1 | use alloc::{borrow::Cow, string::String, sync::Arc, vec::Vec}; |
2 | |
3 | use regex_automata::{meta, util::captures, Input, PatternID}; |
4 | |
5 | use crate::{bytes::RegexBuilder, error::Error}; |
6 | |
7 | /// A compiled regular expression for searching Unicode haystacks. |
8 | /// |
9 | /// A `Regex` can be used to search haystacks, split haystacks into substrings |
10 | /// or replace substrings in a haystack with a different substring. All |
11 | /// searching is done with an implicit `(?s:.)*?` at the beginning and end of |
12 | /// an pattern. To force an expression to match the whole string (or a prefix |
13 | /// or a suffix), you must use an anchor like `^` or `$` (or `\A` and `\z`). |
14 | /// |
15 | /// Like the `Regex` type in the parent module, matches with this regex return |
16 | /// byte offsets into the haystack. **Unlike** the parent `Regex` type, these |
17 | /// byte offsets may not correspond to UTF-8 sequence boundaries since the |
18 | /// regexes in this module can match arbitrary bytes. |
19 | /// |
20 | /// The only methods that allocate new byte strings are the string replacement |
21 | /// methods. All other methods (searching and splitting) return borrowed |
22 | /// references into the haystack given. |
23 | /// |
24 | /// # Example |
25 | /// |
26 | /// Find the offsets of a US phone number: |
27 | /// |
28 | /// ``` |
29 | /// use regex::bytes::Regex; |
30 | /// |
31 | /// let re = Regex::new("[0-9]{3}-[0-9]{3}-[0-9]{4}" ).unwrap(); |
32 | /// let m = re.find(b"phone: 111-222-3333" ).unwrap(); |
33 | /// assert_eq!(7..19, m.range()); |
34 | /// ``` |
35 | /// |
36 | /// # Example: extracting capture groups |
37 | /// |
38 | /// A common way to use regexes is with capture groups. That is, instead of |
39 | /// just looking for matches of an entire regex, parentheses are used to create |
40 | /// groups that represent part of the match. |
41 | /// |
42 | /// For example, consider a haystack with multiple lines, and each line has |
43 | /// three whitespace delimited fields where the second field is expected to be |
44 | /// a number and the third field a boolean. To make this convenient, we use |
45 | /// the [`Captures::extract`] API to put the strings that match each group |
46 | /// into a fixed size array: |
47 | /// |
48 | /// ``` |
49 | /// use regex::bytes::Regex; |
50 | /// |
51 | /// let hay = b" |
52 | /// rabbit 54 true |
53 | /// groundhog 2 true |
54 | /// does not match |
55 | /// fox 109 false |
56 | /// " ; |
57 | /// let re = Regex::new(r"(?m)^\s*(\S+)\s+([0-9]+)\s+(true|false)\s*$" ).unwrap(); |
58 | /// let mut fields: Vec<(&[u8], i64, bool)> = vec![]; |
59 | /// for (_, [f1, f2, f3]) in re.captures_iter(hay).map(|caps| caps.extract()) { |
60 | /// // These unwraps are OK because our pattern is written in a way where |
61 | /// // all matches for f2 and f3 will be valid UTF-8. |
62 | /// let f2 = std::str::from_utf8(f2).unwrap(); |
63 | /// let f3 = std::str::from_utf8(f3).unwrap(); |
64 | /// fields.push((f1, f2.parse()?, f3.parse()?)); |
65 | /// } |
66 | /// assert_eq!(fields, vec![ |
67 | /// (&b"rabbit" [..], 54, true), |
68 | /// (&b"groundhog" [..], 2, true), |
69 | /// (&b"fox" [..], 109, false), |
70 | /// ]); |
71 | /// |
72 | /// # Ok::<(), Box<dyn std::error::Error>>(()) |
73 | /// ``` |
74 | /// |
75 | /// # Example: matching invalid UTF-8 |
76 | /// |
77 | /// One of the reasons for searching `&[u8]` haystacks is that the `&[u8]` |
78 | /// might not be valid UTF-8. Indeed, with a `bytes::Regex`, patterns that |
79 | /// match invalid UTF-8 are explicitly allowed. Here's one example that looks |
80 | /// for valid UTF-8 fields that might be separated by invalid UTF-8. In this |
81 | /// case, we use `(?s-u:.)`, which matches any byte. Attempting to use it in a |
82 | /// top-level `Regex` will result in the regex failing to compile. Notice also |
83 | /// that we use `.` with Unicode mode enabled, in which case, only valid UTF-8 |
84 | /// is matched. In this way, we can build one pattern where some parts only |
85 | /// match valid UTF-8 while other parts are more permissive. |
86 | /// |
87 | /// ``` |
88 | /// use regex::bytes::Regex; |
89 | /// |
90 | /// // F0 9F 92 A9 is the UTF-8 encoding for a Pile of Poo. |
91 | /// let hay = b" \xFF\xFFfoo \xFF\xFF\xFF\xF0\x9F\x92\xA9\xFF" ; |
92 | /// // An equivalent to '(?s-u:.)' is '(?-u:[\x00-\xFF])'. |
93 | /// let re = Regex::new(r"(?s)(?-u:.)*?(?<f1>.+)(?-u:.)*?(?<f2>.+)" ).unwrap(); |
94 | /// let caps = re.captures(hay).unwrap(); |
95 | /// assert_eq!(&caps["f1" ], &b"foo" [..]); |
96 | /// assert_eq!(&caps["f2" ], "💩" .as_bytes()); |
97 | /// ``` |
98 | #[derive(Clone)] |
99 | pub struct Regex { |
100 | pub(crate) meta: meta::Regex, |
101 | pub(crate) pattern: Arc<str>, |
102 | } |
103 | |
104 | impl core::fmt::Display for Regex { |
105 | /// Shows the original regular expression. |
106 | fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result { |
107 | write!(f, "{}" , self.as_str()) |
108 | } |
109 | } |
110 | |
111 | impl core::fmt::Debug for Regex { |
112 | /// Shows the original regular expression. |
113 | fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result { |
114 | f.debug_tuple("Regex" ).field(&self.as_str()).finish() |
115 | } |
116 | } |
117 | |
118 | impl core::str::FromStr for Regex { |
119 | type Err = Error; |
120 | |
121 | /// Attempts to parse a string into a regular expression |
122 | fn from_str(s: &str) -> Result<Regex, Error> { |
123 | Regex::new(s) |
124 | } |
125 | } |
126 | |
127 | impl TryFrom<&str> for Regex { |
128 | type Error = Error; |
129 | |
130 | /// Attempts to parse a string into a regular expression |
131 | fn try_from(s: &str) -> Result<Regex, Error> { |
132 | Regex::new(s) |
133 | } |
134 | } |
135 | |
136 | impl TryFrom<String> for Regex { |
137 | type Error = Error; |
138 | |
139 | /// Attempts to parse a string into a regular expression |
140 | fn try_from(s: String) -> Result<Regex, Error> { |
141 | Regex::new(&s) |
142 | } |
143 | } |
144 | |
145 | /// Core regular expression methods. |
146 | impl Regex { |
147 | /// Compiles a regular expression. Once compiled, it can be used repeatedly |
148 | /// to search, split or replace substrings in a haystack. |
149 | /// |
150 | /// Note that regex compilation tends to be a somewhat expensive process, |
151 | /// and unlike higher level environments, compilation is not automatically |
152 | /// cached for you. One should endeavor to compile a regex once and then |
153 | /// reuse it. For example, it's a bad idea to compile the same regex |
154 | /// repeatedly in a loop. |
155 | /// |
156 | /// # Errors |
157 | /// |
158 | /// If an invalid pattern is given, then an error is returned. |
159 | /// An error is also returned if the pattern is valid, but would |
160 | /// produce a regex that is bigger than the configured size limit via |
161 | /// [`RegexBuilder::size_limit`]. (A reasonable size limit is enabled by |
162 | /// default.) |
163 | /// |
164 | /// # Example |
165 | /// |
166 | /// ``` |
167 | /// use regex::bytes::Regex; |
168 | /// |
169 | /// // An Invalid pattern because of an unclosed parenthesis |
170 | /// assert!(Regex::new(r"foo(bar" ).is_err()); |
171 | /// // An invalid pattern because the regex would be too big |
172 | /// // because Unicode tends to inflate things. |
173 | /// assert!(Regex::new(r"\w{1000}" ).is_err()); |
174 | /// // Disabling Unicode can make the regex much smaller, |
175 | /// // potentially by up to or more than an order of magnitude. |
176 | /// assert!(Regex::new(r"(?-u:\w){1000}" ).is_ok()); |
177 | /// ``` |
178 | pub fn new(re: &str) -> Result<Regex, Error> { |
179 | RegexBuilder::new(re).build() |
180 | } |
181 | |
182 | /// Returns true if and only if there is a match for the regex anywhere |
183 | /// in the haystack given. |
184 | /// |
185 | /// It is recommended to use this method if all you need to do is test |
186 | /// whether a match exists, since the underlying matching engine may be |
187 | /// able to do less work. |
188 | /// |
189 | /// # Example |
190 | /// |
191 | /// Test if some haystack contains at least one word with exactly 13 |
192 | /// Unicode word characters: |
193 | /// |
194 | /// ``` |
195 | /// use regex::bytes::Regex; |
196 | /// |
197 | /// let re = Regex::new(r"\b\w{13}\b" ).unwrap(); |
198 | /// let hay = b"I categorically deny having triskaidekaphobia." ; |
199 | /// assert!(re.is_match(hay)); |
200 | /// ``` |
201 | #[inline ] |
202 | pub fn is_match(&self, haystack: &[u8]) -> bool { |
203 | self.is_match_at(haystack, 0) |
204 | } |
205 | |
206 | /// This routine searches for the first match of this regex in the |
207 | /// haystack given, and if found, returns a [`Match`]. The `Match` |
208 | /// provides access to both the byte offsets of the match and the actual |
209 | /// substring that matched. |
210 | /// |
211 | /// Note that this should only be used if you want to find the entire |
212 | /// match. If instead you just want to test the existence of a match, |
213 | /// it's potentially faster to use `Regex::is_match(hay)` instead of |
214 | /// `Regex::find(hay).is_some()`. |
215 | /// |
216 | /// # Example |
217 | /// |
218 | /// Find the first word with exactly 13 Unicode word characters: |
219 | /// |
220 | /// ``` |
221 | /// use regex::bytes::Regex; |
222 | /// |
223 | /// let re = Regex::new(r"\b\w{13}\b" ).unwrap(); |
224 | /// let hay = b"I categorically deny having triskaidekaphobia." ; |
225 | /// let mat = re.find(hay).unwrap(); |
226 | /// assert_eq!(2..15, mat.range()); |
227 | /// assert_eq!(b"categorically" , mat.as_bytes()); |
228 | /// ``` |
229 | #[inline ] |
230 | pub fn find<'h>(&self, haystack: &'h [u8]) -> Option<Match<'h>> { |
231 | self.find_at(haystack, 0) |
232 | } |
233 | |
234 | /// Returns an iterator that yields successive non-overlapping matches in |
235 | /// the given haystack. The iterator yields values of type [`Match`]. |
236 | /// |
237 | /// # Time complexity |
238 | /// |
239 | /// Note that since `find_iter` runs potentially many searches on the |
240 | /// haystack and since each search has worst case `O(m * n)` time |
241 | /// complexity, the overall worst case time complexity for iteration is |
242 | /// `O(m * n^2)`. |
243 | /// |
244 | /// # Example |
245 | /// |
246 | /// Find every word with exactly 13 Unicode word characters: |
247 | /// |
248 | /// ``` |
249 | /// use regex::bytes::Regex; |
250 | /// |
251 | /// let re = Regex::new(r"\b\w{13}\b" ).unwrap(); |
252 | /// let hay = b"Retroactively relinquishing remunerations is reprehensible." ; |
253 | /// let matches: Vec<_> = re.find_iter(hay).map(|m| m.as_bytes()).collect(); |
254 | /// assert_eq!(matches, vec![ |
255 | /// &b"Retroactively" [..], |
256 | /// &b"relinquishing" [..], |
257 | /// &b"remunerations" [..], |
258 | /// &b"reprehensible" [..], |
259 | /// ]); |
260 | /// ``` |
261 | #[inline ] |
262 | pub fn find_iter<'r, 'h>(&'r self, haystack: &'h [u8]) -> Matches<'r, 'h> { |
263 | Matches { haystack, it: self.meta.find_iter(haystack) } |
264 | } |
265 | |
266 | /// This routine searches for the first match of this regex in the haystack |
267 | /// given, and if found, returns not only the overall match but also the |
268 | /// matches of each capture group in the regex. If no match is found, then |
269 | /// `None` is returned. |
270 | /// |
271 | /// Capture group `0` always corresponds to an implicit unnamed group that |
272 | /// includes the entire match. If a match is found, this group is always |
273 | /// present. Subsequent groups may be named and are numbered, starting |
274 | /// at 1, by the order in which the opening parenthesis appears in the |
275 | /// pattern. For example, in the pattern `(?<a>.(?<b>.))(?<c>.)`, `a`, |
276 | /// `b` and `c` correspond to capture group indices `1`, `2` and `3`, |
277 | /// respectively. |
278 | /// |
279 | /// You should only use `captures` if you need access to the capture group |
280 | /// matches. Otherwise, [`Regex::find`] is generally faster for discovering |
281 | /// just the overall match. |
282 | /// |
283 | /// # Example |
284 | /// |
285 | /// Say you have some haystack with movie names and their release years, |
286 | /// like "'Citizen Kane' (1941)". It'd be nice if we could search for |
287 | /// strings looking like that, while also extracting the movie name and its |
288 | /// release year separately. The example below shows how to do that. |
289 | /// |
290 | /// ``` |
291 | /// use regex::bytes::Regex; |
292 | /// |
293 | /// let re = Regex::new(r"'([^']+)'\s+\((\d{4})\)" ).unwrap(); |
294 | /// let hay = b"Not my favorite movie: 'Citizen Kane' (1941)." ; |
295 | /// let caps = re.captures(hay).unwrap(); |
296 | /// assert_eq!(caps.get(0).unwrap().as_bytes(), b"'Citizen Kane' (1941)" ); |
297 | /// assert_eq!(caps.get(1).unwrap().as_bytes(), b"Citizen Kane" ); |
298 | /// assert_eq!(caps.get(2).unwrap().as_bytes(), b"1941" ); |
299 | /// // You can also access the groups by index using the Index notation. |
300 | /// // Note that this will panic on an invalid index. In this case, these |
301 | /// // accesses are always correct because the overall regex will only |
302 | /// // match when these capture groups match. |
303 | /// assert_eq!(&caps[0], b"'Citizen Kane' (1941)" ); |
304 | /// assert_eq!(&caps[1], b"Citizen Kane" ); |
305 | /// assert_eq!(&caps[2], b"1941" ); |
306 | /// ``` |
307 | /// |
308 | /// Note that the full match is at capture group `0`. Each subsequent |
309 | /// capture group is indexed by the order of its opening `(`. |
