blob: ca01e0ec983108cca698518730eb4fa0cb7cb469 [file] [log] [blame]
Chih-Hung Hsiehe42c5052020-04-16 10:44:21 -07001use std::borrow::Cow;
2use std::collections::HashMap;
3use std::fmt;
4use std::ops::{Index, Range};
5use std::str::FromStr;
6use std::sync::Arc;
7
8use find_byte::find_byte;
9
10use error::Error;
11use exec::{Exec, ExecNoSync};
12use expand::expand_bytes;
13use re_builder::bytes::RegexBuilder;
14use re_trait::{self, RegularExpression, SubCapturesPosIter};
15
16/// Match represents a single match of a regex in a haystack.
17///
18/// The lifetime parameter `'t` refers to the lifetime of the matched text.
19#[derive(Copy, Clone, Debug, Eq, PartialEq)]
20pub struct Match<'t> {
21 text: &'t [u8],
22 start: usize,
23 end: usize,
24}
25
26impl<'t> Match<'t> {
27 /// Returns the starting byte offset of the match in the haystack.
28 #[inline]
29 pub fn start(&self) -> usize {
30 self.start
31 }
32
33 /// Returns the ending byte offset of the match in the haystack.
34 #[inline]
35 pub fn end(&self) -> usize {
36 self.end
37 }
38
39 /// Returns the range over the starting and ending byte offsets of the
40 /// match in the haystack.
41 #[inline]
42 pub fn range(&self) -> Range<usize> {
43 self.start..self.end
44 }
45
46 /// Returns the matched text.
47 #[inline]
48 pub fn as_bytes(&self) -> &'t [u8] {
49 &self.text[self.range()]
50 }
51
52 /// Creates a new match from the given haystack and byte offsets.
53 #[inline]
54 fn new(haystack: &'t [u8], start: usize, end: usize) -> Match<'t> {
55 Match { text: haystack, start: start, end: end }
56 }
57}
58
59impl<'t> From<Match<'t>> for Range<usize> {
60 fn from(m: Match<'t>) -> Range<usize> {
61 m.range()
62 }
63}
64
65/// A compiled regular expression for matching arbitrary bytes.
66///
67/// It can be used to search, split or replace text. All searching is done with
68/// an implicit `.*?` at the beginning and end of an expression. To force an
69/// expression to match the whole string (or a prefix or a suffix), you must
70/// use an anchor like `^` or `$` (or `\A` and `\z`).
71///
72/// Like the `Regex` type in the parent module, matches with this regex return
73/// byte offsets into the search text. **Unlike** the parent `Regex` type,
74/// these byte offsets may not correspond to UTF-8 sequence boundaries since
75/// the regexes in this module can match arbitrary bytes.
76#[derive(Clone)]
77pub struct Regex(Exec);
78
79impl fmt::Display for Regex {
80 /// Shows the original regular expression.
81 fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
82 write!(f, "{}", self.as_str())
83 }
84}
85
86impl fmt::Debug for Regex {
87 /// Shows the original regular expression.
88 fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
89 fmt::Display::fmt(self, f)
90 }
91}
92
93/// A constructor for Regex from an Exec.
94///
95/// This is hidden because Exec isn't actually part of the public API.
96#[doc(hidden)]
97impl From<Exec> for Regex {
98 fn from(exec: Exec) -> Regex {
99 Regex(exec)
100 }
101}
102
103impl FromStr for Regex {
104 type Err = Error;
105
106 /// Attempts to parse a string into a regular expression
107 fn from_str(s: &str) -> Result<Regex, Error> {
108 Regex::new(s)
109 }
110}
111
112/// Core regular expression methods.
113impl Regex {
114 /// Compiles a regular expression. Once compiled, it can be used repeatedly
115 /// to search, split or replace text in a string.
116 ///
117 /// If an invalid expression is given, then an error is returned.
118 pub fn new(re: &str) -> Result<Regex, Error> {
119 RegexBuilder::new(re).build()
120 }
121
Chih-Hung Hsieh849e4452020-10-26 13:16:47 -0700122 /// Returns true if and only if there is a match for the regex in the
123 /// string given.
Chih-Hung Hsiehe42c5052020-04-16 10:44:21 -0700124 ///
125 /// It is recommended to use this method if all you need to do is test
126 /// a match, since the underlying matching engine may be able to do less
127 /// work.
128 ///
129 /// # Example
130 ///
131 /// Test if some text contains at least one word with exactly 13 ASCII word
132 /// bytes:
133 ///
134 /// ```rust
135 /// # extern crate regex; use regex::bytes::Regex;
136 /// # fn main() {
137 /// let text = b"I categorically deny having triskaidekaphobia.";
138 /// assert!(Regex::new(r"\b\w{13}\b").unwrap().is_match(text));
139 /// # }
140 /// ```
141 pub fn is_match(&self, text: &[u8]) -> bool {
142 self.is_match_at(text, 0)
143 }
144
145 /// Returns the start and end byte range of the leftmost-first match in
146 /// `text`. If no match exists, then `None` is returned.
147 ///
148 /// Note that this should only be used if you want to discover the position
149 /// of the match. Testing the existence of a match is faster if you use
150 /// `is_match`.
151 ///
152 /// # Example
153 ///
154 /// Find the start and end location of the first word with exactly 13
155 /// ASCII word bytes:
156 ///
157 /// ```rust
158 /// # extern crate regex; use regex::bytes::Regex;
159 /// # fn main() {
160 /// let text = b"I categorically deny having triskaidekaphobia.";
161 /// let mat = Regex::new(r"\b\w{13}\b").unwrap().find(text).unwrap();
162 /// assert_eq!((mat.start(), mat.end()), (2, 15));
163 /// # }
164 /// ```
165 pub fn find<'t>(&self, text: &'t [u8]) -> Option<Match<'t>> {
166 self.find_at(text, 0)
167 }
168
169 /// Returns an iterator for each successive non-overlapping match in
170 /// `text`, returning the start and end byte indices with respect to
171 /// `text`.
