Commit | Line | Data |
---|---|---|
8ac170f3 PH |
1 | This file contains the PCRE man page that describes the regular expressions |
2 | supported by PCRE version 6.0. Note that not all of the features are relevant | |
495ae4b0 PH |
3 | in the context of Exim. In particular, the version of PCRE that is compiled |
4 | with Exim does not include UTF-8 support, there is no mechanism for changing | |
5 | the options with which the PCRE functions are called, and features such as | |
6 | callout are not accessible. | |
7 | ----------------------------------------------------------------------------- | |
8 | ||
495ae4b0 PH |
9 | |
10 | ||
11 | NAME | |
12 | PCRE - Perl-compatible regular expressions | |
13 | ||
8ac170f3 | 14 | |
495ae4b0 PH |
15 | PCRE REGULAR EXPRESSION DETAILS |
16 | ||
17 | The syntax and semantics of the regular expressions supported by PCRE | |
18 | are described below. Regular expressions are also described in the Perl | |
19 | documentation and in a number of books, some of which have copious | |
20 | examples. Jeffrey Friedl's "Mastering Regular Expressions", published | |
21 | by O'Reilly, covers regular expressions in great detail. This descrip- | |
22 | tion of PCRE's regular expressions is intended as reference material. | |
23 | ||
24 | The original operation of PCRE was on strings of one-byte characters. | |
25 | However, there is now also support for UTF-8 character strings. To use | |
26 | this, you must build PCRE to include UTF-8 support, and then call | |
27 | pcre_compile() with the PCRE_UTF8 option. How this affects pattern | |
28 | matching is mentioned in several places below. There is also a summary | |
29 | of UTF-8 features in the section on UTF-8 support in the main pcre | |
30 | page. | |
31 | ||
8ac170f3 PH |
32 | The remainder of this document discusses the patterns that are sup- |
33 | ported by PCRE when its main matching function, pcre_exec(), is used. | |
34 | From release 6.0, PCRE offers a second matching function, | |
35 | pcre_dfa_exec(), which matches using a different algorithm that is not | |
36 | Perl-compatible. The advantages and disadvantages of the alternative | |
37 | function, and how it differs from the normal function, are discussed in | |
38 | the pcrematching page. | |
39 | ||
495ae4b0 PH |
40 | A regular expression is a pattern that is matched against a subject |
41 | string from left to right. Most characters stand for themselves in a | |
42 | pattern, and match the corresponding characters in the subject. As a | |
43 | trivial example, the pattern | |
44 | ||
45 | The quick brown fox | |
46 | ||
8ac170f3 PH |
47 | matches a portion of a subject string that is identical to itself. When |
48 | caseless matching is specified (the PCRE_CASELESS option), letters are | |
49 | matched independently of case. In UTF-8 mode, PCRE always understands | |
50 | the concept of case for characters whose values are less than 128, so | |
51 | caseless matching is always possible. For characters with higher val- | |
52 | ues, the concept of case is supported if PCRE is compiled with Unicode | |
53 | property support, but not otherwise. If you want to use caseless | |
54 | matching for characters 128 and above, you must ensure that PCRE is | |
55 | compiled with Unicode property support as well as with UTF-8 support. | |
56 | ||
57 | The power of regular expressions comes from the ability to include | |
58 | alternatives and repetitions in the pattern. These are encoded in the | |
59 | pattern by the use of metacharacters, which do not stand for themselves | |
60 | but instead are interpreted in some special way. | |
61 | ||
62 | There are two different sets of metacharacters: those that are recog- | |
63 | nized anywhere in the pattern except within square brackets, and those | |
64 | that are recognized in square brackets. Outside square brackets, the | |
495ae4b0 PH |
65 | metacharacters are as follows: |
66 | ||
67 | \ general escape character with several uses | |
68 | ^ assert start of string (or line, in multiline mode) | |
69 | $ assert end of string (or line, in multiline mode) | |
70 | . match any character except newline (by default) | |
71 | [ start character class definition | |
72 | | start of alternative branch | |
73 | ( start subpattern | |
74 | ) end subpattern | |
75 | ? extends the meaning of ( | |
76 | also 0 or 1 quantifier | |
77 | also quantifier minimizer | |
78 | * 0 or more quantifier | |
79 | + 1 or more quantifier | |
80 | also "possessive quantifier" | |
81 | { start min/max quantifier | |
82 | ||
8ac170f3 | 83 | Part of a pattern that is in square brackets is called a "character |
495ae4b0 PH |
84 | class". In a character class the only metacharacters are: |
85 | ||
86 | \ general escape character | |
87 | ^ negate the class, but only if the first character | |
88 | - indicates character range | |
89 | [ POSIX character class (only if followed by POSIX | |
90 | syntax) | |
91 | ] terminates the character class | |
92 | ||
8ac170f3 | 93 | The following sections describe the use of each of the metacharacters. |
495ae4b0 PH |
94 | |
95 | ||
96 | BACKSLASH | |
97 | ||
98 | The backslash character has several uses. Firstly, if it is followed by | |
8ac170f3 PH |
99 | a non-alphanumeric character, it takes away any special meaning that |
100 | character may have. This use of backslash as an escape character | |
495ae4b0 PH |
101 | applies both inside and outside character classes. |
102 | ||
8ac170f3 PH |
103 | For example, if you want to match a * character, you write \* in the |
104 | pattern. This escaping action applies whether or not the following | |
105 | character would otherwise be interpreted as a metacharacter, so it is | |
106 | always safe to precede a non-alphanumeric with backslash to specify | |
107 | that it stands for itself. In particular, if you want to match a back- | |
495ae4b0 PH |
108 | slash, you write \\. |
109 | ||
8ac170f3 PH |
110 | If a pattern is compiled with the PCRE_EXTENDED option, whitespace in |
111 | the pattern (other than in a character class) and characters between a | |
495ae4b0 | 112 | # outside a character class and the next newline character are ignored. |
8ac170f3 | 113 | An escaping backslash can be used to include a whitespace or # charac- |
495ae4b0 PH |
114 | ter as part of the pattern. |
115 | ||
8ac170f3 PH |
116 | If you want to remove the special meaning from a sequence of charac- |
117 | ters, you can do so by putting them between \Q and \E. This is differ- | |
118 | ent from Perl in that $ and @ are handled as literals in \Q...\E | |
119 | sequences in PCRE, whereas in Perl, $ and @ cause variable interpola- | |
495ae4b0 PH |
120 | tion. Note the following examples: |
121 | ||
122 | Pattern PCRE matches Perl matches | |
123 | ||
124 | \Qabc$xyz\E abc$xyz abc followed by the | |
125 | contents of $xyz | |
126 | \Qabc\$xyz\E abc\$xyz abc\$xyz | |
127 | \Qabc\E\$\Qxyz\E abc$xyz abc$xyz | |
128 | ||
8ac170f3 | 129 | The \Q...\E sequence is recognized both inside and outside character |
495ae4b0 PH |
130 | classes. |
131 | ||
132 | Non-printing characters | |
133 | ||
134 | A second use of backslash provides a way of encoding non-printing char- | |
8ac170f3 PH |
135 | acters in patterns in a visible manner. There is no restriction on the |
136 | appearance of non-printing characters, apart from the binary zero that | |
137 | terminates a pattern, but when a pattern is being prepared by text | |
138 | editing, it is usually easier to use one of the following escape | |
495ae4b0 PH |
139 | sequences than the binary character it represents: |
140 | ||
141 | \a alarm, that is, the BEL character (hex 07) | |
142 | \cx "control-x", where x is any character | |
143 | \e escape (hex 1B) | |
144 | \f formfeed (hex 0C) | |
145 | \n newline (hex 0A) | |
146 | \r carriage return (hex 0D) | |
147 | \t tab (hex 09) | |
148 | \ddd character with octal code ddd, or backreference | |
149 | \xhh character with hex code hh | |
150 | \x{hhh..} character with hex code hhh... (UTF-8 mode only) | |
151 | ||
8ac170f3 PH |
152 | The precise effect of \cx is as follows: if x is a lower case letter, |
153 | it is converted to upper case. Then bit 6 of the character (hex 40) is | |
154 | inverted. Thus \cz becomes hex 1A, but \c{ becomes hex 3B, while \c; | |
495ae4b0 PH |
155 | becomes hex 7B. |
156 | ||
8ac170f3 PH |
157 | After \x, from zero to two hexadecimal digits are read (letters can be |
158 | in upper or lower case). In UTF-8 mode, any number of hexadecimal dig- | |
159 | its may appear between \x{ and }, but the value of the character code | |
160 | must be less than 2**31 (that is, the maximum hexadecimal value is | |
161 | 7FFFFFFF). If characters other than hexadecimal digits appear between | |
162 | \x{ and }, or if there is no terminating }, this form of escape is not | |
163 | recognized. Instead, the initial \x will be interpreted as a basic | |
164 | hexadecimal escape, with no following digits, giving a character whose | |
495ae4b0 PH |
165 | value is zero. |
166 | ||
167 | Characters whose value is less than 256 can be defined by either of the | |
8ac170f3 PH |
168 | two syntaxes for \x when PCRE is in UTF-8 mode. There is no difference |
169 | in the way they are handled. For example, \xdc is exactly the same as | |
495ae4b0 PH |
170 | \x{dc}. |
171 | ||
8ac170f3 PH |
172 | After \0 up to two further octal digits are read. In both cases, if |
