1 | //! Parallel iterator types for [strings][std::str] |
2 | //! |
3 | //! You will rarely need to interact with this module directly unless you need |
4 | //! to name one of the iterator types. |
5 | //! |
6 | //! Note: [`ParallelString::par_split()`] and [`par_split_terminator()`] |
7 | //! reference a `Pattern` trait which is not visible outside this crate. |
8 | //! This trait is intentionally kept private, for use only by Rayon itself. |
9 | //! It is implemented for `char`, `&[char]`, and any function or closure |
10 | //! `F: Fn(char) -> bool + Sync + Send`. |
11 | //! |
12 | //! [`ParallelString::par_split()`]: trait.ParallelString.html#method.par_split |
13 | //! [`par_split_terminator()`]: trait.ParallelString.html#method.par_split_terminator |
14 | //! |
15 | //! [std::str]: https://doc.rust-lang.org/stable/std/str/ |
16 | |
17 | use crate::iter::plumbing::*; |
18 | use crate::iter::*; |
19 | use crate::split_producer::*; |
20 | |
21 | /// Test if a byte is the start of a UTF-8 character. |
22 | /// (extracted from `str::is_char_boundary`) |
23 | #[inline ] |
24 | fn is_char_boundary(b: u8) -> bool { |
25 | // This is bit magic equivalent to: b < 128 || b >= 192 |
26 | (b as i8) >= -0x40 |
27 | } |
28 | |
29 | /// Find the index of a character boundary near the midpoint. |
30 | #[inline ] |
31 | fn find_char_midpoint(chars: &str) -> usize { |
32 | let mid = chars.len() / 2; |
33 | |
34 | // We want to split near the midpoint, but we need to find an actual |
35 | // character boundary. So we look at the raw bytes, first scanning |
36 | // forward from the midpoint for a boundary, then trying backward. |
37 | let (left, right) = chars.as_bytes().split_at(mid); |
38 | match right.iter().copied().position(is_char_boundary) { |
39 | Some(i) => mid + i, |
40 | None => left |
41 | .iter() |
42 | .copied() |
43 | .rposition(is_char_boundary) |
44 | .unwrap_or(0), |
45 | } |
46 | } |
47 | |
48 | /// Try to split a string near the midpoint. |
49 | #[inline ] |
50 | fn split(chars: &str) -> Option<(&str, &str)> { |
51 | let index = find_char_midpoint(chars); |
52 | if index > 0 { |
53 | Some(chars.split_at(index)) |
54 | } else { |
55 | None |
56 | } |
57 | } |
58 | |
59 | /// Parallel extensions for strings. |
60 | pub trait ParallelString { |
61 | /// Returns a plain string slice, which is used to implement the rest of |
62 | /// the parallel methods. |
63 | fn as_parallel_string(&self) -> &str; |
64 | |
65 | /// Returns a parallel iterator over the characters of a string. |
66 | /// |
67 | /// # Examples |
68 | /// |
69 | /// ``` |
70 | /// use rayon::prelude::*; |
71 | /// let max = "hello" .par_chars().max_by_key(|c| *c as i32); |
72 | /// assert_eq!(Some('o' ), max); |
73 | /// ``` |
74 | fn par_chars(&self) -> Chars<'_> { |
75 | Chars { |
76 | chars: self.as_parallel_string(), |
77 | } |
78 | } |
79 | |
80 | /// Returns a parallel iterator over the characters of a string, with their positions. |
81 | /// |
82 | /// # Examples |
83 | /// |
84 | /// ``` |
85 | /// use rayon::prelude::*; |
86 | /// let min = "hello" .par_char_indices().min_by_key(|&(_i, c)| c as i32); |
87 | /// assert_eq!(Some((1, 'e' )), min); |
88 | /// ``` |
89 | fn par_char_indices(&self) -> CharIndices<'_> { |
90 | CharIndices { |
91 | chars: self.as_parallel_string(), |
92 | } |
93 | } |
94 | |
95 | /// Returns a parallel iterator over the bytes of a string. |