172 ///
173 /// # Example
174 ///
175 /// Find the start and end location of every word with exactly 13 ASCII
176 /// word bytes:
177 ///
178 /// ```rust
179 /// # extern crate regex; use regex::bytes::Regex;
180 /// # fn main() {
181 /// let text = b"Retroactively relinquishing remunerations is reprehensible.";
182 /// for mat in Regex::new(r"\b\w{13}\b").unwrap().find_iter(text) {
183 /// println!("{:?}", mat);
184 /// }
185 /// # }
186 /// ```
187 pub fn find_iter<'r, 't>(&'r self, text: &'t [u8]) -> Matches<'r, 't> {
188 Matches(self.0.searcher().find_iter(text))
189 }
190
191 /// Returns the capture groups corresponding to the leftmost-first
192 /// match in `text`. Capture group `0` always corresponds to the entire
193 /// match. If no match is found, then `None` is returned.
194 ///
195 /// You should only use `captures` if you need access to the location of
196 /// capturing group matches. Otherwise, `find` is faster for discovering
197 /// the location of the overall match.
198 ///
199 /// # Examples
200 ///
201 /// Say you have some text with movie names and their release years,
202 /// like "'Citizen Kane' (1941)". It'd be nice if we could search for text
203 /// looking like that, while also extracting the movie name and its release
204 /// year separately.
205 ///
206 /// ```rust
207 /// # extern crate regex; use regex::bytes::Regex;
208 /// # fn main() {
209 /// let re = Regex::new(r"'([^']+)'\s+\((\d{4})\)").unwrap();
210 /// let text = b"Not my favorite movie: 'Citizen Kane' (1941).";
211 /// let caps = re.captures(text).unwrap();
212 /// assert_eq!(caps.get(1).unwrap().as_bytes(), &b"Citizen Kane"[..]);
213 /// assert_eq!(caps.get(2).unwrap().as_bytes(), &b"1941"[..]);
214 /// assert_eq!(caps.get(0).unwrap().as_bytes(), &b"'Citizen Kane' (1941)"[..]);
215 /// // You can also access the groups by index using the Index notation.
216 /// // Note that this will panic on an invalid index.
217 /// assert_eq!(&caps[1], b"Citizen Kane");
218 /// assert_eq!(&caps[2], b"1941");
219 /// assert_eq!(&caps[0], b"'Citizen Kane' (1941)");
220 /// # }
221 /// ```
222 ///
223 /// Note that the full match is at capture group `0`. Each subsequent
224 /// capture group is indexed by the order of its opening `(`.
225 ///
226 /// We can make this example a bit clearer by using *named* capture groups:
227 ///
228 /// ```rust
229 /// # extern crate regex; use regex::bytes::Regex;
230 /// # fn main() {
231 /// let re = Regex::new(r"'(?P<title>[^']+)'\s+\((?P<year>\d{4})\)")
232 /// .unwrap();
233 /// let text = b"Not my favorite movie: 'Citizen Kane' (1941).";
234 /// let caps = re.captures(text).unwrap();
235 /// assert_eq!(caps.name("title").unwrap().as_bytes(), b"Citizen Kane");
236 /// assert_eq!(caps.name("year").unwrap().as_bytes(), b"1941");
237 /// assert_eq!(caps.get(0).unwrap().as_bytes(), &b"'Citizen Kane' (1941)"[..]);
238 /// // You can also access the groups by name using the Index notation.
239 /// // Note that this will panic on an invalid group name.
240 /// assert_eq!(&caps["title"], b"Citizen Kane");
241 /// assert_eq!(&caps["year"], b"1941");
242 /// assert_eq!(&caps[0], b"'Citizen Kane' (1941)");
243 ///
244 /// # }
245 /// ```
246 ///
247 /// Here we name the capture groups, which we can access with the `name`
248 /// method or the `Index` notation with a `&str`. Note that the named
249 /// capture groups are still accessible with `get` or the `Index` notation
250 /// with a `usize`.
251 ///
252 /// The `0`th capture group is always unnamed, so it must always be
253 /// accessed with `get(0)` or `[0]`.
254 pub fn captures<'t>(&self, text: &'t [u8]) -> Option<Captures<'t>> {
255 let mut locs = self.capture_locations();
256 self.captures_read_at(&mut locs, text, 0).map(move |_| Captures {
257 text: text,
258 locs: locs.0,
259 named_groups: self.0.capture_name_idx().clone(),
260 })
261 }
262
263 /// Returns an iterator over all the non-overlapping capture groups matched
264 /// in `text`. This is operationally the same as `find_iter`, except it
265 /// yields information about capturing group matches.
266 ///
267 /// # Example
268 ///
269 /// We can use this to find all movie titles and their release years in
270 /// some text, where the movie is formatted like "'Title' (xxxx)":
271 ///
272 /// ```rust
273 /// # extern crate regex; use std::str; use regex::bytes::Regex;
274 /// # fn main() {
275 /// let re = Regex::new(r"'(?P<title>[^']+)'\s+\((?P<year>\d{4})\)")
276 /// .unwrap();
277 /// let text = b"'Citizen Kane' (1941), 'The Wizard of Oz' (1939), 'M' (1931).";
278 /// for caps in re.captures_iter(text) {
279 /// let title = str::from_utf8(&caps["title"]).unwrap();
280 /// let year = str::from_utf8(&caps["year"]).unwrap();
281 /// println!("Movie: {:?}, Released: {:?}", title, year);
282 /// }
283 /// // Output:
284 /// // Movie: Citizen Kane, Released: 1941
285 /// // Movie: The Wizard of Oz, Released: 1939
286 /// // Movie: M, Released: 1931
287 /// # }
288 /// ```
289 pub fn captures_iter<'r, 't>(
290 &'r self,
291 text: &'t [u8],
292 ) -> CaptureMatches<'r, 't> {
293 CaptureMatches(self.0.searcher().captures_iter(text))
294 }
295
296 /// Returns an iterator of substrings of `text` delimited by a match of the
297 /// regular expression. Namely, each element of the iterator corresponds to
298 /// text that *isn't* matched by the regular expression.