173 | there are fewer than two digits, just those that are present are used. | |
174 | Thus the sequence \0\x\07 specifies two binary zeros followed by a BEL | |
175 | character (code value 7). Make sure you supply two digits after the | |
176 | initial zero if the pattern character that follows is itself an octal | |
495ae4b0 PH |
177 | digit. |
178 | ||
179 | The handling of a backslash followed by a digit other than 0 is compli- | |
180 | cated. Outside a character class, PCRE reads it and any following dig- | |
8ac170f3 | 181 | its as a decimal number. If the number is less than 10, or if there |
495ae4b0 | 182 | have been at least that many previous capturing left parentheses in the |
8ac170f3 PH |
183 | expression, the entire sequence is taken as a back reference. A |
184 | description of how this works is given later, following the discussion | |
495ae4b0 PH |
185 | of parenthesized subpatterns. |
186 | ||
8ac170f3 PH |
187 | Inside a character class, or if the decimal number is greater than 9 |
188 | and there have not been that many capturing subpatterns, PCRE re-reads | |
189 | up to three octal digits following the backslash, and generates a sin- | |
495ae4b0 PH |
190 | gle byte from the least significant 8 bits of the value. Any subsequent |
191 | digits stand for themselves. For example: | |
192 | ||
193 | \040 is another way of writing a space | |
194 | \40 is the same, provided there are fewer than 40 | |
195 | previous capturing subpatterns | |
196 | \7 is always a back reference | |
197 | \11 might be a back reference, or another way of | |
198 | writing a tab | |
199 | \011 is always a tab | |
200 | \0113 is a tab followed by the character "3" | |
201 | \113 might be a back reference, otherwise the | |
202 | character with octal code 113 | |
203 | \377 might be a back reference, otherwise | |
204 | the byte consisting entirely of 1 bits | |
205 | \81 is either a back reference, or a binary zero | |
206 | followed by the two characters "8" and "1" | |
207 | ||
8ac170f3 | 208 | Note that octal values of 100 or greater must not be introduced by a |
495ae4b0 PH |
209 | leading zero, because no more than three octal digits are ever read. |
210 | ||
8ac170f3 | 211 | All the sequences that define a single byte value or a single UTF-8 |
495ae4b0 | 212 | character (in UTF-8 mode) can be used both inside and outside character |
8ac170f3 | 213 | classes. In addition, inside a character class, the sequence \b is |
495ae4b0 | 214 | interpreted as the backspace character (hex 08), and the sequence \X is |
8ac170f3 | 215 | interpreted as the character "X". Outside a character class, these |
495ae4b0 PH |
216 | sequences have different meanings (see below). |
217 | ||
218 | Generic character types | |
219 | ||
8ac170f3 | 220 | The third use of backslash is for specifying generic character types. |
495ae4b0 PH |
221 | The following are always recognized: |
222 | ||
223 | \d any decimal digit | |
224 | \D any character that is not a decimal digit | |
225 | \s any whitespace character | |
226 | \S any character that is not a whitespace character | |
227 | \w any "word" character | |
228 | \W any "non-word" character | |
229 | ||
230 | Each pair of escape sequences partitions the complete set of characters | |
8ac170f3 | 231 | into two disjoint sets. Any given character matches one, and only one, |
495ae4b0 PH |
232 | of each pair. |
233 | ||
234 | These character type sequences can appear both inside and outside char- | |
8ac170f3 PH |
235 | acter classes. They each match one character of the appropriate type. |
236 | If the current matching point is at the end of the subject string, all | |
495ae4b0 PH |
237 | of them fail, since there is no character to match. |
238 | ||
8ac170f3 PH |
239 | For compatibility with Perl, \s does not match the VT character (code |
240 | 11). This makes it different from the the POSIX "space" class. The \s | |
495ae4b0 PH |
241 | characters are HT (9), LF (10), FF (12), CR (13), and space (32). |
242 | ||
243 | A "word" character is an underscore or any character less than 256 that | |
8ac170f3 PH |
244 | is a letter or digit. The definition of letters and digits is con- |
245 | trolled by PCRE's low-valued character tables, and may vary if locale- | |
246 | specific matching is taking place (see "Locale support" in the pcreapi | |
247 | page). For example, in the "fr_FR" (French) locale, some character | |
248 | codes greater than 128 are used for accented letters, and these are | |
495ae4b0 PH |
249 | matched by \w. |
250 | ||
8ac170f3 | 251 | In UTF-8 mode, characters with values greater than 128 never match \d, |
495ae4b0 PH |
252 | \s, or \w, and always match \D, \S, and \W. This is true even when Uni- |
253 | code character property support is available. | |
254 | ||
255 | Unicode character properties | |
256 | ||
257 | When PCRE is built with Unicode character property support, three addi- | |
8ac170f3 | 258 | tional escape sequences to match generic character types are available |
495ae4b0 PH |
259 | when UTF-8 mode is selected. They are: |
260 | ||
261 | \p{xx} a character with the xx property | |
262 | \P{xx} a character without the xx property | |
263 | \X an extended Unicode sequence | |
264 | ||
8ac170f3 PH |
265 | The property names represented by xx above are limited to the Unicode |
266 | general category properties. Each character has exactly one such prop- | |
267 | erty, specified by a two-letter abbreviation. For compatibility with | |
268 | Perl, negation can be specified by including a circumflex between the | |
269 | opening brace and the property name. For example, \p{^Lu} is the same | |
495ae4b0 PH |
270 | as \P{Lu}. |
271 | ||
8ac170f3 | 272 | If only one letter is specified with \p or \P, it includes all the |
495ae4b0 PH |
273 | properties that start with that letter. In this case, in the absence of |
274 | negation, the curly brackets in the escape sequence are optional; these | |
275 | two examples have the same effect: | |
276 | ||
277 | \p{L} | |
278 | \pL | |
279 | ||
280 | The following property codes are supported: | |
281 | ||
282 | C Other | |
283 | Cc Control | |
284 | Cf Format | |
285 | Cn Unassigned | |
286 | Co Private use | |
287 | Cs Surrogate | |
288 | ||
289 | L Letter | |
290 | Ll Lower case letter | |
291 | Lm Modifier letter | |
292 | Lo Other letter | |
293 | Lt Title case letter | |
294 | Lu Upper case letter | |
295 | ||
296 | M Mark | |
297 | Mc Spacing mark | |
298 | Me Enclosing mark | |
299 | Mn Non-spacing mark | |
300 | ||
301 | N Number | |
302 | Nd Decimal number | |
303 | Nl Letter number | |
304 | No Other number | |
305 | ||
306 | P Punctuation | |
307 | Pc Connector punctuation | |
308 | Pd Dash punctuation | |
309 | Pe Close punctuation | |
310 | Pf Final punctuation | |
311 | Pi Initial punctuation | |
312 | Po Other punctuation | |
313 | Ps Open punctuation | |
314 | ||
315 | S Symbol | |
316 | Sc Currency symbol | |
317 | Sk Modifier symbol | |
318 | Sm Mathematical symbol | |
319 | So Other symbol | |
320 | ||
321 | Z Separator | |
322 | Zl Line separator | |
323 | Zp Paragraph separator | |
324 | Zs Space separator | |
325 | ||
8ac170f3 | 326 | Extended properties such as "Greek" or "InMusicalSymbols" are not sup- |
495ae4b0 PH |
327 | ported by PCRE. |
328 | ||
8ac170f3 | 329 | Specifying caseless matching does not affect these escape sequences. |
495ae4b0 PH |
330 | For example, \p{Lu} always matches only upper case letters. |
331 | ||
8ac170f3 | 332 | The \X escape matches any number of Unicode characters that form an |
495ae4b0 PH |
333 | extended Unicode sequence. \X is equivalent to |
334 | ||
335 | (?>\PM\pM*) | |
336 | ||
8ac170f3 PH |
337 | That is, it matches a character without the "mark" property, followed |
338 | by zero or more characters with the "mark" property, and treats the | |
339 | sequence as an atomic group (see below). Characters with the "mark" | |
495ae4b0 PH |
340 | property are typically accents that affect the preceding character. |
341 | ||
8ac170f3 PH |
342 | Matching characters by Unicode property is not fast, because PCRE has |
343 | to search a structure that contains data for over fifteen thousand | |
495ae4b0 PH |
344 | characters. That is why the traditional escape sequences such as \d and |
345 | \w do not use Unicode properties in PCRE. | |
346 | ||
347 | Simple assertions | |
348 | ||
349 | The fourth use of backslash is for certain simple assertions. An asser- | |
8ac170f3 PH |
350 | tion specifies a condition that has to be met at a particular point in |
351 | a match, without consuming any characters from the subject string. The | |
352 | use of subpatterns for more complicated assertions is described below. | |
495ae4b0 PH |
353 | The backslashed assertions are: |
354 | ||
355 | \b matches at a word boundary | |
356 | \B matches when not at a word boundary | |
357 | \A matches at start of subject | |
358 | \Z matches at end of subject or before newline at end | |
359 | \z matches at end of subject | |
360 | \G matches at first matching position in subject | |
361 | ||
8ac170f3 | 362 | These assertions may not appear in character classes (but note that \b |
495ae4b0 PH |
363 | has a different meaning, namely the backspace character, inside a char- |
364 | acter class). | |
365 | ||
8ac170f3 PH |
366 | A word boundary is a position in the subject string where the current |
367 | character and the previous character do not both match \w or \W (i.e. | |
