2386 | let len = written + convert_utf16_to_utf8(&src[read..], &mut dst[written..]); |
2387 | dst.truncate(len); |
2388 | assert_eq!(dst, reference.as_bytes()); |
2389 | } |
2390 | |
2391 | #[test ] |
2392 | fn test_convert_utf16_to_utf8() { |
2393 | let reference = "abcdefghijklmnopqrstu \u{1F4A9}v \u{2603}w \u{00B6}xyzz" ; |
2394 | let src: Vec<u16> = reference.encode_utf16().collect(); |
2395 | let mut dst: Vec<u8> = Vec::with_capacity(src.len() * 3 + 1); |
2396 | dst.resize(src.len() * 3 + 1, 0); |
2397 | let len = convert_utf16_to_utf8(&src[..], &mut dst[..]); |
2398 | dst.truncate(len); |
2399 | assert_eq!(dst, reference.as_bytes()); |
2400 | } |
2401 | |
2402 | #[test ] |
2403 | fn test_convert_latin1_to_utf16() { |
2404 | let mut src: Vec<u8> = Vec::with_capacity(256); |
2405 | src.resize(256, 0); |
2406 | let mut reference: Vec<u16> = Vec::with_capacity(256); |
2407 | reference.resize(256, 0); |
2408 | for i in 0..256 { |
2409 | src[i] = i as u8; |
2410 | reference[i] = i as u16; |
2411 | } |
2412 | let mut dst: Vec<u16> = Vec::with_capacity(src.len()); |
2413 | dst.resize(src.len(), 0); |
2414 | convert_latin1_to_utf16(&src[..], &mut dst[..]); |
2415 | assert_eq!(dst, reference); |
2416 | } |
2417 | |
2418 | #[test ] |
2419 | fn test_convert_latin1_to_utf8_partial() { |
2420 | let mut dst = [0u8, 2]; |
2421 | let (read, written) = convert_latin1_to_utf8_partial(b"a \xFF" , &mut dst[..]); |
2422 | assert_eq!(read, 1); |
2423 | assert_eq!(written, 1); |
2424 | } |
2425 | |
2426 | #[test ] |
2427 | fn test_convert_latin1_to_utf8() { |
2428 | let mut src: Vec<u8> = Vec::with_capacity(256); |
2429 | src.resize(256, 0); |
2430 | let mut reference: Vec<u16> = Vec::with_capacity(256); |
2431 | reference.resize(256, 0); |
2432 | for i in 0..256 { |
2433 | src[i] = i as u8; |
2434 | reference[i] = i as u16; |
2435 | } |
2436 | let s = String::from_utf16(&reference[..]).unwrap(); |
2437 | let mut dst: Vec<u8> = Vec::with_capacity(src.len() * 2); |
2438 | dst.resize(src.len() * 2, 0); |
2439 | let len = convert_latin1_to_utf8(&src[..], &mut dst[..]); |
2440 | dst.truncate(len); |
2441 | assert_eq!(&dst[..], s.as_bytes()); |
2442 | } |
2443 | |
2444 | #[test ] |
2445 | fn test_convert_utf8_to_latin1_lossy() { |
2446 | let mut reference: Vec<u8> = Vec::with_capacity(256); |
2447 | reference.resize(256, 0); |
2448 | let mut src16: Vec<u16> = Vec::with_capacity(256); |
2449 | src16.resize(256, 0); |
2450 | for i in 0..256 { |
2451 | src16[i] = i as u16; |
2452 | reference[i] = i as u8; |
2453 | } |
2454 | let src = String::from_utf16(&src16[..]).unwrap(); |
2455 | let mut dst: Vec<u8> = Vec::with_capacity(src.len()); |
2456 | dst.resize(src.len(), 0); |
2457 | let len = convert_utf8_to_latin1_lossy(src.as_bytes(), &mut dst[..]); |
2458 | dst.truncate(len); |
2459 | assert_eq!(dst, reference); |
2460 | } |
2461 | |
2462 | #[cfg (all(debug_assertions, not(fuzzing)))] |
2463 | #[test ] |
2464 | #[should_panic ] |
2465 | fn test_convert_utf8_to_latin1_lossy_panics() { |
2466 | let mut dst = [0u8; 16]; |
2467 | let _ = convert_utf8_to_latin1_lossy(" \u{100}" .as_bytes(), &mut dst[..]); |
2468 | } |
2469 | |
2470 | #[test ] |
2471 | fn test_convert_utf16_to_latin1_lossy() { |
2472 | let mut src: Vec<u16> = Vec::with_capacity(256); |
2473 | src.resize(256, 0); |
2474 | let mut reference: Vec<u8> = Vec::with_capacity(256); |
2475 | reference.resize(256, 0); |
2476 | for i in 0..256 { |
2477 | src[i] = i as u16; |
2478 | reference[i] = i as u8; |
2479 | } |
2480 | let mut dst: Vec<u8> = Vec::with_capacity(src.len()); |
2481 | dst.resize(src.len(), 0); |
2482 | convert_utf16_to_latin1_lossy(&src[..], &mut dst[..]); |
2483 | assert_eq!(dst, reference); |
2484 | } |
2485 | |
2486 | #[test ] |
2487 | // #[should_panic] |
2488 | fn test_convert_utf16_to_latin1_lossy_panics() { |
2489 | let mut dst = [0u8; 16]; |
2490 | let _ = convert_utf16_to_latin1_lossy(&[0x0100u16], &mut dst[..]); |
2491 | } |
2492 | |
2493 | #[test ] |
2494 | fn test_utf16_valid_up_to() { |
2495 | let valid = vec![ |
2496 | 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0x2603u16, |
2497 | 0xD83Du16, 0xDCA9u16, 0x00B6u16, |
2498 | ]; |
2499 | assert_eq!(utf16_valid_up_to(&valid[..]), 16); |
2500 | let lone_high = vec![ |
2501 | 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, |
2502 | 0x2603u16, 0xD83Du16, 0x00B6u16, |
2503 | ]; |
2504 | assert_eq!(utf16_valid_up_to(&lone_high[..]), 14); |
2505 | let lone_low = vec![ |
2506 | 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, |