96 | /// |
97 | /// Note that multi-byte sequences (for code points greater than `U+007F`) |
98 | /// are produced as separate items, but will not be split across threads. |
99 | /// If you would prefer an indexed iterator without that guarantee, consider |
100 | /// `string.as_bytes().par_iter().copied()` instead. |
101 | /// |
102 | /// # Examples |
103 | /// |
104 | /// ``` |
105 | /// use rayon::prelude::*; |
106 | /// let max = "hello" .par_bytes().max(); |
107 | /// assert_eq!(Some(b'o' ), max); |
108 | /// ``` |
109 | fn par_bytes(&self) -> Bytes<'_> { |
110 | Bytes { |
111 | chars: self.as_parallel_string(), |
112 | } |
113 | } |
114 | |
115 | /// Returns a parallel iterator over a string encoded as UTF-16. |
116 | /// |
117 | /// Note that surrogate pairs (for code points greater than `U+FFFF`) are |
118 | /// produced as separate items, but will not be split across threads. |
119 | /// |
120 | /// # Examples |
121 | /// |
122 | /// ``` |
123 | /// use rayon::prelude::*; |
124 | /// |
125 | /// let max = "hello" .par_encode_utf16().max(); |
126 | /// assert_eq!(Some(b'o' as u16), max); |
127 | /// |
128 | /// let text = "Zażółć gęślą jaźń" ; |
129 | /// let utf8_len = text.len(); |
130 | /// let utf16_len = text.par_encode_utf16().count(); |
131 | /// assert!(utf16_len <= utf8_len); |
132 | /// ``` |
133 | fn par_encode_utf16(&self) -> EncodeUtf16<'_> { |
134 | EncodeUtf16 { |
135 | chars: self.as_parallel_string(), |
136 | } |
137 | } |
138 | |
139 | /// Returns a parallel iterator over substrings separated by a |
140 | /// given character or predicate, similar to `str::split`. |
141 | /// |
142 | /// Note: the `Pattern` trait is private, for use only by Rayon itself. |
143 | /// It is implemented for `char`, `&[char]`, and any function or closure |
144 | /// `F: Fn(char) -> bool + Sync + Send`. |
145 | /// |
146 | /// # Examples |
147 | /// |
148 | /// ``` |
149 | /// use rayon::prelude::*; |
150 | /// let total = "1, 2, buckle, 3, 4, door" |
151 | /// .par_split(',' ) |
152 | /// .filter_map(|s| s.trim().parse::<i32>().ok()) |
153 | /// .sum(); |
154 | /// assert_eq!(10, total); |
155 | /// ``` |
156 | fn par_split<P: Pattern>(&self, separator: P) -> Split<'_, P> { |
157 | Split::new(self.as_parallel_string(), separator) |
158 | } |
159 | |
160 | /// Returns a parallel iterator over substrings terminated by a |
161 | /// given character or predicate, similar to `str::split_terminator`. |
162 | /// It's equivalent to `par_split`, except it doesn't produce an empty |
163 | /// substring after a trailing terminator. |
164 | /// |
165 | /// Note: the `Pattern` trait is private, for use only by Rayon itself. |
166 | /// It is implemented for `char`, `&[char]`, and any function or closure |
167 | /// `F: Fn(char) -> bool + Sync + Send`. |
168 | /// |
169 | /// # Examples |
170 | /// |
171 | /// ``` |
172 | /// use rayon::prelude::*; |
173 | /// let parts: Vec<_> = "((1 + 3) * 2)" |
174 | /// .par_split_terminator(|c| c == '(' || c == ')' ) |
175 | /// .collect(); |
176 | /// assert_eq!(vec!["" , "" , "1 + 3" , " * 2" ], parts); |
177 | /// ``` |
178 | fn par_split_terminator<P: Pattern>(&self, terminator: P) -> SplitTerminator<'_, P> { |
179 | SplitTerminator::new(self.as_parallel_string(), terminator) |
180 | } |
181 | |
182 | /// Returns a parallel iterator over the lines of a string, ending with an |