368 | one matches \w and the other matches \W), or the start or end of the | |
495ae4b0 PH |
369 | string if the first or last character matches \w, respectively. |
370 | ||
8ac170f3 | 371 | The \A, \Z, and \z assertions differ from the traditional circumflex |
495ae4b0 | 372 | and dollar (described in the next section) in that they only ever match |
8ac170f3 PH |
373 | at the very start and end of the subject string, whatever options are |
374 | set. Thus, they are independent of multiline mode. These three asser- | |
495ae4b0 | 375 | tions are not affected by the PCRE_NOTBOL or PCRE_NOTEOL options, which |
8ac170f3 PH |
376 | affect only the behaviour of the circumflex and dollar metacharacters. |
377 | However, if the startoffset argument of pcre_exec() is non-zero, indi- | |
495ae4b0 | 378 | cating that matching is to start at a point other than the beginning of |
8ac170f3 PH |
379 | the subject, \A can never match. The difference between \Z and \z is |
380 | that \Z matches before a newline that is the last character of the | |
381 | string as well as at the end of the string, whereas \z matches only at | |
495ae4b0 PH |
382 | the end. |
383 | ||
8ac170f3 PH |
384 | The \G assertion is true only when the current matching position is at |
385 | the start point of the match, as specified by the startoffset argument | |
386 | of pcre_exec(). It differs from \A when the value of startoffset is | |
387 | non-zero. By calling pcre_exec() multiple times with appropriate argu- | |
495ae4b0 PH |
388 | ments, you can mimic Perl's /g option, and it is in this kind of imple- |
389 | mentation where \G can be useful. | |
390 | ||
8ac170f3 | 391 | Note, however, that PCRE's interpretation of \G, as the start of the |
495ae4b0 | 392 | current match, is subtly different from Perl's, which defines it as the |
8ac170f3 PH |
393 | end of the previous match. In Perl, these can be different when the |
394 | previously matched string was empty. Because PCRE does just one match | |
495ae4b0 PH |
395 | at a time, it cannot reproduce this behaviour. |
396 | ||
8ac170f3 | 397 | If all the alternatives of a pattern begin with \G, the expression is |
495ae4b0 PH |
398 | anchored to the starting match position, and the "anchored" flag is set |
399 | in the compiled regular expression. | |
400 | ||
401 | ||
402 | CIRCUMFLEX AND DOLLAR | |
403 | ||
404 | Outside a character class, in the default matching mode, the circumflex | |
8ac170f3 PH |
405 | character is an assertion that is true only if the current matching |
406 | point is at the start of the subject string. If the startoffset argu- | |
407 | ment of pcre_exec() is non-zero, circumflex can never match if the | |
408 | PCRE_MULTILINE option is unset. Inside a character class, circumflex | |
495ae4b0 PH |
409 | has an entirely different meaning (see below). |
410 | ||
8ac170f3 PH |
411 | Circumflex need not be the first character of the pattern if a number |
412 | of alternatives are involved, but it should be the first thing in each | |
413 | alternative in which it appears if the pattern is ever to match that | |
414 | branch. If all possible alternatives start with a circumflex, that is, | |
415 | if the pattern is constrained to match only at the start of the sub- | |
416 | ject, it is said to be an "anchored" pattern. (There are also other | |
495ae4b0 PH |
417 | constructs that can cause a pattern to be anchored.) |
418 | ||
8ac170f3 PH |
419 | A dollar character is an assertion that is true only if the current |
420 | matching point is at the end of the subject string, or immediately | |
495ae4b0 | 421 | before a newline character that is the last character in the string (by |
8ac170f3 PH |
422 | default). Dollar need not be the last character of the pattern if a |
423 | number of alternatives are involved, but it should be the last item in | |
424 | any branch in which it appears. Dollar has no special meaning in a | |
495ae4b0 PH |
425 | character class. |
426 | ||
8ac170f3 PH |
427 | The meaning of dollar can be changed so that it matches only at the |
428 | very end of the string, by setting the PCRE_DOLLAR_ENDONLY option at | |
495ae4b0 PH |
429 | compile time. This does not affect the \Z assertion. |
430 | ||
431 | The meanings of the circumflex and dollar characters are changed if the | |
432 | PCRE_MULTILINE option is set. When this is the case, they match immedi- | |
8ac170f3 PH |
433 | ately after and immediately before an internal newline character, |
434 | respectively, in addition to matching at the start and end of the sub- | |
435 | ject string. For example, the pattern /^abc$/ matches the subject | |
436 | string "def\nabc" (where \n represents a newline character) in multi- | |
495ae4b0 | 437 | line mode, but not otherwise. Consequently, patterns that are anchored |
8ac170f3 PH |
438 | in single line mode because all branches start with ^ are not anchored |
439 | in multiline mode, and a match for circumflex is possible when the | |
440 | startoffset argument of pcre_exec() is non-zero. The PCRE_DOL- | |
495ae4b0 PH |
441 | LAR_ENDONLY option is ignored if PCRE_MULTILINE is set. |
442 | ||
8ac170f3 PH |
443 | Note that the sequences \A, \Z, and \z can be used to match the start |
444 | and end of the subject in both modes, and if all branches of a pattern | |
445 | start with \A it is always anchored, whether PCRE_MULTILINE is set or | |
495ae4b0 PH |
446 | not. |
447 | ||
448 | ||
449 | FULL STOP (PERIOD, DOT) | |
450 | ||
451 | Outside a character class, a dot in the pattern matches any one charac- | |
8ac170f3 PH |
452 | ter in the subject, including a non-printing character, but not (by |
453 | default) newline. In UTF-8 mode, a dot matches any UTF-8 character, | |
495ae4b0 | 454 | which might be more than one byte long, except (by default) newline. If |
8ac170f3 PH |
455 | the PCRE_DOTALL option is set, dots match newlines as well. The han- |
456 | dling of dot is entirely independent of the handling of circumflex and | |
457 | dollar, the only relationship being that they both involve newline | |
495ae4b0 PH |
458 | characters. Dot has no special meaning in a character class. |
459 | ||
460 | ||
461 | MATCHING A SINGLE BYTE | |
462 | ||
463 | Outside a character class, the escape sequence \C matches any one byte, | |
8ac170f3 PH |
464 | both in and out of UTF-8 mode. Unlike a dot, it can match a newline. |
465 | The feature is provided in Perl in order to match individual bytes in | |
466 | UTF-8 mode. Because it breaks up UTF-8 characters into individual | |
467 | bytes, what remains in the string may be a malformed UTF-8 string. For | |
495ae4b0 PH |
468 | this reason, the \C escape sequence is best avoided. |
469 | ||
8ac170f3 PH |
470 | PCRE does not allow \C to appear in lookbehind assertions (described |
471 | below), because in UTF-8 mode this would make it impossible to calcu- | |
495ae4b0 PH |
472 | late the length of the lookbehind. |
473 | ||
474 | ||
475 | SQUARE BRACKETS AND CHARACTER CLASSES | |
476 | ||
477 | An opening square bracket introduces a character class, terminated by a | |
478 | closing square bracket. A closing square bracket on its own is not spe- | |
479 | cial. If a closing square bracket is required as a member of the class, | |
8ac170f3 | 480 | it should be the first data character in the class (after an initial |
495ae4b0 PH |
481 | circumflex, if present) or escaped with a backslash. |
482 | ||
8ac170f3 PH |
483 | A character class matches a single character in the subject. In UTF-8 |
484 | mode, the character may occupy more than one byte. A matched character | |
495ae4b0 | 485 | must be in the set of characters defined by the class, unless the first |
8ac170f3 PH |
486 | character in the class definition is a circumflex, in which case the |
487 | subject character must not be in the set defined by the class. If a | |
488 | circumflex is actually required as a member of the class, ensure it is | |
495ae4b0 PH |
489 | not the first character, or escape it with a backslash. |
490 | ||
8ac170f3 PH |
491 | For example, the character class [aeiou] matches any lower case vowel, |
492 | while [^aeiou] matches any character that is not a lower case vowel. | |
495ae4b0 | 493 | Note that a circumflex is just a convenient notation for specifying the |
8ac170f3 PH |
494 | characters that are in the class by enumerating those that are not. A |
495 | class that starts with a circumflex is not an assertion: it still con- | |
496 | sumes a character from the subject string, and therefore it fails if | |
495ae4b0 PH |
497 | the current pointer is at the end of the string. |
498 | ||
8ac170f3 PH |
499 | In UTF-8 mode, characters with values greater than 255 can be included |
500 | in a class as a literal string of bytes, or by using the \x{ escaping | |
495ae4b0 PH |
501 | mechanism. |
502 | ||
8ac170f3 PH |
503 | When caseless matching is set, any letters in a class represent both |
504 | their upper case and lower case versions, so for example, a caseless | |
505 | [aeiou] matches "A" as well as "a", and a caseless [^aeiou] does not | |
506 | match "A", whereas a caseful version would. In UTF-8 mode, PCRE always | |
507 | understands the concept of case for characters whose values are less | |
508 | than 128, so caseless matching is always possible. For characters with | |
509 | higher values, the concept of case is supported if PCRE is compiled | |