2507 | 0x2603u16, 0xDCA9u16, 0x00B6u16, |
2508 | ]; |
2509 | assert_eq!(utf16_valid_up_to(&lone_low[..]), 14); |
2510 | let lone_high_at_end = vec![ |
2511 | 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, |
2512 | 0x2603u16, 0x00B6u16, 0xD83Du16, |
2513 | ]; |
2514 | assert_eq!(utf16_valid_up_to(&lone_high_at_end[..]), 15); |
2515 | } |
2516 | |
2517 | #[test ] |
2518 | fn test_ensure_utf16_validity() { |
2519 | let mut src = vec![ |
2520 | 0u16, 0xD83Du16, 0u16, 0u16, 0u16, 0xD83Du16, 0xDCA9u16, 0u16, 0u16, 0u16, 0u16, 0u16, |
2521 | 0u16, 0xDCA9u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, |
2522 | 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, |
2523 | ]; |
2524 | let reference = vec![ |
2525 | 0u16, 0xFFFDu16, 0u16, 0u16, 0u16, 0xD83Du16, 0xDCA9u16, 0u16, 0u16, 0u16, 0u16, 0u16, |
2526 | 0u16, 0xFFFDu16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, |
2527 | 0u16, 0u16, 0u16, 0u16, 0u16, 0u16, |
2528 | ]; |
2529 | ensure_utf16_validity(&mut src[..]); |
2530 | assert_eq!(src, reference); |
2531 | } |
2532 | |
2533 | #[test ] |
2534 | fn test_is_char_bidi() { |
2535 | assert!(!is_char_bidi('a' )); |
2536 | assert!(!is_char_bidi(' \u{03B1}' )); |
2537 | assert!(!is_char_bidi(' \u{3041}' )); |
2538 | assert!(!is_char_bidi(' \u{1F4A9}' )); |
2539 | assert!(!is_char_bidi(' \u{FE00}' )); |
2540 | assert!(!is_char_bidi(' \u{202C}' )); |
2541 | assert!(!is_char_bidi(' \u{FEFF}' )); |
2542 | assert!(is_char_bidi(' \u{0590}' )); |
2543 | assert!(is_char_bidi(' \u{08FF}' )); |
2544 | assert!(is_char_bidi(' \u{061C}' )); |
2545 | assert!(is_char_bidi(' \u{FB50}' )); |
2546 | assert!(is_char_bidi(' \u{FDFF}' )); |
2547 | assert!(is_char_bidi(' \u{FE70}' )); |
2548 | assert!(is_char_bidi(' \u{FEFE}' )); |
2549 | assert!(is_char_bidi(' \u{200F}' )); |
2550 | assert!(is_char_bidi(' \u{202B}' )); |
2551 | assert!(is_char_bidi(' \u{202E}' )); |
2552 | assert!(is_char_bidi(' \u{2067}' )); |
2553 | assert!(is_char_bidi(' \u{10800}' )); |
2554 | assert!(is_char_bidi(' \u{10FFF}' )); |
2555 | assert!(is_char_bidi(' \u{1E800}' )); |
2556 | assert!(is_char_bidi(' \u{1EFFF}' )); |
2557 | } |
2558 | |
2559 | #[test ] |
2560 | fn test_is_utf16_code_unit_bidi() { |
2561 | assert!(!is_utf16_code_unit_bidi(0x0062)); |
2562 | assert!(!is_utf16_code_unit_bidi(0x03B1)); |
2563 | assert!(!is_utf16_code_unit_bidi(0x3041)); |
2564 | assert!(!is_utf16_code_unit_bidi(0xD801)); |
2565 | assert!(!is_utf16_code_unit_bidi(0xFE00)); |
2566 | assert!(!is_utf16_code_unit_bidi(0x202C)); |
2567 | assert!(!is_utf16_code_unit_bidi(0xFEFF)); |
2568 | assert!(is_utf16_code_unit_bidi(0x0590)); |
2569 | assert!(is_utf16_code_unit_bidi(0x08FF)); |
2570 | assert!(is_utf16_code_unit_bidi(0x061C)); |
2571 | assert!(is_utf16_code_unit_bidi(0xFB1D)); |
2572 | assert!(is_utf16_code_unit_bidi(0xFB50)); |
2573 | assert!(is_utf16_code_unit_bidi(0xFDFF)); |
2574 | assert!(is_utf16_code_unit_bidi(0xFE70)); |
2575 | assert!(is_utf16_code_unit_bidi(0xFEFE)); |
2576 | assert!(is_utf16_code_unit_bidi(0x200F)); |
2577 | assert!(is_utf16_code_unit_bidi(0x202B)); |
2578 | assert!(is_utf16_code_unit_bidi(0x202E)); |
2579 | assert!(is_utf16_code_unit_bidi(0x2067)); |
2580 | assert!(is_utf16_code_unit_bidi(0xD802)); |
2581 | assert!(is_utf16_code_unit_bidi(0xD803)); |
2582 | assert!(is_utf16_code_unit_bidi(0xD83A)); |
2583 | assert!(is_utf16_code_unit_bidi(0xD83B)); |
2584 | } |
2585 | |
2586 | #[test ] |
2587 | fn test_is_str_bidi() { |
2588 | assert!(!is_str_bidi("abcdefghijklmnopaabcdefghijklmnop" )); |
2589 | assert!(!is_str_bidi("abcdefghijklmnop \u{03B1}abcdefghijklmnop" )); |
2590 | assert!(!is_str_bidi("abcdefghijklmnop \u{3041}abcdefghijklmnop" )); |
2591 | assert!(!is_str_bidi("abcdefghijklmnop \u{1F4A9}abcdefghijklmnop" )); |
2592 | assert!(!is_str_bidi("abcdefghijklmnop \u{FE00}abcdefghijklmnop" )); |
2593 | assert!(!is_str_bidi("abcdefghijklmnop \u{202C}abcdefghijklmnop" )); |
2594 | assert!(!is_str_bidi("abcdefghijklmnop \u{FEFF}abcdefghijklmnop" )); |
2595 | assert!(is_str_bidi("abcdefghijklmnop \u{0590}abcdefghijklmnop" )); |
2596 | assert!(is_str_bidi("abcdefghijklmnop \u{08FF}abcdefghijklmnop" )); |
2597 | assert!(is_str_bidi("abcdefghijklmnop \u{061C}abcdefghijklmnop" )); |
2598 | assert!(is_str_bidi("abcdefghijklmnop \u{FB50}abcdefghijklmnop" )); |
2599 | assert!(is_str_bidi("abcdefghijklmnop \u{FDFF}abcdefghijklmnop" )); |
2600 | assert!(is_str_bidi("abcdefghijklmnop \u{FE70}abcdefghijklmnop" )); |
