lib.rs - Codebrowser

1	/!*
2	This crate provides routines for searching strings for matches of a [regular
3	expression] (aka "regex"). The regex syntax supported by this crate is similar
4	to other regex engines, but it lacks several features that are not known how to
5	implement efficiently. This includes, but is not limited to, look-around and
6	backreferences. In exchange, all regex searches in this crate have worst case
7	`O(m n)` time complexity, where `m` is proportional to the size of the regex*
8	and `n` is proportional to the size of the string being searched.
9
10	[regular expression]: https://en.wikipedia.org/wiki/Regular_expression
11
12	If you just want API documentation, then skip to the [`Regex`] type. Otherwise,
13	here's a quick example showing one way of parsing the output of a grep-like
14	program:
15
16	```rust
17	use regex::Regex;
18
19	let re = Regex::new(r"(?m)^([^:]+):([0-9]+):(.+)$").unwrap();
20	let hay = "\
21	path/to/foo:54:Blue Harvest
22	path/to/bar:90:Something, Something, Something, Dark Side
23	path/to/baz:3:It's a Trap!
24	";
25
26	let mut results = vec![];
27	for (_, [path, lineno, line]) in re.captures_iter(hay).map(\|c\| c.extract()) {
28	results.push((path, lineno.parse::<u64>()?, line));
29	}
30	assert_eq!(results, vec![
31	("path/to/foo", `54`, "Blue Harvest"),
32	("path/to/bar", `90`, "Something, Something, Something, Dark Side"),
33	("path/to/baz", `3`, "It's a Trap!"),
34	]);
35	# Ok::<(), Box<dyn std::error::Error>>(())
36	```
37
38	# Overview
39
40	The primary type in this crate is a [`Regex`]. Its most important methods are
41	as follows:
42
43	* [`Regex::new`] compiles a regex using the default configuration. A
44	[`RegexBuilder`] permits setting a non-default configuration. (For example,
45	case insensitive matching, verbose mode and others.)
46	* [`Regex::is_match`] reports whether a match exists in a particular haystack.
47	* [`Regex::find`] reports the byte offsets of a match in a haystack, if one
48	exists. [`Regex::find_iter`] returns an iterator over all such matches.
49	* [`Regex::captures`] returns a [`Captures`], which reports both the byte
50	offsets of a match in a haystack and the byte offsets of each matching capture
51	group from the regex in the haystack.
52	[`Regex::captures_iter`] returns an iterator over all such matches.
53
54	There is also a [`RegexSet`], which permits searching for multiple regex
55	patterns simultaneously in a single search. However, it currently only reports
56	which patterns match and not* the byte offsets of a match.*
57
58	Otherwise, this top-level crate documentation is organized as follows:
59
60	* [Usage](#usage) shows how to add the `regex` crate to your Rust project.
61	* [Examples](#examples) provides a limited selection of regex search examples.
62	* [Performance](#performance) provides a brief summary of how to optimize regex
63	searching speed.
64	* [Unicode](#unicode) discusses support for non-ASCII patterns.
65	* [Syntax](#syntax) enumerates the specific regex syntax supported by this
66	crate.
67	* [Untrusted input](#untrusted-input) discusses how this crate deals with regex
68	patterns or haystacks that are untrusted.
69	* [Crate features](#crate-features) documents the Cargo features that can be
70	enabled or disabled for this crate.
71	* [Other crates](#other-crates) links to other crates in the `regex` family.
72
73	# Usage
74
75	The `regex` crate is [on crates.io](https://crates.io/crates/regex) and can be
76	used by adding `regex` to your dependencies in your project's `Cargo.toml`.
77	Or more simply, just run `cargo add regex`.
78
79	Here is a complete example that creates a new Rust project, adds a dependency
80	on `regex`, creates the source code for a regex search and then runs the
81	program.
82
83	First, create the project in a new directory:
84
85	```text
86	$ mkdir regex-example
87	$ cd regex-example
88	$ cargo init
89	```
90
91	Second, add a dependency on `regex`:
92
93	```text
94	$ cargo add regex
95	```
96
97	Third, edit `src/main.rs`. Delete what's there and replace it with this:
98
99	```
100	use regex::Regex;
101
102	fn main() {
103	let re = Regex::new(r"Hello (?<name>\w+)!").unwrap();
104	let Some(caps) = re.captures("Hello Murphy!") else {
105	println!("no match!");
106	return;
107	};
108	println!("The name is: {}", &caps["name"]);
109	}
110	```
111
112	Fourth, run it with `cargo run`:
113
114	```text
115	$ cargo run
116	Compiling memchr v2.5.0
117	Compiling regex-syntax v0.7.1
118	Compiling aho-corasick v1.0.1
119	Compiling regex v1.8.1
120	Compiling regex-example v0.1.0 (/tmp/regex-example)
121	Finished dev [unoptimized + debuginfo] target(s) in 4.22s
122	Running `target/debug/regex-example`
123	The name is: Murphy
124	```
125
126	The first time you run the program will show more output like above. But
127	subsequent runs shouldn't have to re-compile the dependencies.
128
129	# Examples
130
131	This section provides a few examples, in tutorial style, showing how to
132	search a haystack with a regex. There are more examples throughout the API
133	documentation.
134
135	Before starting though, it's worth defining a few terms:
136
137	* A regex is a Rust value whose type is `Regex`. We use `re` as a
138	variable name for a regex.
139	* A pattern is the string that is used to build a regex. We use `pat` as
140	a variable name for a pattern.
141	* A haystack is the string that is searched by a regex. We use `hay` as a
142	variable name for a haystack.
143
144	Sometimes the words "regex" and "pattern" are used interchangeably.
145
146	General use of regular expressions in this crate proceeds by compiling a
147	pattern into a regex, and then using that regex to search, split or
148	replace parts of a haystack.
149
150	### Example: find a middle initial
151
152	We'll start off with a very simple example: a regex that looks for a specific
153	name but uses a wildcard to match a middle initial. Our pattern serves as
154	something like a template that will match a particular name with any* middle*
155	initial.
156
157	```rust
158	use regex::Regex;
159
160	// We use 'unwrap()' here because it would be a bug in our program if the
161	// pattern failed to compile to a regex. Panicking in the presence of a bug
162	// is okay.
163	let re = Regex::new(r"Homer (.)\. Simpson").unwrap();
164	let hay = "Homer J. Simpson";
165	let Some(caps) = re.captures(hay) else { return };
166	assert_eq!("J", &caps[`1`]);
167	```
168
169	There are a few things worth noticing here in our first example:
170
171	* The `.` is a special pattern meta character that means "match any single
172	character except for new lines." (More precisely, in this crate, it means
173	"match any UTF-8 encoding of any Unicode scalar value other than `\n`.")
174	* We can match an actual `.` literally by escaping it, i.e., `\.`.
175	* We use Rust's [raw strings] to avoid needing to deal with escape sequences in
176	both the regex pattern syntax and in Rust's string literal syntax. If we didn't
177	use raw strings here, we would have had to use `\\.` to match a literal `.`
178	character. That is, `r"\."` and `"\\."` are equivalent patterns.
179	* We put our wildcard `.` instruction in parentheses. These parentheses have a
180	special meaning that says, "make whatever part of the haystack matches within
181	these parentheses available as a capturing group." After finding a match, we
182	access this capture group with `&caps[1]`.
183
184	[raw strings]: https://doc.rust-lang.org/stable/reference/tokens.html#raw-string-literals
185
186	Otherwise, we execute a search using `re.captures(hay)` and return from our
187	function if no match occurred. We then reference the middle initial by asking
188	for the part of the haystack that matched the capture group indexed at `1`.
189	(The capture group at index 0 is implicit and always corresponds to the entire
190	match. In this case, that's `Homer J. Simpson`.)