510 | with Unicode property support, but not otherwise. If you want to use | |
511 | caseless matching for characters 128 and above, you must ensure that | |
512 | PCRE is compiled with Unicode property support as well as with UTF-8 | |
513 | support. | |
495ae4b0 PH |
514 | |
515 | The newline character is never treated in any special way in character | |
516 | classes, whatever the setting of the PCRE_DOTALL or PCRE_MULTILINE | |
517 | options is. A class such as [^a] will always match a newline. | |
518 | ||
519 | The minus (hyphen) character can be used to specify a range of charac- | |
520 | ters in a character class. For example, [d-m] matches any letter | |
521 | between d and m, inclusive. If a minus character is required in a | |
522 | class, it must be escaped with a backslash or appear in a position | |
523 | where it cannot be interpreted as indicating a range, typically as the | |
524 | first or last character in the class. | |
525 | ||
526 | It is not possible to have the literal character "]" as the end charac- | |
527 | ter of a range. A pattern such as [W-]46] is interpreted as a class of | |
528 | two characters ("W" and "-") followed by a literal string "46]", so it | |
529 | would match "W46]" or "-46]". However, if the "]" is escaped with a | |
530 | backslash it is interpreted as the end of range, so [W-\]46] is inter- | |
531 | preted as a class containing a range followed by two other characters. | |
532 | The octal or hexadecimal representation of "]" can also be used to end | |
533 | a range. | |
534 | ||
535 | Ranges operate in the collating sequence of character values. They can | |
536 | also be used for characters specified numerically, for example | |
537 | [\000-\037]. In UTF-8 mode, ranges can include characters whose values | |
538 | are greater than 255, for example [\x{100}-\x{2ff}]. | |
539 | ||
540 | If a range that includes letters is used when caseless matching is set, | |
541 | it matches the letters in either case. For example, [W-c] is equivalent | |
542 | to [][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if | |
543 | character tables for the "fr_FR" locale are in use, [\xc8-\xcb] matches | |
544 | accented E characters in both cases. In UTF-8 mode, PCRE supports the | |
545 | concept of case for characters with values greater than 128 only when | |
546 | it is compiled with Unicode property support. | |
547 | ||
548 | The character types \d, \D, \p, \P, \s, \S, \w, and \W may also appear | |
549 | in a character class, and add the characters that they match to the | |
550 | class. For example, [\dABCDEF] matches any hexadecimal digit. A circum- | |
551 | flex can conveniently be used with the upper case character types to | |
552 | specify a more restricted set of characters than the matching lower | |
553 | case type. For example, the class [^\W_] matches any letter or digit, | |
554 | but not underscore. | |
555 | ||
556 | The only metacharacters that are recognized in character classes are | |
557 | backslash, hyphen (only where it can be interpreted as specifying a | |
558 | range), circumflex (only at the start), opening square bracket (only | |
559 | when it can be interpreted as introducing a POSIX class name - see the | |
560 | next section), and the terminating closing square bracket. However, | |
561 | escaping other non-alphanumeric characters does no harm. | |
562 | ||
563 | ||
564 | POSIX CHARACTER CLASSES | |
565 | ||
566 | Perl supports the POSIX notation for character classes. This uses names | |
567 | enclosed by [: and :] within the enclosing square brackets. PCRE also | |
568 | supports this notation. For example, | |
569 | ||
570 | [01[:alpha:]%] | |
571 | ||
572 | matches "0", "1", any alphabetic character, or "%". The supported class | |
573 | names are | |
574 | ||
575 | alnum letters and digits | |
576 | alpha letters | |
577 | ascii character codes 0 - 127 | |
578 | blank space or tab only | |
579 | cntrl control characters | |
580 | digit decimal digits (same as \d) | |
581 | graph printing characters, excluding space | |
582 | lower lower case letters | |
583 | print printing characters, including space | |
584 | punct printing characters, excluding letters and digits | |
585 | space white space (not quite the same as \s) | |
586 | upper upper case letters | |
587 | word "word" characters (same as \w) | |
588 | xdigit hexadecimal digits | |
589 | ||
590 | The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), | |
591 | and space (32). Notice that this list includes the VT character (code | |
592 | 11). This makes "space" different to \s, which does not include VT (for | |
593 | Perl compatibility). | |
594 | ||
595 | The name "word" is a Perl extension, and "blank" is a GNU extension | |
596 | from Perl 5.8. Another Perl extension is negation, which is indicated | |
597 | by a ^ character after the colon. For example, | |
598 | ||
599 | [12[:^digit:]] | |
600 | ||
601 | matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the | |
602 | POSIX syntax [.ch.] and [=ch=] where "ch" is a "collating element", but | |
603 | these are not supported, and an error is given if they are encountered. | |
604 | ||
605 | In UTF-8 mode, characters with values greater than 128 do not match any | |
606 | of the POSIX character classes. | |
607 | ||
608 | ||
609 | VERTICAL BAR | |
610 | ||
611 | Vertical bar characters are used to separate alternative patterns. For | |
612 | example, the pattern | |
613 | ||
614 | gilbert|sullivan | |
615 | ||
616 | matches either "gilbert" or "sullivan". Any number of alternatives may | |
617 | appear, and an empty alternative is permitted (matching the empty | |
618 | string). The matching process tries each alternative in turn, from | |
619 | left to right, and the first one that succeeds is used. If the alterna- | |
620 | tives are within a subpattern (defined below), "succeeds" means match- | |
621 | ing the rest of the main pattern as well as the alternative in the sub- | |
622 | pattern. | |
623 | ||
624 | ||
625 | INTERNAL OPTION SETTING | |
626 | ||
627 | The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and | |
628 | PCRE_EXTENDED options can be changed from within the pattern by a | |
629 | sequence of Perl option letters enclosed between "(?" and ")". The | |
630 | option letters are | |
631 | ||
632 | i for PCRE_CASELESS | |
633 | m for PCRE_MULTILINE | |
634 | s for PCRE_DOTALL | |
635 | x for PCRE_EXTENDED | |
636 | ||
637 | For example, (?im) sets caseless, multiline matching. It is also possi- | |
638 | ble to unset these options by preceding the letter with a hyphen, and a | |
639 | combined setting and unsetting such as (?im-sx), which sets PCRE_CASE- | |
640 | LESS and PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, | |
641 | is also permitted. If a letter appears both before and after the | |
642 | hyphen, the option is unset. | |
643 | ||
644 | When an option change occurs at top level (that is, not inside subpat- | |
645 | tern parentheses), the change applies to the remainder of the pattern | |
646 | that follows. If the change is placed right at the start of a pattern, | |
647 | PCRE extracts it into the global options (and it will therefore show up | |
648 | in data extracted by the pcre_fullinfo() function). | |
649 | ||
650 | An option change within a subpattern affects only that part of the cur- | |
651 | rent pattern that follows it, so | |
652 | ||
653 | (a(?i)b)c | |
654 | ||
655 | matches abc and aBc and no other strings (assuming PCRE_CASELESS is not | |
656 | used). By this means, options can be made to have different settings | |
657 | in different parts of the pattern. Any changes made in one alternative | |
658 | do carry on into subsequent branches within the same subpattern. For | |
659 | example, | |
660 | ||
661 | (a(?i)b|c) | |
662 | ||
663 | matches "ab", "aB", "c", and "C", even though when matching "C" the | |
664 | first branch is abandoned before the option setting. This is because | |
665 | the effects of option settings happen at compile time. There would be | |
666 | some very weird behaviour otherwise. | |
667 | ||
668 | The PCRE-specific options PCRE_UNGREEDY and PCRE_EXTRA can be changed | |
669 | in the same way as the Perl-compatible options by using the characters | |
670 | U and X respectively. The (?X) flag setting is special in that it must | |
671 | always occur earlier in the pattern than any of the additional features | |
672 | it turns on, even when it is at top level. It is best to put it at the | |
673 | start. | |
674 | ||
675 | ||
676 | SUBPATTERNS | |
677 | ||
678 | Subpatterns are delimited by parentheses (round brackets), which can be | |
679 | nested. Turning part of a pattern into a subpattern does two things: | |
680 | ||
681 | 1. It localizes a set of alternatives. For example, the pattern | |
682 | ||
683 | cat(aract|erpillar|) | |
684 | ||
685 | matches one of the words "cat", "cataract", or "caterpillar". Without | |
686 | the parentheses, it would match "cataract", "erpillar" or the empty | |
687 | string. | |
688 | ||
689 | 2. It sets up the subpattern as a capturing subpattern. This means | |
690 | that, when the whole pattern matches, that portion of the subject | |
691 | string that matched the subpattern is passed back to the caller via the | |
692 | ovector argument of pcre_exec(). Opening parentheses are counted from | |
