/[pcre]/code/trunk/doc/html/pcrepattern.html
ViewVC logotype

Contents of /code/trunk/doc/html/pcrepattern.html

Parent Directory Parent Directory | Revision Log Revision Log


Revision 87 - (show annotations)
Sat Feb 24 21:41:21 2007 UTC (8 years, 5 months ago) by nigel
File MIME type: text/html
File size: 67970 byte(s)
Error occurred while calculating annotation data.
Load pcre-6.5 into code/trunk.
1 <html>
2 <head>
3 <title>pcrepattern specification</title>
4 </head>
5 <body bgcolor="#FFFFFF" text="#00005A" link="#0066FF" alink="#3399FF" vlink="#2222BB">
6 <h1>pcrepattern man page</h1>
7 <p>
8 Return to the <a href="index.html">PCRE index page</a>.
9 </p>
10 <p>
11 This page is part of the PCRE HTML documentation. It was generated automatically
12 from the original man page. If there is any nonsense in it, please consult the
13 man page, in case the conversion went wrong.
14 <br>
15 <ul>
16 <li><a name="TOC1" href="#SEC1">PCRE REGULAR EXPRESSION DETAILS</a>
17 <li><a name="TOC2" href="#SEC2">BACKSLASH</a>
18 <li><a name="TOC3" href="#SEC3">CIRCUMFLEX AND DOLLAR</a>
19 <li><a name="TOC4" href="#SEC4">FULL STOP (PERIOD, DOT)</a>
20 <li><a name="TOC5" href="#SEC5">MATCHING A SINGLE BYTE</a>
21 <li><a name="TOC6" href="#SEC6">SQUARE BRACKETS AND CHARACTER CLASSES</a>
22 <li><a name="TOC7" href="#SEC7">POSIX CHARACTER CLASSES</a>
23 <li><a name="TOC8" href="#SEC8">VERTICAL BAR</a>
24 <li><a name="TOC9" href="#SEC9">INTERNAL OPTION SETTING</a>
25 <li><a name="TOC10" href="#SEC10">SUBPATTERNS</a>
26 <li><a name="TOC11" href="#SEC11">NAMED SUBPATTERNS</a>
27 <li><a name="TOC12" href="#SEC12">REPETITION</a>
28 <li><a name="TOC13" href="#SEC13">ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS</a>
29 <li><a name="TOC14" href="#SEC14">BACK REFERENCES</a>
30 <li><a name="TOC15" href="#SEC15">ASSERTIONS</a>
31 <li><a name="TOC16" href="#SEC16">CONDITIONAL SUBPATTERNS</a>
32 <li><a name="TOC17" href="#SEC17">COMMENTS</a>
33 <li><a name="TOC18" href="#SEC18">RECURSIVE PATTERNS</a>
34 <li><a name="TOC19" href="#SEC19">SUBPATTERNS AS SUBROUTINES</a>
35 <li><a name="TOC20" href="#SEC20">CALLOUTS</a>
36 </ul>
37 <br><a name="SEC1" href="#TOC1">PCRE REGULAR EXPRESSION DETAILS</a><br>
38 <P>
39 The syntax and semantics of the regular expressions supported by PCRE are
40 described below. Regular expressions are also described in the Perl
41 documentation and in a number of books, some of which have copious examples.
42 Jeffrey Friedl's "Mastering Regular Expressions", published by O'Reilly, covers
43 regular expressions in great detail. This description of PCRE's regular
44 expressions is intended as reference material.
45 </P>
46 <P>
47 The original operation of PCRE was on strings of one-byte characters. However,
48 there is now also support for UTF-8 character strings. To use this, you must
49 build PCRE to include UTF-8 support, and then call <b>pcre_compile()</b> with
50 the PCRE_UTF8 option. How this affects pattern matching is mentioned in several
51 places below. There is also a summary of UTF-8 features in the
52 <a href="pcre.html#utf8support">section on UTF-8 support</a>
53 in the main
54 <a href="pcre.html"><b>pcre</b></a>
55 page.
56 </P>
57 <P>
58 The remainder of this document discusses the patterns that are supported by
59 PCRE when its main matching function, <b>pcre_exec()</b>, is used.
60 From release 6.0, PCRE offers a second matching function,
61 <b>pcre_dfa_exec()</b>, which matches using a different algorithm that is not
62 Perl-compatible. The advantages and disadvantages of the alternative function,
63 and how it differs from the normal function, are discussed in the
64 <a href="pcrematching.html"><b>pcrematching</b></a>
65 page.
66 </P>
67 <P>
68 A regular expression is a pattern that is matched against a subject string from
69 left to right. Most characters stand for themselves in a pattern, and match the
70 corresponding characters in the subject. As a trivial example, the pattern
71 <pre>
72 The quick brown fox
73 </pre>
74 matches a portion of a subject string that is identical to itself. When
75 caseless matching is specified (the PCRE_CASELESS option), letters are matched
76 independently of case. In UTF-8 mode, PCRE always understands the concept of
77 case for characters whose values are less than 128, so caseless matching is
78 always possible. For characters with higher values, the concept of case is
79 supported if PCRE is compiled with Unicode property support, but not otherwise.
80 If you want to use caseless matching for characters 128 and above, you must
81 ensure that PCRE is compiled with Unicode property support as well as with
82 UTF-8 support.
83 </P>
84 <P>
85 The power of regular expressions comes from the ability to include alternatives
86 and repetitions in the pattern. These are encoded in the pattern by the use of
87 <i>metacharacters</i>, which do not stand for themselves but instead are
88 interpreted in some special way.
89 </P>
90 <P>
91 There are two different sets of metacharacters: those that are recognized
92 anywhere in the pattern except within square brackets, and those that are
93 recognized in square brackets. Outside square brackets, the metacharacters are
94 as follows:
95 <pre>
96 \ general escape character with several uses
97 ^ assert start of string (or line, in multiline mode)
98 $ assert end of string (or line, in multiline mode)
99 . match any character except newline (by default)
100 [ start character class definition
101 | start of alternative branch
102 ( start subpattern
103 ) end subpattern
104 ? extends the meaning of (
105 also 0 or 1 quantifier
106 also quantifier minimizer
107 * 0 or more quantifier
108 + 1 or more quantifier
109 also "possessive quantifier"
110 { start min/max quantifier
111 </pre>
112 Part of a pattern that is in square brackets is called a "character class". In
113 a character class the only metacharacters are:
114 <pre>
115 \ general escape character
116 ^ negate the class, but only if the first character
117 - indicates character range
118 [ POSIX character class (only if followed by POSIX syntax)
119 ] terminates the character class
120 </pre>
121 The following sections describe the use of each of the metacharacters.
122 </P>
123 <br><a name="SEC2" href="#TOC1">BACKSLASH</a><br>
124 <P>
125 The backslash character has several uses. Firstly, if it is followed by a
126 non-alphanumeric character, it takes away any special meaning that character may
127 have. This use of backslash as an escape character applies both inside and
128 outside character classes.
129 </P>
130 <P>
131 For example, if you want to match a * character, you write \* in the pattern.
132 This escaping action applies whether or not the following character would
133 otherwise be interpreted as a metacharacter, so it is always safe to precede a
134 non-alphanumeric with backslash to specify that it stands for itself. In
135 particular, if you want to match a backslash, you write \\.
136 </P>
137 <P>
138 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
139 pattern (other than in a character class) and characters between a # outside
140 a character class and the next newline character are ignored. An escaping
141 backslash can be used to include a whitespace or # character as part of the
142 pattern.