310 | /// |
311 | /// We can make this example a bit clearer by using *named* capture groups: |
312 | /// |
313 | /// ``` |
314 | /// use regex::bytes::Regex; |
315 | /// |
316 | /// let re = Regex::new(r"'(?<title>[^']+)'\s+\((?<year>\d{4})\)" ).unwrap(); |
317 | /// let hay = b"Not my favorite movie: 'Citizen Kane' (1941)." ; |
318 | /// let caps = re.captures(hay).unwrap(); |
319 | /// assert_eq!(caps.get(0).unwrap().as_bytes(), b"'Citizen Kane' (1941)" ); |
320 | /// assert_eq!(caps.name("title" ).unwrap().as_bytes(), b"Citizen Kane" ); |
321 | /// assert_eq!(caps.name("year" ).unwrap().as_bytes(), b"1941" ); |
322 | /// // You can also access the groups by name using the Index notation. |
323 | /// // Note that this will panic on an invalid group name. In this case, |
324 | /// // these accesses are always correct because the overall regex will |
325 | /// // only match when these capture groups match. |
326 | /// assert_eq!(&caps[0], b"'Citizen Kane' (1941)" ); |
327 | /// assert_eq!(&caps["title" ], b"Citizen Kane" ); |
328 | /// assert_eq!(&caps["year" ], b"1941" ); |
329 | /// ``` |
330 | /// |
331 | /// Here we name the capture groups, which we can access with the `name` |
332 | /// method or the `Index` notation with a `&str`. Note that the named |
333 | /// capture groups are still accessible with `get` or the `Index` notation |
334 | /// with a `usize`. |
335 | /// |
336 | /// The `0`th capture group is always unnamed, so it must always be |
337 | /// accessed with `get(0)` or `[0]`. |
338 | /// |
339 | /// Finally, one other way to to get the matched substrings is with the |
340 | /// [`Captures::extract`] API: |
341 | /// |
342 | /// ``` |
343 | /// use regex::bytes::Regex; |
344 | /// |
345 | /// let re = Regex::new(r"'([^']+)'\s+\((\d{4})\)" ).unwrap(); |
346 | /// let hay = b"Not my favorite movie: 'Citizen Kane' (1941)." ; |
347 | /// let (full, [title, year]) = re.captures(hay).unwrap().extract(); |
348 | /// assert_eq!(full, b"'Citizen Kane' (1941)" ); |
349 | /// assert_eq!(title, b"Citizen Kane" ); |
350 | /// assert_eq!(year, b"1941" ); |
351 | /// ``` |
352 | #[inline ] |
353 | pub fn captures<'h>(&self, haystack: &'h [u8]) -> Option<Captures<'h>> { |
354 | self.captures_at(haystack, 0) |
355 | } |
356 | |
357 | /// Returns an iterator that yields successive non-overlapping matches in |
358 | /// the given haystack. The iterator yields values of type [`Captures`]. |
359 | /// |
360 | /// This is the same as [`Regex::find_iter`], but instead of only providing |
361 | /// access to the overall match, each value yield includes access to the |
362 | /// matches of all capture groups in the regex. Reporting this extra match |
363 | /// data is potentially costly, so callers should only use `captures_iter` |
364 | /// over `find_iter` when they actually need access to the capture group |
365 | /// matches. |
366 | /// |
367 | /// # Time complexity |
368 | /// |
369 | /// Note that since `captures_iter` runs potentially many searches on the |
370 | /// haystack and since each search has worst case `O(m * n)` time |
371 | /// complexity, the overall worst case time complexity for iteration is |
372 | /// `O(m * n^2)`. |
373 | /// |
374 | /// # Example |
375 | /// |
376 | /// We can use this to find all movie titles and their release years in |
377 | /// some haystack, where the movie is formatted like "'Title' (xxxx)": |
378 | /// |
379 | /// ``` |
380 | /// use regex::bytes::Regex; |
381 | /// |
382 | /// let re = Regex::new(r"'([^']+)'\s+\(([0-9]{4})\)" ).unwrap(); |
383 | /// let hay = b"'Citizen Kane' (1941), 'The Wizard of Oz' (1939), 'M' (1931)." ; |
384 | /// let mut movies = vec![]; |
385 | /// for (_, [title, year]) in re.captures_iter(hay).map(|c| c.extract()) { |
386 | /// // OK because [0-9]{4} can only match valid UTF-8. |
387 | /// let year = std::str::from_utf8(year).unwrap(); |
388 | /// movies.push((title, year.parse::<i64>()?)); |
389 | /// } |
390 | /// assert_eq!(movies, vec![ |
391 | /// (&b"Citizen Kane" [..], 1941), |
392 | /// (&b"The Wizard of Oz" [..], 1939), |
393 | /// (&b"M" [..], 1931), |
394 | /// ]); |
395 | /// # Ok::<(), Box<dyn std::error::Error>>(()) |
396 | /// ``` |
397 | /// |
398 | /// Or with named groups: |
399 | /// |
400 | /// ``` |
401 | /// use regex::bytes::Regex; |
402 | /// |
403 | /// let re = Regex::new(r"'(?<title>[^']+)'\s+\((?<year>[0-9]{4})\)" ).unwrap(); |
404 | /// let hay = b"'Citizen Kane' (1941), 'The Wizard of Oz' (1939), 'M' (1931)." ; |
405 | /// let mut it = re.captures_iter(hay); |
406 | /// |
407 | /// let caps = it.next().unwrap(); |
408 | /// assert_eq!(&caps["title" ], b"Citizen Kane" ); |
409 | /// assert_eq!(&caps["year" ], b"1941" ); |
410 | /// |
411 | /// let caps = it.next().unwrap(); |
412 | /// assert_eq!(&caps["title" ], b"The Wizard of Oz" ); |
413 | /// assert_eq!(&caps["year" ], b"1939" ); |
414 | /// |
415 | /// let caps = it.next().unwrap(); |
416 | /// assert_eq!(&caps["title" ], b"M" ); |
417 | /// assert_eq!(&caps["year" ], b"1931" ); |
418 | /// ``` |
419 | #[inline ] |
420 | pub fn captures_iter<'r, 'h>( |
421 | &'r self, |
422 | haystack: &'h [u8], |
423 | ) -> CaptureMatches<'r, 'h> { |
424 | CaptureMatches { haystack, it: self.meta.captures_iter(haystack) } |
425 | } |
426 | |
427 | /// Returns an iterator of substrings of the haystack given, delimited by a |
428 | /// match of the regex. Namely, each element of the iterator corresponds to |
429 | /// a part of the haystack that *isn't* matched by the regular expression. |
430 | /// |
431 | /// # Time complexity |
432 | /// |
433 | /// Since iterators over all matches requires running potentially many |
434 | /// searches on the haystack, and since each search has worst case |
435 | /// `O(m * n)` time complexity, the overall worst case time complexity for |
436 | /// this routine is `O(m * n^2)`. |
437 | /// |
438 | /// # Example |
439 | /// |
440 | /// To split a string delimited by arbitrary amounts of spaces or tabs: |
441 | /// |
442 | /// ``` |
443 | /// use regex::bytes::Regex; |
444 | /// |
445 | /// let re = Regex::new(r"[ \t]+" ).unwrap(); |
446 | /// let hay = b"a b \t c \td e" ; |
447 | /// let fields: Vec<&[u8]> = re.split(hay).collect(); |
448 | /// assert_eq!(fields, vec![ |
449 | /// &b"a" [..], &b"b" [..], &b"c" [..], &b"d" [..], &b"e" [..], |
450 | /// ]); |
451 | /// ``` |
452 | /// |
453 | /// # Example: more cases |
454 | /// |
455 | /// Basic usage: |
456 | /// |
457 | /// ``` |
458 | /// use regex::bytes::Regex; |
459 | /// |
460 | /// let re = Regex::new(r" " ).unwrap(); |
461 | /// let hay = b"Mary had a little lamb" ; |
462 | /// let got: Vec<&[u8]> = re.split(hay).collect(); |
463 | /// assert_eq!(got, vec![ |
464 | /// &b"Mary" [..], &b"had" [..], &b"a" [..], &b"little" [..], &b"lamb" [..], |
465 | /// ]); |
466 | /// |
467 | /// let re = Regex::new(r"X" ).unwrap(); |
468 | /// let hay = b"" ; |
469 | /// let got: Vec<&[u8]> = re.split(hay).collect(); |
470 | /// assert_eq!(got, vec![&b"" [..]]); |
471 | /// |
472 | /// let re = Regex::new(r"X" ).unwrap(); |
473 | /// let hay = b"lionXXtigerXleopard" ; |
474 | /// let got: Vec<&[u8]> = re.split(hay).collect(); |
475 | /// assert_eq!(got, vec![ |
476 | /// &b"lion" [..], &b"" [..], &b"tiger" [..], &b"leopard" [..], |
477 | /// ]); |
478 | /// |
479 | /// let re = Regex::new(r"::" ).unwrap(); |
480 | /// let hay = b"lion::tiger::leopard" ; |
481 | /// let got: Vec<&[u8]> = re.split(hay).collect(); |
482 | /// assert_eq!(got, vec![&b"lion" [..], &b"tiger" [..], &b"leopard" [..]]); |
483 | /// ``` |
484 | /// |
485 | /// If a haystack contains multiple contiguous matches, you will end up |
486 | /// with empty spans yielded by the iterator: |
487 | /// |
488 | /// ``` |
489 | /// use regex::bytes::Regex; |
490 | /// |
491 | /// let re = Regex::new(r"X" ).unwrap(); |
492 | /// let hay = b"XXXXaXXbXc" ; |
493 | /// let got: Vec<&[u8]> = re.split(hay).collect(); |
494 | /// assert_eq!(got, vec![ |
495 | /// &b"" [..], &b"" [..], &b"" [..], &b"" [..], |
496 | /// &b"a" [..], &b"" [..], &b"b" [..], &b"c" [..], |
497 | /// ]); |
498 | /// |
499 | /// let re = Regex::new(r"/" ).unwrap(); |
500 | /// let hay = b"(///)" ; |
501 | /// let got: Vec<&[u8]> = re.split(hay).collect(); |
502 | /// assert_eq!(got, vec![&b"(" [..], &b"" [..], &b"" [..], &b")" [..]]); |
503 | /// ``` |
504 | /// |
505 | /// Separators at the start or end of a haystack are neighbored by empty |
506 | /// substring. |
507 | /// |
508 | /// ``` |
509 | /// use regex::bytes::Regex; |
510 | /// |
511 | /// let re = Regex::new(r"0" ).unwrap(); |
512 | /// let hay = b"010" ; |
513 | /// let got: Vec<&[u8]> = re.split(hay).collect(); |
514 | /// assert_eq!(got, vec![&b"" [..], &b"1" [..], &b"" [..]]); |
515 | /// ``` |
516 | /// |
517 | /// When the regex can match the empty string, it splits at every byte |
518 | /// position in the haystack. This includes between all UTF-8 code units. |
519 | /// (The top-level [`Regex::split`](crate::Regex::split) will only split |
520 | /// at valid UTF-8 boundaries.) |
521 | /// |
522 | /// ``` |
523 | /// use regex::bytes::Regex; |
524 | /// |
525 | /// let re = Regex::new(r"" ).unwrap(); |
526 | /// let hay = "☃" .as_bytes(); |
527 | /// let got: Vec<&[u8]> = re.split(hay).collect(); |
528 | /// assert_eq!(got, vec![ |
529 | /// &[][..], &[b' \xE2' ][..], &[b' \x98' ][..], &[b' \x83' ][..], &[][..], |
530 | /// ]); |
531 | /// ``` |
532 | /// |
533 | /// Contiguous separators (commonly shows up with whitespace), can lead to |
534 | /// possibly surprising behavior. For example, this code is correct: |
535 | /// |
536 | /// ``` |
537 | /// use regex::bytes::Regex; |
538 | /// |
539 | /// let re = Regex::new(r" " ).unwrap(); |
540 | /// let hay = b" a b c" ; |
541 | /// let got: Vec<&[u8]> = re.split(hay).collect(); |
542 | /// assert_eq!(got, vec![ |
543 | /// &b"" [..], &b"" [..], &b"" [..], &b"" [..], |
544 | /// &b"a" [..], &b"" [..], &b"b" [..], &b"c" [..], |
545 | /// ]); |
546 | /// ``` |
547 | /// |
548 | /// It does *not* give you `["a", "b", "c"]`. For that behavior, you'd want |
549 | /// to match contiguous space characters: |
550 | /// |
551 | /// ``` |
552 | /// use regex::bytes::Regex; |
553 | /// |
554 | /// let re = Regex::new(r" +" ).unwrap(); |
555 | /// let hay = b" a b c" ; |
556 | /// let got: Vec<&[u8]> = re.split(hay).collect(); |
557 | /// // N.B. This does still include a leading empty span because ' +' |
558 | /// // matches at the beginning of the haystack. |
559 | /// assert_eq!(got, vec![&b"" [..], &b"a" [..], &b"b" [..], &b"c" [..]]); |
560 | /// ``` |
561 | #[inline ] |
562 | pub fn split<'r, 'h>(&'r self, haystack: &'h [u8]) -> Split<'r, 'h> { |
563 | Split { haystack, it: self.meta.split(haystack) } |
564 | } |
565 | |
566 | /// Returns an iterator of at most `limit` substrings of the haystack |
567 | /// given, delimited by a match of the regex. (A `limit` of `0` will return |
568 | /// no substrings.) Namely, each element of the iterator corresponds to a |
569 | /// part of the haystack that *isn't* matched by the regular expression. |
570 | /// The remainder of the haystack that is not split will be the last |
571 | /// element in the iterator. |
572 | /// |
573 | /// # Time complexity |
574 | /// |
575 | /// Since iterators over all matches requires running potentially many |
576 | /// searches on the haystack, and since each search has worst case |
577 | /// `O(m * n)` time complexity, the overall worst case time complexity for |
578 | /// this routine is `O(m * n^2)`. |
579 | /// |
580 | /// Although note that the worst case time here has an upper bound given |
581 | /// by the `limit` parameter. |
582 | /// |
583 | /// # Example |
584 | /// |
585 | /// Get the first two words in some haystack: |
586 | /// |
587 | /// ``` |
588 | /// use regex::bytes::Regex; |
589 | /// |
590 | /// let re = Regex::new(r"\W+" ).unwrap(); |
591 | /// let hay = b"Hey! How are you?" ; |
592 | /// let fields: Vec<&[u8]> = re.splitn(hay, 3).collect(); |
593 | /// assert_eq!(fields, vec![&b"Hey" [..], &b"How" [..], &b"are you?" [..]]); |
594 | /// ``` |
595 | /// |
596 | /// # Examples: more cases |
597 | /// |
598 | /// ``` |
599 | /// use regex::bytes::Regex; |
600 | /// |
601 | /// let re = Regex::new(r" " ).unwrap(); |
602 | /// let hay = b"Mary had a little lamb" ; |
603 | /// let got: Vec<&[u8]> = re.splitn(hay, 3).collect(); |
604 | /// assert_eq!(got, vec![&b"Mary" [..], &b"had" [..], &b"a little lamb" [..]]); |
605 | /// |
606 | /// let re = Regex::new(r"X" ).unwrap(); |
607 | /// let hay = b"" ; |
608 | /// let got: Vec<&[u8]> = re.splitn(hay, 3).collect(); |
609 | /// assert_eq!(got, vec![&b"" [..]]); |
610 | /// |
611 | /// let re = Regex::new(r"X" ).unwrap(); |
612 | /// let hay = b"lionXXtigerXleopard" ; |
613 | /// let got: Vec<&[u8]> = re.splitn(hay, 3).collect(); |
614 | /// assert_eq!(got, vec![&b"lion" [..], &b"" [..], &b"tigerXleopard" [..]]); |
615 | /// |
616 | /// let re = Regex::new(r"::" ).unwrap(); |
617 | /// let hay = b"lion::tiger::leopard" ; |
618 | /// let got: Vec<&[u8]> = re.splitn(hay, 2).collect(); |
619 | /// assert_eq!(got, vec![&b"lion" [..], &b"tiger::leopard" [..]]); |
620 | /// |
621 | /// let re = Regex::new(r"X" ).unwrap(); |
622 | /// let hay = b"abcXdef" ; |
623 | /// let got: Vec<&[u8]> = re.splitn(hay, 1).collect(); |
624 | /// assert_eq!(got, vec![&b"abcXdef" [..]]); |
625 | /// |
626 | /// let re = Regex::new(r"X" ).unwrap(); |
627 | /// let hay = b"abcdef" ; |
628 | /// let got: Vec<&[u8]> = re.splitn(hay, 2).collect(); |