693 | left to right (starting from 1) to obtain numbers for the capturing | |
694 | subpatterns. | |
695 | ||
696 | For example, if the string "the red king" is matched against the pat- | |
697 | tern | |
698 | ||
699 | the ((red|white) (king|queen)) | |
700 | ||
701 | the captured substrings are "red king", "red", and "king", and are num- | |
702 | bered 1, 2, and 3, respectively. | |
703 | ||
704 | The fact that plain parentheses fulfil two functions is not always | |
705 | helpful. There are often times when a grouping subpattern is required | |
706 | without a capturing requirement. If an opening parenthesis is followed | |
707 | by a question mark and a colon, the subpattern does not do any captur- | |
708 | ing, and is not counted when computing the number of any subsequent | |
709 | capturing subpatterns. For example, if the string "the white queen" is | |
710 | matched against the pattern | |
711 | ||
712 | the ((?:red|white) (king|queen)) | |
713 | ||
714 | the captured substrings are "white queen" and "queen", and are numbered | |
715 | 1 and 2. The maximum number of capturing subpatterns is 65535, and the | |
716 | maximum depth of nesting of all subpatterns, both capturing and non- | |
717 | capturing, is 200. | |
718 | ||
719 | As a convenient shorthand, if any option settings are required at the | |
720 | start of a non-capturing subpattern, the option letters may appear | |
721 | between the "?" and the ":". Thus the two patterns | |
722 | ||
723 | (?i:saturday|sunday) | |
724 | (?:(?i)saturday|sunday) | |
725 | ||
726 | match exactly the same set of strings. Because alternative branches are | |
727 | tried from left to right, and options are not reset until the end of | |
728 | the subpattern is reached, an option setting in one branch does affect | |
729 | subsequent branches, so the above patterns match "SUNDAY" as well as | |
730 | "Saturday". | |
731 | ||
732 | ||
733 | NAMED SUBPATTERNS | |
734 | ||
735 | Identifying capturing parentheses by number is simple, but it can be | |
736 | very hard to keep track of the numbers in complicated regular expres- | |
737 | sions. Furthermore, if an expression is modified, the numbers may | |
738 | change. To help with this difficulty, PCRE supports the naming of sub- | |
739 | patterns, something that Perl does not provide. The Python syntax | |
740 | (?P<name>...) is used. Names consist of alphanumeric characters and | |
741 | underscores, and must be unique within a pattern. | |
742 | ||
743 | Named capturing parentheses are still allocated numbers as well as | |
744 | names. The PCRE API provides function calls for extracting the name-to- | |
745 | number translation table from a compiled pattern. There is also a con- | |
746 | venience function for extracting a captured substring by name. For fur- | |
747 | ther details see the pcreapi documentation. | |
748 | ||
749 | ||
750 | REPETITION | |
751 | ||
752 | Repetition is specified by quantifiers, which can follow any of the | |
753 | following items: | |
754 | ||
755 | a literal data character | |
756 | the . metacharacter | |
757 | the \C escape sequence | |
758 | the \X escape sequence (in UTF-8 mode with Unicode properties) | |
759 | an escape such as \d that matches a single character | |
760 | a character class | |
761 | a back reference (see next section) | |
762 | a parenthesized subpattern (unless it is an assertion) | |
763 | ||
764 | The general repetition quantifier specifies a minimum and maximum num- | |
765 | ber of permitted matches, by giving the two numbers in curly brackets | |
766 | (braces), separated by a comma. The numbers must be less than 65536, | |
767 | and the first must be less than or equal to the second. For example: | |
768 | ||
769 | z{2,4} | |
770 | ||
771 | matches "zz", "zzz", or "zzzz". A closing brace on its own is not a | |
772 | special character. If the second number is omitted, but the comma is | |
773 | present, there is no upper limit; if the second number and the comma | |
774 | are both omitted, the quantifier specifies an exact number of required | |
775 | matches. Thus | |
776 | ||
777 | [aeiou]{3,} | |
778 | ||
779 | matches at least 3 successive vowels, but may match many more, while | |
780 | ||
781 | \d{8} | |
782 | ||
783 | matches exactly 8 digits. An opening curly bracket that appears in a | |
784 | position where a quantifier is not allowed, or one that does not match | |
785 | the syntax of a quantifier, is taken as a literal character. For exam- | |
786 | ple, {,6} is not a quantifier, but a literal string of four characters. | |
787 | ||
788 | In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to | |
789 | individual bytes. Thus, for example, \x{100}{2} matches two UTF-8 char- | |
790 | acters, each of which is represented by a two-byte sequence. Similarly, | |
791 | when Unicode property support is available, \X{3} matches three Unicode | |
792 | extended sequences, each of which may be several bytes long (and they | |
793 | may be of different lengths). | |
794 | ||
795 | The quantifier {0} is permitted, causing the expression to behave as if | |
796 | the previous item and the quantifier were not present. | |
797 | ||
798 | For convenience (and historical compatibility) the three most common | |
799 | quantifiers have single-character abbreviations: | |
800 | ||
801 | * is equivalent to {0,} | |
802 | + is equivalent to {1,} | |
803 | ? is equivalent to {0,1} | |
804 | ||
805 | It is possible to construct infinite loops by following a subpattern | |
806 | that can match no characters with a quantifier that has no upper limit, | |
807 | for example: | |
808 | ||
809 | (a?)* | |
810 | ||
811 | Earlier versions of Perl and PCRE used to give an error at compile time | |
812 | for such patterns. However, because there are cases where this can be | |
813 | useful, such patterns are now accepted, but if any repetition of the | |
814 | subpattern does in fact match no characters, the loop is forcibly bro- | |
815 | ken. | |
816 | ||
817 | By default, the quantifiers are "greedy", that is, they match as much | |
818 | as possible (up to the maximum number of permitted times), without | |
819 | causing the rest of the pattern to fail. The classic example of where | |
820 | this gives problems is in trying to match comments in C programs. These | |
821 | appear between /* and */ and within the comment, individual * and / | |
822 | characters may appear. An attempt to match C comments by applying the | |
823 | pattern | |
824 | ||
825 | /\*.*\*/ | |
826 | ||
827 | to the string | |
828 | ||
829 | /* first comment */ not comment /* second comment */ | |
830 | ||
831 | fails, because it matches the entire string owing to the greediness of | |
832 | the .* item. | |
833 | ||
834 | However, if a quantifier is followed by a question mark, it ceases to | |
835 | be greedy, and instead matches the minimum number of times possible, so | |
836 | the pattern | |
837 | ||
838 | /\*.*?\*/ | |
839 | ||
840 | does the right thing with the C comments. The meaning of the various | |
841 | quantifiers is not otherwise changed, just the preferred number of | |
842 | matches. Do not confuse this use of question mark with its use as a | |
843 | quantifier in its own right. Because it has two uses, it can sometimes | |
844 | appear doubled, as in | |
845 | ||
846 | \d??\d | |
847 | ||
848 | which matches one digit by preference, but can match two if that is the | |
849 | only way the rest of the pattern matches. | |
850 | ||
851 | If the PCRE_UNGREEDY option is set (an option which is not available in | |
852 | Perl), the quantifiers are not greedy by default, but individual ones | |
853 | can be made greedy by following them with a question mark. In other | |
854 | words, it inverts the default behaviour. | |
855 | ||
856 | When a parenthesized subpattern is quantified with a minimum repeat | |
857 | count that is greater than 1 or with a limited maximum, more memory is | |
858 | required for the compiled pattern, in proportion to the size of the | |
859 | minimum or maximum. | |
860 | ||
861 | If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equiv- | |
862 | alent to Perl's /s) is set, thus allowing the . to match newlines, the | |
863 | pattern is implicitly anchored, because whatever follows will be tried | |
864 | against every character position in the subject string, so there is no | |
865 | point in retrying the overall match at any position after the first. | |
866 | PCRE normally treats such a pattern as though it were preceded by \A. | |
867 | ||
868 | In cases where it is known that the subject string contains no new- | |
869 | lines, it is worth setting PCRE_DOTALL in order to obtain this opti- | |
870 | mization, or alternatively using ^ to indicate anchoring explicitly. | |
871 | ||
872 | However, there is one situation where the optimization cannot be used. | |
873 | When .* is inside capturing parentheses that are the subject of a | |
874 | backreference elsewhere in the pattern, a match at the start may fail, | |
875 | and a later one succeed. Consider, for example: | |
876 | ||
877 | (.*)abc\1 | |
878 | ||
879 | If the subject is "xyz123abc123" the match point is the fourth charac- | |
880 | ter. For this reason, such a pattern is not implicitly anchored. | |
881 | ||
882 | When a capturing subpattern is repeated, the value captured is the sub- | |
883 | string that matched the final iteration. For example, after | |