191
192	### Example: named capture groups
193
194	Continuing from our middle initial example above, we can tweak the pattern
195	slightly to give a name to the group that matches the middle initial:
196
197	```rust
198	use regex::Regex;
199
200	// Note that (?P<middle>.) is a different way to spell the same thing.
201	let re = Regex::new(r"Homer (?<middle>.)\. Simpson").unwrap();
202	let hay = "Homer J. Simpson";
203	let Some(caps) = re.captures(hay) else { return };
204	assert_eq!("J", &caps["middle"]);
205	```
206
207	Giving a name to a group can be useful when there are multiple groups in
208	a pattern. It makes the code referring to those groups a bit easier to
209	understand.
210
211	### Example: validating a particular date format
212
213	This examples shows how to confirm whether a haystack, in its entirety, matches
214	a particular date format:
215
216	```rust
217	use regex::Regex;
218
219	let re = Regex::new(r"^\d{4}-\d{2}-\d{2}$").unwrap();
220	assert!(re.is_match("2010-03-14"));
221	```
222
223	Notice the use of the `^` and `$` anchors. In this crate, every regex search is
224	run with an implicit `(?s:.)?` at the beginning of its pattern, which allows*
225	the regex to match anywhere in a haystack. Anchors, as above, can be used to
226	ensure that the full haystack matches a pattern.
227
228	This crate is also Unicode aware by default, which means that `\d` might match
229	more than you might expect it to. For example:
230
231	```rust
232	use regex::Regex;
233
234	let re = Regex::new(r"^\d{4}-\d{2}-\d{2}$").unwrap();
235	assert!(re.is_match("𝟚𝟘𝟙𝟘-𝟘𝟛-𝟙𝟜"));
236	```
237
238	To only match an ASCII decimal digit, all of the following are equivalent:
239
240	* `[0-9]`
241	* `(?-u:\d)`
242	* `[[:digit:]]`
243	* `[\d&&\p{ascii}]`
244
245	### Example: finding dates in a haystack
246
247	In the previous example, we showed how one might validate that a haystack,
248	in its entirety, corresponded to a particular date format. But what if we wanted
249	to extract all things that look like dates in a specific format from a haystack?
250	To do this, we can use an iterator API to find all matches (notice that we've
251	removed the anchors and switched to looking for ASCII-only digits):
252
253	```rust
254	use regex::Regex;
255
256	let re = Regex::new(r"[0-9]{4}-[0-9]{2}-[0-9]{2}").unwrap();
257	let hay = "What do 1865-04-14, 1881-07-02, 1901-09-06 and 1963-11-22 have in common?";
258	// 'm' is a 'Match', and 'as_str()' returns the matching part of the haystack.
259	let dates: Vec<&str> = re.find_iter(hay).map(\|m\| m.as_str()).collect();
260	assert_eq!(dates, vec![
261	"1865-04-14",
262	"1881-07-02",
263	"1901-09-06",
264	"1963-11-22",
265	]);
266	```
267
268	We can also iterate over [`Captures`] values instead of [`Match`] values, and
269	that in turn permits accessing each component of the date via capturing groups:
270
271	```rust
272	use regex::Regex;
273
274	let re = Regex::new(r"(?<y>[0-9]{4})-(?<m>[0-9]{2})-(?<d>[0-9]{2})").unwrap();
275	let hay = "What do 1865-04-14, 1881-07-02, 1901-09-06 and 1963-11-22 have in common?";
276	// 'm' is a 'Match', and 'as_str()' returns the matching part of the haystack.
277	let dates: Vec<(&str, &str, &str)> = re.captures_iter(hay).map(\|caps\| {
278	// The unwraps are okay because every capture group must match if the whole
279	// regex matches, and in this context, we know we have a match.
280	//
281	// Note that we use `caps.name("y").unwrap().as_str()` instead of
282	// `&caps["y"]` because the lifetime of the former is the same as the
283	// lifetime of `hay` above, but the lifetime of the latter is tied to the
284	// lifetime of `caps` due to how the `Index` trait is defined.
285	let year = caps.name("y").unwrap().as_str();
286	let month = caps.name("m").unwrap().as_str();
287	let day = caps.name("d").unwrap().as_str();
288	(year, month, day)
289	}).collect();
290	assert_eq!(dates, vec![
291	("1865", "04", "14"),
292	("1881", "07", "02"),
293	("1901", "09", "06"),
294	("1963", "11", "22"),
295	]);
296	```
297
298	### Example: simpler capture group extraction
299
300	One can use [`Captures::extract`] to make the code from the previous example a
301	bit simpler in this case:
302
303	```rust
304	use regex::Regex;
305
306	let re = Regex::new(r"([0-9]{4})-([0-9]{2})-([0-9]{2})").unwrap();
307	let hay = "What do 1865-04-14, 1881-07-02, 1901-09-06 and 1963-11-22 have in common?";
308	let dates: Vec<(&str, &str, &str)> = re.captures_iter(hay).map(\|caps\| {
309	let (_, [year, month, day]) = caps.extract();
310	(year, month, day)
311	}).collect();
312	assert_eq!(dates, vec![
313	("1865", "04", "14"),
314	("1881", "07", "02"),
315	("1901", "09", "06"),
316	("1963", "11", "22"),
317	]);
318	```
319
320	`Captures::extract` works by ensuring that the number of matching groups match
321	the number of groups requested via the `[year, month, day]` syntax. If they do,
322	then the substrings for each corresponding capture group are automatically
323	returned in an appropriately sized array. Rust's syntax for pattern matching
324	arrays does the rest.
325
326	### Example: replacement with named capture groups
327
328	Building on the previous example, perhaps we'd like to rearrange the date
329	formats. This can be done by finding each match and replacing it with
330	something different. The [`Regex::replace_all`] routine provides a convenient
331	way to do this, including by supporting references to named groups in the
332	replacement string:
333
334	```rust
335	use regex::Regex;
336
337	let re = Regex::new(r"(?<y>\d{4})-(?<m>\d{2})-(?<d>\d{2})").unwrap();
338	let before = "1973-01-05, 1975-08-25 and 1980-10-18";
339	let after = re.replace_all(before, "$m/$d/$y");
340	assert_eq!(after, "01/05/1973, 08/25/1975 and 10/18/1980");
341	```
342
343	The replace methods are actually polymorphic in the replacement, which
344	provides more flexibility than is seen here. (See the documentation for
345	[`Regex::replace`] for more details.)
346
347	### Example: verbose mode
348
349	When your regex gets complicated, you might consider using something other
350	than regex. But if you stick with regex, you can use the `x` flag to enable
351	insignificant whitespace mode or "verbose mode." In this mode, whitespace
352	is treated as insignificant and one may write comments. This may make your
353	patterns easier to comprehend.
354
355	```rust
356	use regex::Regex;
357
358	let re = Regex::new(r"(?x)
359	(?P<y>\d{4}) # the year, including all Unicode digits
360	-
361	(?P<m>\d{2}) # the month, including all Unicode digits
362	-
363	(?P<d>\d{2}) # the day, including all Unicode digits
364	").unwrap();
365
366	let before = "1973-01-05, 1975-08-25 and 1980-10-18";
367	let after = re.replace_all(before, "$m/$d/$y");
368	assert_eq!(after, "01/05/1973, 08/25/1975 and 10/18/1980");
369	```
370
371	If you wish to match against whitespace in this mode, you can still use `\s`,
372	`\n`, `\t`, etc. For escaping a single space character, you can escape it
373	directly with `\ `, use its hex character code `\x20` or temporarily disable
374	the `x` flag, e.g., `(?-x: )`.