629 | /// assert_eq!(got, vec![&b"abcdef" [..]]); |
630 | /// |
631 | /// let re = Regex::new(r"X" ).unwrap(); |
632 | /// let hay = b"abcXdef" ; |
633 | /// let got: Vec<&[u8]> = re.splitn(hay, 0).collect(); |
634 | /// assert!(got.is_empty()); |
635 | /// ``` |
636 | #[inline ] |
637 | pub fn splitn<'r, 'h>( |
638 | &'r self, |
639 | haystack: &'h [u8], |
640 | limit: usize, |
641 | ) -> SplitN<'r, 'h> { |
642 | SplitN { haystack, it: self.meta.splitn(haystack, limit) } |
643 | } |
644 | |
645 | /// Replaces the leftmost-first match in the given haystack with the |
646 | /// replacement provided. The replacement can be a regular string (where |
647 | /// `$N` and `$name` are expanded to match capture groups) or a function |
648 | /// that takes a [`Captures`] and returns the replaced string. |
649 | /// |
650 | /// If no match is found, then the haystack is returned unchanged. In that |
651 | /// case, this implementation will likely return a `Cow::Borrowed` value |
652 | /// such that no allocation is performed. |
653 | /// |
654 | /// # Replacement string syntax |
655 | /// |
656 | /// All instances of `$ref` in the replacement string are replaced with |
657 | /// the substring corresponding to the capture group identified by `ref`. |
658 | /// |
659 | /// `ref` may be an integer corresponding to the index of the capture group |
660 | /// (counted by order of opening parenthesis where `0` is the entire match) |
661 | /// or it can be a name (consisting of letters, digits or underscores) |
662 | /// corresponding to a named capture group. |
663 | /// |
664 | /// If `ref` isn't a valid capture group (whether the name doesn't exist or |
665 | /// isn't a valid index), then it is replaced with the empty string. |
666 | /// |
667 | /// The longest possible name is used. For example, `$1a` looks up the |
668 | /// capture group named `1a` and not the capture group at index `1`. To |
669 | /// exert more precise control over the name, use braces, e.g., `${1}a`. |
670 | /// |
671 | /// To write a literal `$` use `$$`. |
672 | /// |
673 | /// # Example |
674 | /// |
675 | /// Note that this function is polymorphic with respect to the replacement. |
676 | /// In typical usage, this can just be a normal string: |
677 | /// |
678 | /// ``` |
679 | /// use regex::bytes::Regex; |
680 | /// |
681 | /// let re = Regex::new(r"[^01]+" ).unwrap(); |
682 | /// assert_eq!(re.replace(b"1078910" , b"" ), &b"1010" [..]); |
683 | /// ``` |
684 | /// |
685 | /// But anything satisfying the [`Replacer`] trait will work. For example, |
686 | /// a closure of type `|&Captures| -> String` provides direct access to the |
687 | /// captures corresponding to a match. This allows one to access capturing |
688 | /// group matches easily: |
689 | /// |
690 | /// ``` |
691 | /// use regex::bytes::{Captures, Regex}; |
692 | /// |
693 | /// let re = Regex::new(r"([^,\s]+),\s+(\S+)" ).unwrap(); |
694 | /// let result = re.replace(b"Springsteen, Bruce" , |caps: &Captures| { |
695 | /// let mut buf = vec![]; |
696 | /// buf.extend_from_slice(&caps[2]); |
697 | /// buf.push(b' ' ); |
698 | /// buf.extend_from_slice(&caps[1]); |
699 | /// buf |
700 | /// }); |
701 | /// assert_eq!(result, &b"Bruce Springsteen" [..]); |
702 | /// ``` |
703 | /// |
704 | /// But this is a bit cumbersome to use all the time. Instead, a simple |
705 | /// syntax is supported (as described above) that expands `$name` into the |
706 | /// corresponding capture group. Here's the last example, but using this |
707 | /// expansion technique with named capture groups: |
708 | /// |
709 | /// ``` |
710 | /// use regex::bytes::Regex; |
711 | /// |
712 | /// let re = Regex::new(r"(?<last>[^,\s]+),\s+(?<first>\S+)" ).unwrap(); |
713 | /// let result = re.replace(b"Springsteen, Bruce" , b"$first $last" ); |
714 | /// assert_eq!(result, &b"Bruce Springsteen" [..]); |
715 | /// ``` |
716 | /// |
717 | /// Note that using `$2` instead of `$first` or `$1` instead of `$last` |
718 | /// would produce the same result. To write a literal `$` use `$$`. |
719 | /// |
720 | /// Sometimes the replacement string requires use of curly braces to |
721 | /// delineate a capture group replacement when it is adjacent to some other |
722 | /// literal text. For example, if we wanted to join two words together with |
723 | /// an underscore: |
724 | /// |
725 | /// ``` |
726 | /// use regex::bytes::Regex; |
727 | /// |
728 | /// let re = Regex::new(r"(?<first>\w+)\s+(?<second>\w+)" ).unwrap(); |
729 | /// let result = re.replace(b"deep fried" , b"${first}_$second" ); |
730 | /// assert_eq!(result, &b"deep_fried" [..]); |
731 | /// ``` |
732 | /// |
733 | /// Without the curly braces, the capture group name `first_` would be |
734 | /// used, and since it doesn't exist, it would be replaced with the empty |
735 | /// string. |
736 | /// |
737 | /// Finally, sometimes you just want to replace a literal string with no |
738 | /// regard for capturing group expansion. This can be done by wrapping a |
739 | /// string with [`NoExpand`]: |
740 | /// |
741 | /// ``` |
742 | /// use regex::bytes::{NoExpand, Regex}; |
743 | /// |
744 | /// let re = Regex::new(r"(?<last>[^,\s]+),\s+(\S+)" ).unwrap(); |
745 | /// let result = re.replace(b"Springsteen, Bruce" , NoExpand(b"$2 $last" )); |
746 | /// assert_eq!(result, &b"$2 $last" [..]); |
747 | /// ``` |
748 | /// |
749 | /// Using `NoExpand` may also be faster, since the replacement string won't |
750 | /// need to be parsed for the `$` syntax. |
751 | #[inline ] |
752 | pub fn replace<'h, R: Replacer>( |
753 | &self, |
754 | haystack: &'h [u8], |
755 | rep: R, |
756 | ) -> Cow<'h, [u8]> { |
757 | self.replacen(haystack, 1, rep) |
758 | } |
759 | |
760 | /// Replaces all non-overlapping matches in the haystack with the |
761 | /// replacement provided. This is the same as calling `replacen` with |
762 | /// `limit` set to `0`. |
763 | /// |
764 | /// The documentation for [`Regex::replace`] goes into more detail about |
765 | /// what kinds of replacement strings are supported. |
766 | /// |
767 | /// # Time complexity |
768 | /// |
769 | /// Since iterators over all matches requires running potentially many |
770 | /// searches on the haystack, and since each search has worst case |
771 | /// `O(m * n)` time complexity, the overall worst case time complexity for |
772 | /// this routine is `O(m * n^2)`. |
773 | /// |
774 | /// # Fallibility |
775 | /// |
776 | /// If you need to write a replacement routine where any individual |
777 | /// replacement might "fail," doing so with this API isn't really feasible |
778 | /// because there's no way to stop the search process if a replacement |
779 | /// fails. Instead, if you need this functionality, you should consider |
780 | /// implementing your own replacement routine: |
781 | /// |
782 | /// ``` |
783 | /// use regex::bytes::{Captures, Regex}; |
784 | /// |
785 | /// fn replace_all<E>( |
786 | /// re: &Regex, |
787 | /// haystack: &[u8], |
788 | /// replacement: impl Fn(&Captures) -> Result<Vec<u8>, E>, |
789 | /// ) -> Result<Vec<u8>, E> { |
790 | /// let mut new = Vec::with_capacity(haystack.len()); |
791 | /// let mut last_match = 0; |
792 | /// for caps in re.captures_iter(haystack) { |
793 | /// let m = caps.get(0).unwrap(); |
794 | /// new.extend_from_slice(&haystack[last_match..m.start()]); |
795 | /// new.extend_from_slice(&replacement(&caps)?); |
796 | /// last_match = m.end(); |
797 | /// } |
798 | /// new.extend_from_slice(&haystack[last_match..]); |
799 | /// Ok(new) |
800 | /// } |
801 | /// |
802 | /// // Let's replace each word with the number of bytes in that word. |
803 | /// // But if we see a word that is "too long," we'll give up. |
804 | /// let re = Regex::new(r"\w+" ).unwrap(); |
805 | /// let replacement = |caps: &Captures| -> Result<Vec<u8>, &'static str> { |
806 | /// if caps[0].len() >= 5 { |
807 | /// return Err("word too long" ); |
808 | /// } |
809 | /// Ok(caps[0].len().to_string().into_bytes()) |
810 | /// }; |
811 | /// assert_eq!( |
812 | /// Ok(b"2 3 3 3?" .to_vec()), |
813 | /// replace_all(&re, b"hi how are you?" , &replacement), |
814 | /// ); |
815 | /// assert!(replace_all(&re, b"hi there" , &replacement).is_err()); |
816 | /// ``` |
817 | /// |
818 | /// # Example |
819 | /// |
820 | /// This example shows how to flip the order of whitespace (excluding line |
821 | /// terminators) delimited fields, and normalizes the whitespace that |
822 | /// delimits the fields: |
823 | /// |
824 | /// ``` |
825 | /// use regex::bytes::Regex; |
826 | /// |
827 | /// let re = Regex::new(r"(?m)^(\S+)[\s--\r\n]+(\S+)$" ).unwrap(); |
828 | /// let hay = b" |
829 | /// Greetings 1973 |
830 | /// Wild \t1973 |
831 | /// BornToRun \t\t\t\t1975 |
832 | /// Darkness 1978 |
833 | /// TheRiver 1980 |
834 | /// " ; |
835 | /// let new = re.replace_all(hay, b"$2 $1" ); |
836 | /// assert_eq!(new, &b" |
837 | /// 1973 Greetings |
838 | /// 1973 Wild |
839 | /// 1975 BornToRun |
840 | /// 1978 Darkness |
841 | /// 1980 TheRiver |
842 | /// " [..]); |
843 | /// ``` |
844 | #[inline ] |
845 | pub fn replace_all<'h, R: Replacer>( |
846 | &self, |
847 | haystack: &'h [u8], |
848 | rep: R, |
849 | ) -> Cow<'h, [u8]> { |
850 | self.replacen(haystack, 0, rep) |
851 | } |
852 | |
853 | /// Replaces at most `limit` non-overlapping matches in the haystack with |
854 | /// the replacement provided. If `limit` is `0`, then all non-overlapping |
855 | /// matches are replaced. That is, `Regex::replace_all(hay, rep)` is |
856 | /// equivalent to `Regex::replacen(hay, 0, rep)`. |
857 | /// |
858 | /// The documentation for [`Regex::replace`] goes into more detail about |
859 | /// what kinds of replacement strings are supported. |
860 | /// |
861 | /// # Time complexity |
862 | /// |
863 | /// Since iterators over all matches requires running potentially many |
864 | /// searches on the haystack, and since each search has worst case |
865 | /// `O(m * n)` time complexity, the overall worst case time complexity for |
866 | /// this routine is `O(m * n^2)`. |
867 | /// |
868 | /// Although note that the worst case time here has an upper bound given |
869 | /// by the `limit` parameter. |
870 | /// |
871 | /// # Fallibility |
872 | /// |
873 | /// See the corresponding section in the docs for [`Regex::replace_all`] |
874 | /// for tips on how to deal with a replacement routine that can fail. |
875 | /// |
876 | /// # Example |
877 | /// |
878 | /// This example shows how to flip the order of whitespace (excluding line |
879 | /// terminators) delimited fields, and normalizes the whitespace that |
880 | /// delimits the fields. But we only do it for the first two matches. |
881 | /// |
882 | /// ``` |
883 | /// use regex::bytes::Regex; |
884 | /// |
885 | /// let re = Regex::new(r"(?m)^(\S+)[\s--\r\n]+(\S+)$" ).unwrap(); |
886 | /// let hay = b" |
887 | /// Greetings 1973 |
888 | /// Wild \t1973 |
889 | /// BornToRun \t\t\t\t1975 |
890 | /// Darkness 1978 |
891 | /// TheRiver 1980 |
892 | /// " ; |
893 | /// let new = re.replacen(hay, 2, b"$2 $1" ); |
894 | /// assert_eq!(new, &b" |
895 | /// 1973 Greetings |
896 | /// 1973 Wild |
897 | /// BornToRun \t\t\t\t1975 |
898 | /// Darkness 1978 |
899 | /// TheRiver 1980 |
900 | /// " [..]); |
901 | /// ``` |
902 | #[inline ] |
903 | pub fn replacen<'h, R: Replacer>( |
904 | &self, |
905 | haystack: &'h [u8], |
906 | limit: usize, |
907 | mut rep: R, |
908 | ) -> Cow<'h, [u8]> { |
909 | // If we know that the replacement doesn't have any capture expansions, |
910 | // then we can use the fast path. The fast path can make a tremendous |
911 | // difference: |
912 | // |
913 | // 1) We use `find_iter` instead of `captures_iter`. Not asking for |
914 | // captures generally makes the regex engines faster. |
915 | // 2) We don't need to look up all of the capture groups and do |
916 | // replacements inside the replacement string. We just push it |
917 | // at each match and be done with it. |
918 | if let Some(rep) = rep.no_expansion() { |
919 | let mut it = self.find_iter(haystack).enumerate().peekable(); |
920 | if it.peek().is_none() { |
921 | return Cow::Borrowed(haystack); |
922 | } |
923 | let mut new = Vec::with_capacity(haystack.len()); |
924 | let mut last_match = 0; |
925 | for (i, m) in it { |
926 | new.extend_from_slice(&haystack[last_match..m.start()]); |
927 | new.extend_from_slice(&rep); |
928 | last_match = m.end(); |
929 | if limit > 0 && i >= limit - 1 { |
930 | break; |
931 | } |
932 | } |
933 | new.extend_from_slice(&haystack[last_match..]); |
934 | return Cow::Owned(new); |
935 | } |
936 | |
937 | // The slower path, which we use if the replacement needs access to |
938 | // capture groups. |
939 | let mut it = self.captures_iter(haystack).enumerate().peekable(); |
940 | if it.peek().is_none() { |
941 | return Cow::Borrowed(haystack); |
942 | } |
943 | let mut new = Vec::with_capacity(haystack.len()); |
944 | let mut last_match = 0; |
945 | for (i, cap) in it { |
946 | // unwrap on 0 is OK because captures only reports matches |
947 | let m = cap.get(0).unwrap(); |
948 | new.extend_from_slice(&haystack[last_match..m.start()]); |
949 | rep.replace_append(&cap, &mut new); |
950 | last_match = m.end(); |
951 | if limit > 0 && i >= limit - 1 { |
952 | break; |
953 | } |
954 | } |
955 | new.extend_from_slice(&haystack[last_match..]); |
956 | Cow::Owned(new) |
957 | } |
958 | } |
959 | |
960 | /// A group of advanced or "lower level" search methods. Some methods permit |
961 | /// starting the search at a position greater than `0` in the haystack. Other |