299 ///
300 /// This method will *not* copy the text given.
301 ///
302 /// # Example
303 ///
304 /// To split a string delimited by arbitrary amounts of spaces or tabs:
305 ///
306 /// ```rust
307 /// # extern crate regex; use regex::bytes::Regex;
308 /// # fn main() {
309 /// let re = Regex::new(r"[ \t]+").unwrap();
310 /// let fields: Vec<&[u8]> = re.split(b"a b \t c\td e").collect();
311 /// assert_eq!(fields, vec![
312 /// &b"a"[..], &b"b"[..], &b"c"[..], &b"d"[..], &b"e"[..],
313 /// ]);
314 /// # }
315 /// ```
316 pub fn split<'r, 't>(&'r self, text: &'t [u8]) -> Split<'r, 't> {
317 Split { finder: self.find_iter(text), last: 0 }
318 }
319
320 /// Returns an iterator of at most `limit` substrings of `text` delimited
321 /// by a match of the regular expression. (A `limit` of `0` will return no
322 /// substrings.) Namely, each element of the iterator corresponds to text
323 /// that *isn't* matched by the regular expression. The remainder of the
324 /// string that is not split will be the last element in the iterator.
325 ///
326 /// This method will *not* copy the text given.
327 ///
328 /// # Example
329 ///
330 /// Get the first two words in some text:
331 ///
332 /// ```rust
333 /// # extern crate regex; use regex::bytes::Regex;
334 /// # fn main() {
335 /// let re = Regex::new(r"\W+").unwrap();
336 /// let fields: Vec<&[u8]> = re.splitn(b"Hey! How are you?", 3).collect();
337 /// assert_eq!(fields, vec![&b"Hey"[..], &b"How"[..], &b"are you?"[..]]);
338 /// # }
339 /// ```
340 pub fn splitn<'r, 't>(
341 &'r self,
342 text: &'t [u8],
343 limit: usize,
344 ) -> SplitN<'r, 't> {
345 SplitN { splits: self.split(text), n: limit }
346 }
347
348 /// Replaces the leftmost-first match with the replacement provided. The
349 /// replacement can be a regular byte string (where `$N` and `$name` are
350 /// expanded to match capture groups) or a function that takes the matches'
351 /// `Captures` and returns the replaced byte string.
352 ///
353 /// If no match is found, then a copy of the byte string is returned
354 /// unchanged.
355 ///
356 /// # Replacement string syntax
357 ///
358 /// All instances of `$name` in the replacement text is replaced with the
359 /// corresponding capture group `name`.
360 ///
361 /// `name` may be an integer corresponding to the index of the
362 /// capture group (counted by order of opening parenthesis where `0` is the
363 /// entire match) or it can be a name (consisting of letters, digits or
364 /// underscores) corresponding to a named capture group.
365 ///
366 /// If `name` isn't a valid capture group (whether the name doesn't exist
367 /// or isn't a valid index), then it is replaced with the empty string.
368 ///
369 /// The longest possible name is used. e.g., `$1a` looks up the capture
370 /// group named `1a` and not the capture group at index `1`. To exert more
371 /// precise control over the name, use braces, e.g., `${1}a`.
372 ///
373 /// To write a literal `$` use `$$`.
374 ///
375 /// # Examples
376 ///
377 /// Note that this function is polymorphic with respect to the replacement.
378 /// In typical usage, this can just be a normal byte string:
379 ///
380 /// ```rust
381 /// # extern crate regex; use regex::bytes::Regex;
382 /// # fn main() {
383 /// let re = Regex::new("[^01]+").unwrap();
384 /// assert_eq!(re.replace(b"1078910", &b""[..]), &b"1010"[..]);
385 /// # }
386 /// ```
387 ///
388 /// But anything satisfying the `Replacer` trait will work. For example, a
389 /// closure of type `|&Captures| -> Vec<u8>` provides direct access to the
390 /// captures corresponding to a match. This allows one to access capturing
391 /// group matches easily:
392 ///
393 /// ```rust
394 /// # extern crate regex; use regex::bytes::Regex;
395 /// # use regex::bytes::Captures; fn main() {
396 /// let re = Regex::new(r"([^,\s]+),\s+(\S+)").unwrap();
397 /// let result = re.replace(b"Springsteen, Bruce", |caps: &Captures| {
398 /// let mut replacement = caps[2].to_owned();
399 /// replacement.push(b' ');
400 /// replacement.extend(&caps[1]);
401 /// replacement
402 /// });
403 /// assert_eq!(result, &b"Bruce Springsteen"[..]);
404 /// # }
405 /// ```
406 ///
407 /// But this is a bit cumbersome to use all the time. Instead, a simple
408 /// syntax is supported that expands `$name` into the corresponding capture
409 /// group. Here's the last example, but using this expansion technique
410 /// with named capture groups:
411 ///
412 /// ```rust
413 /// # extern crate regex; use regex::bytes::Regex;
414 /// # fn main() {
415 /// let re = Regex::new(r"(?P<last>[^,\s]+),\s+(?P<first>\S+)").unwrap();
416 /// let result = re.replace(b"Springsteen, Bruce", &b"$first $last"[..]);
417 /// assert_eq!(result, &b"Bruce Springsteen"[..]);
418 /// # }
419 /// ```
420 ///
421 /// Note that using `$2` instead of `$first` or `$1` instead of `$last`
422 /// would produce the same result. To write a literal `$` use `$$`.