143 </P>
144 <P>
145 If you want to remove the special meaning from a sequence of characters, you
146 can do so by putting them between \Q and \E. This is different from Perl in
147 that $ and @ are handled as literals in \Q...\E sequences in PCRE, whereas in
148 Perl, $ and @ cause variable interpolation. Note the following examples:
149 <pre>
150 Pattern PCRE matches Perl matches
151
152 \Qabc$xyz\E abc$xyz abc followed by the contents of $xyz
153 \Qabc\$xyz\E abc\$xyz abc\$xyz
154 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
155 </pre>
156 The \Q...\E sequence is recognized both inside and outside character classes.
157 <a name="digitsafterbackslash"></a></P>
158 <br><b>
159 Non-printing characters
160 </b><br>
161 <P>
162 A second use of backslash provides a way of encoding non-printing characters
163 in patterns in a visible manner. There is no restriction on the appearance of
164 non-printing characters, apart from the binary zero that terminates a pattern,
165 but when a pattern is being prepared by text editing, it is usually easier to
166 use one of the following escape sequences than the binary character it
167 represents:
168 <pre>
169 \a alarm, that is, the BEL character (hex 07)
170 \cx "control-x", where x is any character
171 \e escape (hex 1B)
172 \f formfeed (hex 0C)
173 \n newline (hex 0A)
174 \r carriage return (hex 0D)
175 \t tab (hex 09)
176 \ddd character with octal code ddd, or backreference
177 \xhh character with hex code hh
178 \x{hhh..} character with hex code hhh..
179 </pre>
180 The precise effect of \cx is as follows: if x is a lower case letter, it
181 is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
182 Thus \cz becomes hex 1A, but \c{ becomes hex 3B, while \c; becomes hex
183 7B.
184 </P>
185 <P>
186 After \x, from zero to two hexadecimal digits are read (letters can be in
187 upper or lower case). Any number of hexadecimal digits may appear between \x{
188 and }, but the value of the character code must be less than 256 in non-UTF-8
189 mode, and less than 2**31 in UTF-8 mode (that is, the maximum hexadecimal value
190 is 7FFFFFFF). If characters other than hexadecimal digits appear between \x{
191 and }, or if there is no terminating }, this form of escape is not recognized.
192 Instead, the initial \x will be interpreted as a basic hexadecimal escape,
193 with no following digits, giving a character whose value is zero.
194 </P>
195 <P>
196 Characters whose value is less than 256 can be defined by either of the two
197 syntaxes for \x. There is no difference in the way they are handled. For
198 example, \xdc is exactly the same as \x{dc}.
199 </P>
200 <P>
201 After \0 up to two further octal digits are read. In both cases, if there
202 are fewer than two digits, just those that are present are used. Thus the
203 sequence \0\x\07 specifies two binary zeros followed by a BEL character
204 (code value 7). Make sure you supply two digits after the initial zero if the
205 pattern character that follows is itself an octal digit.
206 </P>
207 <P>
208 The handling of a backslash followed by a digit other than 0 is complicated.
209 Outside a character class, PCRE reads it and any following digits as a decimal
210 number. If the number is less than 10, or if there have been at least that many
211 previous capturing left parentheses in the expression, the entire sequence is
212 taken as a <i>back reference</i>. A description of how this works is given
213 <a href="#backreferences">later,</a>
214 following the discussion of
215 <a href="#subpattern">parenthesized subpatterns.</a>
216 </P>
217 <P>
218 Inside a character class, or if the decimal number is greater than 9 and there
219 have not been that many capturing subpatterns, PCRE re-reads up to three octal
220 digits following the backslash, and generates a single byte from the least
221 significant 8 bits of the value. Any subsequent digits stand for themselves.
222 For example:
223 <pre>
224 \040 is another way of writing a space
225 \40 is the same, provided there are fewer than 40 previous capturing subpatterns
226 \7 is always a back reference
227 \11 might be a back reference, or another way of writing a tab
228 \011 is always a tab
229 \0113 is a tab followed by the character "3"
230 \113 might be a back reference, otherwise the character with octal code 113
231 \377 might be a back reference, otherwise the byte consisting entirely of 1 bits
232 \81 is either a back reference, or a binary zero followed by the two characters "8" and "1"
233 </pre>
234 Note that octal values of 100 or greater must not be introduced by a leading
235 zero, because no more than three octal digits are ever read.
236 </P>
237 <P>
238 All the sequences that define a single byte value or a single UTF-8 character
239 (in UTF-8 mode) can be used both inside and outside character classes. In
240 addition, inside a character class, the sequence \b is interpreted as the
241 backspace character (hex 08), and the sequence \X is interpreted as the
242 character "X". Outside a character class, these sequences have different
243 meanings
244 <a href="#uniextseq">(see below).</a>
245 </P>
246 <br><b>
247 Generic character types
248 </b><br>
249 <P>
250 The third use of backslash is for specifying generic character types. The
251 following are always recognized:
252 <pre>
253 \d any decimal digit
254 \D any character that is not a decimal digit
255 \s any whitespace character
256 \S any character that is not a whitespace character
257 \w any "word" character
258 \W any "non-word" character
259 </pre>
260 Each pair of escape sequences partitions the complete set of characters into
261 two disjoint sets. Any given character matches one, and only one, of each pair.
262 </P>
263 <P>
264 These character type sequences can appear both inside and outside character
265 classes. They each match one character of the appropriate type. If the current
266 matching point is at the end of the subject string, all of them fail, since
267 there is no character to match.
268 </P>
269 <P>
270 For compatibility with Perl, \s does not match the VT character (code 11).
271 This makes it different from the the POSIX "space" class. The \s characters
272 are HT (9), LF (10), FF (12), CR (13), and space (32).
273 </P>
274 <P>
275 A "word" character is an underscore or any character less than 256 that is a
276 letter or digit. The definition of letters and digits is controlled by PCRE's
277 low-valued character tables, and may vary if locale-specific matching is taking
278 place (see
279 <a href="pcreapi.html#localesupport">"Locale support"</a>
280 in the
281 <a href="pcreapi.html"><b>pcreapi</b></a>
282 page). For example, in the "fr_FR" (French) locale, some character codes
283 greater than 128 are used for accented letters, and these are matched by \w.
284 </P>
285 <P>
286 In UTF-8 mode, characters with values greater than 128 never match \d, \s, or
287 \w, and always match \D, \S, and \W. This is true even when Unicode
288 character property support is available. The use of locales with Unicode is
289 discouraged.
290 <a name="uniextseq"></a></P>
291 <br><b>
292 Unicode character properties
293 </b><br>
294 <P>
295 When PCRE is built with Unicode character property support, three additional
296 escape sequences to match character properties are available when UTF-8 mode
297 is selected. They are:
298 <pre>
299 \p{<i>xx</i>} a character with the <i>xx</i> property
300 \P{<i>xx</i>} a character without the <i>xx</i> property
301 \X an extended Unicode sequence
302 </pre>
303 The property names represented by <i>xx</i> above are limited to the Unicode
304 script names, the general category properties, and "Any", which matches any
305 character (including newline). Other properties such as "InMusicalSymbols" are
306 not currently supported by PCRE. Note that \P{Any} does not match any
307 characters, so always causes a match failure.