2601 | assert!(is_str_bidi("abcdefghijklmnop \u{FEFE}abcdefghijklmnop" )); |
2602 | assert!(is_str_bidi("abcdefghijklmnop \u{200F}abcdefghijklmnop" )); |
2603 | assert!(is_str_bidi("abcdefghijklmnop \u{202B}abcdefghijklmnop" )); |
2604 | assert!(is_str_bidi("abcdefghijklmnop \u{202E}abcdefghijklmnop" )); |
2605 | assert!(is_str_bidi("abcdefghijklmnop \u{2067}abcdefghijklmnop" )); |
2606 | assert!(is_str_bidi("abcdefghijklmnop \u{10800}abcdefghijklmnop" )); |
2607 | assert!(is_str_bidi("abcdefghijklmnop \u{10FFF}abcdefghijklmnop" )); |
2608 | assert!(is_str_bidi("abcdefghijklmnop \u{1E800}abcdefghijklmnop" )); |
2609 | assert!(is_str_bidi("abcdefghijklmnop \u{1EFFF}abcdefghijklmnop" )); |
2610 | } |
2611 | |
2612 | #[test ] |
2613 | fn test_is_utf8_bidi() { |
2614 | assert!(!is_utf8_bidi( |
2615 | "abcdefghijklmnopaabcdefghijklmnop" .as_bytes() |
2616 | )); |
2617 | assert!(!is_utf8_bidi( |
2618 | "abcdefghijklmnop \u{03B1}abcdefghijklmnop" .as_bytes() |
2619 | )); |
2620 | assert!(!is_utf8_bidi( |
2621 | "abcdefghijklmnop \u{3041}abcdefghijklmnop" .as_bytes() |
2622 | )); |
2623 | assert!(!is_utf8_bidi( |
2624 | "abcdefghijklmnop \u{1F4A9}abcdefghijklmnop" .as_bytes() |
2625 | )); |
2626 | assert!(!is_utf8_bidi( |
2627 | "abcdefghijklmnop \u{FE00}abcdefghijklmnop" .as_bytes() |
2628 | )); |
2629 | assert!(!is_utf8_bidi( |
2630 | "abcdefghijklmnop \u{202C}abcdefghijklmnop" .as_bytes() |
2631 | )); |
2632 | assert!(!is_utf8_bidi( |
2633 | "abcdefghijklmnop \u{FEFF}abcdefghijklmnop" .as_bytes() |
2634 | )); |
2635 | assert!(is_utf8_bidi( |
2636 | "abcdefghijklmnop \u{0590}abcdefghijklmnop" .as_bytes() |
2637 | )); |
2638 | assert!(is_utf8_bidi( |
2639 | "abcdefghijklmnop \u{08FF}abcdefghijklmnop" .as_bytes() |
2640 | )); |
2641 | assert!(is_utf8_bidi( |
2642 | "abcdefghijklmnop \u{061C}abcdefghijklmnop" .as_bytes() |
2643 | )); |
2644 | assert!(is_utf8_bidi( |
2645 | "abcdefghijklmnop \u{FB50}abcdefghijklmnop" .as_bytes() |
2646 | )); |
2647 | assert!(is_utf8_bidi( |
2648 | "abcdefghijklmnop \u{FDFF}abcdefghijklmnop" .as_bytes() |
2649 | )); |
2650 | assert!(is_utf8_bidi( |
2651 | "abcdefghijklmnop \u{FE70}abcdefghijklmnop" .as_bytes() |
2652 | )); |
2653 | assert!(is_utf8_bidi( |
2654 | "abcdefghijklmnop \u{FEFE}abcdefghijklmnop" .as_bytes() |
2655 | )); |
2656 | assert!(is_utf8_bidi( |
2657 | "abcdefghijklmnop \u{200F}abcdefghijklmnop" .as_bytes() |
2658 | )); |
2659 | assert!(is_utf8_bidi( |
2660 | "abcdefghijklmnop \u{202B}abcdefghijklmnop" .as_bytes() |
2661 | )); |
2662 | assert!(is_utf8_bidi( |
2663 | "abcdefghijklmnop \u{202E}abcdefghijklmnop" .as_bytes() |
2664 | )); |
2665 | assert!(is_utf8_bidi( |
2666 | "abcdefghijklmnop \u{2067}abcdefghijklmnop" .as_bytes() |
2667 | )); |
2668 | assert!(is_utf8_bidi( |
2669 | "abcdefghijklmnop \u{10800}abcdefghijklmnop" .as_bytes() |
2670 | )); |
2671 | assert!(is_utf8_bidi( |
2672 | "abcdefghijklmnop \u{10FFF}abcdefghijklmnop" .as_bytes() |
2673 | )); |
2674 | assert!(is_utf8_bidi( |
2675 | "abcdefghijklmnop \u{1E800}abcdefghijklmnop" .as_bytes() |
2676 | )); |
2677 | assert!(is_utf8_bidi( |
2678 | "abcdefghijklmnop \u{1EFFF}abcdefghijklmnop" .as_bytes() |
2679 | )); |
2680 | } |
2681 | |
2682 | #[test ] |
2683 | fn test_is_utf16_bidi() { |
2684 | assert!(!is_utf16_bidi(&[ |
2685 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x0062, 0x62, 0x63, 0x64, 0x65, 0x66, |
2686 | 0x67, 0x68, 0x69, |
2687 | ])); |
2688 | assert!(!is_utf16_bidi(&[ |
2689 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x03B1, 0x62, 0x63, 0x64, 0x65, 0x66, |
2690 | 0x67, 0x68, 0x69, |
2691 | ])); |
2692 | assert!(!is_utf16_bidi(&[ |
2693 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x3041, 0x62, 0x63, 0x64, 0x65, 0x66, |
2694 | 0x67, 0x68, 0x69, |
2695 | ])); |
2696 | assert!(!is_utf16_bidi(&[ |
2697 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xD801, 0x62, 0x63, 0x64, 0x65, 0x66, |
2698 | 0x67, 0x68, 0x69, |
2699 | ])); |
2700 | assert!(!is_utf16_bidi(&[ |
2701 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFE00, 0x62, 0x63, 0x64, 0x65, 0x66, |
2702 | 0x67, 0x68, 0x69, |
2703 | ])); |
2704 | assert!(!is_utf16_bidi(&[ |
2705 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x202C, 0x62, 0x63, 0x64, 0x65, 0x66, |
2706 | 0x67, 0x68, 0x69, |
2707 | ])); |
2708 | assert!(!is_utf16_bidi(&[ |
2709 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFEFF, 0x62, 0x63, 0x64, 0x65, 0x66, |
2710 | 0x67, 0x68, 0x69, |
2711 | ])); |
2712 | assert!(is_utf16_bidi(&[ |
2713 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x0590, 0x62, 0x63, 0x64, 0x65, 0x66, |