423 ///
424 /// Sometimes the replacement string requires use of curly braces to
425 /// delineate a capture group replacement and surrounding literal text.
426 /// For example, if we wanted to join two words together with an
427 /// underscore:
428 ///
429 /// ```rust
430 /// # extern crate regex; use regex::bytes::Regex;
431 /// # fn main() {
432 /// let re = Regex::new(r"(?P<first>\w+)\s+(?P<second>\w+)").unwrap();
433 /// let result = re.replace(b"deep fried", &b"${first}_$second"[..]);
434 /// assert_eq!(result, &b"deep_fried"[..]);
435 /// # }
436 /// ```
437 ///
438 /// Without the curly braces, the capture group name `first_` would be
439 /// used, and since it doesn't exist, it would be replaced with the empty
440 /// string.
441 ///
442 /// Finally, sometimes you just want to replace a literal string with no
443 /// regard for capturing group expansion. This can be done by wrapping a
444 /// byte string with `NoExpand`:
445 ///
446 /// ```rust
447 /// # extern crate regex; use regex::bytes::Regex;
448 /// # fn main() {
449 /// use regex::bytes::NoExpand;
450 ///
451 /// let re = Regex::new(r"(?P<last>[^,\s]+),\s+(\S+)").unwrap();
452 /// let result = re.replace(b"Springsteen, Bruce", NoExpand(b"$2 $last"));
453 /// assert_eq!(result, &b"$2 $last"[..]);
454 /// # }
455 /// ```
456 pub fn replace<'t, R: Replacer>(
457 &self,
458 text: &'t [u8],
459 rep: R,
460 ) -> Cow<'t, [u8]> {
461 self.replacen(text, 1, rep)
462 }
463
464 /// Replaces all non-overlapping matches in `text` with the replacement
465 /// provided. This is the same as calling `replacen` with `limit` set to
466 /// `0`.
467 ///
468 /// See the documentation for `replace` for details on how to access
469 /// capturing group matches in the replacement text.
470 pub fn replace_all<'t, R: Replacer>(
471 &self,
472 text: &'t [u8],
473 rep: R,
474 ) -> Cow<'t, [u8]> {
475 self.replacen(text, 0, rep)
476 }
477
478 /// Replaces at most `limit` non-overlapping matches in `text` with the
479 /// replacement provided. If `limit` is 0, then all non-overlapping matches
480 /// are replaced.
481 ///
482 /// See the documentation for `replace` for details on how to access
483 /// capturing group matches in the replacement text.
484 pub fn replacen<'t, R: Replacer>(
485 &self,
486 text: &'t [u8],
487 limit: usize,
488 mut rep: R,
489 ) -> Cow<'t, [u8]> {
490 if let Some(rep) = rep.no_expansion() {
491 let mut it = self.find_iter(text).enumerate().peekable();
492 if it.peek().is_none() {
493 return Cow::Borrowed(text);
494 }
495 let mut new = Vec::with_capacity(text.len());
496 let mut last_match = 0;
497 for (i, m) in it {
498 if limit > 0 && i >= limit {
499 break;
500 }
501 new.extend_from_slice(&text[last_match..m.start()]);
502 new.extend_from_slice(&rep);
503 last_match = m.end();
504 }
505 new.extend_from_slice(&text[last_match..]);
506 return Cow::Owned(new);
507 }
508
509 // The slower path, which we use if the replacement needs access to
510 // capture groups.
511 let mut it = self.captures_iter(text).enumerate().peekable();
512 if it.peek().is_none() {
513 return Cow::Borrowed(text);
514 }
515 let mut new = Vec::with_capacity(text.len());
516 let mut last_match = 0;
517 for (i, cap) in it {
518 if limit > 0 && i >= limit {
519 break;
520 }
521 // unwrap on 0 is OK because captures only reports matches
522 let m = cap.get(0).unwrap();
523 new.extend_from_slice(&text[last_match..m.start()]);
524 rep.replace_append(&cap, &mut new);
525 last_match = m.end();
526 }
527 new.extend_from_slice(&text[last_match..]);
528 Cow::Owned(new)
529 }
530}
531
532/// Advanced or "lower level" search methods.
533impl Regex {
534 /// Returns the end location of a match in the text given.
535 ///
536 /// This method may have the same performance characteristics as
537 /// `is_match`, except it provides an end location for a match. In
538 /// particular, the location returned *may be shorter* than the proper end
539 /// of the leftmost-first match.
540 ///
541 /// # Example
542 ///
543 /// Typically, `a+` would match the entire first sequence of `a` in some
544 /// text, but `shortest_match` can give up as soon as it sees the first
545 /// `a`.
546 ///
547 /// ```rust
548 /// # extern crate regex; use regex::bytes::Regex;
549 /// # fn main() {
550 /// let text = b"aaaaa";
551 /// let pos = Regex::new(r"a+").unwrap().shortest_match(text);
552 /// assert_eq!(pos, Some(1));
553 /// # }
554 /// ```
555 pub fn shortest_match(&self, text: &[u8]) -> Option<usize> {
556 self.shortest_match_at(text, 0)
557 }
558
559 /// Returns the same as shortest_match, but starts the search at the given
560 /// offset.
561 ///
562 /// The significance of the starting point is that it takes the surrounding
563 /// context into consideration. For example, the `\A` anchor can only
564 /// match when `start == 0`.
565 pub fn shortest_match_at(
566 &self,
567 text: &[u8],
568 start: usize,
569 ) -> Option<usize> {
570 self.0.searcher().shortest_match_at(text, start)
571 }
572
573 /// Returns the same as is_match, but starts the search at the given
574 /// offset.
575 ///
576 /// The significance of the starting point is that it takes the surrounding
577 /// context into consideration. For example, the `\A` anchor can only
578 /// match when `start == 0`.