308 </P>
309 <P>
310 Sets of Unicode characters are defined as belonging to certain scripts. A
311 character from one of these sets can be matched using a script name. For
312 example:
313 <pre>
314 \p{Greek}
315 \P{Han}
316 </pre>
317 Those that are not part of an identified script are lumped together as
318 "Common". The current list of scripts is:
319 </P>
320 <P>
321 Arabic,
322 Armenian,
323 Bengali,
324 Bopomofo,
325 Braille,
326 Buginese,
327 Buhid,
328 Canadian_Aboriginal,
329 Cherokee,
330 Common,
331 Coptic,
332 Cypriot,
333 Cyrillic,
334 Deseret,
335 Devanagari,
336 Ethiopic,
337 Georgian,
338 Glagolitic,
339 Gothic,
340 Greek,
341 Gujarati,
342 Gurmukhi,
343 Han,
344 Hangul,
345 Hanunoo,
346 Hebrew,
347 Hiragana,
348 Inherited,
349 Kannada,
350 Katakana,
351 Kharoshthi,
352 Khmer,
353 Lao,
354 Latin,
355 Limbu,
356 Linear_B,
357 Malayalam,
358 Mongolian,
359 Myanmar,
360 New_Tai_Lue,
361 Ogham,
362 Old_Italic,
363 Old_Persian,
364 Oriya,
365 Osmanya,
366 Runic,
367 Shavian,
368 Sinhala,
369 Syloti_Nagri,
370 Syriac,
371 Tagalog,
372 Tagbanwa,
373 Tai_Le,
374 Tamil,
375 Telugu,
376 Thaana,
377 Thai,
378 Tibetan,
379 Tifinagh,
380 Ugaritic,
381 Yi.
382 </P>
383 <P>
384 Each character has exactly one general category property, specified by a
385 two-letter abbreviation. For compatibility with Perl, negation can be specified
386 by including a circumflex between the opening brace and the property name. For
387 example, \p{^Lu} is the same as \P{Lu}.
388 </P>
389 <P>
390 If only one letter is specified with \p or \P, it includes all the general
391 category properties that start with that letter. In this case, in the absence
392 of negation, the curly brackets in the escape sequence are optional; these two
393 examples have the same effect:
394 <pre>
395 \p{L}
396 \pL
397 </pre>
398 The following general category property codes are supported:
399 <pre>
400 C Other
401 Cc Control
402 Cf Format
403 Cn Unassigned
404 Co Private use
405 Cs Surrogate
406
407 L Letter
408 Ll Lower case letter
409 Lm Modifier letter
410 Lo Other letter
411 Lt Title case letter
412 Lu Upper case letter
413
414 M Mark
415 Mc Spacing mark
416 Me Enclosing mark
417 Mn Non-spacing mark
418
419 N Number
420 Nd Decimal number
421 Nl Letter number
422 No Other number
423
424 P Punctuation
425 Pc Connector punctuation
426 Pd Dash punctuation
427 Pe Close punctuation
428 Pf Final punctuation
429 Pi Initial punctuation
430 Po Other punctuation
431 Ps Open punctuation
432
433 S Symbol
434 Sc Currency symbol
435 Sk Modifier symbol
436 Sm Mathematical symbol
437 So Other symbol
438
439 Z Separator
440 Zl Line separator
441 Zp Paragraph separator
442 Zs Space separator
443 </pre>
444 The special property L& is also supported: it matches a character that has
445 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
446 a modifier or "other".
447 </P>
448 <P>
449 The long synonyms for these properties that Perl supports (such as \p{Letter})
450 are not supported by PCRE. Nor is is permitted to prefix any of these
451 properties with "Is".
452 </P>
453 <P>
454 No character that is in the Unicode table has the Cn (unassigned) property.
455 Instead, this property is assumed for any code point that is not in the
456 Unicode table.
457 </P>
458 <P>
459 Specifying caseless matching does not affect these escape sequences. For
460 example, \p{Lu} always matches only upper case letters.
461 </P>
462 <P>
463 The \X escape matches any number of Unicode characters that form an extended
464 Unicode sequence. \X is equivalent to
465 <pre>
466 (?&#62;\PM\pM*)
467 </pre>
468 That is, it matches a character without the "mark" property, followed by zero
469 or more characters with the "mark" property, and treats the sequence as an
470 atomic group
471 <a href="#atomicgroup">(see below).</a>
472 Characters with the "mark" property are typically accents that affect the
473 preceding character.
474 </P>
475 <P>
476 Matching characters by Unicode property is not fast, because PCRE has to search
477 a structure that contains data for over fifteen thousand characters. That is
478 why the traditional escape sequences such as \d and \w do not use Unicode
479 properties in PCRE.
480 <a name="smallassertions"></a></P>
481 <br><b>
482 Simple assertions
483 </b><br>
484 <P>
485 The fourth use of backslash is for certain simple assertions. An assertion
486 specifies a condition that has to be met at a particular point in a match,
487 without consuming any characters from the subject string. The use of
488 subpatterns for more complicated assertions is described
489 <a href="#bigassertions">below.</a>
490 The backslashed
491 assertions are:
492 <pre>
493 \b matches at a word boundary
494 \B matches when not at a word boundary
495 \A matches at start of subject
496 \Z matches at end of subject or before newline at end
497 \z matches at end of subject
498 \G matches at first matching position in subject
499 </pre>
500 These assertions may not appear in character classes (but note that \b has a
501 different meaning, namely the backspace character, inside a character class).
502 </P>
503 <P>
504 A word boundary is a position in the subject string where the current character
505 and the previous character do not both match \w or \W (i.e. one matches
506 \w and the other matches \W), or the start or end of the string if the
507 first or last character matches \w, respectively.
508 </P>
509 <P>
510 The \A, \Z, and \z assertions differ from the traditional circumflex and
511 dollar (described in the next section) in that they only ever match at the very
512 start and end of the subject string, whatever options are set. Thus, they are
513 independent of multiline mode. These three assertions are not affected by the
514 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
515 circumflex and dollar metacharacters. However, if the <i>startoffset</i>
516 argument of <b>pcre_exec()</b> is non-zero, indicating that matching is to start
517 at a point other than the beginning of the subject, \A can never match. The
518 difference between \Z and \z is that \Z matches before a newline that is the
519 last character of the string as well as at the end of the string, whereas \z
520 matches only at the end.
521 </P>
522 <P>
523 The \G assertion is true only when the current matching position is at the
524 start point of the match, as specified by the <i>startoffset</i> argument of
525 <b>pcre_exec()</b>. It differs from \A when the value of <i>startoffset</i> is
526 non-zero. By calling <b>pcre_exec()</b> multiple times with appropriate
527 arguments, you can mimic Perl's /g option, and it is in this kind of
528 implementation where \G can be useful.
529 </P>
530 <P>
531 Note, however, that PCRE's interpretation of \G, as the start of the current
532 match, is subtly different from Perl's, which defines it as the end of the
533 previous match. In Perl, these can be different when the previously matched
534 string was empty. Because PCRE does just one match at a time, it cannot
535 reproduce this behaviour.
536 </P>
537 <P>
538 If all the alternatives of a pattern begin with \G, the expression is anchored
539 to the starting match position, and the "anchored" flag is set in the compiled
540 regular expression.
541 </P>
542 <br><a name="SEC3" href="#TOC1">CIRCUMFLEX AND DOLLAR</a><br>
543 <P>
544 Outside a character class, in the default matching mode, the circumflex
545 character is an assertion that is true only if the current matching point is
546 at the start of the subject string. If the <i>startoffset</i> argument of
547 <b>pcre_exec()</b> is non-zero, circumflex can never match if the PCRE_MULTILINE
548 option is unset. Inside a character class, circumflex has an entirely different
549 meaning
550 <a href="#characterclass">(see below).</a>
551 </P>
552 <P>
553 Circumflex need not be the first character of the pattern if a number of
554 alternatives are involved, but it should be the first thing in each alternative
555 in which it appears if the pattern is ever to match that branch. If all
556 possible alternatives start with a circumflex, that is, if the pattern is
557 constrained to match only at the start of the subject, it is said to be an
558 "anchored" pattern. (There are also other constructs that can cause a pattern
559 to be anchored.)
560 </P>
561 <P>
562 A dollar character is an assertion that is true only if the current matching
563 point is at the end of the subject string, or immediately before a newline
564 character that is the last character in the string (by default). Dollar need
565 not be the last character of the pattern if a number of alternatives are
566 involved, but it should be the last item in any branch in which it appears.