375
376	### Example: match multiple regular expressions simultaneously
377
378	This demonstrates how to use a [`RegexSet`] to match multiple (possibly
379	overlapping) regexes in a single scan of a haystack:
380
381	```rust
382	use regex::RegexSet;
383
384	let set = RegexSet::new(&[
385	r"\w+",
386	r"\d+",
387	r"\pL+",
388	r"foo",
389	r"bar",
390	r"barfoo",
391	r"foobar",
392	]).unwrap();
393
394	// Iterate over and collect all of the matches. Each match corresponds to the
395	// ID of the matching pattern.
396	let matches: Vec<_> = set.matches("foobar").into_iter().collect();
397	assert_eq!(matches, vec![`0`, `2`, `3`, `4`, `6`]);
398
399	// You can also test whether a particular regex matched:
400	let matches = set.matches("foobar");
401	assert!(!matches.matched(`5`));
402	assert!(matches.matched(`6`));
403	```
404
405	# Performance
406
407	This section briefly discusses a few concerns regarding the speed and resource
408	usage of regexes.
409
410	### Only ask for what you need
411
412	When running a search with a regex, there are generally three different types
413	of information one can ask for:
414
415	1. Does a regex match in a haystack?
416	2. Where does a regex match in a haystack?
417	3. Where do each of the capturing groups match in a haystack?
418
419	Generally speaking, this crate could provide a function to answer only #3,
420	which would subsume #1 and #2 automatically. However, it can be significantly
421	more expensive to compute the location of capturing group matches, so it's best
422	not to do it if you don't need to.
423
424	Therefore, only ask for what you need. For example, don't use [`Regex::find`]
425	if you only need to test if a regex matches a haystack. Use [`Regex::is_match`]
426	instead.
427
428	### Unicode can impact memory usage and search speed
429
430	This crate has first class support for Unicode and it is enabled by default.
431	In many cases, the extra memory required to support it will be negligible and
432	it typically won't impact search speed. But it can in some cases.
433
434	With respect to memory usage, the impact of Unicode principally manifests
435	through the use of Unicode character classes. Unicode character classes
436	tend to be quite large. For example, `\w` by default matches around 140,000
437	distinct codepoints. This requires additional memory, and tends to slow down
438	regex compilation. While a `\w` here and there is unlikely to be noticed,
439	writing `\w{100}` will for example result in quite a large regex by default.
440	Indeed, `\w` is considerably larger than its ASCII-only version, so if your
441	requirements are satisfied by ASCII, it's probably a good idea to stick to
442	ASCII classes. The ASCII-only version of `\w` can be spelled in a number of
443	ways. All of the following are equivalent:
444
445	* `[0-9A-Za-z_]`
446	* `(?-u:\w)`
447	* `[[:word:]]`
448	* `[\w&&\p{ascii}]`
449
450	With respect to search speed, Unicode tends to be handled pretty well, even when
451	using large Unicode character classes. However, some of the faster internal
452	regex engines cannot handle a Unicode aware word boundary assertion. So if you
453	don't need Unicode-aware word boundary assertions, you might consider using
454	`(?-u:\b)` instead of `\b`, where the former uses an ASCII-only definition of
455	a word character.
456
457	### Literals might accelerate searches
458
459	This crate tends to be quite good at recognizing literals in a regex pattern
460	and using them to accelerate a search. If it is at all possible to include
461	some kind of literal in your pattern, then it might make search substantially
462	faster. For example, in the regex `\w+@\w+`, the engine will look for
463	occurrences of `@` and then try a reverse match for `\w+` to find the start
464	position.
465
466	### Avoid re-compiling regexes, especially in a loop
467
468	It is an anti-pattern to compile the same pattern in a loop since regex
469	compilation is typically expensive. (It takes anywhere from a few microseconds
470	to a few milliseconds* depending on the size of the pattern.) Not only is*
471	compilation itself expensive, but this also prevents optimizations that reuse
472	allocations internally to the regex engine.
473
474	In Rust, it can sometimes be a pain to pass regexes around if they're used from
475	inside a helper function. Instead, we recommend using crates like [`once_cell`]
476	and [`lazy_static`] to ensure that patterns are compiled exactly once.
477
478	[`once_cell`]: https://crates.io/crates/once_cell
479	[`lazy_static`]: https://crates.io/crates/lazy_static
480
481	This example shows how to use `once_cell`:
482
483	```rust
484	use {
485	once_cell::sync::Lazy,
486	regex::Regex,
487	};
488
489	fn some_helper_function(haystack: &str) -> bool {
490	static RE: Lazy<Regex> = Lazy::new(\|\| Regex::new(r"...").unwrap());
491	RE.is_match(haystack)
492	}
493
494	fn main() {
495	assert!(some_helper_function("abc"));
496	assert!(!some_helper_function("ac"));
497	}
498	```
499
500	Specifically, in this example, the regex will be compiled when it is used for
501	the first time. On subsequent uses, it will reuse the previously built `Regex`.
502	Notice how one can define the `Regex` locally to a specific function.
503
504	### Sharing a regex across threads can result in contention
505
506	While a single `Regex` can be freely used from multiple threads simultaneously,
507	there is a small synchronization cost that must be paid. Generally speaking,
508	one shouldn't expect to observe this unless the principal task in each thread
509	is searching with the regex and* most searches are on short haystacks. In this*
510	case, internal contention on shared resources can spike and increase latency,
511	which in turn may slow down each individual search.
512
513	One can work around this by cloning each `Regex` before sending it to another
514	thread. The cloned regexes will still share the same internal read-only portion
515	of its compiled state (it's reference counted), but each thread will get
516	optimized access to the mutable space that is used to run a search. In general,
517	there is no additional cost in memory to doing this. The only cost is the added
518	code complexity required to explicitly clone the regex. (If you share the same
519	`Regex` across multiple threads, each thread still gets its own mutable space,
520	but accessing that space is slower.)
521
522	# Unicode
523
524	This section discusses what kind of Unicode support this regex library has.
525	Before showing some examples, we'll summarize the relevant points:
526
527	* This crate almost fully implements "Basic Unicode Support" (Level 1) as
528	specified by the [Unicode Technical Standard #18][UTS18]. The full details
529	of what is supported are documented in [UNICODE.md] in the root of the regex
530	crate repository. There is virtually no support for "Extended Unicode Support"
531	(Level 2) from UTS#18.
532	* The top-level [`Regex`] runs searches as if* iterating over each of the*
533	codepoints in the haystack. That is, the fundamental atom of matching is a
534	single codepoint.
535	* [`bytes::Regex`], in contrast, permits disabling Unicode mode for part of all
536	of your pattern in all cases. When Unicode mode is disabled, then a search is
537	run as if* iterating over each byte in the haystack. That is, the fundamental*
538	atom of matching is a single byte. (A top-level `Regex` also permits disabling
539	Unicode and thus matching as if* it were one byte at a time, but only when*
540	doing so wouldn't permit matching invalid UTF-8.)
541	* When Unicode mode is enabled (the default), `.` will match an entire Unicode
542	scalar value, even when it is encoded using multiple bytes. When Unicode mode
543	is disabled (e.g., `(?-u:.)`), then `.` will match a single byte in all cases.
544	* The character classes `\w`, `\d` and `\s` are all Unicode-aware by default.
545	Use `(?-u:\w)`, `(?-u:\d)` and `(?-u:\s)` to get their ASCII-only definitions.
546	* Similarly, `\b` and `\B` use a Unicode definition of a "word" character.
547	To get ASCII-only word boundaries, use `(?-u:\b)` and `(?-u:\B)`. This also
548	applies to the special word boundary assertions. (That is, `\b{start}`,
549	`\b{end}`, `\b{start-half}`, `\b{end-half}`.)
550	* `^` and `$` are not Unicode-aware in multi-line mode. Namely, they only
551	recognize `\n` (assuming CRLF mode is not enabled) and not any of the other
552	forms of line terminators defined by Unicode.