884 | ||
885 | (tweedle[dume]{3}\s*)+ | |
886 | ||
887 | has matched "tweedledum tweedledee" the value of the captured substring | |
888 | is "tweedledee". However, if there are nested capturing subpatterns, | |
889 | the corresponding captured values may have been set in previous itera- | |
890 | tions. For example, after | |
891 | ||
892 | /(a|(b))+/ | |
893 | ||
894 | matches "aba" the value of the second captured substring is "b". | |
895 | ||
896 | ||
897 | ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS | |
898 | ||
899 | With both maximizing and minimizing repetition, failure of what follows | |
900 | normally causes the repeated item to be re-evaluated to see if a dif- | |
901 | ferent number of repeats allows the rest of the pattern to match. Some- | |
902 | times it is useful to prevent this, either to change the nature of the | |
903 | match, or to cause it fail earlier than it otherwise might, when the | |
904 | author of the pattern knows there is no point in carrying on. | |
905 | ||
906 | Consider, for example, the pattern \d+foo when applied to the subject | |
907 | line | |
908 | ||
909 | 123456bar | |
910 | ||
911 | After matching all 6 digits and then failing to match "foo", the normal | |
912 | action of the matcher is to try again with only 5 digits matching the | |
913 | \d+ item, and then with 4, and so on, before ultimately failing. | |
914 | "Atomic grouping" (a term taken from Jeffrey Friedl's book) provides | |
915 | the means for specifying that once a subpattern has matched, it is not | |
916 | to be re-evaluated in this way. | |
917 | ||
918 | If we use atomic grouping for the previous example, the matcher would | |
919 | give up immediately on failing to match "foo" the first time. The nota- | |
920 | tion is a kind of special parenthesis, starting with (?> as in this | |
921 | example: | |
922 | ||
923 | (?>\d+)foo | |
924 | ||
925 | This kind of parenthesis "locks up" the part of the pattern it con- | |
926 | tains once it has matched, and a failure further into the pattern is | |
927 | prevented from backtracking into it. Backtracking past it to previous | |
928 | items, however, works as normal. | |
929 | ||
930 | An alternative description is that a subpattern of this type matches | |
931 | the string of characters that an identical standalone pattern would | |
932 | match, if anchored at the current point in the subject string. | |
933 | ||
934 | Atomic grouping subpatterns are not capturing subpatterns. Simple cases | |
935 | such as the above example can be thought of as a maximizing repeat that | |
936 | must swallow everything it can. So, while both \d+ and \d+? are pre- | |
937 | pared to adjust the number of digits they match in order to make the | |
938 | rest of the pattern match, (?>\d+) can only match an entire sequence of | |
939 | digits. | |
940 | ||
941 | Atomic groups in general can of course contain arbitrarily complicated | |
942 | subpatterns, and can be nested. However, when the subpattern for an | |
943 | atomic group is just a single repeated item, as in the example above, a | |
944 | simpler notation, called a "possessive quantifier" can be used. This | |
945 | consists of an additional + character following a quantifier. Using | |
946 | this notation, the previous example can be rewritten as | |
947 | ||
948 | \d++foo | |
949 | ||
950 | Possessive quantifiers are always greedy; the setting of the | |
951 | PCRE_UNGREEDY option is ignored. They are a convenient notation for the | |
952 | simpler forms of atomic group. However, there is no difference in the | |
953 | meaning or processing of a possessive quantifier and the equivalent | |
954 | atomic group. | |
955 | ||
956 | The possessive quantifier syntax is an extension to the Perl syntax. It | |
957 | originates in Sun's Java package. | |
958 | ||
959 | When a pattern contains an unlimited repeat inside a subpattern that | |
960 | can itself be repeated an unlimited number of times, the use of an | |
961 | atomic group is the only way to avoid some failing matches taking a | |
962 | very long time indeed. The pattern | |
963 | ||
964 | (\D+|<\d+>)*[!?] | |
965 | ||
966 | matches an unlimited number of substrings that either consist of non- | |
967 | digits, or digits enclosed in <>, followed by either ! or ?. When it | |
968 | matches, it runs quickly. However, if it is applied to | |
969 | ||
970 | aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa | |
971 | ||
972 | it takes a long time before reporting failure. This is because the | |
973 | string can be divided between the internal \D+ repeat and the external | |
974 | * repeat in a large number of ways, and all have to be tried. (The | |
975 | example uses [!?] rather than a single character at the end, because | |
976 | both PCRE and Perl have an optimization that allows for fast failure | |
977 | when a single character is used. They remember the last single charac- | |
978 | ter that is required for a match, and fail early if it is not present | |
979 | in the string.) If the pattern is changed so that it uses an atomic | |
980 | group, like this: | |
981 | ||
982 | ((?>\D+)|<\d+>)*[!?] | |
983 | ||
984 | sequences of non-digits cannot be broken, and failure happens quickly. | |
985 | ||
986 | ||
987 | BACK REFERENCES | |
988 | ||
989 | Outside a character class, a backslash followed by a digit greater than | |
990 | 0 (and possibly further digits) is a back reference to a capturing sub- | |
991 | pattern earlier (that is, to its left) in the pattern, provided there | |
992 | have been that many previous capturing left parentheses. | |
993 | ||
994 | However, if the decimal number following the backslash is less than 10, | |
995 | it is always taken as a back reference, and causes an error only if | |
996 | there are not that many capturing left parentheses in the entire pat- | |
997 | tern. In other words, the parentheses that are referenced need not be | |
998 | to the left of the reference for numbers less than 10. See the subsec- | |
999 | tion entitled "Non-printing characters" above for further details of | |
1000 | the handling of digits following a backslash. | |
1001 | ||
1002 | A back reference matches whatever actually matched the capturing sub- | |
1003 | pattern in the current subject string, rather than anything matching | |
1004 | the subpattern itself (see "Subpatterns as subroutines" below for a way | |
1005 | of doing that). So the pattern | |
1006 | ||
1007 | (sens|respons)e and \1ibility | |
1008 | ||
1009 | matches "sense and sensibility" and "response and responsibility", but | |
1010 | not "sense and responsibility". If caseful matching is in force at the | |
1011 | time of the back reference, the case of letters is relevant. For exam- | |
1012 | ple, | |
1013 | ||
1014 | ((?i)rah)\s+\1 | |
1015 | ||
1016 | matches "rah rah" and "RAH RAH", but not "RAH rah", even though the | |
1017 | original capturing subpattern is matched caselessly. | |
1018 | ||
1019 | Back references to named subpatterns use the Python syntax (?P=name). | |
1020 | We could rewrite the above example as follows: | |
1021 | ||
1022 | (?<p1>(?i)rah)\s+(?P=p1) | |
1023 | ||
1024 | There may be more than one back reference to the same subpattern. If a | |
1025 | subpattern has not actually been used in a particular match, any back | |
1026 | references to it always fail. For example, the pattern | |
1027 | ||
1028 | (a|(bc))\2 | |
1029 | ||
1030 | always fails if it starts to match "a" rather than "bc". Because there | |
1031 | may be many capturing parentheses in a pattern, all digits following | |
1032 | the backslash are taken as part of a potential back reference number. | |
1033 | If the pattern continues with a digit character, some delimiter must be | |
1034 | used to terminate the back reference. If the PCRE_EXTENDED option is | |
1035 | set, this can be whitespace. Otherwise an empty comment (see "Com- | |
1036 | ments" below) can be used. | |
1037 | ||
1038 | A back reference that occurs inside the parentheses to which it refers | |
1039 | fails when the subpattern is first used, so, for example, (a\1) never | |
1040 | matches. However, such references can be useful inside repeated sub- | |
1041 | patterns. For example, the pattern | |
1042 | ||
1043 | (a|b\1)+ | |
1044 | ||
1045 | matches any number of "a"s and also "aba", "ababbaa" etc. At each iter- | |
1046 | ation of the subpattern, the back reference matches the character | |
1047 | string corresponding to the previous iteration. In order for this to | |
1048 | work, the pattern must be such that the first iteration does not need | |
1049 | to match the back reference. This can be done using alternation, as in | |
1050 | the example above, or by a quantifier with a minimum of zero. | |
1051 | ||
1052 | ||
1053 | ASSERTIONS | |
1054 | ||
1055 | An assertion is a test on the characters following or preceding the | |
1056 | current matching point that does not actually consume any characters. | |
1057 | The simple assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are | |
1058 | described above. | |
1059 | ||
1060 | More complicated assertions are coded as subpatterns. There are two | |
1061 | kinds: those that look ahead of the current position in the subject | |
1062 | string, and those that look behind it. An assertion subpattern is | |
1063 | matched in the normal way, except that it does not cause the current | |