567 Dollar has no special meaning in a character class.
568 </P>
569 <P>
570 The meaning of dollar can be changed so that it matches only at the very end of
571 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
572 does not affect the \Z assertion.
573 </P>
574 <P>
575 The meanings of the circumflex and dollar characters are changed if the
576 PCRE_MULTILINE option is set. When this is the case, they match immediately
577 after and immediately before an internal newline character, respectively, in
578 addition to matching at the start and end of the subject string. For example,
579 the pattern /^abc$/ matches the subject string "def\nabc" (where \n
580 represents a newline character) in multiline mode, but not otherwise.
581 Consequently, patterns that are anchored in single line mode because all
582 branches start with ^ are not anchored in multiline mode, and a match for
583 circumflex is possible when the <i>startoffset</i> argument of <b>pcre_exec()</b>
584 is non-zero. The PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is
585 set.
586 </P>
587 <P>
588 Note that the sequences \A, \Z, and \z can be used to match the start and
589 end of the subject in both modes, and if all branches of a pattern start with
590 \A it is always anchored, whether PCRE_MULTILINE is set or not.
591 </P>
592 <br><a name="SEC4" href="#TOC1">FULL STOP (PERIOD, DOT)</a><br>
593 <P>
594 Outside a character class, a dot in the pattern matches any one character in
595 the subject, including a non-printing character, but not (by default) newline.
596 In UTF-8 mode, a dot matches any UTF-8 character, which might be more than one
597 byte long, except (by default) newline. If the PCRE_DOTALL option is set,
598 dots match newlines as well. The handling of dot is entirely independent of the
599 handling of circumflex and dollar, the only relationship being that they both
600 involve newline characters. Dot has no special meaning in a character class.
601 </P>
602 <br><a name="SEC5" href="#TOC1">MATCHING A SINGLE BYTE</a><br>
603 <P>
604 Outside a character class, the escape sequence \C matches any one byte, both
605 in and out of UTF-8 mode. Unlike a dot, it can match a newline. The feature is
606 provided in Perl in order to match individual bytes in UTF-8 mode. Because it
607 breaks up UTF-8 characters into individual bytes, what remains in the string
608 may be a malformed UTF-8 string. For this reason, the \C escape sequence is
609 best avoided.
610 </P>
611 <P>
612 PCRE does not allow \C to appear in lookbehind assertions
613 <a href="#lookbehind">(described below),</a>
614 because in UTF-8 mode this would make it impossible to calculate the length of
615 the lookbehind.
616 <a name="characterclass"></a></P>
617 <br><a name="SEC6" href="#TOC1">SQUARE BRACKETS AND CHARACTER CLASSES</a><br>
618 <P>
619 An opening square bracket introduces a character class, terminated by a closing
620 square bracket. A closing square bracket on its own is not special. If a
621 closing square bracket is required as a member of the class, it should be the
622 first data character in the class (after an initial circumflex, if present) or
623 escaped with a backslash.
624 </P>
625 <P>
626 A character class matches a single character in the subject. In UTF-8 mode, the
627 character may occupy more than one byte. A matched character must be in the set
628 of characters defined by the class, unless the first character in the class
629 definition is a circumflex, in which case the subject character must not be in
630 the set defined by the class. If a circumflex is actually required as a member
631 of the class, ensure it is not the first character, or escape it with a
632 backslash.
633 </P>
634 <P>
635 For example, the character class [aeiou] matches any lower case vowel, while
636 [^aeiou] matches any character that is not a lower case vowel. Note that a
637 circumflex is just a convenient notation for specifying the characters that
638 are in the class by enumerating those that are not. A class that starts with a
639 circumflex is not an assertion: it still consumes a character from the subject
640 string, and therefore it fails if the current pointer is at the end of the
641 string.
642 </P>
643 <P>
644 In UTF-8 mode, characters with values greater than 255 can be included in a
645 class as a literal string of bytes, or by using the \x{ escaping mechanism.
646 </P>
647 <P>
648 When caseless matching is set, any letters in a class represent both their
649 upper case and lower case versions, so for example, a caseless [aeiou] matches
650 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
651 caseful version would. In UTF-8 mode, PCRE always understands the concept of
652 case for characters whose values are less than 128, so caseless matching is
653 always possible. For characters with higher values, the concept of case is
654 supported if PCRE is compiled with Unicode property support, but not otherwise.
655 If you want to use caseless matching for characters 128 and above, you must
656 ensure that PCRE is compiled with Unicode property support as well as with
657 UTF-8 support.
658 </P>
659 <P>
660 The newline character is never treated in any special way in character classes,
661 whatever the setting of the PCRE_DOTALL or PCRE_MULTILINE options is. A class
662 such as [^a] will always match a newline.
663 </P>
664 <P>
665 The minus (hyphen) character can be used to specify a range of characters in a
666 character class. For example, [d-m] matches any letter between d and m,
667 inclusive. If a minus character is required in a class, it must be escaped with
668 a backslash or appear in a position where it cannot be interpreted as
669 indicating a range, typically as the first or last character in the class.
670 </P>
671 <P>
672 It is not possible to have the literal character "]" as the end character of a
673 range. A pattern such as [W-]46] is interpreted as a class of two characters
674 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
675 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
676 the end of range, so [W-\]46] is interpreted as a class containing a range
677 followed by two other characters. The octal or hexadecimal representation of
678 "]" can also be used to end a range.
679 </P>
680 <P>
681 Ranges operate in the collating sequence of character values. They can also be
682 used for characters specified numerically, for example [\000-\037]. In UTF-8
683 mode, ranges can include characters whose values are greater than 255, for
684 example [\x{100}-\x{2ff}].
685 </P>
686 <P>
687 If a range that includes letters is used when caseless matching is set, it
688 matches the letters in either case. For example, [W-c] is equivalent to
689 [][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
690 tables for the "fr_FR" locale are in use, [\xc8-\xcb] matches accented E
691 characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
692 characters with values greater than 128 only when it is compiled with Unicode
693 property support.
694 </P>
695 <P>
696 The character types \d, \D, \p, \P, \s, \S, \w, and \W may also appear
697 in a character class, and add the characters that they match to the class. For
698 example, [\dABCDEF] matches any hexadecimal digit. A circumflex can
699 conveniently be used with the upper case character types to specify a more
700 restricted set of characters than the matching lower case type. For example,
701 the class [^\W_] matches any letter or digit, but not underscore.
702 </P>
703 <P>
704 The only metacharacters that are recognized in character classes are backslash,
705 hyphen (only where it can be interpreted as specifying a range), circumflex
706 (only at the start), opening square bracket (only when it can be interpreted as
707 introducing a POSIX class name - see the next section), and the terminating
708 closing square bracket. However, escaping other non-alphanumeric characters
709 does no harm.
710 </P>
711 <br><a name="SEC7" href="#TOC1">POSIX CHARACTER CLASSES</a><br>
712 <P>
713 Perl supports the POSIX notation for character classes. This uses names
714 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
715 this notation. For example,
716 <pre>
717 [01[:alpha:]%]
718 </pre>
719 matches "0", "1", any alphabetic character, or "%". The supported class names
720 are
721 <pre>
722 alnum letters and digits
723 alpha letters
724 ascii character codes 0 - 127
725 blank space or tab only
726 cntrl control characters
727 digit decimal digits (same as \d)
728 graph printing characters, excluding space
729 lower lower case letters
730 print printing characters, including space
731 punct printing characters, excluding letters and digits
732 space white space (not quite the same as \s)
733 upper upper case letters
734 word "word" characters (same as \w)
735 xdigit hexadecimal digits
736 </pre>
737 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
738 space (32). Notice that this list includes the VT character (code 11). This
739 makes "space" different to \s, which does not include VT (for Perl
740 compatibility).