962 | /// methods permit reusing allocations, for example, when extracting the |
963 | /// matches for capture groups. |
964 | impl Regex { |
965 | /// Returns the end byte offset of the first match in the haystack given. |
966 | /// |
967 | /// This method may have the same performance characteristics as |
968 | /// `is_match`. Behaviorlly, it doesn't just report whether it match |
969 | /// occurs, but also the end offset for a match. In particular, the offset |
970 | /// returned *may be shorter* than the proper end of the leftmost-first |
971 | /// match that you would find via [`Regex::find`]. |
972 | /// |
973 | /// Note that it is not guaranteed that this routine finds the shortest or |
974 | /// "earliest" possible match. Instead, the main idea of this API is that |
975 | /// it returns the offset at the point at which the internal regex engine |
976 | /// has determined that a match has occurred. This may vary depending on |
977 | /// which internal regex engine is used, and thus, the offset itself may |
978 | /// change based on internal heuristics. |
979 | /// |
980 | /// # Example |
981 | /// |
982 | /// Typically, `a+` would match the entire first sequence of `a` in some |
983 | /// haystack, but `shortest_match` *may* give up as soon as it sees the |
984 | /// first `a`. |
985 | /// |
986 | /// ``` |
987 | /// use regex::bytes::Regex; |
988 | /// |
989 | /// let re = Regex::new(r"a+" ).unwrap(); |
990 | /// let offset = re.shortest_match(b"aaaaa" ).unwrap(); |
991 | /// assert_eq!(offset, 1); |
992 | /// ``` |
993 | #[inline ] |
994 | pub fn shortest_match(&self, haystack: &[u8]) -> Option<usize> { |
995 | self.shortest_match_at(haystack, 0) |
996 | } |
997 | |
998 | /// Returns the same as `shortest_match`, but starts the search at the |
999 | /// given offset. |
1000 | /// |
1001 | /// The significance of the starting point is that it takes the surrounding |
1002 | /// context into consideration. For example, the `\A` anchor can only match |
1003 | /// when `start == 0`. |
1004 | /// |
1005 | /// If a match is found, the offset returned is relative to the beginning |
1006 | /// of the haystack, not the beginning of the search. |
1007 | /// |
1008 | /// # Panics |
1009 | /// |
1010 | /// This panics when `start >= haystack.len() + 1`. |
1011 | /// |
1012 | /// # Example |
1013 | /// |
1014 | /// This example shows the significance of `start` by demonstrating how it |
1015 | /// can be used to permit look-around assertions in a regex to take the |
1016 | /// surrounding context into account. |
1017 | /// |
1018 | /// ``` |
1019 | /// use regex::bytes::Regex; |
1020 | /// |
1021 | /// let re = Regex::new(r"\bchew\b" ).unwrap(); |
1022 | /// let hay = b"eschew" ; |
1023 | /// // We get a match here, but it's probably not intended. |
1024 | /// assert_eq!(re.shortest_match(&hay[2..]), Some(4)); |
1025 | /// // No match because the assertions take the context into account. |
1026 | /// assert_eq!(re.shortest_match_at(hay, 2), None); |
1027 | /// ``` |
1028 | #[inline ] |
1029 | pub fn shortest_match_at( |
1030 | &self, |
1031 | haystack: &[u8], |
1032 | start: usize, |
1033 | ) -> Option<usize> { |
1034 | let input = |
1035 | Input::new(haystack).earliest(true).span(start..haystack.len()); |
1036 | self.meta.search_half(&input).map(|hm| hm.offset()) |
1037 | } |
1038 | |
1039 | /// Returns the same as [`Regex::is_match`], but starts the search at the |
1040 | /// given offset. |
1041 | /// |
1042 | /// The significance of the starting point is that it takes the surrounding |
1043 | /// context into consideration. For example, the `\A` anchor can only |
1044 | /// match when `start == 0`. |
1045 | /// |
1046 | /// # Panics |
1047 | /// |
1048 | /// This panics when `start >= haystack.len() + 1`. |
1049 | /// |
1050 | /// # Example |
1051 | /// |
1052 | /// This example shows the significance of `start` by demonstrating how it |
1053 | /// can be used to permit look-around assertions in a regex to take the |
1054 | /// surrounding context into account. |
1055 | /// |
1056 | /// ``` |
1057 | /// use regex::bytes::Regex; |
1058 | /// |
1059 | /// let re = Regex::new(r"\bchew\b" ).unwrap(); |
1060 | /// let hay = b"eschew" ; |
1061 | /// // We get a match here, but it's probably not intended. |
1062 | /// assert!(re.is_match(&hay[2..])); |
1063 | /// // No match because the assertions take the context into account. |
1064 | /// assert!(!re.is_match_at(hay, 2)); |
1065 | /// ``` |
1066 | #[inline ] |
1067 | pub fn is_match_at(&self, haystack: &[u8], start: usize) -> bool { |
1068 | self.meta.is_match(Input::new(haystack).span(start..haystack.len())) |
1069 | } |
1070 | |
1071 | /// Returns the same as [`Regex::find`], but starts the search at the given |
1072 | /// offset. |
1073 | /// |
1074 | /// The significance of the starting point is that it takes the surrounding |
1075 | /// context into consideration. For example, the `\A` anchor can only |
1076 | /// match when `start == 0`. |
1077 | /// |
1078 | /// # Panics |
1079 | /// |
1080 | /// This panics when `start >= haystack.len() + 1`. |
1081 | /// |
1082 | /// # Example |
1083 | /// |
1084 | /// This example shows the significance of `start` by demonstrating how it |
1085 | /// can be used to permit look-around assertions in a regex to take the |
1086 | /// surrounding context into account. |
1087 | /// |
1088 | /// ``` |
1089 | /// use regex::bytes::Regex; |
1090 | /// |
1091 | /// let re = Regex::new(r"\bchew\b" ).unwrap(); |
1092 | /// let hay = b"eschew" ; |
1093 | /// // We get a match here, but it's probably not intended. |
1094 | /// assert_eq!(re.find(&hay[2..]).map(|m| m.range()), Some(0..4)); |
1095 | /// // No match because the assertions take the context into account. |
1096 | /// assert_eq!(re.find_at(hay, 2), None); |
1097 | /// ``` |
1098 | #[inline ] |
1099 | pub fn find_at<'h>( |
1100 | &self, |
1101 | haystack: &'h [u8], |
1102 | start: usize, |
1103 | ) -> Option<Match<'h>> { |
1104 | let input = Input::new(haystack).span(start..haystack.len()); |
1105 | self.meta.find(input).map(|m| Match::new(haystack, m.start(), m.end())) |
1106 | } |
1107 | |
1108 | /// Returns the same as [`Regex::captures`], but starts the search at the |
1109 | /// given offset. |
1110 | /// |
1111 | /// The significance of the starting point is that it takes the surrounding |
1112 | /// context into consideration. For example, the `\A` anchor can only |
1113 | /// match when `start == 0`. |
1114 | /// |
1115 | /// # Panics |
1116 | /// |
1117 | /// This panics when `start >= haystack.len() + 1`. |
1118 | /// |
1119 | /// # Example |
1120 | /// |
1121 | /// This example shows the significance of `start` by demonstrating how it |
1122 | /// can be used to permit look-around assertions in a regex to take the |
1123 | /// surrounding context into account. |
1124 | /// |
1125 | /// ``` |
1126 | /// use regex::bytes::Regex; |
1127 | /// |
1128 | /// let re = Regex::new(r"\bchew\b" ).unwrap(); |
1129 | /// let hay = b"eschew" ; |
1130 | /// // We get a match here, but it's probably not intended. |
1131 | /// assert_eq!(&re.captures(&hay[2..]).unwrap()[0], b"chew" ); |
1132 | /// // No match because the assertions take the context into account. |
1133 | /// assert!(re.captures_at(hay, 2).is_none()); |
1134 | /// ``` |
1135 | #[inline ] |
1136 | pub fn captures_at<'h>( |
1137 | &self, |
1138 | haystack: &'h [u8], |
1139 | start: usize, |
1140 | ) -> Option<Captures<'h>> { |
1141 | let input = Input::new(haystack).span(start..haystack.len()); |
1142 | let mut caps = self.meta.create_captures(); |
1143 | self.meta.captures(input, &mut caps); |
1144 | if caps.is_match() { |
1145 | let static_captures_len = self.static_captures_len(); |
1146 | Some(Captures { haystack, caps, static_captures_len }) |
1147 | } else { |
1148 | None |
1149 | } |
1150 | } |
1151 | |
1152 | /// This is like [`Regex::captures`], but writes the byte offsets of each |
1153 | /// capture group match into the locations given. |
1154 | /// |
1155 | /// A [`CaptureLocations`] stores the same byte offsets as a [`Captures`], |
1156 | /// but does *not* store a reference to the haystack. This makes its API |
1157 | /// a bit lower level and less convenient. But in exchange, callers |
1158 | /// may allocate their own `CaptureLocations` and reuse it for multiple |
1159 | /// searches. This may be helpful if allocating a `Captures` shows up in a |
1160 | /// profile as too costly. |
1161 | /// |
1162 | /// To create a `CaptureLocations` value, use the |
1163 | /// [`Regex::capture_locations`] method. |
1164 | /// |
1165 | /// This also returns the overall match if one was found. When a match is |
1166 | /// found, its offsets are also always stored in `locs` at index `0`. |
1167 | /// |
1168 | /// # Example |
1169 | /// |
1170 | /// ``` |
1171 | /// use regex::bytes::Regex; |
1172 | /// |
1173 | /// let re = Regex::new(r"^([a-z]+)=(\S*)$" ).unwrap(); |
1174 | /// let mut locs = re.capture_locations(); |
1175 | /// assert!(re.captures_read(&mut locs, b"id=foo123" ).is_some()); |
1176 | /// assert_eq!(Some((0, 9)), locs.get(0)); |
1177 | /// assert_eq!(Some((0, 2)), locs.get(1)); |
1178 | /// assert_eq!(Some((3, 9)), locs.get(2)); |
1179 | /// ``` |
1180 | #[inline ] |
1181 | pub fn captures_read<'h>( |
1182 | &self, |
1183 | locs: &mut CaptureLocations, |
1184 | haystack: &'h [u8], |
1185 | ) -> Option<Match<'h>> { |
1186 | self.captures_read_at(locs, haystack, 0) |
1187 | } |
1188 | |
1189 | /// Returns the same as [`Regex::captures_read`], but starts the search at |
1190 | /// the given offset. |
1191 | /// |
1192 | /// The significance of the starting point is that it takes the surrounding |
1193 | /// context into consideration. For example, the `\A` anchor can only |
1194 | /// match when `start == 0`. |
1195 | /// |
1196 | /// # Panics |
1197 | /// |
1198 | /// This panics when `start >= haystack.len() + 1`. |
1199 | /// |
1200 | /// # Example |
1201 | /// |
1202 | /// This example shows the significance of `start` by demonstrating how it |
1203 | /// can be used to permit look-around assertions in a regex to take the |
1204 | /// surrounding context into account. |
1205 | /// |
1206 | /// ``` |
1207 | /// use regex::bytes::Regex; |
1208 | /// |
1209 | /// let re = Regex::new(r"\bchew\b" ).unwrap(); |
1210 | /// let hay = b"eschew" ; |
1211 | /// let mut locs = re.capture_locations(); |
1212 | /// // We get a match here, but it's probably not intended. |
1213 | /// assert!(re.captures_read(&mut locs, &hay[2..]).is_some()); |
1214 | /// // No match because the assertions take the context into account. |
1215 | /// assert!(re.captures_read_at(&mut locs, hay, 2).is_none()); |
1216 | /// ``` |
1217 | #[inline ] |
1218 | pub fn captures_read_at<'h>( |
1219 | &self, |
1220 | locs: &mut CaptureLocations, |
1221 | haystack: &'h [u8], |
1222 | start: usize, |
1223 | ) -> Option<Match<'h>> { |
1224 | let input = Input::new(haystack).span(start..haystack.len()); |
1225 | self.meta.search_captures(&input, &mut locs.0); |
1226 | locs.0.get_match().map(|m| Match::new(haystack, m.start(), m.end())) |
1227 | } |
1228 | |
1229 | /// An undocumented alias for `captures_read_at`. |
1230 | /// |
1231 | /// The `regex-capi` crate previously used this routine, so to avoid |
1232 | /// breaking that crate, we continue to provide the name as an undocumented |
1233 | /// alias. |
1234 | #[doc (hidden)] |
1235 | #[inline ] |
1236 | pub fn read_captures_at<'h>( |
1237 | &self, |
1238 | locs: &mut CaptureLocations, |
1239 | haystack: &'h [u8], |
1240 | start: usize, |
1241 | ) -> Option<Match<'h>> { |
1242 | self.captures_read_at(locs, haystack, start) |
1243 | } |
1244 | } |
1245 | |
1246 | /// Auxiliary methods. |
1247 | impl Regex { |
1248 | /// Returns the original string of this regex. |
1249 | /// |
1250 | /// # Example |
1251 | /// |
1252 | /// ``` |
1253 | /// use regex::bytes::Regex; |
1254 | /// |
1255 | /// let re = Regex::new(r"foo\w+bar" ).unwrap(); |
1256 | /// assert_eq!(re.as_str(), r"foo\w+bar" ); |
1257 | /// ``` |
1258 | #[inline ] |
1259 | pub fn as_str(&self) -> &str { |
1260 | &self.pattern |
1261 | } |
1262 | |
1263 | /// Returns an iterator over the capture names in this regex. |
1264 | /// |
1265 | /// The iterator returned yields elements of type `Option<&str>`. That is, |
1266 | /// the iterator yields values for all capture groups, even ones that are |
1267 | /// unnamed. The order of the groups corresponds to the order of the group's |
1268 | /// corresponding opening parenthesis. |
1269 | /// |
1270 | /// The first element of the iterator always yields the group corresponding |
1271 | /// to the overall match, and this group is always unnamed. Therefore, the |
1272 | /// iterator always yields at least one group. |
1273 | /// |
1274 | /// # Example |
1275 | /// |
1276 | /// This shows basic usage with a mix of named and unnamed capture groups: |
1277 | /// |
1278 | /// ``` |
1279 | /// use regex::bytes::Regex; |
1280 | /// |
1281 | /// let re = Regex::new(r"(?<a>.(?<b>.))(.)(?:.)(?<c>.)" ).unwrap(); |
1282 | /// let mut names = re.capture_names(); |
1283 | /// assert_eq!(names.next(), Some(None)); |
1284 | /// assert_eq!(names.next(), Some(Some("a" ))); |
1285 | /// assert_eq!(names.next(), Some(Some("b" ))); |
1286 | /// assert_eq!(names.next(), Some(None)); |
1287 | /// // the '(?:.)' group is non-capturing and so doesn't appear here! |
1288 | /// assert_eq!(names.next(), Some(Some("c" ))); |
1289 | /// assert_eq!(names.next(), None); |
1290 | /// ``` |
1291 | /// |
1292 | /// The iterator always yields at least one element, even for regexes with |