1064 | matching position to be changed. | |
1065 | ||
1066 | Assertion subpatterns are not capturing subpatterns, and may not be | |
1067 | repeated, because it makes no sense to assert the same thing several | |
1068 | times. If any kind of assertion contains capturing subpatterns within | |
1069 | it, these are counted for the purposes of numbering the capturing sub- | |
1070 | patterns in the whole pattern. However, substring capturing is carried | |
1071 | out only for positive assertions, because it does not make sense for | |
1072 | negative assertions. | |
1073 | ||
1074 | Lookahead assertions | |
1075 | ||
1076 | Lookahead assertions start with (?= for positive assertions and (?! for | |
1077 | negative assertions. For example, | |
1078 | ||
1079 | \w+(?=;) | |
1080 | ||
1081 | matches a word followed by a semicolon, but does not include the semi- | |
1082 | colon in the match, and | |
1083 | ||
1084 | foo(?!bar) | |
1085 | ||
1086 | matches any occurrence of "foo" that is not followed by "bar". Note | |
1087 | that the apparently similar pattern | |
1088 | ||
1089 | (?!foo)bar | |
1090 | ||
1091 | does not find an occurrence of "bar" that is preceded by something | |
1092 | other than "foo"; it finds any occurrence of "bar" whatsoever, because | |
1093 | the assertion (?!foo) is always true when the next three characters are | |
1094 | "bar". A lookbehind assertion is needed to achieve the other effect. | |
1095 | ||
1096 | If you want to force a matching failure at some point in a pattern, the | |
1097 | most convenient way to do it is with (?!) because an empty string | |
1098 | always matches, so an assertion that requires there not to be an empty | |
1099 | string must always fail. | |
1100 | ||
1101 | Lookbehind assertions | |
1102 | ||
1103 | Lookbehind assertions start with (?<= for positive assertions and (?<! | |
1104 | for negative assertions. For example, | |
1105 | ||
1106 | (?<!foo)bar | |
1107 | ||
1108 | does find an occurrence of "bar" that is not preceded by "foo". The | |
1109 | contents of a lookbehind assertion are restricted such that all the | |
1110 | strings it matches must have a fixed length. However, if there are sev- | |
1111 | eral alternatives, they do not all have to have the same fixed length. | |
1112 | Thus | |
1113 | ||
1114 | (?<=bullock|donkey) | |
1115 | ||
1116 | is permitted, but | |
1117 | ||
1118 | (?<!dogs?|cats?) | |
1119 | ||
1120 | causes an error at compile time. Branches that match different length | |
1121 | strings are permitted only at the top level of a lookbehind assertion. | |
1122 | This is an extension compared with Perl (at least for 5.8), which | |
1123 | requires all branches to match the same length of string. An assertion | |
1124 | such as | |
1125 | ||
1126 | (?<=ab(c|de)) | |
1127 | ||
1128 | is not permitted, because its single top-level branch can match two | |
1129 | different lengths, but it is acceptable if rewritten to use two top- | |
1130 | level branches: | |
1131 | ||
1132 | (?<=abc|abde) | |
1133 | ||
1134 | The implementation of lookbehind assertions is, for each alternative, | |
1135 | to temporarily move the current position back by the fixed width and | |
1136 | then try to match. If there are insufficient characters before the cur- | |
1137 | rent position, the match is deemed to fail. | |
1138 | ||
1139 | PCRE does not allow the \C escape (which matches a single byte in UTF-8 | |
1140 | mode) to appear in lookbehind assertions, because it makes it impossi- | |
1141 | ble to calculate the length of the lookbehind. The \X escape, which can | |
1142 | match different numbers of bytes, is also not permitted. | |
1143 | ||
1144 | Atomic groups can be used in conjunction with lookbehind assertions to | |
1145 | specify efficient matching at the end of the subject string. Consider a | |
1146 | simple pattern such as | |
1147 | ||
1148 | abcd$ | |
1149 | ||
1150 | when applied to a long string that does not match. Because matching | |
1151 | proceeds from left to right, PCRE will look for each "a" in the subject | |
1152 | and then see if what follows matches the rest of the pattern. If the | |
1153 | pattern is specified as | |
1154 | ||
1155 | ^.*abcd$ | |
1156 | ||
1157 | the initial .* matches the entire string at first, but when this fails | |
1158 | (because there is no following "a"), it backtracks to match all but the | |
1159 | last character, then all but the last two characters, and so on. Once | |
1160 | again the search for "a" covers the entire string, from right to left, | |
1161 | so we are no better off. However, if the pattern is written as | |
1162 | ||
1163 | ^(?>.*)(?<=abcd) | |
1164 | ||
1165 | or, equivalently, using the possessive quantifier syntax, | |
1166 | ||
1167 | ^.*+(?<=abcd) | |
1168 | ||
1169 | there can be no backtracking for the .* item; it can match only the | |
1170 | entire string. The subsequent lookbehind assertion does a single test | |
1171 | on the last four characters. If it fails, the match fails immediately. | |
1172 | For long strings, this approach makes a significant difference to the | |
1173 | processing time. | |
1174 | ||
1175 | Using multiple assertions | |
1176 | ||
1177 | Several assertions (of any sort) may occur in succession. For example, | |
1178 | ||
1179 | (?<=\d{3})(?<!999)foo | |
1180 | ||
1181 | matches "foo" preceded by three digits that are not "999". Notice that | |
1182 | each of the assertions is applied independently at the same point in | |
1183 | the subject string. First there is a check that the previous three | |
1184 | characters are all digits, and then there is a check that the same | |
1185 | three characters are not "999". This pattern does not match "foo" pre- | |
1186 | ceded by six characters, the first of which are digits and the last | |
1187 | three of which are not "999". For example, it doesn't match "123abc- | |
1188 | foo". A pattern to do that is | |
1189 | ||
1190 | (?<=\d{3}...)(?<!999)foo | |
1191 | ||
1192 | This time the first assertion looks at the preceding six characters, | |
1193 | checking that the first three are digits, and then the second assertion | |
1194 | checks that the preceding three characters are not "999". | |
1195 | ||
1196 | Assertions can be nested in any combination. For example, | |
1197 | ||
1198 | (?<=(?<!foo)bar)baz | |
1199 | ||
1200 | matches an occurrence of "baz" that is preceded by "bar" which in turn | |
1201 | is not preceded by "foo", while | |
1202 | ||
1203 | (?<=\d{3}(?!999)...)foo | |
1204 | ||
1205 | is another pattern that matches "foo" preceded by three digits and any | |
1206 | three characters that are not "999". | |
1207 | ||
1208 | ||
1209 | CONDITIONAL SUBPATTERNS | |
1210 | ||
1211 | It is possible to cause the matching process to obey a subpattern con- | |
1212 | ditionally or to choose between two alternative subpatterns, depending | |
1213 | on the result of an assertion, or whether a previous capturing subpat- | |
1214 | tern matched or not. The two possible forms of conditional subpattern | |
1215 | are | |
1216 | ||
1217 | (?(condition)yes-pattern) | |
1218 | (?(condition)yes-pattern|no-pattern) | |
1219 | ||
1220 | If the condition is satisfied, the yes-pattern is used; otherwise the | |
1221 | no-pattern (if present) is used. If there are more than two alterna- | |
1222 | tives in the subpattern, a compile-time error occurs. | |
1223 | ||
1224 | There are three kinds of condition. If the text between the parentheses | |
1225 | consists of a sequence of digits, the condition is satisfied if the | |
1226 | capturing subpattern of that number has previously matched. The number | |
1227 | must be greater than zero. Consider the following pattern, which con- | |
1228 | tains non-significant white space to make it more readable (assume the | |
1229 | PCRE_EXTENDED option) and to divide it into three parts for ease of | |
1230 | discussion: | |
1231 | ||
1232 | ( \( )? [^()]+ (?(1) \) ) | |
1233 | ||
1234 | The first part matches an optional opening parenthesis, and if that | |
1235 | character is present, sets it as the first captured substring. The sec- | |
1236 | ond part matches one or more characters that are not parentheses. The | |
1237 | third part is a conditional subpattern that tests whether the first set | |
1238 | of parentheses matched or not. If they did, that is, if subject started | |
1239 | with an opening parenthesis, the condition is true, and so the yes-pat- | |
1240 | tern is executed and a closing parenthesis is required. Otherwise, | |
1241 | since no-pattern is not present, the subpattern matches nothing. In | |
1242 | other words, this pattern matches a sequence of non-parentheses, | |
1243 | optionally enclosed in parentheses. | |
1244 | ||
1245 | If the condition is the string (R), it is satisfied if a recursive call | |
1246 | to the pattern or subpattern has been made. At "top level", the condi- | |
1247 | tion is false. This is a PCRE extension. Recursive patterns are | |
1248 | described in the next section. | |
1249 | ||
1250 | If the condition is not a sequence of digits or (R), it must be an | |
1251 | assertion. This may be a positive or negative lookahead or lookbehind | |
1252 | assertion. Consider this pattern, again containing non-significant | |
1253 | white space, and with the two alternatives on the second line: | |
1254 | ||
1255 | (?(?=[^a-z]*[a-z]) | |