741 </P>
742 <P>
743 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
744 5.8. Another Perl extension is negation, which is indicated by a ^ character
745 after the colon. For example,
746 <pre>
747 [12[:^digit:]]
748 </pre>
749 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
750 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
751 supported, and an error is given if they are encountered.
752 </P>
753 <P>
754 In UTF-8 mode, characters with values greater than 128 do not match any of
755 the POSIX character classes.
756 </P>
757 <br><a name="SEC8" href="#TOC1">VERTICAL BAR</a><br>
758 <P>
759 Vertical bar characters are used to separate alternative patterns. For example,
760 the pattern
761 <pre>
762 gilbert|sullivan
763 </pre>
764 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
765 and an empty alternative is permitted (matching the empty string).
766 The matching process tries each alternative in turn, from left to right,
767 and the first one that succeeds is used. If the alternatives are within a
768 subpattern
769 <a href="#subpattern">(defined below),</a>
770 "succeeds" means matching the rest of the main pattern as well as the
771 alternative in the subpattern.
772 </P>
773 <br><a name="SEC9" href="#TOC1">INTERNAL OPTION SETTING</a><br>
774 <P>
775 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
776 PCRE_EXTENDED options can be changed from within the pattern by a sequence of
777 Perl option letters enclosed between "(?" and ")". The option letters are
778 <pre>
779 i for PCRE_CASELESS
780 m for PCRE_MULTILINE
781 s for PCRE_DOTALL
782 x for PCRE_EXTENDED
783 </pre>
784 For example, (?im) sets caseless, multiline matching. It is also possible to
785 unset these options by preceding the letter with a hyphen, and a combined
786 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
787 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
788 permitted. If a letter appears both before and after the hyphen, the option is
789 unset.
790 </P>
791 <P>
792 When an option change occurs at top level (that is, not inside subpattern
793 parentheses), the change applies to the remainder of the pattern that follows.
794 If the change is placed right at the start of a pattern, PCRE extracts it into
795 the global options (and it will therefore show up in data extracted by the
796 <b>pcre_fullinfo()</b> function).
797 </P>
798 <P>
799 An option change within a subpattern affects only that part of the current
800 pattern that follows it, so
801 <pre>
802 (a(?i)b)c
803 </pre>
804 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
805 By this means, options can be made to have different settings in different
806 parts of the pattern. Any changes made in one alternative do carry on
807 into subsequent branches within the same subpattern. For example,
808 <pre>
809 (a(?i)b|c)
810 </pre>
811 matches "ab", "aB", "c", and "C", even though when matching "C" the first
812 branch is abandoned before the option setting. This is because the effects of
813 option settings happen at compile time. There would be some very weird
814 behaviour otherwise.
815 </P>
816 <P>
817 The PCRE-specific options PCRE_UNGREEDY and PCRE_EXTRA can be changed in the
818 same way as the Perl-compatible options by using the characters U and X
819 respectively. The (?X) flag setting is special in that it must always occur
820 earlier in the pattern than any of the additional features it turns on, even
821 when it is at top level. It is best to put it at the start.
822 <a name="subpattern"></a></P>
823 <br><a name="SEC10" href="#TOC1">SUBPATTERNS</a><br>
824 <P>
825 Subpatterns are delimited by parentheses (round brackets), which can be nested.
826 Turning part of a pattern into a subpattern does two things:
827 <br>
828 <br>
829 1. It localizes a set of alternatives. For example, the pattern
830 <pre>
831 cat(aract|erpillar|)
832 </pre>
833 matches one of the words "cat", "cataract", or "caterpillar". Without the
834 parentheses, it would match "cataract", "erpillar" or the empty string.
835 <br>
836 <br>
837 2. It sets up the subpattern as a capturing subpattern. This means that, when
838 the whole pattern matches, that portion of the subject string that matched the
839 subpattern is passed back to the caller via the <i>ovector</i> argument of
840 <b>pcre_exec()</b>. Opening parentheses are counted from left to right (starting
841 from 1) to obtain numbers for the capturing subpatterns.
842 </P>
843 <P>
844 For example, if the string "the red king" is matched against the pattern
845 <pre>
846 the ((red|white) (king|queen))
847 </pre>
848 the captured substrings are "red king", "red", and "king", and are numbered 1,
849 2, and 3, respectively.
850 </P>
851 <P>
852 The fact that plain parentheses fulfil two functions is not always helpful.
853 There are often times when a grouping subpattern is required without a
854 capturing requirement. If an opening parenthesis is followed by a question mark
855 and a colon, the subpattern does not do any capturing, and is not counted when
856 computing the number of any subsequent capturing subpatterns. For example, if
857 the string "the white queen" is matched against the pattern
858 <pre>
859 the ((?:red|white) (king|queen))
860 </pre>
861 the captured substrings are "white queen" and "queen", and are numbered 1 and
862 2. The maximum number of capturing subpatterns is 65535, and the maximum depth
863 of nesting of all subpatterns, both capturing and non-capturing, is 200.
864 </P>
865 <P>
866 As a convenient shorthand, if any option settings are required at the start of
867 a non-capturing subpattern, the option letters may appear between the "?" and
868 the ":". Thus the two patterns
869 <pre>
870 (?i:saturday|sunday)
871 (?:(?i)saturday|sunday)
872 </pre>
873 match exactly the same set of strings. Because alternative branches are tried
874 from left to right, and options are not reset until the end of the subpattern
875 is reached, an option setting in one branch does affect subsequent branches, so
876 the above patterns match "SUNDAY" as well as "Saturday".
877 </P>
878 <br><a name="SEC11" href="#TOC1">NAMED SUBPATTERNS</a><br>
879 <P>
880 Identifying capturing parentheses by number is simple, but it can be very hard
881 to keep track of the numbers in complicated regular expressions. Furthermore,
882 if an expression is modified, the numbers may change. To help with this
883 difficulty, PCRE supports the naming of subpatterns, something that Perl does
884 not provide. The Python syntax (?P&#60;name&#62;...) is used. Names consist of
885 alphanumeric characters and underscores, and must be unique within a pattern.
886 </P>
887 <P>
888 Named capturing parentheses are still allocated numbers as well as names. The
889 PCRE API provides function calls for extracting the name-to-number translation
890 table from a compiled pattern. There is also a convenience function for
891 extracting a captured substring by name. For further details see the
892 <a href="pcreapi.html"><b>pcreapi</b></a>
893 documentation.
894 </P>
895 <br><a name="SEC12" href="#TOC1">REPETITION</a><br>
896 <P>
897 Repetition is specified by quantifiers, which can follow any of the following
898 items:
899 <pre>
900 a literal data character
901 the . metacharacter
902 the \C escape sequence
903 the \X escape sequence (in UTF-8 mode with Unicode properties)
904 an escape such as \d that matches a single character
905 a character class
906 a back reference (see next section)
907 a parenthesized subpattern (unless it is an assertion)
908 </pre>
909 The general repetition quantifier specifies a minimum and maximum number of
910 permitted matches, by giving the two numbers in curly brackets (braces),
911 separated by a comma. The numbers must be less than 65536, and the first must
912 be less than or equal to the second. For example:
913 <pre>
914 z{2,4}
915 </pre>
916 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
917 character. If the second number is omitted, but the comma is present, there is
918 no upper limit; if the second number and the comma are both omitted, the
919 quantifier specifies an exact number of required matches. Thus
920 <pre>
921 [aeiou]{3,}
922 </pre>
923 matches at least 3 successive vowels, but may match many more, while
924 <pre>
925 \d{8}
926 </pre>
927 matches exactly 8 digits. An opening curly bracket that appears in a position
928 where a quantifier is not allowed, or one that does not match the syntax of a
929 quantifier, is taken as a literal character. For example, {,6} is not a
930 quantifier, but a literal string of four characters.