553	* Case insensitive searching is Unicode-aware and uses simple case folding.
554	* Unicode general categories, scripts and many boolean properties are available
555	by default via the `\p{property name}` syntax.
556	* In all cases, matches are reported using byte offsets. Or more precisely,
557	UTF-8 code unit offsets. This permits constant time indexing and slicing of the
558	haystack.
559
560	[UTS18]: https://unicode.org/reports/tr18/
561	[UNICODE.md]: https://github.com/rust-lang/regex/blob/master/UNICODE.md
562
563	Patterns themselves are only* interpreted as a sequence of Unicode scalar*
564	values. This means you can use Unicode characters directly in your pattern:
565
566	```rust
567	use regex::Regex;
568
569	let re = Regex::new(r"(?i)Δ+").unwrap();
570	let m = re.find("ΔδΔ").unwrap();
571	assert_eq!((`0`, `6`), (m.start(), m.end()));
572	// alternatively:
573	assert_eq!(`0`..`6`, m.range());
574	```
575
576	As noted above, Unicode general categories, scripts, script extensions, ages
577	and a smattering of boolean properties are available as character classes. For
578	example, you can match a sequence of numerals, Greek or Cherokee letters:
579
580	```rust
581	use regex::Regex;
582
583	let re = Regex::new(r"[\pN\p{Greek}\p{Cherokee}]+").unwrap();
584	let m = re.find("abcΔᎠβⅠᏴγδⅡxyz").unwrap();
585	assert_eq!(`3`..`23`, m.range());
586	```
587
588	While not specific to Unicode, this library also supports character class set
589	operations. Namely, one can nest character classes arbitrarily and perform set
590	operations on them. Those set operations are union (the default), intersection,
591	difference and symmetric difference. These set operations tend to be most
592	useful with Unicode character classes. For example, to match any codepoint
593	that is both in the `Greek` script and in the `Letter` general category:
594
595	```rust
596	use regex::Regex;
597
598	let re = Regex::new(r"[\p{Greek}&&\pL]+").unwrap();
599	let subs: Vec<&str> = re.find_iter("ΔδΔ𐅌ΔδΔ").map(\|m\| m.as_str()).collect();
600	assert_eq!(subs, vec!["ΔδΔ", "ΔδΔ"]);
601
602	// If we just matches on Greek, then all codepoints would match!
603	let re = Regex::new(r"\p{Greek}+").unwrap();
604	let subs: Vec<&str> = re.find_iter("ΔδΔ𐅌ΔδΔ").map(\|m\| m.as_str()).collect();
605	assert_eq!(subs, vec!["ΔδΔ𐅌ΔδΔ"]);
606	```
607
608	### Opt out of Unicode support
609
610	The [`bytes::Regex`] type that can be used to search `&[u8]` haystacks. By
611	default, haystacks are conventionally treated as UTF-8 just like it is with the
612	main `Regex` type. However, this behavior can be disabled by turning off the
613	`u` flag, even if doing so could result in matching invalid UTF-8. For example,
614	when the `u` flag is disabled, `.` will match any byte instead of any Unicode
615	scalar value.
616
617	Disabling the `u` flag is also possible with the standard `&str`-based `Regex`
618	type, but it is only allowed where the UTF-8 invariant is maintained. For
619	example, `(?-u:\w)` is an ASCII-only `\w` character class and is legal in an
620	`&str`-based `Regex`, but `(?-u:\W)` will attempt to match any byte* that*
621	isn't in `(?-u:\w)`, which in turn includes bytes that are invalid UTF-8.
622	Similarly, `(?-u:\xFF)` will attempt to match the raw byte `\xFF` (instead of
623	`U+00FF`), which is invalid UTF-8 and therefore is illegal in `&str`-based
624	regexes.
625
626	Finally, since Unicode support requires bundling large Unicode data
627	tables, this crate exposes knobs to disable the compilation of those
628	data tables, which can be useful for shrinking binary size and reducing
629	compilation times. For details on how to do that, see the section on [crate
630	features](#crate-features).
631
632	# Syntax
633
634	The syntax supported in this crate is documented below.
635
636	Note that the regular expression parser and abstract syntax are exposed in
637	a separate crate, [`regex-syntax`](https://docs.rs/regex-syntax).
638
639	### Matching one character
640
641	<pre class="rust">
642	. any character except new line (includes new line with s flag)
643	[0-9] any ASCII digit
644	\d digit (\p{Nd})
645	\D not digit
646	\pX Unicode character class identified by a one-letter name
647	\p{Greek} Unicode character class (general category or script)
648	\PX Negated Unicode character class identified by a one-letter name
649	\P{Greek} negated Unicode character class (general category or script)
650	</pre>
651
652	### Character classes
653
654	<pre class="rust">
655	[xyz] A character class matching either x, y or z (union).
656	[^xyz] A character class matching any character except x, y and z.
657	[a-z] A character class matching any character in range a-z.
658	[[:alpha:]] ASCII character class ([A-Za-z])
659	[[:^alpha:]] Negated ASCII character class ([^A-Za-z])
660	[x[^xyz]] Nested/grouping character class (matching any character except y and z)
661	[a-y&&xyz] Intersection (matching x or y)
662	[0-9&&[^4]] Subtraction using intersection and negation (matching 0-9 except 4)
663	[0-9--4] Direct subtraction (matching 0-9 except 4)
664	[a-g~~b-h] Symmetric difference (matching `a` and `h` only)
665	[\[\]] Escaping in character classes (matching [ or ])
666	[a&&b] An empty character class matching nothing
667	</pre>
668
669	Any named character class may appear inside a bracketed `[...]` character
670	class. For example, `[\p{Greek}[:digit:]]` matches any ASCII digit or any
671	codepoint in the `Greek` script. `[\p{Greek}&&\pL]` matches Greek letters.
672
673	Precedence in character classes, from most binding to least:
674
675	1. Ranges: `[a-cd]` == `[[a-c]d]`
676	2. Union: `[ab&&bc]` == `[[ab]&&[bc]]`
677	3. Intersection, difference, symmetric difference. All three have equivalent
678	precedence, and are evaluated in left-to-right order. For example,
679	`[\pL--\p{Greek}&&\p{Uppercase}]` == `[[\pL--\p{Greek}]&&\p{Uppercase}]`.
680	4. Negation: `[^a-z&&b]` == `[^[a-z&&b]]`.
681
682	### Composites
683
684	<pre class="rust">
685	xy concatenation (x followed by y)
686	x\|y alternation (x or y, prefer x)
687	</pre>
688
689	This example shows how an alternation works, and what it means to prefer a
690	branch in the alternation over subsequent branches.
691
692	```
693	use regex::Regex;
694
695	let haystack = "samwise";
696	// If 'samwise' comes first in our alternation, then it is
697	// preferred as a match, even if the regex engine could
698	// technically detect that 'sam' led to a match earlier.
699	let re = Regex::new(r"samwise\|sam").unwrap();
700	assert_eq!("samwise", re.find(haystack).unwrap().as_str());
701	// But if 'sam' comes first, then it will match instead.
702	// In this case, it is impossible for 'samwise' to match
703	// because 'sam' is a prefix of it.