2714 | 0x67, 0x68, 0x69, |
2715 | ])); |
2716 | assert!(is_utf16_bidi(&[ |
2717 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x08FF, 0x62, 0x63, 0x64, 0x65, 0x66, |
2718 | 0x67, 0x68, 0x69, |
2719 | ])); |
2720 | assert!(is_utf16_bidi(&[ |
2721 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x061C, 0x62, 0x63, 0x64, 0x65, 0x66, |
2722 | 0x67, 0x68, 0x69, |
2723 | ])); |
2724 | assert!(is_utf16_bidi(&[ |
2725 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFB1D, 0x62, 0x63, 0x64, 0x65, 0x66, |
2726 | 0x67, 0x68, 0x69, |
2727 | ])); |
2728 | assert!(is_utf16_bidi(&[ |
2729 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFB50, 0x62, 0x63, 0x64, 0x65, 0x66, |
2730 | 0x67, 0x68, 0x69, |
2731 | ])); |
2732 | assert!(is_utf16_bidi(&[ |
2733 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFDFF, 0x62, 0x63, 0x64, 0x65, 0x66, |
2734 | 0x67, 0x68, 0x69, |
2735 | ])); |
2736 | assert!(is_utf16_bidi(&[ |
2737 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFE70, 0x62, 0x63, 0x64, 0x65, 0x66, |
2738 | 0x67, 0x68, 0x69, |
2739 | ])); |
2740 | assert!(is_utf16_bidi(&[ |
2741 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFEFE, 0x62, 0x63, 0x64, 0x65, 0x66, |
2742 | 0x67, 0x68, 0x69, |
2743 | ])); |
2744 | assert!(is_utf16_bidi(&[ |
2745 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x200F, 0x62, 0x63, 0x64, 0x65, 0x66, |
2746 | 0x67, 0x68, 0x69, |
2747 | ])); |
2748 | assert!(is_utf16_bidi(&[ |
2749 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x202B, 0x62, 0x63, 0x64, 0x65, 0x66, |
2750 | 0x67, 0x68, 0x69, |
2751 | ])); |
2752 | assert!(is_utf16_bidi(&[ |
2753 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x202E, 0x62, 0x63, 0x64, 0x65, 0x66, |
2754 | 0x67, 0x68, 0x69, |
2755 | ])); |
2756 | assert!(is_utf16_bidi(&[ |
2757 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x2067, 0x62, 0x63, 0x64, 0x65, 0x66, |
2758 | 0x67, 0x68, 0x69, |
2759 | ])); |
2760 | assert!(is_utf16_bidi(&[ |
2761 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xD802, 0x62, 0x63, 0x64, 0x65, 0x66, |
2762 | 0x67, 0x68, 0x69, |
2763 | ])); |
2764 | assert!(is_utf16_bidi(&[ |
2765 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xD803, 0x62, 0x63, 0x64, 0x65, 0x66, |
2766 | 0x67, 0x68, 0x69, |
2767 | ])); |
2768 | assert!(is_utf16_bidi(&[ |
2769 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xD83A, 0x62, 0x63, 0x64, 0x65, 0x66, |
2770 | 0x67, 0x68, 0x69, |
2771 | ])); |
2772 | assert!(is_utf16_bidi(&[ |
2773 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xD83B, 0x62, 0x63, 0x64, 0x65, 0x66, |
2774 | 0x67, 0x68, 0x69, |
2775 | ])); |
2776 | |
2777 | assert!(is_utf16_bidi(&[ |
2778 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x0590, 0x3041, 0x62, 0x63, 0x64, 0x65, |
2779 | 0x66, 0x67, 0x68, 0x69, |
2780 | ])); |
2781 | } |
2782 | |
2783 | #[test ] |
2784 | fn test_check_str_for_latin1_and_bidi() { |
2785 | assert_ne!( |
2786 | check_str_for_latin1_and_bidi("abcdefghijklmnopaabcdefghijklmnop" ), |
2787 | Latin1Bidi::Bidi |
2788 | ); |
2789 | assert_ne!( |
2790 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{03B1}abcdefghijklmnop" ), |
2791 | Latin1Bidi::Bidi |
2792 | ); |
2793 | assert_ne!( |
2794 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{3041}abcdefghijklmnop" ), |
2795 | Latin1Bidi::Bidi |
2796 | ); |
2797 | assert_ne!( |
2798 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{1F4A9}abcdefghijklmnop" ), |
2799 | Latin1Bidi::Bidi |
2800 | ); |
2801 | assert_ne!( |
2802 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{FE00}abcdefghijklmnop" ), |
2803 | Latin1Bidi::Bidi |
2804 | ); |
2805 | assert_ne!( |
2806 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{202C}abcdefghijklmnop" ), |
2807 | Latin1Bidi::Bidi |
2808 | ); |
2809 | assert_ne!( |
2810 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{FEFF}abcdefghijklmnop" ), |
2811 | Latin1Bidi::Bidi |
2812 | ); |
2813 | assert_eq!( |
2814 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{0590}abcdefghijklmnop" ), |
2815 | Latin1Bidi::Bidi |
2816 | ); |
2817 | assert_eq!( |
2818 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{08FF}abcdefghijklmnop" ), |
2819 | Latin1Bidi::Bidi |
2820 | ); |
2821 | assert_eq!( |
2822 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{061C}abcdefghijklmnop" ), |
2823 | Latin1Bidi::Bidi |
2824 | ); |
2825 | assert_eq!( |
2826 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{FB50}abcdefghijklmnop" ), |
2827 | Latin1Bidi::Bidi |