579 pub fn is_match_at(&self, text: &[u8], start: usize) -> bool {
580 self.shortest_match_at(text, start).is_some()
581 }
582
583 /// Returns the same as find, but starts the search at the given
584 /// offset.
585 ///
586 /// The significance of the starting point is that it takes the surrounding
587 /// context into consideration. For example, the `\A` anchor can only
588 /// match when `start == 0`.
589 pub fn find_at<'t>(
590 &self,
591 text: &'t [u8],
592 start: usize,
593 ) -> Option<Match<'t>> {
594 self.0
595 .searcher()
596 .find_at(text, start)
597 .map(|(s, e)| Match::new(text, s, e))
598 }
599
600 /// This is like `captures`, but uses
601 /// [`CaptureLocations`](struct.CaptureLocations.html)
602 /// instead of
603 /// [`Captures`](struct.Captures.html) in order to amortize allocations.
604 ///
605 /// To create a `CaptureLocations` value, use the
606 /// `Regex::capture_locations` method.
607 ///
608 /// This returns the overall match if this was successful, which is always
609 /// equivalence to the `0`th capture group.
610 pub fn captures_read<'t>(
611 &self,
612 locs: &mut CaptureLocations,
613 text: &'t [u8],
614 ) -> Option<Match<'t>> {
615 self.captures_read_at(locs, text, 0)
616 }
617
618 /// Returns the same as `captures_read`, but starts the search at the given
619 /// offset and populates the capture locations given.
620 ///
621 /// The significance of the starting point is that it takes the surrounding
622 /// context into consideration. For example, the `\A` anchor can only
623 /// match when `start == 0`.
624 pub fn captures_read_at<'t>(
625 &self,
626 locs: &mut CaptureLocations,
627 text: &'t [u8],
628 start: usize,
629 ) -> Option<Match<'t>> {
630 self.0
631 .searcher()
632 .captures_read_at(&mut locs.0, text, start)
633 .map(|(s, e)| Match::new(text, s, e))
634 }
635
636 /// An undocumented alias for `captures_read_at`.
637 ///
638 /// The `regex-capi` crate previously used this routine, so to avoid
639 /// breaking that crate, we continue to provide the name as an undocumented
640 /// alias.
641 #[doc(hidden)]
642 pub fn read_captures_at<'t>(
643 &self,
644 locs: &mut CaptureLocations,
645 text: &'t [u8],
646 start: usize,
647 ) -> Option<Match<'t>> {
648 self.captures_read_at(locs, text, start)
649 }
650}
651
652/// Auxiliary methods.
653impl Regex {
654 /// Returns the original string of this regex.
655 pub fn as_str(&self) -> &str {
656 &self.0.regex_strings()[0]
657 }
658
659 /// Returns an iterator over the capture names.
660 pub fn capture_names(&self) -> CaptureNames {
661 CaptureNames(self.0.capture_names().iter())
662 }
663
664 /// Returns the number of captures.
665 pub fn captures_len(&self) -> usize {
666 self.0.capture_names().len()
667 }
668
669 /// Returns an empty set of capture locations that can be reused in
670 /// multiple calls to `captures_read` or `captures_read_at`.
671 pub fn capture_locations(&self) -> CaptureLocations {
672 CaptureLocations(self.0.searcher().locations())
673 }
674
675 /// An alias for `capture_locations` to preserve backward compatibility.
676 ///
677 /// The `regex-capi` crate uses this method, so to avoid breaking that
678 /// crate, we continue to export it as an undocumented API.
679 #[doc(hidden)]
680 pub fn locations(&self) -> CaptureLocations {
681 CaptureLocations(self.0.searcher().locations())
682 }
683}
684
685/// An iterator over all non-overlapping matches for a particular string.
686///
687/// The iterator yields a tuple of integers corresponding to the start and end
688/// of the match. The indices are byte offsets. The iterator stops when no more
689/// matches can be found.
690///
691/// `'r` is the lifetime of the compiled regular expression and `'t` is the
692/// lifetime of the matched byte string.
693pub struct Matches<'r, 't>(re_trait::Matches<'t, ExecNoSync<'r>>);
694
695impl<'r, 't> Iterator for Matches<'r, 't> {
696 type Item = Match<'t>;
697
698 fn next(&mut self) -> Option<Match<'t>> {
699 let text = self.0.text();
700 self.0.next().map(|(s, e)| Match::new(text, s, e))
701 }
702}
703
704/// An iterator that yields all non-overlapping capture groups matching a
705/// particular regular expression.
706///
707/// The iterator stops when no more matches can be found.
708///
709/// `'r` is the lifetime of the compiled regular expression and `'t` is the
710/// lifetime of the matched byte string.
711pub struct CaptureMatches<'r, 't>(
712 re_trait::CaptureMatches<'t, ExecNoSync<'r>>,
713);
714
715impl<'r, 't> Iterator for CaptureMatches<'r, 't> {
716 type Item = Captures<'t>;
717
718 fn next(&mut self) -> Option<Captures<'t>> {
719 self.0.next().map(|locs| Captures {
720 text: self.0.text(),
721 locs: locs,
722 named_groups: self.0.regex().capture_name_idx().clone(),
723 })
724 }
725}
726
727/// Yields all substrings delimited by a regular expression match.
728///
729/// `'r` is the lifetime of the compiled regular expression and `'t` is the
730/// lifetime of the byte string being split.
731pub struct Split<'r, 't> {
732 finder: Matches<'r, 't>,
733 last: usize,
734}
735
736impl<'r, 't> Iterator for Split<'r, 't> {
737 type Item = &'t [u8];
738
739 fn next(&mut self) -> Option<&'t [u8]> {
740 let text = self.finder.0.text();
741 match self.finder.next() {
742 None => {
743 if self.last > text.len() {
744 None
745 } else {
746 let s = &text[self.last..];
747 self.last = text.len() + 1; // Next call will return None
748 Some(s)
749 }
750 }
751 Some(m) => {
752 let matched = &text[self.last..m.start()];
753 self.last = m.end();
754 Some(matched)
755 }
756 }
757 }
758}
759
760/// Yields at most `N` substrings delimited by a regular expression match.