1293 | /// no capture groups and even for regexes that can never match: |
1294 | /// |
1295 | /// ``` |
1296 | /// use regex::bytes::Regex; |
1297 | /// |
1298 | /// let re = Regex::new(r"" ).unwrap(); |
1299 | /// let mut names = re.capture_names(); |
1300 | /// assert_eq!(names.next(), Some(None)); |
1301 | /// assert_eq!(names.next(), None); |
1302 | /// |
1303 | /// let re = Regex::new(r"[a&&b]" ).unwrap(); |
1304 | /// let mut names = re.capture_names(); |
1305 | /// assert_eq!(names.next(), Some(None)); |
1306 | /// assert_eq!(names.next(), None); |
1307 | /// ``` |
1308 | #[inline ] |
1309 | pub fn capture_names(&self) -> CaptureNames<'_> { |
1310 | CaptureNames(self.meta.group_info().pattern_names(PatternID::ZERO)) |
1311 | } |
1312 | |
1313 | /// Returns the number of captures groups in this regex. |
1314 | /// |
1315 | /// This includes all named and unnamed groups, including the implicit |
1316 | /// unnamed group that is always present and corresponds to the entire |
1317 | /// match. |
1318 | /// |
1319 | /// Since the implicit unnamed group is always included in this length, the |
1320 | /// length returned is guaranteed to be greater than zero. |
1321 | /// |
1322 | /// # Example |
1323 | /// |
1324 | /// ``` |
1325 | /// use regex::bytes::Regex; |
1326 | /// |
1327 | /// let re = Regex::new(r"foo" ).unwrap(); |
1328 | /// assert_eq!(1, re.captures_len()); |
1329 | /// |
1330 | /// let re = Regex::new(r"(foo)" ).unwrap(); |
1331 | /// assert_eq!(2, re.captures_len()); |
1332 | /// |
1333 | /// let re = Regex::new(r"(?<a>.(?<b>.))(.)(?:.)(?<c>.)" ).unwrap(); |
1334 | /// assert_eq!(5, re.captures_len()); |
1335 | /// |
1336 | /// let re = Regex::new(r"[a&&b]" ).unwrap(); |
1337 | /// assert_eq!(1, re.captures_len()); |
1338 | /// ``` |
1339 | #[inline ] |
1340 | pub fn captures_len(&self) -> usize { |
1341 | self.meta.group_info().group_len(PatternID::ZERO) |
1342 | } |
1343 | |
1344 | /// Returns the total number of capturing groups that appear in every |
1345 | /// possible match. |
1346 | /// |
1347 | /// If the number of capture groups can vary depending on the match, then |
1348 | /// this returns `None`. That is, a value is only returned when the number |
1349 | /// of matching groups is invariant or "static." |
1350 | /// |
1351 | /// Note that like [`Regex::captures_len`], this **does** include the |
1352 | /// implicit capturing group corresponding to the entire match. Therefore, |
1353 | /// when a non-None value is returned, it is guaranteed to be at least `1`. |
1354 | /// Stated differently, a return value of `Some(0)` is impossible. |
1355 | /// |
1356 | /// # Example |
1357 | /// |
1358 | /// This shows a few cases where a static number of capture groups is |
1359 | /// available and a few cases where it is not. |
1360 | /// |
1361 | /// ``` |
1362 | /// use regex::bytes::Regex; |
1363 | /// |
1364 | /// let len = |pattern| { |
1365 | /// Regex::new(pattern).map(|re| re.static_captures_len()) |
1366 | /// }; |
1367 | /// |
1368 | /// assert_eq!(Some(1), len("a" )?); |
1369 | /// assert_eq!(Some(2), len("(a)" )?); |
1370 | /// assert_eq!(Some(2), len("(a)|(b)" )?); |
1371 | /// assert_eq!(Some(3), len("(a)(b)|(c)(d)" )?); |
1372 | /// assert_eq!(None, len("(a)|b" )?); |
1373 | /// assert_eq!(None, len("a|(b)" )?); |
1374 | /// assert_eq!(None, len("(b)*" )?); |
1375 | /// assert_eq!(Some(2), len("(b)+" )?); |
1376 | /// |
1377 | /// # Ok::<(), Box<dyn std::error::Error>>(()) |
1378 | /// ``` |
1379 | #[inline ] |
1380 | pub fn static_captures_len(&self) -> Option<usize> { |
1381 | self.meta.static_captures_len() |
1382 | } |
1383 | |
1384 | /// Returns a fresh allocated set of capture locations that can |
1385 | /// be reused in multiple calls to [`Regex::captures_read`] or |
1386 | /// [`Regex::captures_read_at`]. |
1387 | /// |
1388 | /// # Example |
1389 | /// |
1390 | /// ``` |
1391 | /// use regex::bytes::Regex; |
1392 | /// |
1393 | /// let re = Regex::new(r"(.)(.)(\w+)" ).unwrap(); |
1394 | /// let mut locs = re.capture_locations(); |
1395 | /// assert!(re.captures_read(&mut locs, b"Padron" ).is_some()); |
1396 | /// assert_eq!(locs.get(0), Some((0, 6))); |
1397 | /// assert_eq!(locs.get(1), Some((0, 1))); |
1398 | /// assert_eq!(locs.get(2), Some((1, 2))); |
1399 | /// assert_eq!(locs.get(3), Some((2, 6))); |
1400 | /// ``` |
1401 | #[inline ] |
1402 | pub fn capture_locations(&self) -> CaptureLocations { |
1403 | CaptureLocations(self.meta.create_captures()) |
1404 | } |
1405 | |
1406 | /// An alias for `capture_locations` to preserve backward compatibility. |
1407 | /// |
1408 | /// The `regex-capi` crate uses this method, so to avoid breaking that |
1409 | /// crate, we continue to export it as an undocumented API. |
1410 | #[doc (hidden)] |
1411 | #[inline ] |
1412 | pub fn locations(&self) -> CaptureLocations { |
1413 | self.capture_locations() |
1414 | } |
1415 | } |
1416 | |
1417 | /// Represents a single match of a regex in a haystack. |
1418 | /// |
1419 | /// A `Match` contains both the start and end byte offsets of the match and the |
1420 | /// actual substring corresponding to the range of those byte offsets. It is |
1421 | /// guaranteed that `start <= end`. When `start == end`, the match is empty. |
1422 | /// |
1423 | /// Unlike the top-level `Match` type, this `Match` type is produced by APIs |
1424 | /// that search `&[u8]` haystacks. This means that the offsets in a `Match` can |
1425 | /// point to anywhere in the haystack, including in a place that splits the |
1426 | /// UTF-8 encoding of a Unicode scalar value. |
1427 | /// |
1428 | /// The lifetime parameter `'h` refers to the lifetime of the matched of the |
1429 | /// haystack that this match was produced from. |
1430 | /// |
1431 | /// # Numbering |
1432 | /// |
1433 | /// The byte offsets in a `Match` form a half-open interval. That is, the |
1434 | /// start of the range is inclusive and the end of the range is exclusive. |
1435 | /// For example, given a haystack `abcFOOxyz` and a match of `FOO`, its byte |
1436 | /// offset range starts at `3` and ends at `6`. `3` corresponds to `F` and |
1437 | /// `6` corresponds to `x`, which is one past the end of the match. This |
1438 | /// corresponds to the same kind of slicing that Rust uses. |
1439 | /// |
1440 | /// For more on why this was chosen over other schemes (aside from being |
1441 | /// consistent with how Rust the language works), see [this discussion] and |
1442 | /// [Dijkstra's note on a related topic][note]. |
1443 | /// |
1444 | /// [this discussion]: https://github.com/rust-lang/regex/discussions/866 |
1445 | /// [note]: https://www.cs.utexas.edu/users/EWD/transcriptions/EWD08xx/EWD831.html |
1446 | /// |
1447 | /// # Example |
1448 | /// |
1449 | /// This example shows the value of each of the methods on `Match` for a |
1450 | /// particular search. |
1451 | /// |
1452 | /// ``` |
1453 | /// use regex::bytes::Regex; |
1454 | /// |
1455 | /// let re = Regex::new(r"\p{Greek}+" ).unwrap(); |
1456 | /// let hay = "Greek: αβγδ" .as_bytes(); |
1457 | /// let m = re.find(hay).unwrap(); |
1458 | /// assert_eq!(7, m.start()); |
1459 | /// assert_eq!(15, m.end()); |
1460 | /// assert!(!m.is_empty()); |
1461 | /// assert_eq!(8, m.len()); |
1462 | /// assert_eq!(7..15, m.range()); |
1463 | /// assert_eq!("αβγδ" .as_bytes(), m.as_bytes()); |
1464 | /// ``` |
1465 | #[derive(Copy, Clone, Eq, PartialEq)] |
1466 | pub struct Match<'h> { |
1467 | haystack: &'h [u8], |
1468 | start: usize, |
1469 | end: usize, |
1470 | } |
1471 | |
1472 | impl<'h> Match<'h> { |
1473 | /// Returns the byte offset of the start of the match in the haystack. The |
1474 | /// start of the match corresponds to the position where the match begins |
1475 | /// and includes the first byte in the match. |
1476 | /// |
1477 | /// It is guaranteed that `Match::start() <= Match::end()`. |
1478 | /// |
1479 | /// Unlike the top-level `Match` type, the start offset may appear anywhere |
1480 | /// in the haystack. This includes between the code units of a UTF-8 |
1481 | /// encoded Unicode scalar value. |
1482 | #[inline ] |
1483 | pub fn start(&self) -> usize { |
1484 | self.start |
1485 | } |
1486 | |
1487 | /// Returns the byte offset of the end of the match in the haystack. The |
1488 | /// end of the match corresponds to the byte immediately following the last |
1489 | /// byte in the match. This means that `&slice[start..end]` works as one |
1490 | /// would expect. |
1491 | /// |
1492 | /// It is guaranteed that `Match::start() <= Match::end()`. |
1493 | /// |
1494 | /// Unlike the top-level `Match` type, the start offset may appear anywhere |
1495 | /// in the haystack. This includes between the code units of a UTF-8 |
1496 | /// encoded Unicode scalar value. |
1497 | #[inline ] |
1498 | pub fn end(&self) -> usize { |
1499 | self.end |
1500 | } |
1501 | |
1502 | /// Returns true if and only if this match has a length of zero. |
1503 | /// |
1504 | /// Note that an empty match can only occur when the regex itself can |
1505 | /// match the empty string. Here are some examples of regexes that can |
1506 | /// all match the empty string: `^`, `^$`, `\b`, `a?`, `a*`, `a{0}`, |
1507 | /// `(foo|\d+|quux)?`. |
1508 | #[inline ] |
1509 | pub fn is_empty(&self) -> bool { |
1510 | self.start == self.end |
1511 | } |
1512 | |
1513 | /// Returns the length, in bytes, of this match. |
1514 | #[inline ] |
1515 | pub fn len(&self) -> usize { |
1516 | self.end - self.start |
1517 | } |
1518 | |
1519 | /// Returns the range over the starting and ending byte offsets of the |
1520 | /// match in the haystack. |
1521 | #[inline ] |
1522 | pub fn range(&self) -> core::ops::Range<usize> { |
1523 | self.start..self.end |
1524 | } |
1525 | |
1526 | /// Returns the substring of the haystack that matched. |
1527 | #[inline ] |
1528 | pub fn as_bytes(&self) -> &'h [u8] { |
1529 | &self.haystack[self.range()] |
1530 | } |
1531 | |
1532 | /// Creates a new match from the given haystack and byte offsets. |
1533 | #[inline ] |
1534 | fn new(haystack: &'h [u8], start: usize, end: usize) -> Match<'h> { |
1535 | Match { haystack, start, end } |
1536 | } |
1537 | } |
1538 | |
1539 | impl<'h> core::fmt::Debug for Match<'h> { |
1540 | fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result { |
1541 | let mut fmt = f.debug_struct("Match" ); |
1542 | fmt.field("start" , &self.start).field("end" , &self.end); |
1543 | if let Ok(s) = core::str::from_utf8(self.as_bytes()) { |
1544 | fmt.field("bytes" , &s); |
1545 | } else { |
1546 | // FIXME: It would be nice if this could be printed as a string |
1547 | // with invalid UTF-8 replaced with hex escapes. A alloc would |
1548 | // probably okay if that makes it easier, but regex-automata does |
1549 | // (at time of writing) have internal routines that do this. So |
1550 | // maybe we should expose them. |
1551 | fmt.field("bytes" , &self.as_bytes()); |
1552 | } |
1553 | fmt.finish() |
1554 | } |
1555 | } |
1556 | |
1557 | impl<'h> From<Match<'h>> for &'h [u8] { |
1558 | fn from(m: Match<'h>) -> &'h [u8] { |
1559 | m.as_bytes() |
1560 | } |
1561 | } |
1562 | |
1563 | impl<'h> From<Match<'h>> for core::ops::Range<usize> { |
1564 | fn from(m: Match<'h>) -> core::ops::Range<usize> { |
1565 | m.range() |
1566 | } |
1567 | } |
1568 | |
1569 | /// Represents the capture groups for a single match. |
1570 | /// |
1571 | /// Capture groups refer to parts of a regex enclosed in parentheses. They can |
1572 | /// be optionally named. The purpose of capture groups is to be able to |
1573 | /// reference different parts of a match based on the original pattern. For |
1574 | /// example, say you want to match the individual letters in a 5-letter word: |
1575 | /// |
1576 | /// ```text |
1577 | /// (?<first>\w)(\w)(?:\w)\w(?<last>\w) |
1578 | /// ``` |
1579 | /// |
1580 | /// This regex has 4 capture groups: |
1581 | /// |
1582 | /// * The group at index `0` corresponds to the overall match. It is always |
1583 | /// present in every match and never has a name. |
1584 | /// * The group at index `1` with name `first` corresponding to the first |
1585 | /// letter. |
1586 | /// * The group at index `2` with no name corresponding to the second letter. |
1587 | /// * The group at index `3` with name `last` corresponding to the fifth and |
1588 | /// last letter. |
1589 | /// |
1590 | /// Notice that `(?:\w)` was not listed above as a capture group despite it |
1591 | /// being enclosed in parentheses. That's because `(?:pattern)` is a special |
1592 | /// syntax that permits grouping but *without* capturing. The reason for not |
1593 | /// treating it as a capture is that tracking and reporting capture groups |
1594 | /// requires additional state that may lead to slower searches. So using as few |
1595 | /// capture groups as possible can help performance. (Although the difference |
1596 | /// in performance of a couple of capture groups is likely immaterial.) |
1597 | /// |
1598 | /// Values with this type are created by [`Regex::captures`] or |
1599 | /// [`Regex::captures_iter`]. |
1600 | /// |
1601 | /// `'h` is the lifetime of the haystack that these captures were matched from. |
1602 | /// |
1603 | /// # Example |
1604 | /// |
1605 | /// ``` |
1606 | /// use regex::bytes::Regex; |
1607 | /// |
1608 | /// let re = Regex::new(r"(?<first>\w)(\w)(?:\w)\w(?<last>\w)" ).unwrap(); |
1609 | /// let caps = re.captures(b"toady" ).unwrap(); |
1610 | /// assert_eq!(b"toady" , &caps[0]); |
1611 | /// assert_eq!(b"t" , &caps["first" ]); |
1612 | /// assert_eq!(b"o" , &caps[2]); |