931 </P>
932 <P>
933 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
934 bytes. Thus, for example, \x{100}{2} matches two UTF-8 characters, each of
935 which is represented by a two-byte sequence. Similarly, when Unicode property
936 support is available, \X{3} matches three Unicode extended sequences, each of
937 which may be several bytes long (and they may be of different lengths).
938 </P>
939 <P>
940 The quantifier {0} is permitted, causing the expression to behave as if the
941 previous item and the quantifier were not present.
942 </P>
943 <P>
944 For convenience (and historical compatibility) the three most common
945 quantifiers have single-character abbreviations:
946 <pre>
947 * is equivalent to {0,}
948 + is equivalent to {1,}
949 ? is equivalent to {0,1}
950 </pre>
951 It is possible to construct infinite loops by following a subpattern that can
952 match no characters with a quantifier that has no upper limit, for example:
953 <pre>
954 (a?)*
955 </pre>
956 Earlier versions of Perl and PCRE used to give an error at compile time for
957 such patterns. However, because there are cases where this can be useful, such
958 patterns are now accepted, but if any repetition of the subpattern does in fact
959 match no characters, the loop is forcibly broken.
960 </P>
961 <P>
962 By default, the quantifiers are "greedy", that is, they match as much as
963 possible (up to the maximum number of permitted times), without causing the
964 rest of the pattern to fail. The classic example of where this gives problems
965 is in trying to match comments in C programs. These appear between /* and */
966 and within the comment, individual * and / characters may appear. An attempt to
967 match C comments by applying the pattern
968 <pre>
969 /\*.*\*/
970 </pre>
971 to the string
972 <pre>
973 /* first comment */ not comment /* second comment */
974 </pre>
975 fails, because it matches the entire string owing to the greediness of the .*
976 item.
977 </P>
978 <P>
979 However, if a quantifier is followed by a question mark, it ceases to be
980 greedy, and instead matches the minimum number of times possible, so the
981 pattern
982 <pre>
983 /\*.*?\*/
984 </pre>
985 does the right thing with the C comments. The meaning of the various
986 quantifiers is not otherwise changed, just the preferred number of matches.
987 Do not confuse this use of question mark with its use as a quantifier in its
988 own right. Because it has two uses, it can sometimes appear doubled, as in
989 <pre>
990 \d??\d
991 </pre>
992 which matches one digit by preference, but can match two if that is the only
993 way the rest of the pattern matches.
994 </P>
995 <P>
996 If the PCRE_UNGREEDY option is set (an option which is not available in Perl),
997 the quantifiers are not greedy by default, but individual ones can be made
998 greedy by following them with a question mark. In other words, it inverts the
999 default behaviour.
1000 </P>
1001 <P>
1002 When a parenthesized subpattern is quantified with a minimum repeat count that
1003 is greater than 1 or with a limited maximum, more memory is required for the
1004 compiled pattern, in proportion to the size of the minimum or maximum.
1005 </P>
1006 <P>
1007 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1008 to Perl's /s) is set, thus allowing the . to match newlines, the pattern is
1009 implicitly anchored, because whatever follows will be tried against every
1010 character position in the subject string, so there is no point in retrying the
1011 overall match at any position after the first. PCRE normally treats such a
1012 pattern as though it were preceded by \A.
1013 </P>
1014 <P>
1015 In cases where it is known that the subject string contains no newlines, it is
1016 worth setting PCRE_DOTALL in order to obtain this optimization, or
1017 alternatively using ^ to indicate anchoring explicitly.
1018 </P>
1019 <P>
1020 However, there is one situation where the optimization cannot be used. When .*
1021 is inside capturing parentheses that are the subject of a backreference
1022 elsewhere in the pattern, a match at the start may fail, and a later one
1023 succeed. Consider, for example:
1024 <pre>
1025 (.*)abc\1
1026 </pre>
1027 If the subject is "xyz123abc123" the match point is the fourth character. For
1028 this reason, such a pattern is not implicitly anchored.
1029 </P>
1030 <P>
1031 When a capturing subpattern is repeated, the value captured is the substring
1032 that matched the final iteration. For example, after
1033 <pre>
1034 (tweedle[dume]{3}\s*)+
1035 </pre>
1036 has matched "tweedledum tweedledee" the value of the captured substring is
1037 "tweedledee". However, if there are nested capturing subpatterns, the
1038 corresponding captured values may have been set in previous iterations. For
1039 example, after
1040 <pre>
1041 /(a|(b))+/
1042 </pre>
1043 matches "aba" the value of the second captured substring is "b".
1044 <a name="atomicgroup"></a></P>
1045 <br><a name="SEC13" href="#TOC1">ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS</a><br>
1046 <P>
1047 With both maximizing and minimizing repetition, failure of what follows
1048 normally causes the repeated item to be re-evaluated to see if a different
1049 number of repeats allows the rest of the pattern to match. Sometimes it is
1050 useful to prevent this, either to change the nature of the match, or to cause
1051 it fail earlier than it otherwise might, when the author of the pattern knows
1052 there is no point in carrying on.
1053 </P>
1054 <P>
1055 Consider, for example, the pattern \d+foo when applied to the subject line
1056 <pre>
1057 123456bar
1058 </pre>
1059 After matching all 6 digits and then failing to match "foo", the normal
1060 action of the matcher is to try again with only 5 digits matching the \d+
1061 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1062 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1063 that once a subpattern has matched, it is not to be re-evaluated in this way.
1064 </P>
1065 <P>
1066 If we use atomic grouping for the previous example, the matcher would give up
1067 immediately on failing to match "foo" the first time. The notation is a kind of
1068 special parenthesis, starting with (?&#62; as in this example:
1069 <pre>
1070 (?&#62;\d+)foo
1071 </pre>
1072 This kind of parenthesis "locks up" the part of the pattern it contains once
1073 it has matched, and a failure further into the pattern is prevented from
1074 backtracking into it. Backtracking past it to previous items, however, works as
1075 normal.
1076 </P>
1077 <P>
1078 An alternative description is that a subpattern of this type matches the string
1079 of characters that an identical standalone pattern would match, if anchored at
1080 the current point in the subject string.
1081 </P>
1082 <P>
1083 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1084 the above example can be thought of as a maximizing repeat that must swallow
1085 everything it can. So, while both \d+ and \d+? are prepared to adjust the
1086 number of digits they match in order to make the rest of the pattern match,
1087 (?&#62;\d+) can only match an entire sequence of digits.
1088 </P>
1089 <P>
1090 Atomic groups in general can of course contain arbitrarily complicated
1091 subpatterns, and can be nested. However, when the subpattern for an atomic
1092 group is just a single repeated item, as in the example above, a simpler
1093 notation, called a "possessive quantifier" can be used. This consists of an
1094 additional + character following a quantifier. Using this notation, the
1095 previous example can be rewritten as
1096 <pre>
1097 \d++foo
1098 </pre>
1099 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1100 option is ignored. They are a convenient notation for the simpler forms of
1101 atomic group. However, there is no difference in the meaning or processing of a
1102 possessive quantifier and the equivalent atomic group.
1103 </P>
1104 <P>
1105 The possessive quantifier syntax is an extension to the Perl syntax. It
1106 originates in Sun's Java package.
1107 </P>
1108 <P>
1109 When a pattern contains an unlimited repeat inside a subpattern that can itself
1110 be repeated an unlimited number of times, the use of an atomic group is the
1111 only way to avoid some failing matches taking a very long time indeed. The
1112 pattern
1113 <pre>
1114 (\D+|&#60;\d+&#62;)*[!?]