2828 | ); |
2829 | assert_eq!( |
2830 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{FDFF}abcdefghijklmnop" ), |
2831 | Latin1Bidi::Bidi |
2832 | ); |
2833 | assert_eq!( |
2834 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{FE70}abcdefghijklmnop" ), |
2835 | Latin1Bidi::Bidi |
2836 | ); |
2837 | assert_eq!( |
2838 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{FEFE}abcdefghijklmnop" ), |
2839 | Latin1Bidi::Bidi |
2840 | ); |
2841 | assert_eq!( |
2842 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{200F}abcdefghijklmnop" ), |
2843 | Latin1Bidi::Bidi |
2844 | ); |
2845 | assert_eq!( |
2846 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{202B}abcdefghijklmnop" ), |
2847 | Latin1Bidi::Bidi |
2848 | ); |
2849 | assert_eq!( |
2850 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{202E}abcdefghijklmnop" ), |
2851 | Latin1Bidi::Bidi |
2852 | ); |
2853 | assert_eq!( |
2854 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{2067}abcdefghijklmnop" ), |
2855 | Latin1Bidi::Bidi |
2856 | ); |
2857 | assert_eq!( |
2858 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{10800}abcdefghijklmnop" ), |
2859 | Latin1Bidi::Bidi |
2860 | ); |
2861 | assert_eq!( |
2862 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{10FFF}abcdefghijklmnop" ), |
2863 | Latin1Bidi::Bidi |
2864 | ); |
2865 | assert_eq!( |
2866 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{1E800}abcdefghijklmnop" ), |
2867 | Latin1Bidi::Bidi |
2868 | ); |
2869 | assert_eq!( |
2870 | check_str_for_latin1_and_bidi("abcdefghijklmnop \u{1EFFF}abcdefghijklmnop" ), |
2871 | Latin1Bidi::Bidi |
2872 | ); |
2873 | } |
2874 | |
2875 | #[test ] |
2876 | fn test_check_utf8_for_latin1_and_bidi() { |
2877 | assert_ne!( |
2878 | check_utf8_for_latin1_and_bidi("abcdefghijklmnopaabcdefghijklmnop" .as_bytes()), |
2879 | Latin1Bidi::Bidi |
2880 | ); |
2881 | assert_ne!( |
2882 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{03B1}abcdefghijklmnop" .as_bytes()), |
2883 | Latin1Bidi::Bidi |
2884 | ); |
2885 | assert_ne!( |
2886 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{3041}abcdefghijklmnop" .as_bytes()), |
2887 | Latin1Bidi::Bidi |
2888 | ); |
2889 | assert_ne!( |
2890 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{1F4A9}abcdefghijklmnop" .as_bytes()), |
2891 | Latin1Bidi::Bidi |
2892 | ); |
2893 | assert_ne!( |
2894 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{FE00}abcdefghijklmnop" .as_bytes()), |
2895 | Latin1Bidi::Bidi |
2896 | ); |
2897 | assert_ne!( |
2898 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{202C}abcdefghijklmnop" .as_bytes()), |
2899 | Latin1Bidi::Bidi |
2900 | ); |
2901 | assert_ne!( |
2902 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{FEFF}abcdefghijklmnop" .as_bytes()), |
2903 | Latin1Bidi::Bidi |
2904 | ); |
2905 | assert_eq!( |
2906 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{0590}abcdefghijklmnop" .as_bytes()), |
2907 | Latin1Bidi::Bidi |
2908 | ); |
2909 | assert_eq!( |
2910 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{08FF}abcdefghijklmnop" .as_bytes()), |
2911 | Latin1Bidi::Bidi |
2912 | ); |
2913 | assert_eq!( |
2914 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{061C}abcdefghijklmnop" .as_bytes()), |
2915 | Latin1Bidi::Bidi |
2916 | ); |
2917 | assert_eq!( |
2918 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{FB50}abcdefghijklmnop" .as_bytes()), |
2919 | Latin1Bidi::Bidi |
2920 | ); |
2921 | assert_eq!( |
2922 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{FDFF}abcdefghijklmnop" .as_bytes()), |
2923 | Latin1Bidi::Bidi |
2924 | ); |
2925 | assert_eq!( |
2926 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{FE70}abcdefghijklmnop" .as_bytes()), |
2927 | Latin1Bidi::Bidi |
2928 | ); |
2929 | assert_eq!( |
2930 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{FEFE}abcdefghijklmnop" .as_bytes()), |
2931 | Latin1Bidi::Bidi |
2932 | ); |
2933 | assert_eq!( |
2934 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{200F}abcdefghijklmnop" .as_bytes()), |
2935 | Latin1Bidi::Bidi |
2936 | ); |
2937 | assert_eq!( |
2938 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{202B}abcdefghijklmnop" .as_bytes()), |
2939 | Latin1Bidi::Bidi |
2940 | ); |
2941 | assert_eq!( |
2942 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{202E}abcdefghijklmnop" .as_bytes()), |
2943 | Latin1Bidi::Bidi |
2944 | ); |
2945 | assert_eq!( |
2946 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{2067}abcdefghijklmnop" .as_bytes()), |
2947 | Latin1Bidi::Bidi |