704	let re = Regex::new(r"sam\|samwise").unwrap();
705	assert_eq!("sam", re.find(haystack).unwrap().as_str());
706	```
707
708	### Repetitions
709
710	<pre class="rust">
711	x zero or more of x (greedy)*
712	x+ one or more of x (greedy)
713	x? zero or one of x (greedy)
714	x? zero or more of x (ungreedy/lazy)*
715	x+? one or more of x (ungreedy/lazy)
716	x?? zero or one of x (ungreedy/lazy)
717	x{n,m} at least n x and at most m x (greedy)
718	x{n,} at least n x (greedy)
719	x{n} exactly n x
720	x{n,m}? at least n x and at most m x (ungreedy/lazy)
721	x{n,}? at least n x (ungreedy/lazy)
722	x{n}? exactly n x
723	</pre>
724
725	### Empty matches
726
727	<pre class="rust">
728	^ the beginning of a haystack (or start-of-line with multi-line mode)
729	$ the end of a haystack (or end-of-line with multi-line mode)
730	\A only the beginning of a haystack (even with multi-line mode enabled)
731	\z only the end of a haystack (even with multi-line mode enabled)
732	\b a Unicode word boundary (\w on one side and \W, \A, or \z on other)
733	\B not a Unicode word boundary
734	\b{start}, \< a Unicode start-of-word boundary (\W\|\A on the left, \w on the right)
735	\b{end}, \> a Unicode end-of-word boundary (\w on the left, \W\|\z on the right))
736	\b{start-half} half of a Unicode start-of-word boundary (\W\|\A on the left)
737	\b{end-half} half of a Unicode end-of-word boundary (\W\|\z on the right)
738	</pre>
739
740	The empty regex is valid and matches the empty string. For example, the
741	empty regex matches `abc` at positions `0`, `1`, `2` and `3`. When using the
742	top-level [`Regex`] on `&str` haystacks, an empty match that splits a codepoint
743	is guaranteed to never be returned. However, such matches are permitted when
744	using a [`bytes::Regex`]. For example:
745
746	```rust
747	let re = regex::Regex::new(r"").unwrap();
748	let ranges: Vec<_> = re.find_iter("💩").map(\|m\| m.range()).collect();
749	assert_eq!(ranges, vec![`0`..`0`, `4`..`4`]);
750
751	let re = regex::bytes::Regex::new(r"").unwrap();
752	let ranges: Vec<_> = re.find_iter("💩".as_bytes()).map(\|m\| m.range()).collect();
753	assert_eq!(ranges, vec![`0`..`0`, `1`..`1`, `2`..`2`, `3`..`3`, `4`..`4`]);
754	```
755
756	Note that an empty regex is distinct from a regex that can never match.
757	For example, the regex `[a&&b]` is a character class that represents the
758	intersection of `a` and `b`. That intersection is empty, which means the
759	character class is empty. Since nothing is in the empty set, `[a&&b]` matches
760	nothing, not even the empty string.
761
762	### Grouping and flags
763
764	<pre class="rust">
765	(exp) numbered capture group (indexed by opening parenthesis)
766	(?P<name>exp) named (also numbered) capture group (names must be alpha-numeric)
767	(?<name>exp) named (also numbered) capture group (names must be alpha-numeric)
768	(?:exp) non-capturing group
769	(?flags) set flags within current group
770	(?flags:exp) set flags for exp (non-capturing)
771	</pre>
772
773	Capture group names must be any sequence of alpha-numeric Unicode codepoints,
774	in addition to `.`, `_`, `[` and `]`. Names must start with either an `_` or
775	an alphabetic codepoint. Alphabetic codepoints correspond to the `Alphabetic`
776	Unicode property, while numeric codepoints correspond to the union of the
777	`Decimal_Number`, `Letter_Number` and `Other_Number` general categories.
778
779	Flags are each a single character. For example, `(?x)` sets the flag `x`
780	and `(?-x)` clears the flag `x`. Multiple flags can be set or cleared at
781	the same time: `(?xy)` sets both the `x` and `y` flags and `(?x-y)` sets
782	the `x` flag and clears the `y` flag.
783
784	All flags are by default disabled unless stated otherwise. They are:
785
786	<pre class="rust">
787	i case-insensitive: letters match both upper and lower case
788	m multi-line mode: ^ and $ match begin/end of line
789	s allow . to match \n
790	R enables CRLF mode: when multi-line mode is enabled, \r\n is used
791	U swap the meaning of x and x?
792	u Unicode support (enabled by default)
793	x verbose mode, ignores whitespace and allow line comments (starting with `#`)
794	</pre>
795
796	Note that in verbose mode, whitespace is ignored everywhere, including within
797	character classes. To insert whitespace, use its escaped form or a hex literal.
798	For example, `\ ` or `\x20` for an ASCII space.
799
800	Flags can be toggled within a pattern. Here's an example that matches
801	case-insensitively for the first part but case-sensitively for the second part:
802
803	```rust
804	use regex::Regex;
805
806	let re = Regex::new(r"(?i)a+(?-i)b+").unwrap();
807	let m = re.find("AaAaAbbBBBb").unwrap();
808	assert_eq!(m.as_str(), "AaAaAbb");
809	```
810
811	Notice that the `a+` matches either `a` or `A`, but the `b+` only matches
812	`b`.
813
814	Multi-line mode means `^` and `$` no longer match just at the beginning/end of
815	the input, but also at the beginning/end of lines:
816
817	```
818	use regex::Regex;
819
820	let re = Regex::new(r"(?m)^line \d+").unwrap();
821	let m = re.find("line one`\n`line 2`\n`").unwrap();
822	assert_eq!(m.as_str(), "line 2");
823	```
824
825	Note that `^` matches after new lines, even at the end of input:
826
827	```
828	use regex::Regex;
829
830	let re = Regex::new(r"(?m)^").unwrap();
831	let m = re.find_iter("test`\n`").last().unwrap();
832	assert_eq!((m.start(), m.end()), (`5`, `5`));
833	```
834
835	When both CRLF mode and multi-line mode are enabled, then `^` and `$` will
836	match either `\r` and `\n`, but never in the middle of a `\r\n`:
837
838	```
839	use regex::Regex;
840
841	let re = Regex::new(r"(?mR)^foo$").unwrap();
842	let m = re.find("`\r\n`foo`\r\n`").unwrap();
843	assert_eq!(m.as_str(), "foo");
844	```
845
846	Unicode mode can also be selectively disabled, although only when the result
847	would not match invalid UTF-8. One good example of this is using an ASCII
848	word boundary instead of a Unicode word boundary, which might make some regex
849	searches run faster:
850
851	```rust
852	use regex::Regex;
853
854	let re = Regex::new(r"(?-u:\b).+(?-u:\b)").unwrap();
855	let m = re.find("$$abc$$").unwrap();
856	assert_eq!(m.as_str(), "abc");
857	```
858
859	### Escape sequences
860
861	Note that this includes all possible escape sequences, even ones that are
862	documented elsewhere.