761///
762/// The last substring will be whatever remains after splitting.
763///
764/// `'r` is the lifetime of the compiled regular expression and `'t` is the
765/// lifetime of the byte string being split.
766pub struct SplitN<'r, 't> {
767 splits: Split<'r, 't>,
768 n: usize,
769}
770
771impl<'r, 't> Iterator for SplitN<'r, 't> {
772 type Item = &'t [u8];
773
774 fn next(&mut self) -> Option<&'t [u8]> {
775 if self.n == 0 {
776 return None;
777 }
778
779 self.n -= 1;
780 if self.n > 0 {
781 return self.splits.next();
782 }
783
784 let text = self.splits.finder.0.text();
785 if self.splits.last > text.len() {
786 // We've already returned all substrings.
787 None
788 } else {
789 // self.n == 0, so future calls will return None immediately
790 Some(&text[self.splits.last..])
791 }
792 }
793}
794
795/// An iterator over the names of all possible captures.
796///
797/// `None` indicates an unnamed capture; the first element (capture 0, the
798/// whole matched region) is always unnamed.
799///
800/// `'r` is the lifetime of the compiled regular expression.
801pub struct CaptureNames<'r>(::std::slice::Iter<'r, Option<String>>);
802
803impl<'r> Iterator for CaptureNames<'r> {
804 type Item = Option<&'r str>;
805
806 fn next(&mut self) -> Option<Option<&'r str>> {
807 self.0
808 .next()
809 .as_ref()
810 .map(|slot| slot.as_ref().map(|name| name.as_ref()))
811 }
812
813 fn size_hint(&self) -> (usize, Option<usize>) {
814 self.0.size_hint()
815 }
816}
817
818/// CaptureLocations is a low level representation of the raw offsets of each
819/// submatch.
820///
821/// You can think of this as a lower level
822/// [`Captures`](struct.Captures.html), where this type does not support
823/// named capturing groups directly and it does not borrow the text that these
824/// offsets were matched on.
825///
826/// Primarily, this type is useful when using the lower level `Regex` APIs
827/// such as `read_captures`, which permits amortizing the allocation in which
828/// capture match locations are stored.
829///
830/// In order to build a value of this type, you'll need to call the
831/// `capture_locations` method on the `Regex` being used to execute the search.
832/// The value returned can then be reused in subsequent searches.
833#[derive(Clone, Debug)]
834pub struct CaptureLocations(re_trait::Locations);
835
836/// A type alias for `CaptureLocations` for backwards compatibility.
837///
838/// Previously, we exported `CaptureLocations` as `Locations` in an
839/// undocumented API. To prevent breaking that code (e.g., in `regex-capi`),
840/// we continue re-exporting the same undocumented API.
841#[doc(hidden)]
842pub type Locations = CaptureLocations;
843
844impl CaptureLocations {
845 /// Returns the start and end positions of the Nth capture group. Returns
846 /// `None` if `i` is not a valid capture group or if the capture group did
847 /// not match anything. The positions returned are *always* byte indices
848 /// with respect to the original string matched.
849 #[inline]
850 pub fn get(&self, i: usize) -> Option<(usize, usize)> {
851 self.0.pos(i)
852 }
853
854 /// Returns the total number of capturing groups.
855 ///
856 /// This is always at least `1` since every regex has at least `1`
857 /// capturing group that corresponds to the entire match.
858 #[inline]
859 pub fn len(&self) -> usize {
860 self.0.len()
861 }
862
863 /// An alias for the `get` method for backwards compatibility.
864 ///
865 /// Previously, we exported `get` as `pos` in an undocumented API. To
866 /// prevent breaking that code (e.g., in `regex-capi`), we continue
867 /// re-exporting the same undocumented API.
868 #[doc(hidden)]
869 #[inline]
870 pub fn pos(&self, i: usize) -> Option<(usize, usize)> {
871 self.get(i)
872 }
873}
874
875/// Captures represents a group of captured byte strings for a single match.
876///
877/// The 0th capture always corresponds to the entire match. Each subsequent
878/// index corresponds to the next capture group in the regex. If a capture
879/// group is named, then the matched byte string is *also* available via the
880/// `name` method. (Note that the 0th capture is always unnamed and so must be
881/// accessed with the `get` method.)
882///
883/// Positions returned from a capture group are always byte indices.
884///
885/// `'t` is the lifetime of the matched text.
886pub struct Captures<'t> {
887 text: &'t [u8],
888 locs: re_trait::Locations,
889 named_groups: Arc<HashMap<String, usize>>,
890}
891
892impl<'t> Captures<'t> {
893 /// Returns the match associated with the capture group at index `i`. If
894 /// `i` does not correspond to a capture group, or if the capture group
895 /// did not participate in the match, then `None` is returned.
896 ///
897 /// # Examples
898 ///
899 /// Get the text of the match with a default of an empty string if this
900 /// group didn't participate in the match:
901 ///
902 /// ```rust
903 /// # use regex::bytes::Regex;
904 /// let re = Regex::new(r"[a-z]+(?:([0-9]+)|([A-Z]+))").unwrap();
905 /// let caps = re.captures(b"abc123").unwrap();
906 ///
907 /// let text1 = caps.get(1).map_or(&b""[..], |m| m.as_bytes());
908 /// let text2 = caps.get(2).map_or(&b""[..], |m| m.as_bytes());
909 /// assert_eq!(text1, &b"123"[..]);
910 /// assert_eq!(text2, &b""[..]);
911 /// ```
912 pub fn get(&self, i: usize) -> Option<Match<'t>> {
913 self.locs.pos(i).map(|(s, e)| Match::new(self.text, s, e))
914 }
915
916 /// Returns the match for the capture group named `name`. If `name` isn't a
917 /// valid capture group or didn't match anything, then `None` is returned.