2948 | ); |
2949 | assert_eq!( |
2950 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{10800}abcdefghijklmnop" .as_bytes()), |
2951 | Latin1Bidi::Bidi |
2952 | ); |
2953 | assert_eq!( |
2954 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{10FFF}abcdefghijklmnop" .as_bytes()), |
2955 | Latin1Bidi::Bidi |
2956 | ); |
2957 | assert_eq!( |
2958 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{1E800}abcdefghijklmnop" .as_bytes()), |
2959 | Latin1Bidi::Bidi |
2960 | ); |
2961 | assert_eq!( |
2962 | check_utf8_for_latin1_and_bidi("abcdefghijklmnop \u{1EFFF}abcdefghijklmnop" .as_bytes()), |
2963 | Latin1Bidi::Bidi |
2964 | ); |
2965 | } |
2966 | |
2967 | #[test ] |
2968 | fn test_check_utf16_for_latin1_and_bidi() { |
2969 | assert_ne!( |
2970 | check_utf16_for_latin1_and_bidi(&[ |
2971 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x0062, 0x62, 0x63, 0x64, 0x65, |
2972 | 0x66, 0x67, 0x68, 0x69, |
2973 | ]), |
2974 | Latin1Bidi::Bidi |
2975 | ); |
2976 | assert_ne!( |
2977 | check_utf16_for_latin1_and_bidi(&[ |
2978 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x03B1, 0x62, 0x63, 0x64, 0x65, |
2979 | 0x66, 0x67, 0x68, 0x69, |
2980 | ]), |
2981 | Latin1Bidi::Bidi |
2982 | ); |
2983 | assert_ne!( |
2984 | check_utf16_for_latin1_and_bidi(&[ |
2985 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x3041, 0x62, 0x63, 0x64, 0x65, |
2986 | 0x66, 0x67, 0x68, 0x69, |
2987 | ]), |
2988 | Latin1Bidi::Bidi |
2989 | ); |
2990 | assert_ne!( |
2991 | check_utf16_for_latin1_and_bidi(&[ |
2992 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xD801, 0x62, 0x63, 0x64, 0x65, |
2993 | 0x66, 0x67, 0x68, 0x69, |
2994 | ]), |
2995 | Latin1Bidi::Bidi |
2996 | ); |
2997 | assert_ne!( |
2998 | check_utf16_for_latin1_and_bidi(&[ |
2999 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFE00, 0x62, 0x63, 0x64, 0x65, |
3000 | 0x66, 0x67, 0x68, 0x69, |
3001 | ]), |
3002 | Latin1Bidi::Bidi |
3003 | ); |
3004 | assert_ne!( |
3005 | check_utf16_for_latin1_and_bidi(&[ |
3006 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x202C, 0x62, 0x63, 0x64, 0x65, |
3007 | 0x66, 0x67, 0x68, 0x69, |
3008 | ]), |
3009 | Latin1Bidi::Bidi |
3010 | ); |
3011 | assert_ne!( |
3012 | check_utf16_for_latin1_and_bidi(&[ |
3013 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFEFF, 0x62, 0x63, 0x64, 0x65, |
3014 | 0x66, 0x67, 0x68, 0x69, |
3015 | ]), |
3016 | Latin1Bidi::Bidi |
3017 | ); |
3018 | assert_eq!( |
3019 | check_utf16_for_latin1_and_bidi(&[ |
3020 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x0590, 0x62, 0x63, 0x64, 0x65, |
3021 | 0x66, 0x67, 0x68, 0x69, |
3022 | ]), |
3023 | Latin1Bidi::Bidi |
3024 | ); |
3025 | assert_eq!( |
3026 | check_utf16_for_latin1_and_bidi(&[ |
3027 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x08FF, 0x62, 0x63, 0x64, 0x65, |
3028 | 0x66, 0x67, 0x68, 0x69, |
3029 | ]), |
3030 | Latin1Bidi::Bidi |
3031 | ); |
3032 | assert_eq!( |
3033 | check_utf16_for_latin1_and_bidi(&[ |
3034 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x061C, 0x62, 0x63, 0x64, 0x65, |
3035 | 0x66, 0x67, 0x68, 0x69, |
3036 | ]), |
3037 | Latin1Bidi::Bidi |
3038 | ); |
3039 | assert_eq!( |
3040 | check_utf16_for_latin1_and_bidi(&[ |
3041 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFB1D, 0x62, 0x63, 0x64, 0x65, |
3042 | 0x66, 0x67, 0x68, 0x69, |
3043 | ]), |
3044 | Latin1Bidi::Bidi |
3045 | ); |
3046 | assert_eq!( |
3047 | check_utf16_for_latin1_and_bidi(&[ |
3048 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFB50, 0x62, 0x63, 0x64, 0x65, |
3049 | 0x66, 0x67, 0x68, 0x69, |
3050 | ]), |
3051 | Latin1Bidi::Bidi |
3052 | ); |
3053 | assert_eq!( |
3054 | check_utf16_for_latin1_and_bidi(&[ |
3055 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFDFF, 0x62, 0x63, 0x64, 0x65, |
3056 | 0x66, 0x67, 0x68, 0x69, |
3057 | ]), |
3058 | Latin1Bidi::Bidi |
3059 | ); |
3060 | assert_eq!( |
3061 | check_utf16_for_latin1_and_bidi(&[ |
3062 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFE70, 0x62, 0x63, 0x64, 0x65, |
3063 | 0x66, 0x67, 0x68, 0x69, |
3064 | ]), |
3065 | Latin1Bidi::Bidi |
3066 | ); |
3067 | assert_eq!( |
3068 | check_utf16_for_latin1_and_bidi(&[ |
3069 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xFEFE, 0x62, 0x63, 0x64, 0x65, |
3070 | 0x66, 0x67, 0x68, 0x69, |
3071 | ]), |
3072 | Latin1Bidi::Bidi |
3073 | ); |
3074 | assert_eq!( |
3075 | check_utf16_for_latin1_and_bidi(&[ |
3076 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x200F, 0x62, 0x63, 0x64, 0x65, |
3077 | 0x66, 0x67, 0x68, 0x69, |