863
864	<pre class="rust">
865	\ literal , applies to all ASCII except [0-9A-Za-z<>]
866	\a bell (\x07)
867	\f form feed (\x0C)
868	\t horizontal tab
869	\n new line
870	\r carriage return
871	\v vertical tab (\x0B)
872	\A matches at the beginning of a haystack
873	\z matches at the end of a haystack
874	\b word boundary assertion
875	\B negated word boundary assertion
876	\b{start}, \< start-of-word boundary assertion
877	\b{end}, \> end-of-word boundary assertion
878	\b{start-half} half of a start-of-word boundary assertion
879	\b{end-half} half of a end-of-word boundary assertion
880	\123 octal character code, up to three digits (when enabled)
881	\x7F hex character code (exactly two digits)
882	\x{10FFFF} any hex character code corresponding to a Unicode code point
883	\u007F hex character code (exactly four digits)
884	\u{7F} any hex character code corresponding to a Unicode code point
885	\U0000007F hex character code (exactly eight digits)
886	\U{7F} any hex character code corresponding to a Unicode code point
887	\p{Letter} Unicode character class
888	\P{Letter} negated Unicode character class
889	\d, \s, \w Perl character class
890	\D, \S, \W negated Perl character class
891	</pre>
892
893	### Perl character classes (Unicode friendly)
894
895	These classes are based on the definitions provided in
896	[UTS#18](https://www.unicode.org/reports/tr18/#Compatibility_Properties):
897
898	<pre class="rust">
899	\d digit (\p{Nd})
900	\D not digit
901	\s whitespace (\p{White_Space})
902	\S not whitespace
903	\w word character (\p{Alphabetic} + \p{M} + \d + \p{Pc} + \p{Join_Control})
904	\W not word character
905	</pre>
906
907	### ASCII character classes
908
909	These classes are based on the definitions provided in
910	[UTS#18](https://www.unicode.org/reports/tr18/#Compatibility_Properties):
911
912	<pre class="rust">
913	[[:alnum:]] alphanumeric ([0-9A-Za-z])
914	[[:alpha:]] alphabetic ([A-Za-z])
915	[[:ascii:]] ASCII ([\x00-\x7F])
916	[[:blank:]] blank ([\t ])
917	[[:cntrl:]] control ([\x00-\x1F\x7F])
918	[[:digit:]] digits ([0-9])
919	[[:graph:]] graphical ([!-~])
920	[[:lower:]] lower case ([a-z])
921	[[:print:]] printable ([ -~])
922	[[:punct:]] punctuation ([!-/:-@\[-`{-~])
923	[[:space:]] whitespace ([\t\n\v\f\r ])
924	[[:upper:]] upper case ([A-Z])
925	[[:word:]] word characters ([0-9A-Za-z_])
926	[[:xdigit:]] hex digit ([0-9A-Fa-f])
927	</pre>
928
929	# Untrusted input
930
931	This crate is meant to be able to run regex searches on untrusted haystacks
932	without fear of [ReDoS]. This crate also, to a certain extent, supports
933	untrusted patterns.
934
935	[ReDoS]: https://en.wikipedia.org/wiki/ReDoS
936
937	This crate differs from most (but not all) other regex engines in that it
938	doesn't use unbounded backtracking to run a regex search. In those cases,
939	one generally cannot use untrusted patterns or* untrusted haystacks because*
940	it can be very difficult to know whether a particular pattern will result in
941	catastrophic backtracking or not.
942
943	We'll first discuss how this crate deals with untrusted inputs and then wrap
944	it up with a realistic discussion about what practice really looks like.
945
946	### Panics
947
948	Outside of clearly documented cases, most APIs in this crate are intended to
949	never panic regardless of the inputs given to them. For example, `Regex::new`,
950	`Regex::is_match`, `Regex::find` and `Regex::captures` should never panic. That
951	is, it is an API promise that those APIs will never panic no matter what inputs
952	are given to them. With that said, regex engines are complicated beasts, and
953	providing a rock solid guarantee that these APIs literally never panic is
954	essentially equivalent to saying, "there are no bugs in this library." That is
955	a bold claim, and not really one that can be feasibly made with a straight
956	face.
957
958	Don't get the wrong impression here. This crate is extensively tested, not just
959	with unit and integration tests, but also via fuzz testing. For example, this
960	crate is part of the [OSS-fuzz project]. Panics should be incredibly rare, but
961	it is possible for bugs to exist, and thus possible for a panic to occur. If
962	you need a rock solid guarantee against panics, then you should wrap calls into
963	this library with [`std::panic::catch_unwind`].
964
965	It's also worth pointing out that this library will generally* panic when*
966	other regex engines would commit undefined behavior. When undefined behavior
967	occurs, your program might continue as if nothing bad has happened, but it also
968	might mean your program is open to the worst kinds of exploits. In contrast,
969	the worst thing a panic can do is a denial of service.
970
971	[OSS-fuzz project]: https://android.googlesource.com/platform/external/oss-fuzz/+/refs/tags/android-t-preview-1/projects/rust-regex/
972	[`std::panic::catch_unwind`]: https://doc.rust-lang.org/std/panic/fn.catch_unwind.html
973
974	### Untrusted patterns
975
976	The principal way this crate deals with them is by limiting their size by
977	default. The size limit can be configured via [`RegexBuilder::size_limit`]. The
978	idea of a size limit is that compiling a pattern into a `Regex` will fail if it
979	becomes "too big." Namely, while most* resources consumed by compiling a regex*
980	are approximately proportional (albeit with some high constant factors in some
981	cases, such as with Unicode character classes) to the length of the pattern
982	itself, there is one particular exception to this: counted repetitions. Namely,
983	this pattern:
984
985	```text
986	a{5}{5}{5}{5}{5}{5}
987	```
988
989	Is equivalent to this pattern:
990
991	```text
992	a{15625}
993	```
994
995	In both of these cases, the actual pattern string is quite small, but the
996	resulting `Regex` value is quite large. Indeed, as the first pattern shows,
997	it isn't enough to locally limit the size of each repetition because they can
998	be stacked in a way that results in exponential growth.
999
1000	To provide a bit more context, a simplified view of regex compilation looks
1001	like this:
1002
1003	* The pattern string is parsed into a structured representation called an AST.
1004	Counted repetitions are not expanded and Unicode character classes are not
1005	looked up in this stage. That is, the size of the AST is proportional to the
1006	size of the pattern with "reasonable" constant factors. In other words, one
1007	can reasonably limit the memory used by an AST by limiting the length of the
1008	pattern string.
1009	* The AST is translated into an HIR. Counted repetitions are still not
1010	expanded at this stage, but Unicode character classes are embedded into the
1011	HIR. The memory usage of a HIR is still proportional to the length of the
1012	original pattern string, but the constant factors---mostly as a result of
1013	Unicode character classes---can be quite high. Still though, the memory used by
1014	an HIR can be reasonably limited by limiting the length of the pattern string.
1015	* The HIR is compiled into a [Thompson NFA]. This is the stage at which
1016	something like `\w{5}` is rewritten to `\w\w\w\w\w`. Thus, this is the stage
1017	at which [`RegexBuilder::size_limit`] is enforced. If the NFA exceeds the
1018	configured size, then this stage will fail.
1019
1020	[Thompson NFA]: https://en.wikipedia.org/wiki/Thompson%27s_construction
1021
1022	The size limit helps avoid two different kinds of exorbitant resource usage:
1023
1024	* It avoids permitting exponential memory usage based on the size of the
1025	pattern string.
1026	* It avoids long search times. This will be discussed in more detail in the
1027	next section, but worst case search time is* dependent on the size of the*
1028	regex. So keeping regexes limited to a reasonable size is also a way of keeping
1029	search times reasonable.
1030
1031	Finally, it's worth pointing out that regex compilation is guaranteed to take
1032	worst case `O(m)` time, where `m` is proportional to the size of regex. The
1033	size of the regex here is after* the counted repetitions have been expanded.*
1034
1035	Advice for those using untrusted regexes: limit the pattern length to
1036	something small and expand it as needed. Configure [`RegexBuilder::size_limit`]
1037	to something small and then expand it as needed.
1038
1039	### Untrusted haystacks
1040
1041	The main way this crate guards against searches from taking a long time is by
1042	using algorithms that guarantee a `O(m n)` worst case time and space bound.*
1043	Namely:
1044
1045	* `m` is proportional to the size of the regex, where the size of the regex
1046	includes the expansion of all counted repetitions. (See the previous section on
1047	untrusted patterns.)
1048	* `n` is proportional to the length, in bytes, of the haystack.
1049
1050	In other words, if you consider `m` to be a constant (for example, the regex
1051	pattern is a literal in the source code), then the search can be said to run
1052	in "linear time." Or equivalently, "linear time with respect to the size of the
1053	haystack."
1054
1055	But the `m` factor here is important not to ignore. If a regex is
1056	particularly big, the search times can get quite slow. This is why, in part,
1057	[`RegexBuilder::size_limit`] exists.