183 | /// optional carriage return and with a newline (`\r\n` or just `\n`). |
184 | /// The final line ending is optional, and line endings are not included in |
185 | /// the output strings. |
186 | /// |
187 | /// # Examples |
188 | /// |
189 | /// ``` |
190 | /// use rayon::prelude::*; |
191 | /// let lengths: Vec<_> = "hello world \nfizbuzz" |
192 | /// .par_lines() |
193 | /// .map(|l| l.len()) |
194 | /// .collect(); |
195 | /// assert_eq!(vec![11, 7], lengths); |
196 | /// ``` |
197 | fn par_lines(&self) -> Lines<'_> { |
198 | Lines(self.as_parallel_string()) |
199 | } |
200 | |
201 | /// Returns a parallel iterator over the sub-slices of a string that are |
202 | /// separated by any amount of whitespace. |
203 | /// |
204 | /// As with `str::split_whitespace`, 'whitespace' is defined according to |
205 | /// the terms of the Unicode Derived Core Property `White_Space`. |
206 | /// |
207 | /// # Examples |
208 | /// |
209 | /// ``` |
210 | /// use rayon::prelude::*; |
211 | /// let longest = "which is the longest word?" |
212 | /// .par_split_whitespace() |
213 | /// .max_by_key(|word| word.len()); |
214 | /// assert_eq!(Some("longest" ), longest); |
215 | /// ``` |
216 | fn par_split_whitespace(&self) -> SplitWhitespace<'_> { |
217 | SplitWhitespace(self.as_parallel_string()) |
218 | } |
219 | |
220 | /// Returns a parallel iterator over substrings that match a |
221 | /// given character or predicate, similar to `str::matches`. |
222 | /// |
223 | /// Note: the `Pattern` trait is private, for use only by Rayon itself. |
224 | /// It is implemented for `char`, `&[char]`, and any function or closure |
225 | /// `F: Fn(char) -> bool + Sync + Send`. |
226 | /// |
227 | /// # Examples |
228 | /// |
229 | /// ``` |
230 | /// use rayon::prelude::*; |
231 | /// let total = "1, 2, buckle, 3, 4, door" |
232 | /// .par_matches(char::is_numeric) |
233 | /// .map(|s| s.parse::<i32>().expect("digit" )) |
234 | /// .sum(); |
235 | /// assert_eq!(10, total); |
236 | /// ``` |
237 | fn par_matches<P: Pattern>(&self, pattern: P) -> Matches<'_, P> { |
238 | Matches { |
239 | chars: self.as_parallel_string(), |
240 | pattern, |
241 | } |
242 | } |
243 | |
244 | /// Returns a parallel iterator over substrings that match a given character |
245 | /// or predicate, with their positions, similar to `str::match_indices`. |
246 | /// |
247 | /// Note: the `Pattern` trait is private, for use only by Rayon itself. |
248 | /// It is implemented for `char`, `&[char]`, and any function or closure |
249 | /// `F: Fn(char) -> bool + Sync + Send`. |
250 | /// |
251 | /// # Examples |
252 | /// |
253 | /// ``` |
254 | /// use rayon::prelude::*; |
255 | /// let digits: Vec<_> = "1, 2, buckle, 3, 4, door" |
256 | /// .par_match_indices(char::is_numeric) |
257 | /// .collect(); |
258 | /// assert_eq!(digits, vec![(0, "1" ), (3, "2" ), (14, "3" ), (17, "4" )]); |
259 | /// ``` |
260 | fn par_match_indices<P: Pattern>(&self, pattern: P) -> MatchIndices<'_, P> { |
261 | MatchIndices { |
262 | chars: self.as_parallel_string(), |
263 | pattern, |
264 | } |
265 | } |
266 | } |
267 | |
268 | impl ParallelString for str { |
269 | #[inline ] |
270 | fn as_parallel_string(&self) -> &str { |
271 | self |
272 | } |
273 | } |
274 | |
275 | // ///////////////////////////////////////////////////////////////////////// |