918 pub fn name(&self, name: &str) -> Option<Match<'t>> {
919 self.named_groups.get(name).and_then(|&i| self.get(i))
920 }
921
922 /// An iterator that yields all capturing matches in the order in which
923 /// they appear in the regex. If a particular capture group didn't
924 /// participate in the match, then `None` is yielded for that capture.
925 ///
926 /// The first match always corresponds to the overall match of the regex.
927 pub fn iter<'c>(&'c self) -> SubCaptureMatches<'c, 't> {
928 SubCaptureMatches { caps: self, it: self.locs.iter() }
929 }
930
931 /// Expands all instances of `$name` in `replacement` to the corresponding
932 /// capture group `name`, and writes them to the `dst` buffer given.
933 ///
Chih-Hung Hsieh849e4452020-10-26 13:16:47 -0700934 /// `name` may be an integer corresponding to the index of the capture
935 /// group (counted by order of opening parenthesis where `0` is the
Chih-Hung Hsiehe42c5052020-04-16 10:44:21 -0700936 /// entire match) or it can be a name (consisting of letters, digits or
937 /// underscores) corresponding to a named capture group.
938 ///
939 /// If `name` isn't a valid capture group (whether the name doesn't exist
940 /// or isn't a valid index), then it is replaced with the empty string.
941 ///
Chih-Hung Hsieh849e4452020-10-26 13:16:47 -0700942 /// The longest possible name consisting of the characters `[_0-9A-Za-z]`
943 /// is used. e.g., `$1a` looks up the capture group named `1a` and not the
944 /// capture group at index `1`. To exert more precise control over the
945 /// name, or to refer to a capture group name that uses characters outside
946 /// of `[_0-9A-Za-z]`, use braces, e.g., `${1}a` or `${foo[bar].baz}`. When
947 /// using braces, any sequence of valid UTF-8 bytes is permitted. If the
948 /// sequence does not refer to a capture group name in the corresponding
949 /// regex, then it is replaced with an empty string.
Chih-Hung Hsiehe42c5052020-04-16 10:44:21 -0700950 ///
951 /// To write a literal `$` use `$$`.
952 pub fn expand(&self, replacement: &[u8], dst: &mut Vec<u8>) {
953 expand_bytes(self, replacement, dst)
954 }
955
956 /// Returns the number of captured groups.
957 ///
958 /// This is always at least `1`, since every regex has at least one capture
959 /// group that corresponds to the full match.
960 #[inline]
961 pub fn len(&self) -> usize {
962 self.locs.len()
963 }
964}
965
966impl<'t> fmt::Debug for Captures<'t> {
967 fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
968 f.debug_tuple("Captures").field(&CapturesDebug(self)).finish()
969 }
970}
971
972struct CapturesDebug<'c, 't: 'c>(&'c Captures<'t>);
973
974impl<'c, 't> fmt::Debug for CapturesDebug<'c, 't> {
975 fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
976 fn escape_bytes(bytes: &[u8]) -> String {
977 let mut s = String::new();
978 for &b in bytes {
979 s.push_str(&escape_byte(b));
980 }
981 s
982 }
983
984 fn escape_byte(byte: u8) -> String {
985 use std::ascii::escape_default;
986
987 let escaped: Vec<u8> = escape_default(byte).collect();
988 String::from_utf8_lossy(&escaped).into_owned()
989 }
990
991 // We'd like to show something nice here, even if it means an
992 // allocation to build a reverse index.
993 let slot_to_name: HashMap<&usize, &String> =
994 self.0.named_groups.iter().map(|(a, b)| (b, a)).collect();
995 let mut map = f.debug_map();
996 for (slot, m) in self.0.locs.iter().enumerate() {
997 let m = m.map(|(s, e)| escape_bytes(&self.0.text[s..e]));
998 if let Some(name) = slot_to_name.get(&slot) {
999 map.entry(&name, &m);
1000 } else {
1001 map.entry(&slot, &m);
1002 }
1003 }
1004 map.finish()
1005 }
1006}
1007
1008/// Get a group by index.
1009///
1010/// `'t` is the lifetime of the matched text.
1011///
1012/// The text can't outlive the `Captures` object if this method is
1013/// used, because of how `Index` is defined (normally `a[i]` is part
1014/// of `a` and can't outlive it); to do that, use `get()` instead.
1015///
1016/// # Panics
1017///
1018/// If there is no group at the given index.
1019impl<'t> Index<usize> for Captures<'t> {
1020 type Output = [u8];
1021
1022 fn index(&self, i: usize) -> &[u8] {
1023 self.get(i)
1024 .map(|m| m.as_bytes())
1025 .unwrap_or_else(|| panic!("no group at index '{}'", i))
1026 }
1027}
1028
1029/// Get a group by name.
1030///
1031/// `'t` is the lifetime of the matched text and `'i` is the lifetime
1032/// of the group name (the index).
1033///
1034/// The text can't outlive the `Captures` object if this method is
1035/// used, because of how `Index` is defined (normally `a[i]` is part
1036/// of `a` and can't outlive it); to do that, use `name` instead.
1037///
1038/// # Panics
1039///
1040/// If there is no group named by the given value.
1041impl<'t, 'i> Index<&'i str> for Captures<'t> {
1042 type Output = [u8];
1043
1044 fn index<'a>(&'a self, name: &'i str) -> &'a [u8] {
1045 self.name(name)
1046 .map(|m| m.as_bytes())
1047 .unwrap_or_else(|| panic!("no group named '{}'", name))
1048 }
1049}
1050
1051/// An iterator that yields all capturing matches in the order in which they
1052/// appear in the regex.