1115 </pre>
1116 matches an unlimited number of substrings that either consist of non-digits, or
1117 digits enclosed in &#60;&#62;, followed by either ! or ?. When it matches, it runs
1118 quickly. However, if it is applied to
1119 <pre>
1120 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1121 </pre>
1122 it takes a long time before reporting failure. This is because the string can
1123 be divided between the internal \D+ repeat and the external * repeat in a
1124 large number of ways, and all have to be tried. (The example uses [!?] rather
1125 than a single character at the end, because both PCRE and Perl have an
1126 optimization that allows for fast failure when a single character is used. They
1127 remember the last single character that is required for a match, and fail early
1128 if it is not present in the string.) If the pattern is changed so that it uses
1129 an atomic group, like this:
1130 <pre>
1131 ((?&#62;\D+)|&#60;\d+&#62;)*[!?]
1132 </pre>
1133 sequences of non-digits cannot be broken, and failure happens quickly.
1134 <a name="backreferences"></a></P>
1135 <br><a name="SEC14" href="#TOC1">BACK REFERENCES</a><br>
1136 <P>
1137 Outside a character class, a backslash followed by a digit greater than 0 (and
1138 possibly further digits) is a back reference to a capturing subpattern earlier
1139 (that is, to its left) in the pattern, provided there have been that many
1140 previous capturing left parentheses.
1141 </P>
1142 <P>
1143 However, if the decimal number following the backslash is less than 10, it is
1144 always taken as a back reference, and causes an error only if there are not
1145 that many capturing left parentheses in the entire pattern. In other words, the
1146 parentheses that are referenced need not be to the left of the reference for
1147 numbers less than 10. See the subsection entitled "Non-printing characters"
1148 <a href="#digitsafterbackslash">above</a>
1149 for further details of the handling of digits following a backslash.
1150 </P>
1151 <P>
1152 A back reference matches whatever actually matched the capturing subpattern in
1153 the current subject string, rather than anything matching the subpattern
1154 itself (see
1155 <a href="#subpatternsassubroutines">"Subpatterns as subroutines"</a>
1156 below for a way of doing that). So the pattern
1157 <pre>
1158 (sens|respons)e and \1ibility
1159 </pre>
1160 matches "sense and sensibility" and "response and responsibility", but not
1161 "sense and responsibility". If caseful matching is in force at the time of the
1162 back reference, the case of letters is relevant. For example,
1163 <pre>
1164 ((?i)rah)\s+\1
1165 </pre>
1166 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1167 capturing subpattern is matched caselessly.
1168 </P>
1169 <P>
1170 Back references to named subpatterns use the Python syntax (?P=name). We could
1171 rewrite the above example as follows:
1172 <pre>
1173 (?&#60;p1&#62;(?i)rah)\s+(?P=p1)
1174 </pre>
1175 There may be more than one back reference to the same subpattern. If a
1176 subpattern has not actually been used in a particular match, any back
1177 references to it always fail. For example, the pattern
1178 <pre>
1179 (a|(bc))\2
1180 </pre>
1181 always fails if it starts to match "a" rather than "bc". Because there may be
1182 many capturing parentheses in a pattern, all digits following the backslash are
1183 taken as part of a potential back reference number. If the pattern continues
1184 with a digit character, some delimiter must be used to terminate the back
1185 reference. If the PCRE_EXTENDED option is set, this can be whitespace.
1186 Otherwise an empty comment (see
1187 <a href="#comments">"Comments"</a>
1188 below) can be used.
1189 </P>
1190 <P>
1191 A back reference that occurs inside the parentheses to which it refers fails
1192 when the subpattern is first used, so, for example, (a\1) never matches.
1193 However, such references can be useful inside repeated subpatterns. For
1194 example, the pattern
1195 <pre>
1196 (a|b\1)+
1197 </pre>
1198 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1199 the subpattern, the back reference matches the character string corresponding
1200 to the previous iteration. In order for this to work, the pattern must be such
1201 that the first iteration does not need to match the back reference. This can be
1202 done using alternation, as in the example above, or by a quantifier with a
1203 minimum of zero.
1204 <a name="bigassertions"></a></P>
1205 <br><a name="SEC15" href="#TOC1">ASSERTIONS</a><br>
1206 <P>
1207 An assertion is a test on the characters following or preceding the current
1208 matching point that does not actually consume any characters. The simple
1209 assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are described
1210 <a href="#smallassertions">above.</a>
1211 </P>
1212 <P>
1213 More complicated assertions are coded as subpatterns. There are two kinds:
1214 those that look ahead of the current position in the subject string, and those
1215 that look behind it. An assertion subpattern is matched in the normal way,
1216 except that it does not cause the current matching position to be changed.
1217 </P>
1218 <P>
1219 Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1220 because it makes no sense to assert the same thing several times. If any kind
1221 of assertion contains capturing subpatterns within it, these are counted for
1222 the purposes of numbering the capturing subpatterns in the whole pattern.
1223 However, substring capturing is carried out only for positive assertions,
1224 because it does not make sense for negative assertions.
1225 </P>
1226 <br><b>
1227 Lookahead assertions
1228 </b><br>
1229 <P>
1230 Lookahead assertions start
1231 with (?= for positive assertions and (?! for negative assertions. For example,
1232 <pre>
1233 \w+(?=;)
1234 </pre>
1235 matches a word followed by a semicolon, but does not include the semicolon in
1236 the match, and
1237 <pre>
1238 foo(?!bar)
1239 </pre>
1240 matches any occurrence of "foo" that is not followed by "bar". Note that the
1241 apparently similar pattern
1242 <pre>
1243 (?!foo)bar
1244 </pre>
1245 does not find an occurrence of "bar" that is preceded by something other than
1246 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1247 (?!foo) is always true when the next three characters are "bar". A
1248 lookbehind assertion is needed to achieve the other effect.
1249 </P>
1250 <P>
1251 If you want to force a matching failure at some point in a pattern, the most
1252 convenient way to do it is with (?!) because an empty string always matches, so
1253 an assertion that requires there not to be an empty string must always fail.
1254 <a name="lookbehind"></a></P>
1255 <br><b>
1256 Lookbehind assertions
1257 </b><br>
1258 <P>
1259 Lookbehind assertions start with (?&#60;= for positive assertions and (?&#60;! for
1260 negative assertions. For example,
1261 <pre>
1262 (?&#60;!foo)bar
1263 </pre>
1264 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1265 a lookbehind assertion are restricted such that all the strings it matches must
1266 have a fixed length. However, if there are several alternatives, they do not
1267 all have to have the same fixed length. Thus
1268 <pre>
1269 (?&#60;=bullock|donkey)
1270 </pre>
1271 is permitted, but
1272 <pre>
1273 (?&#60;!dogs?|cats?)
1274 </pre>
1275 causes an error at compile time. Branches that match different length strings
1276 are permitted only at the top level of a lookbehind assertion. This is an
1277 extension compared with Perl (at least for 5.8), which requires all branches to
1278 match the same length of string. An assertion such as
1279 <pre>
1280 (?&#60;=ab(c|de))
1281 </pre>
1282 is not permitted, because its single top-level branch can match two different
1283 lengths, but it is acceptable if rewritten to use two top-level branches:
1284 <pre>
1285 (?&#60;=abc|abde)
1286 </pre>
1287 The implementation of lookbehind assertions is, for each alternative, to
1288 temporarily move the current position back by the fixed width and then try to
1289 match. If there are insufficient characters before the current position, the
1290 match is deemed to fail.