276 | |
277 | /// We hide the `Pattern` trait in a private module, as its API is not meant |
278 | /// for general consumption. If we could have privacy on trait items, then it |
279 | /// would be nicer to have its basic existence and implementors public while |
280 | /// keeping all of the methods private. |
281 | mod private { |
282 | use crate::iter::plumbing::Folder; |
283 | |
284 | /// Pattern-matching trait for `ParallelString`, somewhat like a mix of |
285 | /// `std::str::pattern::{Pattern, Searcher}`. |
286 | /// |
287 | /// Implementing this trait is not permitted outside of `rayon`. |
288 | pub trait Pattern: Sized + Sync + Send { |
289 | private_decl! {} |
290 | fn find_in(&self, haystack: &str) -> Option<usize>; |
291 | fn rfind_in(&self, haystack: &str) -> Option<usize>; |
292 | fn is_suffix_of(&self, haystack: &str) -> bool; |
293 | fn fold_splits<'ch, F>(&self, haystack: &'ch str, folder: F, skip_last: bool) -> F |
294 | where |
295 | F: Folder<&'ch str>; |
296 | fn fold_matches<'ch, F>(&self, haystack: &'ch str, folder: F) -> F |
297 | where |
298 | F: Folder<&'ch str>; |
299 | fn fold_match_indices<'ch, F>(&self, haystack: &'ch str, folder: F, base: usize) -> F |
300 | where |
301 | F: Folder<(usize, &'ch str)>; |
302 | } |
303 | } |
304 | use self::private::Pattern; |
305 | |
306 | #[inline ] |
307 | fn offset<T>(base: usize) -> impl Fn((usize, T)) -> (usize, T) { |
308 | move |(i, x)| (base + i, x) |
309 | } |
310 | |
311 | macro_rules! impl_pattern { |
312 | (&$self:ident => $pattern:expr) => { |
313 | private_impl! {} |
314 | |
315 | #[inline] |
316 | fn find_in(&$self, chars: &str) -> Option<usize> { |
317 | chars.find($pattern) |
318 | } |
319 | |
320 | #[inline] |
321 | fn rfind_in(&$self, chars: &str) -> Option<usize> { |
322 | chars.rfind($pattern) |
323 | } |
324 | |
325 | #[inline] |
326 | fn is_suffix_of(&$self, chars: &str) -> bool { |
327 | chars.ends_with($pattern) |
328 | } |
329 | |
330 | fn fold_splits<'ch, F>(&$self, chars: &'ch str, folder: F, skip_last: bool) -> F |
331 | where |
332 | F: Folder<&'ch str>, |
333 | { |
334 | let mut split = chars.split($pattern); |
335 | if skip_last { |
336 | split.next_back(); |
337 | } |
338 | folder.consume_iter(split) |
339 | } |
340 | |
341 | fn fold_matches<'ch, F>(&$self, chars: &'ch str, folder: F) -> F |
342 | where |
343 | F: Folder<&'ch str>, |
344 | { |
345 | folder.consume_iter(chars.matches($pattern)) |
346 | } |
347 | |
348 | fn fold_match_indices<'ch, F>(&$self, chars: &'ch str, folder: F, base: usize) -> F |
349 | where |
350 | F: Folder<(usize, &'ch str)>, |
351 | { |
352 | folder.consume_iter(chars.match_indices($pattern).map(offset(base))) |
353 | } |
354 | } |
355 | } |
356 | |
357 | impl Pattern for char { |
358 | impl_pattern!(&self => *self); |
359 | } |
360 | |
361 | impl Pattern for &[char] { |
362 | impl_pattern!(&self => *self); |
363 | } |
364 | |
365 | impl<FN: Sync + Send + Fn(char) -> bool> Pattern for FN { |
366 | impl_pattern!(&self => self); |
367 | } |
368 | |
369 | // ///////////////////////////////////////////////////////////////////////// |
370 | |
371 | /// Parallel iterator over the characters of a string |
372 | #[derive(Debug, Clone)] |
373 | pub struct Chars<'ch> { |
374 | chars: &'ch str, |
375 | } |
376 | |
377 | struct CharsProducer<'ch> { |
378 | chars: &'ch str, |
379 | } |
380 | |