1053///
1054/// If a particular capture group didn't participate in the match, then `None`
1055/// is yielded for that capture. The first match always corresponds to the
1056/// overall match of the regex.
1057///
1058/// The lifetime `'c` corresponds to the lifetime of the `Captures` value, and
1059/// the lifetime `'t` corresponds to the originally matched text.
Chih-Hung Hsieh849e4452020-10-26 13:16:47 -07001060#[derive(Clone)]
Chih-Hung Hsiehe42c5052020-04-16 10:44:21 -07001061pub struct SubCaptureMatches<'c, 't: 'c> {
1062 caps: &'c Captures<'t>,
1063 it: SubCapturesPosIter<'c>,
1064}
1065
1066impl<'c, 't> Iterator for SubCaptureMatches<'c, 't> {
1067 type Item = Option<Match<'t>>;
1068
1069 fn next(&mut self) -> Option<Option<Match<'t>>> {
1070 self.it
1071 .next()
1072 .map(|cap| cap.map(|(s, e)| Match::new(self.caps.text, s, e)))
1073 }
1074}
1075
1076/// Replacer describes types that can be used to replace matches in a byte
1077/// string.
1078///
1079/// In general, users of this crate shouldn't need to implement this trait,
1080/// since implementations are already provided for `&[u8]` and
1081/// `FnMut(&Captures) -> Vec<u8>` (or any `FnMut(&Captures) -> T`
1082/// where `T: AsRef<[u8]>`), which covers most use cases.
1083pub trait Replacer {
1084 /// Appends text to `dst` to replace the current match.
1085 ///
1086 /// The current match is represented by `caps`, which is guaranteed to
1087 /// have a match at capture group `0`.
1088 ///
1089 /// For example, a no-op replacement would be
1090 /// `dst.extend(&caps[0])`.
1091 fn replace_append(&mut self, caps: &Captures, dst: &mut Vec<u8>);
1092
1093 /// Return a fixed unchanging replacement byte string.
1094 ///
1095 /// When doing replacements, if access to `Captures` is not needed (e.g.,
1096 /// the replacement byte string does not need `$` expansion), then it can
1097 /// be beneficial to avoid finding sub-captures.
1098 ///
1099 /// In general, this is called once for every call to `replacen`.
1100 fn no_expansion<'r>(&'r mut self) -> Option<Cow<'r, [u8]>> {
1101 None
1102 }
1103
1104 /// Return a `Replacer` that borrows and wraps this `Replacer`.
1105 ///
1106 /// This is useful when you want to take a generic `Replacer` (which might
1107 /// not be cloneable) and use it without consuming it, so it can be used
1108 /// more than once.
1109 ///
1110 /// # Example
1111 ///
1112 /// ```
1113 /// use regex::bytes::{Regex, Replacer};
1114 ///
1115 /// fn replace_all_twice<R: Replacer>(
1116 /// re: Regex,
1117 /// src: &[u8],
1118 /// mut rep: R,
1119 /// ) -> Vec<u8> {
1120 /// let dst = re.replace_all(src, rep.by_ref());
1121 /// let dst = re.replace_all(&dst, rep.by_ref());
1122 /// dst.into_owned()
1123 /// }
1124 /// ```
1125 fn by_ref<'r>(&'r mut self) -> ReplacerRef<'r, Self> {
1126 ReplacerRef(self)
1127 }
1128}
1129
1130/// By-reference adaptor for a `Replacer`
1131///
1132/// Returned by [`Replacer::by_ref`](trait.Replacer.html#method.by_ref).
1133#[derive(Debug)]
1134pub struct ReplacerRef<'a, R: ?Sized + 'a>(&'a mut R);
1135
1136impl<'a, R: Replacer + ?Sized + 'a> Replacer for ReplacerRef<'a, R> {
1137 fn replace_append(&mut self, caps: &Captures, dst: &mut Vec<u8>) {
1138 self.0.replace_append(caps, dst)
1139 }
1140 fn no_expansion<'r>(&'r mut self) -> Option<Cow<'r, [u8]>> {
1141 self.0.no_expansion()
1142 }
1143}
1144
1145impl<'a> Replacer for &'a [u8] {
1146 fn replace_append(&mut self, caps: &Captures, dst: &mut Vec<u8>) {
1147 caps.expand(*self, dst);
1148 }
1149
1150 fn no_expansion(&mut self) -> Option<Cow<[u8]>> {
1151 match find_byte(b'$', *self) {
1152 Some(_) => None,
1153 None => Some(Cow::Borrowed(*self)),
1154 }
1155 }
1156}
1157
1158impl<F, T> Replacer for F
1159where
1160 F: FnMut(&Captures) -> T,
1161 T: AsRef<[u8]>,
1162{
1163 fn replace_append(&mut self, caps: &Captures, dst: &mut Vec<u8>) {
1164 dst.extend_from_slice((*self)(caps).as_ref());
1165 }
1166}
1167
1168/// `NoExpand` indicates literal byte string replacement.
1169///
1170/// It can be used with `replace` and `replace_all` to do a literal byte string
1171/// replacement without expanding `$name` to their corresponding capture
1172/// groups. This can be both convenient (to avoid escaping `$`, for example)
1173/// and performant (since capture groups don't need to be found).
1174///
1175/// `'t` is the lifetime of the literal text.
1176pub struct NoExpand<'t>(pub &'t [u8]);
1177
1178impl<'t> Replacer for NoExpand<'t> {
1179 fn replace_append(&mut self, _: &Captures, dst: &mut Vec<u8>) {
1180 dst.extend_from_slice(self.0);
1181 }
1182
1183 fn no_expansion(&mut self) -> Option<Cow<[u8]>> {
1184 Some(Cow::Borrowed(self.0))
1185 }
1186}