3078 | ]), |
3079 | Latin1Bidi::Bidi |
3080 | ); |
3081 | assert_eq!( |
3082 | check_utf16_for_latin1_and_bidi(&[ |
3083 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x202B, 0x62, 0x63, 0x64, 0x65, |
3084 | 0x66, 0x67, 0x68, 0x69, |
3085 | ]), |
3086 | Latin1Bidi::Bidi |
3087 | ); |
3088 | assert_eq!( |
3089 | check_utf16_for_latin1_and_bidi(&[ |
3090 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x202E, 0x62, 0x63, 0x64, 0x65, |
3091 | 0x66, 0x67, 0x68, 0x69, |
3092 | ]), |
3093 | Latin1Bidi::Bidi |
3094 | ); |
3095 | assert_eq!( |
3096 | check_utf16_for_latin1_and_bidi(&[ |
3097 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x2067, 0x62, 0x63, 0x64, 0x65, |
3098 | 0x66, 0x67, 0x68, 0x69, |
3099 | ]), |
3100 | Latin1Bidi::Bidi |
3101 | ); |
3102 | assert_eq!( |
3103 | check_utf16_for_latin1_and_bidi(&[ |
3104 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xD802, 0x62, 0x63, 0x64, 0x65, |
3105 | 0x66, 0x67, 0x68, 0x69, |
3106 | ]), |
3107 | Latin1Bidi::Bidi |
3108 | ); |
3109 | assert_eq!( |
3110 | check_utf16_for_latin1_and_bidi(&[ |
3111 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xD803, 0x62, 0x63, 0x64, 0x65, |
3112 | 0x66, 0x67, 0x68, 0x69, |
3113 | ]), |
3114 | Latin1Bidi::Bidi |
3115 | ); |
3116 | assert_eq!( |
3117 | check_utf16_for_latin1_and_bidi(&[ |
3118 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xD83A, 0x62, 0x63, 0x64, 0x65, |
3119 | 0x66, 0x67, 0x68, 0x69, |
3120 | ]), |
3121 | Latin1Bidi::Bidi |
3122 | ); |
3123 | assert_eq!( |
3124 | check_utf16_for_latin1_and_bidi(&[ |
3125 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0xD83B, 0x62, 0x63, 0x64, 0x65, |
3126 | 0x66, 0x67, 0x68, 0x69, |
3127 | ]), |
3128 | Latin1Bidi::Bidi |
3129 | ); |
3130 | |
3131 | assert_eq!( |
3132 | check_utf16_for_latin1_and_bidi(&[ |
3133 | 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x0590, 0x3041, 0x62, 0x63, 0x64, |
3134 | 0x65, 0x66, 0x67, 0x68, 0x69, |
3135 | ]), |
3136 | Latin1Bidi::Bidi |
3137 | ); |
3138 | } |
3139 | |
3140 | #[inline (always)] |
3141 | pub fn reference_is_char_bidi(c: char) -> bool { |
3142 | match c { |
3143 | ' \u{0590}' ..=' \u{08FF}' |
3144 | | ' \u{FB1D}' ..=' \u{FDFF}' |
3145 | | ' \u{FE70}' ..=' \u{FEFE}' |
3146 | | ' \u{10800}' ..=' \u{10FFF}' |
3147 | | ' \u{1E800}' ..=' \u{1EFFF}' |
3148 | | ' \u{200F}' |
3149 | | ' \u{202B}' |
3150 | | ' \u{202E}' |
3151 | | ' \u{2067}' => true, |
3152 | _ => false, |
3153 | } |
3154 | } |
3155 | |
3156 | #[inline (always)] |
3157 | pub fn reference_is_utf16_code_unit_bidi(u: u16) -> bool { |
3158 | match u { |
3159 | 0x0590..=0x08FF |
3160 | | 0xFB1D..=0xFDFF |
3161 | | 0xFE70..=0xFEFE |
3162 | | 0xD802 |
3163 | | 0xD803 |
3164 | | 0xD83A |
3165 | | 0xD83B |
3166 | | 0x200F |
3167 | | 0x202B |
3168 | | 0x202E |
3169 | | 0x2067 => true, |
3170 | _ => false, |
3171 | } |
3172 | } |
3173 | |
3174 | #[test ] |
3175 | #[cfg_attr (miri, ignore)] // Miri is too slow |
3176 | fn test_is_char_bidi_thoroughly() { |
3177 | for i in 0..0xD800u32 { |
3178 | let c: char = ::core::char::from_u32(i).unwrap(); |
3179 | assert_eq!(is_char_bidi(c), reference_is_char_bidi(c)); |
3180 | } |
3181 | for i in 0xE000..0x110000u32 { |
3182 | let c: char = ::core::char::from_u32(i).unwrap(); |
3183 | assert_eq!(is_char_bidi(c), reference_is_char_bidi(c)); |
3184 | } |
3185 | } |
3186 | |
3187 | #[test ] |
3188 | #[cfg_attr (miri, ignore)] // Miri is too slow |
3189 | fn test_is_utf16_code_unit_bidi_thoroughly() { |
3190 | for i in 0..0x10000u32 { |
3191 | let u = i as u16; |
3192 | assert_eq!( |
3193 | is_utf16_code_unit_bidi(u), |
3194 | reference_is_utf16_code_unit_bidi(u) |
3195 | ); |
3196 | } |
3197 | } |
3198 | |
3199 | #[test ] |
3200 | #[cfg_attr (miri, ignore)] // Miri is too slow |
3201 | fn test_is_str_bidi_thoroughly() { |
3202 | let mut buf = [0; 4]; |
3203 | for i in 0..0xD800u32 { |
3204 | let c: char = ::core::char::from_u32(i).unwrap(); |
3205 | assert_eq!( |
3206 | is_str_bidi(c.encode_utf8(&mut buf[..])), |
3207 | reference_is_char_bidi(c) |
3208 | ); |
3209 | } |
3210 | for i in 0xE000..0x110000u32 { |
3211 | let c: char = ::core::char::from_u32(i).unwrap(); |
3212 | assert_eq!( |
3213 | is_str_bidi(c.encode_utf8(&mut buf[..])), |
3214 | reference_is_char_bidi(c) |
3215 | ); |
3216 | } |
3217 | } |
3218 | |
3219 | #[test ] |
3220 | #[cfg_attr (miri, ignore)] // Miri is too slow |
3221 | fn test_is_utf8_bidi_thoroughly() { |
3222 | let mut buf = [0; 8]; |
3223 | for i in 0..0xD800u32 { |