1058
1059	Advice for those searching untrusted haystacks: As long as your regexes
1060	are not enormous, you should expect to be able to search untrusted haystacks
1061	without fear. If you aren't sure, you should benchmark it. Unlike backtracking
1062	engines, if your regex is so big that it's likely to result in slow searches,
1063	this is probably something you'll be able to observe regardless of what the
1064	haystack is made up of.
1065
1066	### Iterating over matches
1067
1068	One thing that is perhaps easy to miss is that the worst case time
1069	complexity bound of `O(m n)` applies to methods like [`Regex::is_match`],*
1070	[`Regex::find`] and [`Regex::captures`]. It does not* apply to*
1071	[`Regex::find_iter`] or [`Regex::captures_iter`]. Namely, since iterating over
1072	all matches can execute many searches, and each search can scan the entire
1073	haystack, the worst case time complexity for iterators is `O(m n^2)`.*
1074
1075	One example of where this occurs is when a pattern consists of an alternation,
1076	where an earlier branch of the alternation requires scanning the entire
1077	haystack only to discover that there is no match. It also requires a later
1078	branch of the alternation to have matched at the beginning of the search. For
1079	example, consider the pattern `.[^A-Z]\|[A-Z]` and the haystack `AAAAA`. The*
1080	first search will scan to the end looking for matches of `.[^A-Z]` even though*
1081	a finite automata engine (as in this crate) knows that `[A-Z]` has already
1082	matched the first character of the haystack. This is due to the greedy nature
1083	of regex searching. That first search will report a match at the first `A` only
1084	after scanning to the end to discover that no other match exists. The next
1085	search then begins at the second `A` and the behavior repeats.
1086
1087	There is no way to avoid this. This means that if both patterns and haystacks
1088	are untrusted and you're iterating over all matches, you're susceptible to
1089	worst case quadratic time complexity. One possible way to mitigate this
1090	is to drop down to the lower level `regex-automata` crate and use its
1091	`meta::Regex` iterator APIs. There, you can configure the search to operate
1092	in "earliest" mode by passing a `Input::new(haystack).earliest(true)` to
1093	`meta::Regex::find_iter` (for example). By enabling this mode, you give up
1094	the normal greedy match semantics of regex searches and instead ask the regex
1095	engine to immediately stop as soon as a match has been found. Enabling this
1096	mode will thus restore the worst case `O(m n)` time complexity bound, but at*
1097	the cost of different semantics.
1098
1099	### Untrusted inputs in practice
1100
1101	While providing a `O(m n)` worst case time bound on all searches goes a long*
1102	way toward preventing [ReDoS], that doesn't mean every search you can possibly
1103	run will complete without burning CPU time. In general, there are a few ways
1104	for the `m n` time bound to still bite you:*
1105
1106	* You are searching an exceptionally long haystack. No matter how you slice
1107	it, a longer haystack will take more time to search. This crate may often make
1108	very quick work of even long haystacks because of its literal optimizations,
1109	but those aren't available for all regexes.
1110	* Unicode character classes can cause searches to be quite slow in some cases.
1111	This is especially true when they are combined with counted repetitions. While
1112	the regex size limit above will protect you from the most egregious cases,
1113	the default size limit still permits pretty big regexes that can execute more
1114	slowly than one might expect.
1115	* While routines like [`Regex::find`] and [`Regex::captures`] guarantee
1116	worst case `O(m n)` search time, routines like [`Regex::find_iter`] and*
1117	[`Regex::captures_iter`] actually have worst case `O(m n^2)` search time.*
1118	This is because `find_iter` runs many searches, and each search takes worst
1119	case `O(m n)` time. Thus, iteration of all matches in a haystack has*
1120	worst case `O(m n^2)`. A good example of a pattern that exhibits this is*
1121	`(?:A+){1000}\|` or even `.[^A-Z]\|[A-Z]`.*
1122
1123	In general, unstrusted haystacks are easier to stomach than untrusted patterns.
1124	Untrusted patterns give a lot more control to the caller to impact the
1125	performance of a search. In many cases, a regex search will actually execute in
1126	average case `O(n)` time (i.e., not dependent on the size of the regex), but
1127	this can't be guaranteed in general. Therefore, permitting untrusted patterns
1128	means that your only line of defense is to put a limit on how big `m` (and
1129	perhaps also `n`) can be in `O(m n)`. `n` is limited by simply inspecting*
1130	the length of the haystack while `m` is limited by both* applying a limit to*
1131	the length of the pattern and* a limit on the compiled size of the regex via*
1132	[`RegexBuilder::size_limit`].
1133
1134	It bears repeating: if you're accepting untrusted patterns, it would be a good
1135	idea to start with conservative limits on `m` and `n`, and then carefully
1136	increase them as needed.
1137
1138	# Crate features
1139
1140	By default, this crate tries pretty hard to make regex matching both as fast
1141	as possible and as correct as it can be. This means that there is a lot of
1142	code dedicated to performance, the handling of Unicode data and the Unicode
1143	data itself. Overall, this leads to more dependencies, larger binaries and
1144	longer compile times. This trade off may not be appropriate in all cases, and
1145	indeed, even when all Unicode and performance features are disabled, one is
1146	still left with a perfectly serviceable regex engine that will work well in
1147	many cases. (Note that code is not arbitrarily reducible, and for this reason,
1148	the [`regex-lite`](https://docs.rs/regex-lite) crate exists to provide an even
1149	more minimal experience by cutting out Unicode and performance, but still
1150	maintaining the linear search time bound.)
1151
1152	This crate exposes a number of features for controlling that trade off. Some
1153	of these features are strictly performance oriented, such that disabling them
1154	won't result in a loss of functionality, but may result in worse performance.
1155	Other features, such as the ones controlling the presence or absence of Unicode
1156	data, can result in a loss of functionality. For example, if one disables the
1157	`unicode-case` feature (described below), then compiling the regex `(?i)a`
1158	will fail since Unicode case insensitivity is enabled by default. Instead,
1159	callers must use `(?i-u)a` to disable Unicode case folding. Stated differently,
1160	enabling or disabling any of the features below can only add or subtract from
1161	the total set of valid regular expressions. Enabling or disabling a feature
1162	will never modify the match semantics of a regular expression.
1163
1164	Most features below are enabled by default. Features that aren't enabled by
1165	default are noted.
1166
1167	### Ecosystem features
1168
1169	* std -
1170	When enabled, this will cause `regex` to use the standard library. In terms
1171	of APIs, `std` causes error types to implement the `std::error::Error`
1172	trait. Enabling `std` will also result in performance optimizations,
1173	including SIMD and faster synchronization primitives. Notably, disabling
1174	the `std` feature will result in the use of spin locks. To use a regex
1175	engine without `std` and without spin locks, you'll need to drop down to
1176	the [`regex-automata`](https://docs.rs/regex-automata) crate.
1177	* logging -
1178	When enabled, the `log` crate is used to emit messages about regex
1179	compilation and search strategies. This is disabled by default. This is
1180	typically only useful to someone working on this crate's internals, but might
1181	be useful if you're doing some rabbit hole performance hacking. Or if you're
1182	just interested in the kinds of decisions being made by the regex engine.
1183
1184	### Performance features
1185
1186	* perf -
1187	Enables all performance related features except for `perf-dfa-full`. This
1188	feature is enabled by default is intended to cover all reasonable features
1189	that improve performance, even if more are added in the future.
1190	* perf-dfa -
1191	Enables the use of a lazy DFA for matching. The lazy DFA is used to compile
1192	portions of a regex to a very fast DFA on an as-needed basis. This can
1193	result in substantial speedups, usually by an order of magnitude on large
1194	haystacks. The lazy DFA does not bring in any new dependencies, but it can
1195	make compile times longer.