1613 | /// assert_eq!(b"y" , &caps["last" ]); |
1614 | /// ``` |
1615 | pub struct Captures<'h> { |
1616 | haystack: &'h [u8], |
1617 | caps: captures::Captures, |
1618 | static_captures_len: Option<usize>, |
1619 | } |
1620 | |
1621 | impl<'h> Captures<'h> { |
1622 | /// Returns the `Match` associated with the capture group at index `i`. If |
1623 | /// `i` does not correspond to a capture group, or if the capture group did |
1624 | /// not participate in the match, then `None` is returned. |
1625 | /// |
1626 | /// When `i == 0`, this is guaranteed to return a non-`None` value. |
1627 | /// |
1628 | /// # Examples |
1629 | /// |
1630 | /// Get the substring that matched with a default of an empty string if the |
1631 | /// group didn't participate in the match: |
1632 | /// |
1633 | /// ``` |
1634 | /// use regex::bytes::Regex; |
1635 | /// |
1636 | /// let re = Regex::new(r"[a-z]+(?:([0-9]+)|([A-Z]+))" ).unwrap(); |
1637 | /// let caps = re.captures(b"abc123" ).unwrap(); |
1638 | /// |
1639 | /// let substr1 = caps.get(1).map_or(&b"" [..], |m| m.as_bytes()); |
1640 | /// let substr2 = caps.get(2).map_or(&b"" [..], |m| m.as_bytes()); |
1641 | /// assert_eq!(substr1, b"123" ); |
1642 | /// assert_eq!(substr2, b"" ); |
1643 | /// ``` |
1644 | #[inline ] |
1645 | pub fn get(&self, i: usize) -> Option<Match<'h>> { |
1646 | self.caps |
1647 | .get_group(i) |
1648 | .map(|sp| Match::new(self.haystack, sp.start, sp.end)) |
1649 | } |
1650 | |
1651 | /// Returns the `Match` associated with the capture group named `name`. If |
1652 | /// `name` isn't a valid capture group or it refers to a group that didn't |
1653 | /// match, then `None` is returned. |
1654 | /// |
1655 | /// Note that unlike `caps["name"]`, this returns a `Match` whose lifetime |
1656 | /// matches the lifetime of the haystack in this `Captures` value. |
1657 | /// Conversely, the substring returned by `caps["name"]` has a lifetime |
1658 | /// of the `Captures` value, which is likely shorter than the lifetime of |
1659 | /// the haystack. In some cases, it may be necessary to use this method to |
1660 | /// access the matching substring instead of the `caps["name"]` notation. |
1661 | /// |
1662 | /// # Examples |
1663 | /// |
1664 | /// Get the substring that matched with a default of an empty string if the |
1665 | /// group didn't participate in the match: |
1666 | /// |
1667 | /// ``` |
1668 | /// use regex::bytes::Regex; |
1669 | /// |
1670 | /// let re = Regex::new( |
1671 | /// r"[a-z]+(?:(?<numbers>[0-9]+)|(?<letters>[A-Z]+))" , |
1672 | /// ).unwrap(); |
1673 | /// let caps = re.captures(b"abc123" ).unwrap(); |
1674 | /// |
1675 | /// let numbers = caps.name("numbers" ).map_or(&b"" [..], |m| m.as_bytes()); |
1676 | /// let letters = caps.name("letters" ).map_or(&b"" [..], |m| m.as_bytes()); |
1677 | /// assert_eq!(numbers, b"123" ); |
1678 | /// assert_eq!(letters, b"" ); |
1679 | /// ``` |
1680 | #[inline ] |
1681 | pub fn name(&self, name: &str) -> Option<Match<'h>> { |
1682 | self.caps |
1683 | .get_group_by_name(name) |
1684 | .map(|sp| Match::new(self.haystack, sp.start, sp.end)) |
1685 | } |
1686 | |
1687 | /// This is a convenience routine for extracting the substrings |
1688 | /// corresponding to matching capture groups. |
1689 | /// |
1690 | /// This returns a tuple where the first element corresponds to the full |
1691 | /// substring of the haystack that matched the regex. The second element is |
1692 | /// an array of substrings, with each corresponding to the to the substring |
1693 | /// that matched for a particular capture group. |
1694 | /// |
1695 | /// # Panics |
1696 | /// |
1697 | /// This panics if the number of possible matching groups in this |
1698 | /// `Captures` value is not fixed to `N` in all circumstances. |
1699 | /// More precisely, this routine only works when `N` is equivalent to |
1700 | /// [`Regex::static_captures_len`]. |
1701 | /// |
1702 | /// Stated more plainly, if the number of matching capture groups in a |
1703 | /// regex can vary from match to match, then this function always panics. |
1704 | /// |
1705 | /// For example, `(a)(b)|(c)` could produce two matching capture groups |
1706 | /// or one matching capture group for any given match. Therefore, one |
1707 | /// cannot use `extract` with such a pattern. |
1708 | /// |
1709 | /// But a pattern like `(a)(b)|(c)(d)` can be used with `extract` because |
1710 | /// the number of capture groups in every match is always equivalent, |
1711 | /// even if the capture _indices_ in each match are not. |
1712 | /// |
1713 | /// # Example |
1714 | /// |
1715 | /// ``` |
1716 | /// use regex::bytes::Regex; |
1717 | /// |
1718 | /// let re = Regex::new(r"([0-9]{4})-([0-9]{2})-([0-9]{2})" ).unwrap(); |
1719 | /// let hay = b"On 2010-03-14, I became a Tenneessee lamb." ; |
1720 | /// let Some((full, [year, month, day])) = |
1721 | /// re.captures(hay).map(|caps| caps.extract()) else { return }; |
1722 | /// assert_eq!(b"2010-03-14" , full); |
1723 | /// assert_eq!(b"2010" , year); |
1724 | /// assert_eq!(b"03" , month); |
1725 | /// assert_eq!(b"14" , day); |
1726 | /// ``` |
1727 | /// |
1728 | /// # Example: iteration |
1729 | /// |
1730 | /// This example shows how to use this method when iterating over all |
1731 | /// `Captures` matches in a haystack. |
1732 | /// |
1733 | /// ``` |
1734 | /// use regex::bytes::Regex; |
1735 | /// |
1736 | /// let re = Regex::new(r"([0-9]{4})-([0-9]{2})-([0-9]{2})" ).unwrap(); |
1737 | /// let hay = b"1973-01-05, 1975-08-25 and 1980-10-18" ; |
1738 | /// |
1739 | /// let mut dates: Vec<(&[u8], &[u8], &[u8])> = vec![]; |
1740 | /// for (_, [y, m, d]) in re.captures_iter(hay).map(|c| c.extract()) { |
1741 | /// dates.push((y, m, d)); |
1742 | /// } |
1743 | /// assert_eq!(dates, vec![ |
1744 | /// (&b"1973" [..], &b"01" [..], &b"05" [..]), |
1745 | /// (&b"1975" [..], &b"08" [..], &b"25" [..]), |
1746 | /// (&b"1980" [..], &b"10" [..], &b"18" [..]), |
1747 | /// ]); |
1748 | /// ``` |
1749 | /// |
1750 | /// # Example: parsing different formats |
1751 | /// |
1752 | /// This API is particularly useful when you need to extract a particular |
1753 | /// value that might occur in a different format. Consider, for example, |
1754 | /// an identifier that might be in double quotes or single quotes: |
1755 | /// |
1756 | /// ``` |
1757 | /// use regex::bytes::Regex; |
1758 | /// |
1759 | /// let re = Regex::new(r#"id:(?:"([^"]+)"|'([^']+)')"# ).unwrap(); |
1760 | /// let hay = br#"The first is id:"foo" and the second is id:'bar'."# ; |
1761 | /// let mut ids = vec![]; |
1762 | /// for (_, [id]) in re.captures_iter(hay).map(|c| c.extract()) { |
1763 | /// ids.push(id); |
1764 | /// } |
1765 | /// assert_eq!(ids, vec![b"foo" , b"bar" ]); |
1766 | /// ``` |
1767 | pub fn extract<const N: usize>(&self) -> (&'h [u8], [&'h [u8]; N]) { |
1768 | let len = self |
1769 | .static_captures_len |
1770 | .expect("number of capture groups can vary in a match" ) |
1771 | .checked_sub(1) |
1772 | .expect("number of groups is always greater than zero" ); |
1773 | assert_eq!(N, len, "asked for {} groups, but must ask for {}" , N, len); |
1774 | // The regex-automata variant of extract is a bit more permissive. |
1775 | // It doesn't require the number of matching capturing groups to be |
1776 | // static, and you can even request fewer groups than what's there. So |
1777 | // this is guaranteed to never panic because we've asserted above that |
1778 | // the user has requested precisely the number of groups that must be |
1779 | // present in any match for this regex. |
1780 | self.caps.extract_bytes(self.haystack) |
1781 | } |
1782 | |
1783 | /// Expands all instances of `$ref` in `replacement` to the corresponding |
1784 | /// capture group, and writes them to the `dst` buffer given. A `ref` can |
1785 | /// be a capture group index or a name. If `ref` doesn't refer to a capture |
1786 | /// group that participated in the match, then it is replaced with the |
1787 | /// empty string. |
1788 | /// |
1789 | /// # Format |
1790 | /// |
1791 | /// The format of the replacement string supports two different kinds of |
1792 | /// capture references: unbraced and braced. |
1793 | /// |
1794 | /// For the unbraced format, the format supported is `$ref` where `name` |
1795 | /// can be any character in the class `[0-9A-Za-z_]`. `ref` is always |
1796 | /// the longest possible parse. So for example, `$1a` corresponds to the |
1797 | /// capture group named `1a` and not the capture group at index `1`. If |
1798 | /// `ref` matches `^[0-9]+$`, then it is treated as a capture group index |
1799 | /// itself and not a name. |
1800 | /// |
1801 | /// For the braced format, the format supported is `${ref}` where `ref` can |
1802 | /// be any sequence of bytes except for `}`. If no closing brace occurs, |
1803 | /// then it is not considered a capture reference. As with the unbraced |
1804 | /// format, if `ref` matches `^[0-9]+$`, then it is treated as a capture |
1805 | /// group index and not a name. |
1806 | /// |
1807 | /// The braced format is useful for exerting precise control over the name |
1808 | /// of the capture reference. For example, `${1}a` corresponds to the |
1809 | /// capture group reference `1` followed by the letter `a`, where as `$1a` |
1810 | /// (as mentioned above) corresponds to the capture group reference `1a`. |
1811 | /// The braced format is also useful for expressing capture group names |
1812 | /// that use characters not supported by the unbraced format. For example, |
1813 | /// `${foo[bar].baz}` refers to the capture group named `foo[bar].baz`. |
1814 | /// |
1815 | /// If a capture group reference is found and it does not refer to a valid |
1816 | /// capture group, then it will be replaced with the empty string. |
1817 | /// |
1818 | /// To write a literal `$`, use `$$`. |
1819 | /// |
1820 | /// # Example |
1821 | /// |
1822 | /// ``` |
1823 | /// use regex::bytes::Regex; |
1824 | /// |
1825 | /// let re = Regex::new( |
1826 | /// r"(?<day>[0-9]{2})-(?<month>[0-9]{2})-(?<year>[0-9]{4})" , |
1827 | /// ).unwrap(); |
1828 | /// let hay = b"On 14-03-2010, I became a Tenneessee lamb." ; |
1829 | /// let caps = re.captures(hay).unwrap(); |
1830 | /// |
1831 | /// let mut dst = vec![]; |
1832 | /// caps.expand(b"year=$year, month=$month, day=$day" , &mut dst); |
1833 | /// assert_eq!(dst, b"year=2010, month=03, day=14" ); |
1834 | /// ``` |
1835 | #[inline ] |
1836 | pub fn expand(&self, replacement: &[u8], dst: &mut Vec<u8>) { |
1837 | self.caps.interpolate_bytes_into(self.haystack, replacement, dst); |
1838 | } |
1839 | |
1840 | /// Returns an iterator over all capture groups. This includes both |
1841 | /// matching and non-matching groups. |
1842 | /// |
1843 | /// The iterator always yields at least one matching group: the first group |
1844 | /// (at index `0`) with no name. Subsequent groups are returned in the order |
1845 | /// of their opening parenthesis in the regex. |
1846 | /// |
1847 | /// The elements yielded have type `Option<Match<'h>>`, where a non-`None` |
1848 | /// value is present if the capture group matches. |
1849 | /// |
1850 | /// # Example |
1851 | /// |
1852 | /// ``` |
1853 | /// use regex::bytes::Regex; |
1854 | /// |
1855 | /// let re = Regex::new(r"(\w)(\d)?(\w)" ).unwrap(); |
1856 | /// let caps = re.captures(b"AZ" ).unwrap(); |
1857 | /// |
1858 | /// let mut it = caps.iter(); |
1859 | /// assert_eq!(it.next().unwrap().map(|m| m.as_bytes()), Some(&b"AZ" [..])); |
1860 | /// assert_eq!(it.next().unwrap().map(|m| m.as_bytes()), Some(&b"A" [..])); |
1861 | /// assert_eq!(it.next().unwrap().map(|m| m.as_bytes()), None); |
1862 | /// assert_eq!(it.next().unwrap().map(|m| m.as_bytes()), Some(&b"Z" [..])); |
1863 | /// assert_eq!(it.next(), None); |
1864 | /// ``` |
1865 | #[inline ] |
1866 | pub fn iter<'c>(&'c self) -> SubCaptureMatches<'c, 'h> { |
1867 | SubCaptureMatches { haystack: self.haystack, it: self.caps.iter() } |
1868 | } |
1869 | |
1870 | /// Returns the total number of capture groups. This includes both |
1871 | /// matching and non-matching groups. |
1872 | /// |
1873 | /// The length returned is always equivalent to the number of elements |
1874 | /// yielded by [`Captures::iter`]. Consequently, the length is always |
1875 | /// greater than zero since every `Captures` value always includes the |
1876 | /// match for the entire regex. |
1877 | /// |
1878 | /// # Example |
1879 | /// |
1880 | /// ``` |
1881 | /// use regex::bytes::Regex; |
1882 | /// |
1883 | /// let re = Regex::new(r"(\w)(\d)?(\w)" ).unwrap(); |
1884 | /// let caps = re.captures(b"AZ" ).unwrap(); |
1885 | /// assert_eq!(caps.len(), 4); |
1886 | /// ``` |
1887 | #[inline ] |
1888 | pub fn len(&self) -> usize { |
1889 | self.caps.group_len() |
1890 | } |
1891 | } |
1892 | |
1893 | impl<'h> core::fmt::Debug for Captures<'h> { |
1894 | fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result { |
1895 | /// A little helper type to provide a nice map-like debug |
1896 | /// representation for our capturing group spans. |
1897 | /// |
1898 | /// regex-automata has something similar, but it includes the pattern |
1899 | /// ID in its debug output, which is confusing. It also doesn't include |
1900 | /// that strings that match because a regex-automata `Captures` doesn't |
1901 | /// borrow the haystack. |
1902 | struct CapturesDebugMap<'a> { |
1903 | caps: &'a Captures<'a>, |
1904 | } |
1905 | |
1906 | impl<'a> core::fmt::Debug for CapturesDebugMap<'a> { |
1907 | fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result { |