1256 | \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} ) | |
1257 | ||
1258 | The condition is a positive lookahead assertion that matches an | |
1259 | optional sequence of non-letters followed by a letter. In other words, | |
1260 | it tests for the presence of at least one letter in the subject. If a | |
1261 | letter is found, the subject is matched against the first alternative; | |
1262 | otherwise it is matched against the second. This pattern matches | |
1263 | strings in one of the two forms dd-aaa-dd or dd-dd-dd, where aaa are | |
1264 | letters and dd are digits. | |
1265 | ||
1266 | ||
1267 | COMMENTS | |
1268 | ||
1269 | The sequence (?# marks the start of a comment that continues up to the | |
1270 | next closing parenthesis. Nested parentheses are not permitted. The | |
1271 | characters that make up a comment play no part in the pattern matching | |
1272 | at all. | |
1273 | ||
1274 | If the PCRE_EXTENDED option is set, an unescaped # character outside a | |
1275 | character class introduces a comment that continues up to the next new- | |
1276 | line character in the pattern. | |
1277 | ||
1278 | ||
1279 | RECURSIVE PATTERNS | |
1280 | ||
1281 | Consider the problem of matching a string in parentheses, allowing for | |
1282 | unlimited nested parentheses. Without the use of recursion, the best | |
1283 | that can be done is to use a pattern that matches up to some fixed | |
1284 | depth of nesting. It is not possible to handle an arbitrary nesting | |
1285 | depth. Perl provides a facility that allows regular expressions to | |
1286 | recurse (amongst other things). It does this by interpolating Perl code | |
1287 | in the expression at run time, and the code can refer to the expression | |
1288 | itself. A Perl pattern to solve the parentheses problem can be created | |
1289 | like this: | |
1290 | ||
1291 | $re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x; | |
1292 | ||
1293 | The (?p{...}) item interpolates Perl code at run time, and in this case | |
1294 | refers recursively to the pattern in which it appears. Obviously, PCRE | |
1295 | cannot support the interpolation of Perl code. Instead, it supports | |
1296 | some special syntax for recursion of the entire pattern, and also for | |
1297 | individual subpattern recursion. | |
1298 | ||
1299 | The special item that consists of (? followed by a number greater than | |
1300 | zero and a closing parenthesis is a recursive call of the subpattern of | |
1301 | the given number, provided that it occurs inside that subpattern. (If | |
1302 | not, it is a "subroutine" call, which is described in the next sec- | |
1303 | tion.) The special item (?R) is a recursive call of the entire regular | |
1304 | expression. | |
1305 | ||
1306 | For example, this PCRE pattern solves the nested parentheses problem | |
1307 | (assume the PCRE_EXTENDED option is set so that white space is | |
1308 | ignored): | |
1309 | ||
1310 | \( ( (?>[^()]+) | (?R) )* \) | |
1311 | ||
1312 | First it matches an opening parenthesis. Then it matches any number of | |
1313 | substrings which can either be a sequence of non-parentheses, or a | |
1314 | recursive match of the pattern itself (that is a correctly parenthe- | |
1315 | sized substring). Finally there is a closing parenthesis. | |
1316 | ||
1317 | If this were part of a larger pattern, you would not want to recurse | |
1318 | the entire pattern, so instead you could use this: | |
1319 | ||
1320 | ( \( ( (?>[^()]+) | (?1) )* \) ) | |
1321 | ||
1322 | We have put the pattern into parentheses, and caused the recursion to | |
1323 | refer to them instead of the whole pattern. In a larger pattern, keep- | |
1324 | ing track of parenthesis numbers can be tricky. It may be more conve- | |
1325 | nient to use named parentheses instead. For this, PCRE uses (?P>name), | |
1326 | which is an extension to the Python syntax that PCRE uses for named | |
1327 | parentheses (Perl does not provide named parentheses). We could rewrite | |
1328 | the above example as follows: | |
1329 | ||
1330 | (?P<pn> \( ( (?>[^()]+) | (?P>pn) )* \) ) | |
1331 | ||
1332 | This particular example pattern contains nested unlimited repeats, and | |
1333 | so the use of atomic grouping for matching strings of non-parentheses | |
1334 | is important when applying the pattern to strings that do not match. | |
1335 | For example, when this pattern is applied to | |
1336 | ||
1337 | (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa() | |
1338 | ||
1339 | it yields "no match" quickly. However, if atomic grouping is not used, | |
1340 | the match runs for a very long time indeed because there are so many | |
1341 | different ways the + and * repeats can carve up the subject, and all | |
1342 | have to be tested before failure can be reported. | |
1343 | ||
1344 | At the end of a match, the values set for any capturing subpatterns are | |
1345 | those from the outermost level of the recursion at which the subpattern | |
1346 | value is set. If you want to obtain intermediate values, a callout | |
1347 | function can be used (see the next section and the pcrecallout documen- | |
1348 | tation). If the pattern above is matched against | |
1349 | ||
1350 | (ab(cd)ef) | |
1351 | ||
1352 | the value for the capturing parentheses is "ef", which is the last | |
1353 | value taken on at the top level. If additional parentheses are added, | |
1354 | giving | |
1355 | ||
1356 | \( ( ( (?>[^()]+) | (?R) )* ) \) | |
1357 | ^ ^ | |
1358 | ^ ^ | |
1359 | ||
1360 | the string they capture is "ab(cd)ef", the contents of the top level | |
1361 | parentheses. If there are more than 15 capturing parentheses in a pat- | |
1362 | tern, PCRE has to obtain extra memory to store data during a recursion, | |
1363 | which it does by using pcre_malloc, freeing it via pcre_free after- | |
1364 | wards. If no memory can be obtained, the match fails with the | |
1365 | PCRE_ERROR_NOMEMORY error. | |
1366 | ||
1367 | Do not confuse the (?R) item with the condition (R), which tests for | |
1368 | recursion. Consider this pattern, which matches text in angle brack- | |
1369 | ets, allowing for arbitrary nesting. Only digits are allowed in nested | |
1370 | brackets (that is, when recursing), whereas any characters are permit- | |
1371 | ted at the outer level. | |
1372 | ||
1373 | < (?: (?(R) \d++ | [^<>]*+) | (?R)) * > | |
1374 | ||
1375 | In this pattern, (?(R) is the start of a conditional subpattern, with | |
1376 | two different alternatives for the recursive and non-recursive cases. | |
1377 | The (?R) item is the actual recursive call. | |
1378 | ||
1379 | ||
1380 | SUBPATTERNS AS SUBROUTINES | |
1381 | ||
1382 | If the syntax for a recursive subpattern reference (either by number or | |
1383 | by name) is used outside the parentheses to which it refers, it oper- | |
1384 | ates like a subroutine in a programming language. An earlier example | |
1385 | pointed out that the pattern | |
1386 | ||
1387 | (sens|respons)e and \1ibility | |
1388 | ||
1389 | matches "sense and sensibility" and "response and responsibility", but | |
1390 | not "sense and responsibility". If instead the pattern | |
1391 | ||
1392 | (sens|respons)e and (?1)ibility | |
1393 | ||
1394 | is used, it does match "sense and responsibility" as well as the other | |
1395 | two strings. Such references must, however, follow the subpattern to | |
1396 | which they refer. | |
1397 | ||
1398 | ||
1399 | CALLOUTS | |
1400 | ||
1401 | Perl has a feature whereby using the sequence (?{...}) causes arbitrary | |
1402 | Perl code to be obeyed in the middle of matching a regular expression. | |
1403 | This makes it possible, amongst other things, to extract different sub- | |
1404 | strings that match the same pair of parentheses when there is a repeti- | |
1405 | tion. | |
1406 | ||
1407 | PCRE provides a similar feature, but of course it cannot obey arbitrary | |
1408 | Perl code. The feature is called "callout". The caller of PCRE provides | |
1409 | an external function by putting its entry point in the global variable | |
1410 | pcre_callout. By default, this variable contains NULL, which disables | |
1411 | all calling out. | |
1412 | ||
1413 | Within a regular expression, (?C) indicates the points at which the | |
1414 | external function is to be called. If you want to identify different | |
1415 | callout points, you can put a number less than 256 after the letter C. | |
1416 | The default value is zero. For example, this pattern has two callout | |
1417 | points: | |
1418 | ||
1419 | (?C1)abc(?C2)def | |
1420 | ||
1421 | If the PCRE_AUTO_CALLOUT flag is passed to pcre_compile(), callouts are | |
1422 | automatically installed before each item in the pattern. They are all | |
1423 | numbered 255. | |
1424 | ||
1425 | During matching, when PCRE reaches a callout point (and pcre_callout is | |
1426 | set), the external function is called. It is provided with the number | |
1427 | of the callout, the position in the pattern, and, optionally, one item | |
1428 | of data originally supplied by the caller of pcre_exec(). The callout | |
1429 | function may cause matching to proceed, to backtrack, or to fail alto- | |
1430 | gether. A complete description of the interface to the callout function | |
1431 | is given in the pcrecallout documentation. | |
1432 | ||
8ac170f3 PH |
1433 | Last updated: 28 February 2005 |
1434 | Copyright (c) 1997-2005 University of Cambridge. |