381 | impl<'ch> ParallelIterator for Chars<'ch> { |
382 | type Item = char; |
383 | |
384 | fn drive_unindexed<C>(self, consumer: C) -> C::Result |
385 | where |
386 | C: UnindexedConsumer<Self::Item>, |
387 | { |
388 | bridge_unindexed(CharsProducer { chars: self.chars }, consumer) |
389 | } |
390 | } |
391 | |
392 | impl<'ch> UnindexedProducer for CharsProducer<'ch> { |
393 | type Item = char; |
394 | |
395 | fn split(self) -> (Self, Option<Self>) { |
396 | match split(self.chars) { |
397 | Some((left, right)) => ( |
398 | CharsProducer { chars: left }, |
399 | Some(CharsProducer { chars: right }), |
400 | ), |
401 | None => (self, None), |
402 | } |
403 | } |
404 | |
405 | fn fold_with<F>(self, folder: F) -> F |
406 | where |
407 | F: Folder<Self::Item>, |
408 | { |
409 | folder.consume_iter(self.chars.chars()) |
410 | } |
411 | } |
412 | |
413 | // ///////////////////////////////////////////////////////////////////////// |
414 | |
415 | /// Parallel iterator over the characters of a string, with their positions |
416 | #[derive(Debug, Clone)] |
417 | pub struct CharIndices<'ch> { |
418 | chars: &'ch str, |
419 | } |
420 | |
421 | struct CharIndicesProducer<'ch> { |
422 | index: usize, |
423 | chars: &'ch str, |
424 | } |
425 | |
426 | impl<'ch> ParallelIterator for CharIndices<'ch> { |
427 | type Item = (usize, char); |
428 | |
429 | fn drive_unindexed<C>(self, consumer: C) -> C::Result |
430 | where |
431 | C: UnindexedConsumer<Self::Item>, |
432 | { |
433 | let producer = CharIndicesProducer { |
434 | index: 0, |
435 | chars: self.chars, |
436 | }; |
437 | bridge_unindexed(producer, consumer) |
438 | } |
439 | } |
440 | |
441 | impl<'ch> UnindexedProducer for CharIndicesProducer<'ch> { |
442 | type Item = (usize, char); |
443 | |
444 | fn split(self) -> (Self, Option<Self>) { |
445 | match split(self.chars) { |
446 | Some((left, right)) => ( |
447 | CharIndicesProducer { |
448 | chars: left, |
449 | ..self |
450 | }, |
451 | Some(CharIndicesProducer { |
452 | chars: right, |
453 | index: self.index + left.len(), |
454 | }), |
455 | ), |
456 | None => (self, None), |
457 | } |
458 | } |
459 | |
460 | fn fold_with<F>(self, folder: F) -> F |
461 | where |
462 | F: Folder<Self::Item>, |
463 | { |
464 | let base = self.index; |
465 | folder.consume_iter(self.chars.char_indices().map(offset(base))) |
466 | } |
467 | } |
468 | |
469 | // ///////////////////////////////////////////////////////////////////////// |
470 | |
471 | /// Parallel iterator over the bytes of a string |
472 | #[derive(Debug, Clone)] |
473 | pub struct Bytes<'ch> { |
474 | chars: &'ch str, |
475 | } |
476 | |
477 | struct BytesProducer<'ch> { |
478 | chars: &'ch str, |
479 | } |
480 | |
481 | impl<'ch> ParallelIterator for Bytes<'ch> { |
482 | type Item = u8; |
483 | |
484 | fn drive_unindexed<C>(self, consumer: C) -> C::Result |
485 | where |
486 | C: UnindexedConsumer<Self::Item>, |
487 | { |
488 | bridge_unindexed(BytesProducer { chars: self.chars }, consumer) |
489 | } |
490 | } |
491 | |
492 | impl<'ch> UnindexedProducer for BytesProducer<'ch> { |
493 | type Item = u8; |
494 | |
495 | fn split(self) -> (Self, Option<Self>) { |
496 | match split(self.chars) { |
497 | Some((left, right)) => ( |
498 | BytesProducer { chars: left }, |
499 | Some(BytesProducer { chars: right }), |
500 | ), |
501 | None => (self, None), |
502 | } |
503 | } |
504 | |