1908 | let mut map = f.debug_map(); |
1909 | let names = |
1910 | self.caps.caps.group_info().pattern_names(PatternID::ZERO); |
1911 | for (group_index, maybe_name) in names.enumerate() { |
1912 | let key = Key(group_index, maybe_name); |
1913 | match self.caps.get(group_index) { |
1914 | None => map.entry(&key, &None::<()>), |
1915 | Some(mat) => map.entry(&key, &Value(mat)), |
1916 | }; |
1917 | } |
1918 | map.finish() |
1919 | } |
1920 | } |
1921 | |
1922 | struct Key<'a>(usize, Option<&'a str>); |
1923 | |
1924 | impl<'a> core::fmt::Debug for Key<'a> { |
1925 | fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result { |
1926 | write!(f, "{}" , self.0)?; |
1927 | if let Some(name) = self.1 { |
1928 | write!(f, "/{:?}" , name)?; |
1929 | } |
1930 | Ok(()) |
1931 | } |
1932 | } |
1933 | |
1934 | struct Value<'a>(Match<'a>); |
1935 | |
1936 | impl<'a> core::fmt::Debug for Value<'a> { |
1937 | fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result { |
1938 | use regex_automata::util::escape::DebugHaystack; |
1939 | |
1940 | write!( |
1941 | f, |
1942 | "{}..{}/{:?}" , |
1943 | self.0.start(), |
1944 | self.0.end(), |
1945 | DebugHaystack(self.0.as_bytes()) |
1946 | ) |
1947 | } |
1948 | } |
1949 | |
1950 | f.debug_tuple("Captures" ) |
1951 | .field(&CapturesDebugMap { caps: self }) |
1952 | .finish() |
1953 | } |
1954 | } |
1955 | |
1956 | /// Get a matching capture group's haystack substring by index. |
1957 | /// |
1958 | /// The haystack substring returned can't outlive the `Captures` object if this |
1959 | /// method is used, because of how `Index` is defined (normally `a[i]` is part |
1960 | /// of `a` and can't outlive it). To work around this limitation, do that, use |
1961 | /// [`Captures::get`] instead. |
1962 | /// |
1963 | /// `'h` is the lifetime of the matched haystack, but the lifetime of the |
1964 | /// `&str` returned by this implementation is the lifetime of the `Captures` |
1965 | /// value itself. |
1966 | /// |
1967 | /// # Panics |
1968 | /// |
1969 | /// If there is no matching group at the given index. |
1970 | impl<'h> core::ops::Index<usize> for Captures<'h> { |
1971 | type Output = [u8]; |
1972 | |
1973 | // The lifetime is written out to make it clear that the &str returned |
1974 | // does NOT have a lifetime equivalent to 'h. |
1975 | fn index<'a>(&'a self, i: usize) -> &'a [u8] { |
1976 | self.get(i) |
1977 | .map(|m| m.as_bytes()) |
1978 | .unwrap_or_else(|| panic!("no group at index '{}'" , i)) |
1979 | } |
1980 | } |
1981 | |
1982 | /// Get a matching capture group's haystack substring by name. |
1983 | /// |
1984 | /// The haystack substring returned can't outlive the `Captures` object if this |
1985 | /// method is used, because of how `Index` is defined (normally `a[i]` is part |
1986 | /// of `a` and can't outlive it). To work around this limitation, do that, use |
1987 | /// [`Captures::get`] instead. |
1988 | /// |
1989 | /// `'h` is the lifetime of the matched haystack, but the lifetime of the |
1990 | /// `&str` returned by this implementation is the lifetime of the `Captures` |
1991 | /// value itself. |
1992 | /// |
1993 | /// `'n` is the lifetime of the group name used to index the `Captures` value. |
1994 | /// |
1995 | /// # Panics |
1996 | /// |
1997 | /// If there is no matching group at the given name. |
1998 | impl<'h, 'n> core::ops::Index<&'n str> for Captures<'h> { |
1999 | type Output = [u8]; |
2000 | |
2001 | fn index<'a>(&'a self, name: &'n str) -> &'a [u8] { |
2002 | self.name(name) |
2003 | .map(|m| m.as_bytes()) |
2004 | .unwrap_or_else(|| panic!("no group named '{}'" , name)) |
2005 | } |
2006 | } |
2007 | |
2008 | /// A low level representation of the byte offsets of each capture group. |
2009 | /// |
2010 | /// You can think of this as a lower level [`Captures`], where this type does |
2011 | /// not support named capturing groups directly and it does not borrow the |
2012 | /// haystack that these offsets were matched on. |
2013 | /// |
2014 | /// Primarily, this type is useful when using the lower level `Regex` APIs such |
2015 | /// as [`Regex::captures_read`], which permits amortizing the allocation in |
2016 | /// which capture match offsets are stored. |
2017 | /// |
2018 | /// In order to build a value of this type, you'll need to call the |
2019 | /// [`Regex::capture_locations`] method. The value returned can then be reused |
2020 | /// in subsequent searches for that regex. Using it for other regexes may |
2021 | /// result in a panic or otherwise incorrect results. |
2022 | /// |
2023 | /// # Example |
2024 | /// |
2025 | /// This example shows how to create and use `CaptureLocations` in a search. |
2026 | /// |
2027 | /// ``` |
2028 | /// use regex::bytes::Regex; |
2029 | /// |
2030 | /// let re = Regex::new(r"(?<first>\w+)\s+(?<last>\w+)" ).unwrap(); |
2031 | /// let mut locs = re.capture_locations(); |
2032 | /// let m = re.captures_read(&mut locs, b"Bruce Springsteen" ).unwrap(); |
2033 | /// assert_eq!(0..17, m.range()); |
2034 | /// assert_eq!(Some((0, 17)), locs.get(0)); |
2035 | /// assert_eq!(Some((0, 5)), locs.get(1)); |
2036 | /// assert_eq!(Some((6, 17)), locs.get(2)); |
2037 | /// |
2038 | /// // Asking for an invalid capture group always returns None. |
2039 | /// assert_eq!(None, locs.get(3)); |
2040 | /// # // literals are too big for 32-bit usize: #1041 |
2041 | /// # #[cfg (target_pointer_width = "64" )] |
2042 | /// assert_eq!(None, locs.get(34973498648)); |
2043 | /// # #[cfg (target_pointer_width = "64" )] |
2044 | /// assert_eq!(None, locs.get(9944060567225171988)); |
2045 | /// ``` |
2046 | #[derive(Clone, Debug)] |
2047 | pub struct CaptureLocations(captures::Captures); |
2048 | |
2049 | /// A type alias for `CaptureLocations` for backwards compatibility. |
2050 | /// |
2051 | /// Previously, we exported `CaptureLocations` as `Locations` in an |
2052 | /// undocumented API. To prevent breaking that code (e.g., in `regex-capi`), |
2053 | /// we continue re-exporting the same undocumented API. |
2054 | #[doc (hidden)] |
2055 | pub type Locations = CaptureLocations; |
2056 | |
2057 | impl CaptureLocations { |
2058 | /// Returns the start and end byte offsets of the capture group at index |
2059 | /// `i`. This returns `None` if `i` is not a valid capture group or if the |
2060 | /// capture group did not match. |
2061 | /// |
2062 | /// # Example |
2063 | /// |
2064 | /// ``` |
2065 | /// use regex::bytes::Regex; |
2066 | /// |
2067 | /// let re = Regex::new(r"(?<first>\w+)\s+(?<last>\w+)" ).unwrap(); |
2068 | /// let mut locs = re.capture_locations(); |
2069 | /// re.captures_read(&mut locs, b"Bruce Springsteen" ).unwrap(); |
2070 | /// assert_eq!(Some((0, 17)), locs.get(0)); |
2071 | /// assert_eq!(Some((0, 5)), locs.get(1)); |
2072 | /// assert_eq!(Some((6, 17)), locs.get(2)); |
2073 | /// ``` |
2074 | #[inline ] |
2075 | pub fn get(&self, i: usize) -> Option<(usize, usize)> { |
2076 | self.0.get_group(i).map(|sp| (sp.start, sp.end)) |
2077 | } |
2078 | |
2079 | /// Returns the total number of capture groups (even if they didn't match). |
2080 | /// That is, the length returned is unaffected by the result of a search. |
2081 | /// |
2082 | /// This is always at least `1` since every regex has at least `1` |
2083 | /// capturing group that corresponds to the entire match. |
2084 | /// |
2085 | /// # Example |
2086 | /// |
2087 | /// ``` |
2088 | /// use regex::bytes::Regex; |
2089 | /// |
2090 | /// let re = Regex::new(r"(?<first>\w+)\s+(?<last>\w+)" ).unwrap(); |
2091 | /// let mut locs = re.capture_locations(); |
2092 | /// assert_eq!(3, locs.len()); |
2093 | /// re.captures_read(&mut locs, b"Bruce Springsteen" ).unwrap(); |
2094 | /// assert_eq!(3, locs.len()); |
2095 | /// ``` |
2096 | /// |
2097 | /// Notice that the length is always at least `1`, regardless of the regex: |
2098 | /// |
2099 | /// ``` |
2100 | /// use regex::bytes::Regex; |
2101 | /// |
2102 | /// let re = Regex::new(r"" ).unwrap(); |
2103 | /// let locs = re.capture_locations(); |
2104 | /// assert_eq!(1, locs.len()); |
2105 | /// |
2106 | /// // [a&&b] is a regex that never matches anything. |
2107 | /// let re = Regex::new(r"[a&&b]" ).unwrap(); |
2108 | /// let locs = re.capture_locations(); |
2109 | /// assert_eq!(1, locs.len()); |
2110 | /// ``` |
2111 | #[inline ] |
2112 | pub fn len(&self) -> usize { |
2113 | // self.0.group_len() returns 0 if the underlying captures doesn't |
2114 | // represent a match, but the behavior guaranteed for this method is |
2115 | // that the length doesn't change based on a match or not. |
2116 | self.0.group_info().group_len(PatternID::ZERO) |
2117 | } |
2118 | |
2119 | /// An alias for the `get` method for backwards compatibility. |
2120 | /// |
2121 | /// Previously, we exported `get` as `pos` in an undocumented API. To |
2122 | /// prevent breaking that code (e.g., in `regex-capi`), we continue |
2123 | /// re-exporting the same undocumented API. |
2124 | #[doc (hidden)] |
2125 | #[inline ] |
2126 | pub fn pos(&self, i: usize) -> Option<(usize, usize)> { |
2127 | self.get(i) |
2128 | } |
2129 | } |
2130 | |
2131 | /// An iterator over all non-overlapping matches in a haystack. |
2132 | /// |
2133 | /// This iterator yields [`Match`] values. The iterator stops when no more |
2134 | /// matches can be found. |
2135 | /// |
2136 | /// `'r` is the lifetime of the compiled regular expression and `'h` is the |
2137 | /// lifetime of the haystack. |
2138 | /// |
2139 | /// This iterator is created by [`Regex::find_iter`]. |
2140 | /// |
2141 | /// # Time complexity |
2142 | /// |
2143 | /// Note that since an iterator runs potentially many searches on the haystack |
2144 | /// and since each search has worst case `O(m * n)` time complexity, the |
2145 | /// overall worst case time complexity for iteration is `O(m * n^2)`. |
2146 | #[derive(Debug)] |
2147 | pub struct Matches<'r, 'h> { |
2148 | haystack: &'h [u8], |
2149 | it: meta::FindMatches<'r, 'h>, |
2150 | } |
2151 | |
2152 | impl<'r, 'h> Iterator for Matches<'r, 'h> { |
2153 | type Item = Match<'h>; |
2154 | |
2155 | #[inline ] |
2156 | fn next(&mut self) -> Option<Match<'h>> { |
2157 | self.it |
2158 | .next() |
2159 | .map(|sp| Match::new(self.haystack, sp.start(), sp.end())) |
2160 | } |
2161 | |
2162 | #[inline ] |
2163 | fn count(self) -> usize { |
2164 | // This can actually be up to 2x faster than calling `next()` until |
2165 | // completion, because counting matches when using a DFA only requires |
2166 | // finding the end of each match. But returning a `Match` via `next()` |
2167 | // requires the start of each match which, with a DFA, requires a |
2168 | // reverse forward scan to find it. |
2169 | self.it.count() |
2170 | } |
2171 | } |
2172 | |
2173 | impl<'r, 'h> core::iter::FusedIterator for Matches<'r, 'h> {} |
2174 | |
2175 | /// An iterator over all non-overlapping capture matches in a haystack. |
2176 | /// |
2177 | /// This iterator yields [`Captures`] values. The iterator stops when no more |
2178 | /// matches can be found. |
2179 | /// |
2180 | /// `'r` is the lifetime of the compiled regular expression and `'h` is the |
2181 | /// lifetime of the matched string. |
2182 | /// |
2183 | /// This iterator is created by [`Regex::captures_iter`]. |
2184 | /// |
2185 | /// # Time complexity |
2186 | /// |
2187 | /// Note that since an iterator runs potentially many searches on the haystack |
2188 | /// and since each search has worst case `O(m * n)` time complexity, the |
2189 | /// overall worst case time complexity for iteration is `O(m * n^2)`. |
2190 | #[derive(Debug)] |
2191 | pub struct CaptureMatches<'r, 'h> { |
2192 | haystack: &'h [u8], |
2193 | it: meta::CapturesMatches<'r, 'h>, |
2194 | } |
2195 | |
2196 | impl<'r, 'h> Iterator for CaptureMatches<'r, 'h> { |
2197 | type Item = Captures<'h>; |
2198 | |
2199 | #[inline ] |
2200 | fn next(&mut self) -> Option<Captures<'h>> { |
2201 | let static_captures_len = self.it.regex().static_captures_len(); |
2202 | self.it.next().map(|caps| Captures { |
2203 | haystack: self.haystack, |
2204 | caps, |
2205 | static_captures_len, |
2206 | }) |
2207 | } |
2208 | |
2209 | #[inline ] |
2210 | fn count(self) -> usize { |
2211 | // This can actually be up to 2x faster than calling `next()` until |
2212 | // completion, because counting matches when using a DFA only requires |
2213 | // finding the end of each match. But returning a `Match` via `next()` |
2214 | // requires the start of each match which, with a DFA, requires a |
2215 | // reverse forward scan to find it. |
2216 | self.it.count() |
2217 | } |
2218 | } |
2219 | |
2220 | impl<'r, 'h> core::iter::FusedIterator for CaptureMatches<'r, 'h> {} |
2221 | |
2222 | /// An iterator over all substrings delimited by a regex match. |
2223 | /// |
2224 | /// `'r` is the lifetime of the compiled regular expression and `'h` is the |
2225 | /// lifetime of the byte string being split. |
2226 | /// |
2227 | /// This iterator is created by [`Regex::split`]. |
2228 | /// |
2229 | /// # Time complexity |
2230 | /// |
2231 | /// Note that since an iterator runs potentially many searches on the haystack |
2232 | /// and since each search has worst case `O(m * n)` time complexity, the |
2233 | /// overall worst case time complexity for iteration is `O(m * n^2)`. |
2234 | #[derive(Debug)] |
2235 | pub struct Split<'r, 'h> { |
2236 | haystack: &'h [u8], |
2237 | it: meta::Split<'r, 'h>, |
2238 | } |
2239 | |
2240 | impl<'r, 'h> Iterator for Split<'r, 'h> { |
2241 | type Item = &'h [u8]; |
2242 | |
2243 | #[inline ] |
2244 | fn next(&mut self) -> Option<&'h [u8]> { |