3224 | let c: char = ::core::char::from_u32(i).unwrap(); |
3225 | let expect = reference_is_char_bidi(c); |
3226 | { |
3227 | let len = { |
3228 | let bytes = c.encode_utf8(&mut buf[..]).as_bytes(); |
3229 | assert_eq!(is_utf8_bidi(bytes), expect); |
3230 | bytes.len() |
3231 | }; |
3232 | { |
3233 | let tail = &mut buf[len..]; |
3234 | for b in tail.iter_mut() { |
3235 | *b = 0; |
3236 | } |
3237 | } |
3238 | } |
3239 | assert_eq!(is_utf8_bidi(&buf[..]), expect); |
3240 | } |
3241 | for i in 0xE000..0x110000u32 { |
3242 | let c: char = ::core::char::from_u32(i).unwrap(); |
3243 | let expect = reference_is_char_bidi(c); |
3244 | { |
3245 | let len = { |
3246 | let bytes = c.encode_utf8(&mut buf[..]).as_bytes(); |
3247 | assert_eq!(is_utf8_bidi(bytes), expect); |
3248 | bytes.len() |
3249 | }; |
3250 | { |
3251 | let tail = &mut buf[len..]; |
3252 | for b in tail.iter_mut() { |
3253 | *b = 0; |
3254 | } |
3255 | } |
3256 | } |
3257 | assert_eq!(is_utf8_bidi(&buf[..]), expect); |
3258 | } |
3259 | } |
3260 | |
3261 | #[test ] |
3262 | #[cfg_attr (miri, ignore)] // Miri is too slow |
3263 | fn test_is_utf16_bidi_thoroughly() { |
3264 | let mut buf = [0; 32]; |
3265 | for i in 0..0x10000u32 { |
3266 | let u = i as u16; |
3267 | buf[15] = u; |
3268 | assert_eq!( |
3269 | is_utf16_bidi(&buf[..]), |
3270 | reference_is_utf16_code_unit_bidi(u) |
3271 | ); |
3272 | } |
3273 | } |
3274 | |
3275 | #[test ] |
3276 | fn test_is_utf8_bidi_edge_cases() { |
3277 | assert!(!is_utf8_bidi(b" \xD5\xBF\x61" )); |
3278 | assert!(!is_utf8_bidi(b" \xD6\x80\x61" )); |
3279 | assert!(!is_utf8_bidi(b"abc" )); |
3280 | assert!(is_utf8_bidi(b" \xD5\xBF\xC2" )); |
3281 | assert!(is_utf8_bidi(b" \xD6\x80\xC2" )); |
3282 | assert!(is_utf8_bidi(b"ab \xC2" )); |
3283 | } |
3284 | |
3285 | #[test ] |
3286 | fn test_decode_latin1() { |
3287 | match decode_latin1(b"ab" ) { |
3288 | Cow::Borrowed(s) => { |
3289 | assert_eq!(s, "ab" ); |
3290 | } |
3291 | Cow::Owned(_) => { |
3292 | unreachable!("Should have borrowed" ); |
3293 | } |
3294 | } |
3295 | assert_eq!(decode_latin1(b"a \xE4" ), "a \u{E4}" ); |
3296 | } |
3297 | |
3298 | #[test ] |
3299 | fn test_encode_latin1_lossy() { |
3300 | match encode_latin1_lossy("ab" ) { |
3301 | Cow::Borrowed(s) => { |
3302 | assert_eq!(s, b"ab" ); |
3303 | } |
3304 | Cow::Owned(_) => { |
3305 | unreachable!("Should have borrowed" ); |
3306 | } |
3307 | } |
3308 | assert_eq!(encode_latin1_lossy("a \u{E4}" ), &(b"a \xE4" )[..]); |
3309 | } |
3310 | |
3311 | #[test ] |
3312 | fn test_convert_utf8_to_utf16_without_replacement() { |
3313 | let mut buf = [0u16; 5]; |
3314 | assert_eq!( |
3315 | convert_utf8_to_utf16_without_replacement(b"ab" , &mut buf[..2]), |
3316 | Some(2) |
3317 | ); |
3318 | assert_eq!(buf[0], u16::from(b'a' )); |
3319 | assert_eq!(buf[1], u16::from(b'b' )); |
3320 | assert_eq!(buf[2], 0); |
3321 | assert_eq!( |
3322 | convert_utf8_to_utf16_without_replacement(b" \xC3\xA4c" , &mut buf[..3]), |
3323 | Some(2) |
3324 | ); |
3325 | assert_eq!(buf[0], 0xE4); |
3326 | assert_eq!(buf[1], u16::from(b'c' )); |
3327 | assert_eq!(buf[2], 0); |
3328 | assert_eq!( |
3329 | convert_utf8_to_utf16_without_replacement(b" \xE2\x98\x83" , &mut buf[..3]), |
3330 | Some(1) |
3331 | ); |
3332 | assert_eq!(buf[0], 0x2603); |
3333 | assert_eq!(buf[1], u16::from(b'c' )); |
3334 | assert_eq!(buf[2], 0); |
3335 | assert_eq!( |
3336 | convert_utf8_to_utf16_without_replacement(b" \xE2\x98\x83d" , &mut buf[..4]), |
3337 | Some(2) |
3338 | ); |
3339 | assert_eq!(buf[0], 0x2603); |
3340 | assert_eq!(buf[1], u16::from(b'd' )); |
3341 | assert_eq!(buf[2], 0); |
3342 | assert_eq!( |
3343 | convert_utf8_to_utf16_without_replacement(b" \xE2\x98\x83\xC3\xA4" , &mut buf[..5]), |
3344 | Some(2) |
3345 | ); |
3346 | assert_eq!(buf[0], 0x2603); |
3347 | assert_eq!(buf[1], 0xE4); |
3348 | assert_eq!(buf[2], 0); |
3349 | assert_eq!( |
3350 | convert_utf8_to_utf16_without_replacement(b" \xF0\x9F\x93\x8E" , &mut buf[..4]), |
3351 | Some(2) |
3352 | ); |
3353 | assert_eq!(buf[0], 0xD83D); |
3354 | assert_eq!(buf[1], 0xDCCE); |
3355 | assert_eq!(buf[2], 0); |
3356 | assert_eq!( |
3357 | convert_utf8_to_utf16_without_replacement(b" \xF0\x9F\x93\x8Ee" , &mut buf[..5]), |
3358 | Some(3) |
3359 | ); |
3360 | assert_eq!(buf[0], 0xD83D); |
3361 | assert_eq!(buf[1], 0xDCCE); |
3362 | assert_eq!(buf[2], u16::from(b'e' )); |
3363 | assert_eq!( |
3364 | convert_utf8_to_utf16_without_replacement(b" \xF0\x9F\x93" , &mut buf[..5]), |
3365 | None |
3366 | ); |
3367 | } |
3368 | } |
3369 | |