505 | fn fold_with<F>(self, folder: F) -> F |
506 | where |
507 | F: Folder<Self::Item>, |
508 | { |
509 | folder.consume_iter(self.chars.bytes()) |
510 | } |
511 | } |
512 | |
513 | // ///////////////////////////////////////////////////////////////////////// |
514 | |
515 | /// Parallel iterator over a string encoded as UTF-16 |
516 | #[derive(Debug, Clone)] |
517 | pub struct EncodeUtf16<'ch> { |
518 | chars: &'ch str, |
519 | } |
520 | |
521 | struct EncodeUtf16Producer<'ch> { |
522 | chars: &'ch str, |
523 | } |
524 | |
525 | impl<'ch> ParallelIterator for EncodeUtf16<'ch> { |
526 | type Item = u16; |
527 | |
528 | fn drive_unindexed<C>(self, consumer: C) -> C::Result |
529 | where |
530 | C: UnindexedConsumer<Self::Item>, |
531 | { |
532 | bridge_unindexed(EncodeUtf16Producer { chars: self.chars }, consumer) |
533 | } |
534 | } |
535 | |
536 | impl<'ch> UnindexedProducer for EncodeUtf16Producer<'ch> { |
537 | type Item = u16; |
538 | |
539 | fn split(self) -> (Self, Option<Self>) { |
540 | match split(self.chars) { |
541 | Some((left, right)) => ( |
542 | EncodeUtf16Producer { chars: left }, |
543 | Some(EncodeUtf16Producer { chars: right }), |
544 | ), |
545 | None => (self, None), |
546 | } |
547 | } |
548 | |
549 | fn fold_with<F>(self, folder: F) -> F |
550 | where |
551 | F: Folder<Self::Item>, |
552 | { |
553 | folder.consume_iter(self.chars.encode_utf16()) |
554 | } |
555 | } |
556 | |
557 | // ///////////////////////////////////////////////////////////////////////// |
558 | |
559 | /// Parallel iterator over substrings separated by a pattern |
560 | #[derive(Debug, Clone)] |
561 | pub struct Split<'ch, P: Pattern> { |
562 | chars: &'ch str, |
563 | separator: P, |
564 | } |
565 | |
566 | impl<'ch, P: Pattern> Split<'ch, P> { |
567 | fn new(chars: &'ch str, separator: P) -> Self { |
568 | Split { chars, separator } |
569 | } |
570 | } |
571 | |
572 | impl<'ch, P: Pattern> ParallelIterator for Split<'ch, P> { |
573 | type Item = &'ch str; |
574 | |
575 | fn drive_unindexed<C>(self, consumer: C) -> C::Result |
576 | where |
577 | C: UnindexedConsumer<Self::Item>, |
578 | { |
579 | let producer = SplitProducer::new(self.chars, &self.separator); |
580 | bridge_unindexed(producer, consumer) |
581 | } |
582 | } |
583 | |
584 | /// Implement support for `SplitProducer`. |
585 | impl<'ch, P: Pattern> Fissile<P> for &'ch str { |
586 | fn length(&self) -> usize { |
587 | self.len() |
588 | } |
589 | |
590 | fn midpoint(&self, end: usize) -> usize { |
591 | // First find a suitable UTF-8 boundary. |
592 | find_char_midpoint(&self[..end]) |
593 | } |
594 | |
595 | fn find(&self, separator: &P, start: usize, end: usize) -> Option<usize> { |
596 | separator.find_in(&self[start..end]) |
597 | } |
598 | |
599 | fn rfind(&self, separator: &P, end: usize) -> Option<usize> { |
600 | separator.rfind_in(&self[..end]) |
601 | } |
602 | |
603 | fn split_once(self, index: usize) -> (Self, Self) { |
604 | let (left, right) = self.split_at(index); |
605 | let mut right_iter = right.chars(); |
606 | right_iter.next(); // skip the separator |
607 | (left, right_iter.as_str()) |
608 | } |
609 | |
610 | fn fold_splits<F>(self, separator: &P, folder: F, skip_last: bool) -> F |
611 | where |
612 | F: Folder<Self>, |
613 | { |
614 | separator.fold_splits(self, folder, skip_last) |
615 | } |
616 | } |
617 | |