1196	* perf-dfa-full -
1197	Enables the use of a full DFA for matching. Full DFAs are problematic because
1198	they have worst case `O(2^n)` construction time. For this reason, when this
1199	feature is enabled, full DFAs are only used for very small regexes and a
1200	very small space bound is used during determinization to avoid the DFA
1201	from blowing up. This feature is not enabled by default, even as part of
1202	`perf`, because it results in fairly sizeable increases in binary size and
1203	compilation time. It can result in faster search times, but they tend to be
1204	more modest and limited to non-Unicode regexes.
1205	* perf-onepass -
1206	Enables the use of a one-pass DFA for extracting the positions of capture
1207	groups. This optimization applies to a subset of certain types of NFAs and
1208	represents the fastest engine in this crate for dealing with capture groups.
1209	* perf-backtrack -
1210	Enables the use of a bounded backtracking algorithm for extracting the
1211	positions of capture groups. This usually sits between the slowest engine
1212	(the PikeVM) and the fastest engine (one-pass DFA) for extracting capture
1213	groups. It's used whenever the regex is not one-pass and is small enough.
1214	* perf-inline -
1215	Enables the use of aggressive inlining inside match routines. This reduces
1216	the overhead of each match. The aggressive inlining, however, increases
1217	compile times and binary size.
1218	* perf-literal -
1219	Enables the use of literal optimizations for speeding up matches. In some
1220	cases, literal optimizations can result in speedups of _several_ orders of
1221	magnitude. Disabling this drops the `aho-corasick` and `memchr` dependencies.
1222	* perf-cache -
1223	This feature used to enable a faster internal cache at the cost of using
1224	additional dependencies, but this is no longer an option. A fast internal
1225	cache is now used unconditionally with no additional dependencies. This may
1226	change in the future.
1227
1228	### Unicode features
1229
1230	* unicode -
1231	Enables all Unicode features. This feature is enabled by default, and will
1232	always cover all Unicode features, even if more are added in the future.
1233	* unicode-age -
1234	Provide the data for the
1235	[Unicode `Age` property](https://www.unicode.org/reports/tr44/tr44-24.html#Character_Age).
1236	This makes it possible to use classes like `\p{Age:6.0}` to refer to all
1237	codepoints first introduced in Unicode 6.0
1238	* unicode-bool -
1239	Provide the data for numerous Unicode boolean properties. The full list
1240	is not included here, but contains properties like `Alphabetic`, `Emoji`,
1241	`Lowercase`, `Math`, `Uppercase` and `White_Space`.
1242	* unicode-case -
1243	Provide the data for case insensitive matching using
1244	[Unicode's "simple loose matches" specification](https://www.unicode.org/reports/tr18/#Simple_Loose_Matches).
1245	* unicode-gencat -
1246	Provide the data for
1247	[Unicode general categories](https://www.unicode.org/reports/tr44/tr44-24.html#General_Category_Values).
1248	This includes, but is not limited to, `Decimal_Number`, `Letter`,
1249	`Math_Symbol`, `Number` and `Punctuation`.
1250	* unicode-perl -
1251	Provide the data for supporting the Unicode-aware Perl character classes,
1252	corresponding to `\w`, `\s` and `\d`. This is also necessary for using
1253	Unicode-aware word boundary assertions. Note that if this feature is
1254	disabled, the `\s` and `\d` character classes are still available if the
1255	`unicode-bool` and `unicode-gencat` features are enabled, respectively.
1256	* unicode-script -
1257	Provide the data for
1258	[Unicode scripts and script extensions](https://www.unicode.org/reports/tr24/).
1259	This includes, but is not limited to, `Arabic`, `Cyrillic`, `Hebrew`,
1260	`Latin` and `Thai`.
1261	* unicode-segment -
1262	Provide the data necessary to provide the properties used to implement the
1263	[Unicode text segmentation algorithms](https://www.unicode.org/reports/tr29/).
1264	This enables using classes like `\p{gcb=Extend}`, `\p{wb=Katakana}` and
1265	`\p{sb=ATerm}`.
1266
1267	# Other crates
1268
1269	This crate has two required dependencies and several optional dependencies.
1270	This section briefly describes them with the goal of raising awareness of how
1271	different components of this crate may be used independently.
1272
1273	It is somewhat unusual for a regex engine to have dependencies, as most regex
1274	libraries are self contained units with no dependencies other than a particular
1275	environment's standard library. Indeed, for other similarly optimized regex
1276	engines, most or all of the code in the dependencies of this crate would
1277	normally just be unseparable or coupled parts of the crate itself. But since
1278	Rust and its tooling ecosystem make the use of dependencies so easy, it made
1279	sense to spend some effort de-coupling parts of this crate and making them
1280	independently useful.
1281
1282	We only briefly describe each crate here.
1283
1284	* [`regex-lite`](https://docs.rs/regex-lite) is not a dependency of `regex`,
1285	but rather, a standalone zero-dependency simpler version of `regex` that
1286	prioritizes compile times and binary size. In exchange, it eschews Unicode
1287	support and performance. Its match semantics are as identical as possible to
1288	the `regex` crate, and for the things it supports, its APIs are identical to
1289	the APIs in this crate. In other words, for a lot of use cases, it is a drop-in
1290	replacement.
1291	* [`regex-syntax`](https://docs.rs/regex-syntax) provides a regular expression
1292	parser via `Ast` and `Hir` types. It also provides routines for extracting
1293	literals from a pattern. Folks can use this crate to do analysis, or even to
1294	build their own regex engine without having to worry about writing a parser.
1295	* [`regex-automata`](https://docs.rs/regex-automata) provides the regex engines
1296	themselves. One of the downsides of finite automata based regex engines is that
1297	they often need multiple internal engines in order to have similar or better
1298	performance than an unbounded backtracking engine in practice. `regex-automata`
1299	in particular provides public APIs for a PikeVM, a bounded backtracker, a
1300	one-pass DFA, a lazy DFA, a fully compiled DFA and a meta regex engine that
1301	combines all them together. It also has native multi-pattern support and
1302	provides a way to compile and serialize full DFAs such that they can be loaded
1303	and searched in a no-std no-alloc environment. `regex-automata` itself doesn't
1304	even have a required dependency on `regex-syntax`!
1305	* [`memchr`](https://docs.rs/memchr) provides low level SIMD vectorized
1306	routines for quickly finding the location of single bytes or even substrings
1307	in a haystack. In other words, it provides fast `memchr` and `memmem` routines.
1308	These are used by this crate in literal optimizations.
1309	* [`aho-corasick`](https://docs.rs/aho-corasick) provides multi-substring
1310	search. It also provides SIMD vectorized routines in the case where the number
1311	of substrings to search for is relatively small. The `regex` crate also uses
1312	this for literal optimizations.
1313	*/
1314
1315	#![no_std]
1316	#![deny(missing_docs)]
1317	#![cfg_attr(feature = "pattern", feature(pattern))]
1318	#![warn(missing_debug_implementations)]
1319
1320	#[cfg(doctest)]
1321	doc_comment::doctest!("../README.md");
1322
1323	extern crate alloc;
1324	#[cfg(any(test, feature = "std"))]
1325	extern crate std;
1326
1327	pub use crate::error::Error;
1328
1329	pub use crate::{builders::string::, regex::string::, regexset::string::*};
1330
1331	mod builders;
1332	pub mod bytes;
1333	mod error;
1334	mod find_byte;
1335	#[cfg(feature = "pattern")]
1336	mod pattern;
1337	mod regex;
1338	mod regexset;
1339
1340	/// Escapes all regular expression meta characters in `pattern`.
1341	///
1342	/// The string returned may be safely used as a literal in a regular
1343	/// expression.
1344	pub fn escape(pattern: &str) -> alloc::string::String {
1345	regex_syntax::escape(pattern)
1346	}
1347