2245 | self.it.next().map(|span| &self.haystack[span]) |
2246 | } |
2247 | } |
2248 | |
2249 | impl<'r, 'h> core::iter::FusedIterator for Split<'r, 'h> {} |
2250 | |
2251 | /// An iterator over at most `N` substrings delimited by a regex match. |
2252 | /// |
2253 | /// The last substring yielded by this iterator will be whatever remains after |
2254 | /// `N-1` splits. |
2255 | /// |
2256 | /// `'r` is the lifetime of the compiled regular expression and `'h` is the |
2257 | /// lifetime of the byte string being split. |
2258 | /// |
2259 | /// This iterator is created by [`Regex::splitn`]. |
2260 | /// |
2261 | /// # Time complexity |
2262 | /// |
2263 | /// Note that since an iterator runs potentially many searches on the haystack |
2264 | /// and since each search has worst case `O(m * n)` time complexity, the |
2265 | /// overall worst case time complexity for iteration is `O(m * n^2)`. |
2266 | /// |
2267 | /// Although note that the worst case time here has an upper bound given |
2268 | /// by the `limit` parameter to [`Regex::splitn`]. |
2269 | #[derive(Debug)] |
2270 | pub struct SplitN<'r, 'h> { |
2271 | haystack: &'h [u8], |
2272 | it: meta::SplitN<'r, 'h>, |
2273 | } |
2274 | |
2275 | impl<'r, 'h> Iterator for SplitN<'r, 'h> { |
2276 | type Item = &'h [u8]; |
2277 | |
2278 | #[inline ] |
2279 | fn next(&mut self) -> Option<&'h [u8]> { |
2280 | self.it.next().map(|span| &self.haystack[span]) |
2281 | } |
2282 | |
2283 | #[inline ] |
2284 | fn size_hint(&self) -> (usize, Option<usize>) { |
2285 | self.it.size_hint() |
2286 | } |
2287 | } |
2288 | |
2289 | impl<'r, 'h> core::iter::FusedIterator for SplitN<'r, 'h> {} |
2290 | |
2291 | /// An iterator over the names of all capture groups in a regex. |
2292 | /// |
2293 | /// This iterator yields values of type `Option<&str>` in order of the opening |
2294 | /// capture group parenthesis in the regex pattern. `None` is yielded for |
2295 | /// groups with no name. The first element always corresponds to the implicit |
2296 | /// and unnamed group for the overall match. |
2297 | /// |
2298 | /// `'r` is the lifetime of the compiled regular expression. |
2299 | /// |
2300 | /// This iterator is created by [`Regex::capture_names`]. |
2301 | #[derive(Clone, Debug)] |
2302 | pub struct CaptureNames<'r>(captures::GroupInfoPatternNames<'r>); |
2303 | |
2304 | impl<'r> Iterator for CaptureNames<'r> { |
2305 | type Item = Option<&'r str>; |
2306 | |
2307 | #[inline ] |
2308 | fn next(&mut self) -> Option<Option<&'r str>> { |
2309 | self.0.next() |
2310 | } |
2311 | |
2312 | #[inline ] |
2313 | fn size_hint(&self) -> (usize, Option<usize>) { |
2314 | self.0.size_hint() |
2315 | } |
2316 | |
2317 | #[inline ] |
2318 | fn count(self) -> usize { |
2319 | self.0.count() |
2320 | } |
2321 | } |
2322 | |
2323 | impl<'r> ExactSizeIterator for CaptureNames<'r> {} |
2324 | |
2325 | impl<'r> core::iter::FusedIterator for CaptureNames<'r> {} |
2326 | |
2327 | /// An iterator over all group matches in a [`Captures`] value. |
2328 | /// |
2329 | /// This iterator yields values of type `Option<Match<'h>>`, where `'h` is the |
2330 | /// lifetime of the haystack that the matches are for. The order of elements |
2331 | /// yielded corresponds to the order of the opening parenthesis for the group |
2332 | /// in the regex pattern. `None` is yielded for groups that did not participate |
2333 | /// in the match. |
2334 | /// |
2335 | /// The first element always corresponds to the implicit group for the overall |
2336 | /// match. Since this iterator is created by a [`Captures`] value, and a |
2337 | /// `Captures` value is only created when a match occurs, it follows that the |
2338 | /// first element yielded by this iterator is guaranteed to be non-`None`. |
2339 | /// |
2340 | /// The lifetime `'c` corresponds to the lifetime of the `Captures` value that |
2341 | /// created this iterator, and the lifetime `'h` corresponds to the originally |
2342 | /// matched haystack. |
2343 | #[derive(Clone, Debug)] |
2344 | pub struct SubCaptureMatches<'c, 'h> { |
2345 | haystack: &'h [u8], |
2346 | it: captures::CapturesPatternIter<'c>, |
2347 | } |
2348 | |
2349 | impl<'c, 'h> Iterator for SubCaptureMatches<'c, 'h> { |
2350 | type Item = Option<Match<'h>>; |
2351 | |
2352 | #[inline ] |
2353 | fn next(&mut self) -> Option<Option<Match<'h>>> { |
2354 | self.it.next().map(|group| { |
2355 | group.map(|sp| Match::new(self.haystack, sp.start, sp.end)) |
2356 | }) |
2357 | } |
2358 | |
2359 | #[inline ] |
2360 | fn size_hint(&self) -> (usize, Option<usize>) { |
2361 | self.it.size_hint() |
2362 | } |
2363 | |
2364 | #[inline ] |
2365 | fn count(self) -> usize { |
2366 | self.it.count() |
2367 | } |
2368 | } |
2369 | |
2370 | impl<'c, 'h> ExactSizeIterator for SubCaptureMatches<'c, 'h> {} |
2371 | |
2372 | impl<'c, 'h> core::iter::FusedIterator for SubCaptureMatches<'c, 'h> {} |
2373 | |
2374 | /// A trait for types that can be used to replace matches in a haystack. |
2375 | /// |
2376 | /// In general, users of this crate shouldn't need to implement this trait, |
2377 | /// since implementations are already provided for `&[u8]` along with other |
2378 | /// variants of byte string types, as well as `FnMut(&Captures) -> Vec<u8>` (or |
2379 | /// any `FnMut(&Captures) -> T` where `T: AsRef<[u8]>`). Those cover most use |
2380 | /// cases, but callers can implement this trait directly if necessary. |
2381 | /// |
2382 | /// # Example |
2383 | /// |
2384 | /// This example shows a basic implementation of the `Replacer` trait. This can |
2385 | /// be done much more simply using the replacement byte string interpolation |
2386 | /// support (e.g., `$first $last`), but this approach avoids needing to parse |
2387 | /// the replacement byte string at all. |
2388 | /// |
2389 | /// ``` |
2390 | /// use regex::bytes::{Captures, Regex, Replacer}; |
2391 | /// |
2392 | /// struct NameSwapper; |
2393 | /// |
2394 | /// impl Replacer for NameSwapper { |
2395 | /// fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>) { |
2396 | /// dst.extend_from_slice(&caps["first" ]); |
2397 | /// dst.extend_from_slice(b" " ); |
2398 | /// dst.extend_from_slice(&caps["last" ]); |
2399 | /// } |
2400 | /// } |
2401 | /// |
2402 | /// let re = Regex::new(r"(?<last>[^,\s]+),\s+(?<first>\S+)" ).unwrap(); |
2403 | /// let result = re.replace(b"Springsteen, Bruce" , NameSwapper); |
2404 | /// assert_eq!(result, &b"Bruce Springsteen" [..]); |
2405 | /// ``` |
2406 | pub trait Replacer { |
2407 | /// Appends possibly empty data to `dst` to replace the current match. |
2408 | /// |
2409 | /// The current match is represented by `caps`, which is guaranteed to have |
2410 | /// a match at capture group `0`. |
2411 | /// |
2412 | /// For example, a no-op replacement would be |
2413 | /// `dst.extend_from_slice(&caps[0])`. |
2414 | fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>); |
2415 | |
2416 | /// Return a fixed unchanging replacement byte string. |
2417 | /// |
2418 | /// When doing replacements, if access to [`Captures`] is not needed (e.g., |
2419 | /// the replacement byte string does not need `$` expansion), then it can |
2420 | /// be beneficial to avoid finding sub-captures. |
2421 | /// |
2422 | /// In general, this is called once for every call to a replacement routine |
2423 | /// such as [`Regex::replace_all`]. |
2424 | fn no_expansion<'r>(&'r mut self) -> Option<Cow<'r, [u8]>> { |
2425 | None |
2426 | } |
2427 | |
2428 | /// Returns a type that implements `Replacer`, but that borrows and wraps |
2429 | /// this `Replacer`. |
2430 | /// |
2431 | /// This is useful when you want to take a generic `Replacer` (which might |
2432 | /// not be cloneable) and use it without consuming it, so it can be used |
2433 | /// more than once. |
2434 | /// |
2435 | /// # Example |
2436 | /// |
2437 | /// ``` |
2438 | /// use regex::bytes::{Regex, Replacer}; |
2439 | /// |
2440 | /// fn replace_all_twice<R: Replacer>( |
2441 | /// re: Regex, |
2442 | /// src: &[u8], |
2443 | /// mut rep: R, |
2444 | /// ) -> Vec<u8> { |
2445 | /// let dst = re.replace_all(src, rep.by_ref()); |
2446 | /// let dst = re.replace_all(&dst, rep.by_ref()); |
2447 | /// dst.into_owned() |
2448 | /// } |
2449 | /// ``` |
2450 | fn by_ref<'r>(&'r mut self) -> ReplacerRef<'r, Self> { |
2451 | ReplacerRef(self) |
2452 | } |
2453 | } |
2454 | |
2455 | impl<'a, const N: usize> Replacer for &'a [u8; N] { |
2456 | fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>) { |
2457 | caps.expand(&**self, dst); |
2458 | } |
2459 | |
2460 | fn no_expansion(&mut self) -> Option<Cow<'_, [u8]>> { |
2461 | no_expansion(self) |
2462 | } |
2463 | } |
2464 | |
2465 | impl<const N: usize> Replacer for [u8; N] { |
2466 | fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>) { |
2467 | caps.expand(&*self, dst); |
2468 | } |
2469 | |
2470 | fn no_expansion(&mut self) -> Option<Cow<'_, [u8]>> { |
2471 | no_expansion(self) |
2472 | } |
2473 | } |
2474 | |
2475 | impl<'a> Replacer for &'a [u8] { |
2476 | fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>) { |
2477 | caps.expand(*self, dst); |
2478 | } |
2479 | |
2480 | fn no_expansion(&mut self) -> Option<Cow<'_, [u8]>> { |
2481 | no_expansion(self) |
2482 | } |
2483 | } |
2484 | |
2485 | impl<'a> Replacer for &'a Vec<u8> { |
2486 | fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>) { |
2487 | caps.expand(*self, dst); |
2488 | } |
2489 | |
2490 | fn no_expansion(&mut self) -> Option<Cow<'_, [u8]>> { |
2491 | no_expansion(self) |
2492 | } |
2493 | } |
2494 | |
2495 | impl Replacer for Vec<u8> { |
2496 | fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>) { |
2497 | caps.expand(self, dst); |
2498 | } |
2499 | |
2500 | fn no_expansion(&mut self) -> Option<Cow<'_, [u8]>> { |
2501 | no_expansion(self) |
2502 | } |
2503 | } |
2504 | |
2505 | impl<'a> Replacer for Cow<'a, [u8]> { |
2506 | fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>) { |
2507 | caps.expand(self.as_ref(), dst); |
2508 | } |
2509 | |
2510 | fn no_expansion(&mut self) -> Option<Cow<'_, [u8]>> { |
2511 | no_expansion(self) |
2512 | } |
2513 | } |
2514 | |
2515 | impl<'a> Replacer for &'a Cow<'a, [u8]> { |
2516 | fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>) { |
2517 | caps.expand(self.as_ref(), dst); |
2518 | } |
2519 | |
2520 | fn no_expansion(&mut self) -> Option<Cow<'_, [u8]>> { |
2521 | no_expansion(self) |
2522 | } |
2523 | } |
2524 | |
2525 | impl<F, T> Replacer for F |
2526 | where |
2527 | F: FnMut(&Captures<'_>) -> T, |
2528 | T: AsRef<[u8]>, |
2529 | { |
2530 | fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>) { |
2531 | dst.extend_from_slice((*self)(caps).as_ref()); |
2532 | } |
2533 | } |
2534 | |
2535 | /// A by-reference adaptor for a [`Replacer`]. |
2536 | /// |
2537 | /// This permits reusing the same `Replacer` value in multiple calls to a |
2538 | /// replacement routine like [`Regex::replace_all`]. |
2539 | /// |
2540 | /// This type is created by [`Replacer::by_ref`]. |
2541 | #[derive(Debug)] |
2542 | pub struct ReplacerRef<'a, R: ?Sized>(&'a mut R); |
2543 | |
2544 | impl<'a, R: Replacer + ?Sized + 'a> Replacer for ReplacerRef<'a, R> { |
2545 | fn replace_append(&mut self, caps: &Captures<'_>, dst: &mut Vec<u8>) { |
2546 | self.0.replace_append(caps, dst) |
2547 | } |
2548 | |
2549 | fn no_expansion<'r>(&'r mut self) -> Option<Cow<'r, [u8]>> { |
2550 | self.0.no_expansion() |
2551 | } |
2552 | } |
2553 | |
2554 | /// A helper type for forcing literal string replacement. |
2555 | /// |
2556 | /// It can be used with routines like [`Regex::replace`] and |
2557 | /// [`Regex::replace_all`] to do a literal string replacement without expanding |
2558 | /// `$name` to their corresponding capture groups. This can be both convenient |
2559 | /// (to avoid escaping `$`, for example) and faster (since capture groups |
2560 | /// don't need to be found). |
2561 | /// |
2562 | /// `'s` is the lifetime of the literal string to use. |
2563 | /// |
2564 | /// # Example |
2565 | /// |
2566 | /// ``` |
2567 | /// use regex::bytes::{NoExpand, Regex}; |
2568 | /// |
2569 | /// let re = Regex::new(r"(?<last>[^,\s]+),\s+(\S+)" ).unwrap(); |
2570 | /// let result = re.replace(b"Springsteen, Bruce" , NoExpand(b"$2 $last" )); |
2571 | /// assert_eq!(result, &b"$2 $last" [..]); |
2572 | /// ``` |
2573 | #[derive(Clone, Debug)] |
2574 | pub struct NoExpand<'s>(pub &'s [u8]); |
2575 | |
2576 | impl<'s> Replacer for NoExpand<'s> { |
2577 | fn replace_append(&mut self, _: &Captures<'_>, dst: &mut Vec<u8>) { |
2578 | dst.extend_from_slice(self.0); |
2579 | } |
2580 | |
2581 | fn no_expansion(&mut self) -> Option<Cow<'_, [u8]>> { |
2582 | Some(Cow::Borrowed(self.0)) |
2583 | } |
2584 | } |
2585 | |
2586 | /// Quickly checks the given replacement string for whether interpolation |
2587 | /// should be done on it. It returns `None` if a `$` was found anywhere in the |
2588 | /// given string, which suggests interpolation needs to be done. But if there's |
2589 | /// no `$` anywhere, then interpolation definitely does not need to be done. In |
2590 | /// that case, the given string is returned as a borrowed `Cow`. |
2591 | /// |
2592 | /// This is meant to be used to implement the `Replacer::no_expandsion` method |
2593 | /// in its various trait impls. |
2594 | fn no_expansion<T: AsRef<[u8]>>(replacement: &T) -> Option<Cow<'_, [u8]>> { |
2595 | let replacement = replacement.as_ref(); |
2596 | match crate::find_byte::find_byte(b'$' , replacement) { |
2597 | Some(_) => None, |
2598 | None => Some(Cow::Borrowed(replacement)), |
2599 | } |
2600 | } |
2601 | |