618 | // ///////////////////////////////////////////////////////////////////////// |
619 | |
620 | /// Parallel iterator over substrings separated by a terminator pattern |
621 | #[derive(Debug, Clone)] |
622 | pub struct SplitTerminator<'ch, P: Pattern> { |
623 | chars: &'ch str, |
624 | terminator: P, |
625 | } |
626 | |
627 | struct SplitTerminatorProducer<'ch, 'sep, P: Pattern> { |
628 | splitter: SplitProducer<'sep, P, &'ch str>, |
629 | skip_last: bool, |
630 | } |
631 | |
632 | impl<'ch, P: Pattern> SplitTerminator<'ch, P> { |
633 | fn new(chars: &'ch str, terminator: P) -> Self { |
634 | SplitTerminator { chars, terminator } |
635 | } |
636 | } |
637 | |
638 | impl<'ch, 'sep, P: Pattern + 'sep> SplitTerminatorProducer<'ch, 'sep, P> { |
639 | fn new(chars: &'ch str, terminator: &'sep P) -> Self { |
640 | SplitTerminatorProducer { |
641 | splitter: SplitProducer::new(chars, terminator), |
642 | skip_last: chars.is_empty() || terminator.is_suffix_of(chars), |
643 | } |
644 | } |
645 | } |
646 | |
647 | impl<'ch, P: Pattern> ParallelIterator for SplitTerminator<'ch, P> { |
648 | type Item = &'ch str; |
649 | |
650 | fn drive_unindexed<C>(self, consumer: C) -> C::Result |
651 | where |
652 | C: UnindexedConsumer<Self::Item>, |
653 | { |
654 | let producer = SplitTerminatorProducer::new(self.chars, &self.terminator); |
655 | bridge_unindexed(producer, consumer) |
656 | } |
657 | } |
658 | |
659 | impl<'ch, 'sep, P: Pattern + 'sep> UnindexedProducer for SplitTerminatorProducer<'ch, 'sep, P> { |
660 | type Item = &'ch str; |
661 | |
662 | fn split(mut self) -> (Self, Option<Self>) { |
663 | let (left, right) = self.splitter.split(); |
664 | self.splitter = left; |
665 | let right = right.map(|right| { |
666 | let skip_last = self.skip_last; |
667 | self.skip_last = false; |
668 | SplitTerminatorProducer { |
669 | splitter: right, |
670 | skip_last, |
671 | } |
672 | }); |
673 | (self, right) |
674 | } |
675 | |
676 | fn fold_with<F>(self, folder: F) -> F |
677 | where |
678 | F: Folder<Self::Item>, |
679 | { |
680 | self.splitter.fold_with(folder, self.skip_last) |
681 | } |
682 | } |
683 | |
684 | // ///////////////////////////////////////////////////////////////////////// |
685 | |
686 | /// Parallel iterator over lines in a string |
687 | #[derive(Debug, Clone)] |
688 | pub struct Lines<'ch>(&'ch str); |
689 | |
690 | #[inline ] |
691 | fn no_carriage_return(line: &str) -> &str { |
692 | line.strip_suffix(' \r' ).unwrap_or(line) |
693 | } |
694 | |
695 | impl<'ch> ParallelIterator for Lines<'ch> { |
696 | type Item = &'ch str; |
697 | |
698 | fn drive_unindexed<C>(self, consumer: C) -> C::Result |
699 | where |
700 | C: UnindexedConsumer<Self::Item>, |
701 | { |
702 | self.0 |
703 | .par_split_terminator(' \n' ) |
704 | .map(no_carriage_return) |
705 | .drive_unindexed(consumer) |
706 | } |
707 | } |
708 | |
709 | // ///////////////////////////////////////////////////////////////////////// |
710 | |
711 | /// Parallel iterator over substrings separated by whitespace |
712 | #[derive(Debug, Clone)] |
713 | pub struct SplitWhitespace<'ch>(&'ch str); |
714 | |
715 | #[inline ] |
716 | fn not_empty(s: &&str) -> bool { |
717 | !s.is_empty() |
718 | } |
719 | |
720 | impl<'ch> ParallelIterator for SplitWhitespace<'ch> { |
721 | type Item = &'ch str; |
722 | |
723 | fn drive_unindexed<C>(self, consumer: C) -> C::Result |
724 | where |
725 | C: UnindexedConsumer<Self::Item>, |
726 | { |
727 | self.0 |
728 | .par_split(char::is_whitespace) |
729 | .filter(not_empty) |
730 | .drive_unindexed(consumer) |
731 | } |
732 | } |
733 | |
734 | // ///////////////////////////////////////////////////////////////////////// |
735 | |
736 | /// Parallel iterator over substrings that match a pattern |
737 | #[derive(Debug, Clone)] |
738 | pub struct Matches<'ch, P: Pattern> { |
739 | chars: &'ch str, |
740 | pattern: P, |
741 | } |
742 | |
743 | struct MatchesProducer<'ch, 'pat, P: Pattern> { |
744 | chars: &'ch str, |
745 | pattern: &'pat P, |
746 | } |
747 | |
748 | impl<'ch, P: Pattern> ParallelIterator for Matches<'ch, P> { |
749 | type Item = &'ch str; |
750 | |
751 | fn drive_unindexed<C>(self, consumer: C) -> C::Result |
752 | where |
753 | C: UnindexedConsumer<Self::Item>, |
754 | { |
755 | let producer = MatchesProducer { |
756 | chars: self.chars, |
757 | pattern: &self.pattern, |
758 | }; |
759 | bridge_unindexed(producer, consumer) |
760 | } |
761 | } |
762 | |
763 | impl<'ch, 'pat, P: Pattern> UnindexedProducer for MatchesProducer<'ch, 'pat, P> { |
764 | type Item = &'ch str; |
765 | |
766 | fn split(self) -> (Self, Option<Self>) { |
767 | match split(self.chars) { |
768 | Some((left, right)) => ( |
769 | MatchesProducer { |
770 | chars: left, |
771 | ..self |
772 | }, |
773 | Some(MatchesProducer { |
774 | chars: right, |
775 | ..self |
776 | }), |
777 | ), |
778 | None => (self, None), |
779 | } |
780 | } |
781 | |
782 | fn fold_with<F>(self, folder: F) -> F |
783 | where |
784 | F: Folder<Self::Item>, |
785 | { |
786 | self.pattern.fold_matches(self.chars, folder) |
787 | } |
788 | } |
789 | |
790 | // ///////////////////////////////////////////////////////////////////////// |
791 | |
792 | /// Parallel iterator over substrings that match a pattern, with their positions |
793 | #[derive(Debug, Clone)] |
794 | pub struct MatchIndices<'ch, P: Pattern> { |
795 | chars: &'ch str, |
796 | pattern: P, |
797 | } |
798 | |
799 | struct MatchIndicesProducer<'ch, 'pat, P: Pattern> { |
800 | index: usize, |
801 | chars: &'ch str, |
802 | pattern: &'pat P, |
803 | } |
804 | |
805 | impl<'ch, P: Pattern> ParallelIterator for MatchIndices<'ch, P> { |
806 | type Item = (usize, &'ch str); |
807 | |
808 | fn drive_unindexed<C>(self, consumer: C) -> C::Result |
809 | where |
810 | C: UnindexedConsumer<Self::Item>, |
811 | { |
812 | let producer = MatchIndicesProducer { |
813 | index: 0, |
814 | chars: self.chars, |
815 | pattern: &self.pattern, |
816 | }; |
817 | bridge_unindexed(producer, consumer) |
818 | } |
819 | } |
820 | |
821 | impl<'ch, 'pat, P: Pattern> UnindexedProducer for MatchIndicesProducer<'ch, 'pat, P> { |
822 | type Item = (usize, &'ch str); |
823 | |
824 | fn split(self) -> (Self, Option<Self>) { |
825 | match split(self.chars) { |
826 | Some((left, right)) => ( |
827 | MatchIndicesProducer { |
828 | chars: left, |
829 | ..self |
830 | }, |
831 | Some(MatchIndicesProducer { |
832 | chars: right, |
833 | index: self.index + left.len(), |
834 | ..self |
835 | }), |
836 | ), |
837 | None => (self, None), |
838 | } |
839 | } |
840 | |
841 | fn fold_with<F>(self, folder: F) -> F |
842 | where |
843 | F: Folder<Self::Item>, |
844 | { |
845 | self.pattern |
846 | .fold_match_indices(self.chars, folder, self.index) |
847 | } |
848 | } |
849 | |