1291 </P>
1292 <P>
1293 PCRE does not allow the \C escape (which matches a single byte in UTF-8 mode)
1294 to appear in lookbehind assertions, because it makes it impossible to calculate
1295 the length of the lookbehind. The \X escape, which can match different numbers
1296 of bytes, is also not permitted.
1297 </P>
1298 <P>
1299 Atomic groups can be used in conjunction with lookbehind assertions to specify
1300 efficient matching at the end of the subject string. Consider a simple pattern
1301 such as
1302 <pre>
1303 abcd$
1304 </pre>
1305 when applied to a long string that does not match. Because matching proceeds
1306 from left to right, PCRE will look for each "a" in the subject and then see if
1307 what follows matches the rest of the pattern. If the pattern is specified as
1308 <pre>
1309 ^.*abcd$
1310 </pre>
1311 the initial .* matches the entire string at first, but when this fails (because
1312 there is no following "a"), it backtracks to match all but the last character,
1313 then all but the last two characters, and so on. Once again the search for "a"
1314 covers the entire string, from right to left, so we are no better off. However,
1315 if the pattern is written as
1316 <pre>
1317 ^(?&#62;.*)(?&#60;=abcd)
1318 </pre>
1319 or, equivalently, using the possessive quantifier syntax,
1320 <pre>
1321 ^.*+(?&#60;=abcd)
1322 </pre>
1323 there can be no backtracking for the .* item; it can match only the entire
1324 string. The subsequent lookbehind assertion does a single test on the last four
1325 characters. If it fails, the match fails immediately. For long strings, this
1326 approach makes a significant difference to the processing time.
1327 </P>
1328 <br><b>
1329 Using multiple assertions
1330 </b><br>
1331 <P>
1332 Several assertions (of any sort) may occur in succession. For example,
1333 <pre>
1334 (?&#60;=\d{3})(?&#60;!999)foo
1335 </pre>
1336 matches "foo" preceded by three digits that are not "999". Notice that each of
1337 the assertions is applied independently at the same point in the subject
1338 string. First there is a check that the previous three characters are all
1339 digits, and then there is a check that the same three characters are not "999".
1340 This pattern does <i>not</i> match "foo" preceded by six characters, the first
1341 of which are digits and the last three of which are not "999". For example, it
1342 doesn't match "123abcfoo". A pattern to do that is
1343 <pre>
1344 (?&#60;=\d{3}...)(?&#60;!999)foo
1345 </pre>
1346 This time the first assertion looks at the preceding six characters, checking
1347 that the first three are digits, and then the second assertion checks that the
1348 preceding three characters are not "999".
1349 </P>
1350 <P>
1351 Assertions can be nested in any combination. For example,
1352 <pre>
1353 (?&#60;=(?&#60;!foo)bar)baz
1354 </pre>
1355 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1356 preceded by "foo", while
1357 <pre>
1358 (?&#60;=\d{3}(?!999)...)foo
1359 </pre>
1360 is another pattern that matches "foo" preceded by three digits and any three
1361 characters that are not "999".
1362 </P>
1363 <br><a name="SEC16" href="#TOC1">CONDITIONAL SUBPATTERNS</a><br>
1364 <P>
1365 It is possible to cause the matching process to obey a subpattern
1366 conditionally or to choose between two alternative subpatterns, depending on
1367 the result of an assertion, or whether a previous capturing subpattern matched
1368 or not. The two possible forms of conditional subpattern are
1369 <pre>
1370 (?(condition)yes-pattern)
1371 (?(condition)yes-pattern|no-pattern)
1372 </pre>
1373 If the condition is satisfied, the yes-pattern is used; otherwise the
1374 no-pattern (if present) is used. If there are more than two alternatives in the
1375 subpattern, a compile-time error occurs.
1376 </P>
1377 <P>
1378 There are three kinds of condition. If the text between the parentheses
1379 consists of a sequence of digits, the condition is satisfied if the capturing
1380 subpattern of that number has previously matched. The number must be greater
1381 than zero. Consider the following pattern, which contains non-significant white
1382 space to make it more readable (assume the PCRE_EXTENDED option) and to divide
1383 it into three parts for ease of discussion:
1384 <pre>
1385 ( \( )? [^()]+ (?(1) \) )
1386 </pre>
1387 The first part matches an optional opening parenthesis, and if that
1388 character is present, sets it as the first captured substring. The second part
1389 matches one or more characters that are not parentheses. The third part is a
1390 conditional subpattern that tests whether the first set of parentheses matched
1391 or not. If they did, that is, if subject started with an opening parenthesis,
1392 the condition is true, and so the yes-pattern is executed and a closing
1393 parenthesis is required. Otherwise, since no-pattern is not present, the
1394 subpattern matches nothing. In other words, this pattern matches a sequence of
1395 non-parentheses, optionally enclosed in parentheses.
1396 </P>
1397 <P>
1398 If the condition is the string (R), it is satisfied if a recursive call to the
1399 pattern or subpattern has been made. At "top level", the condition is false.
1400 This is a PCRE extension. Recursive patterns are described in the next section.
1401 </P>
1402 <P>
1403 If the condition is not a sequence of digits or (R), it must be an assertion.
1404 This may be a positive or negative lookahead or lookbehind assertion. Consider
1405 this pattern, again containing non-significant white space, and with the two
1406 alternatives on the second line:
1407 <pre>
1408 (?(?=[^a-z]*[a-z])
1409 \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )
1410 </pre>
1411 The condition is a positive lookahead assertion that matches an optional
1412 sequence of non-letters followed by a letter. In other words, it tests for the
1413 presence of at least one letter in the subject. If a letter is found, the
1414 subject is matched against the first alternative; otherwise it is matched
1415 against the second. This pattern matches strings in one of the two forms
1416 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
1417 <a name="comments"></a></P>
1418 <br><a name="SEC17" href="#TOC1">COMMENTS</a><br>
1419 <P>
1420 The sequence (?# marks the start of a comment that continues up to the next
1421 closing parenthesis. Nested parentheses are not permitted. The characters
1422 that make up a comment play no part in the pattern matching at all.
1423 </P>
1424 <P>
1425 If the PCRE_EXTENDED option is set, an unescaped # character outside a
1426 character class introduces a comment that continues up to the next newline
1427 character in the pattern.
1428 </P>
1429 <br><a name="SEC18" href="#TOC1">RECURSIVE PATTERNS</a><br>
1430 <P>
1431 Consider the problem of matching a string in parentheses, allowing for
1432 unlimited nested parentheses. Without the use of recursion, the best that can
1433 be done is to use a pattern that matches up to some fixed depth of nesting. It
1434 is not possible to handle an arbitrary nesting depth. Perl provides a facility
1435 that allows regular expressions to recurse (amongst other things). It does this
1436 by interpolating Perl code in the expression at run time, and the code can
1437 refer to the expression itself. A Perl pattern to solve the parentheses problem
1438 can be created like this:
1439 <pre>
1440 $re = qr{\( (?: (?&#62;[^()]+) | (?p{$re}) )* \)}x;
1441 </pre>
1442 The (?p{...}) item interpolates Perl code at run time, and in this case refers
1443 recursively to the pattern in which it appears. Obviously, PCRE cannot support
1444 the interpolation of Perl code. Instead, it supports some special syntax for
1445 recursion of the entire pattern, and also for individual subpattern recursion.
1446 </P>
1447 <P>
1448 The special item that consists of (? followed by a number greater than zero and
1449 a closing parenthesis is a recursive call of the subpattern of the given
1450 number, provided that it occurs inside that subpattern. (If not, it is a
1451 "subroutine" call, which is described in the next section.) The special item
1452 (?R) is a recursive call of the entire regular expression.
1453 </P>
1454 <P>
1455 A recursive subpattern call is always treated as an atomic group. That is, once
1456 it has matched some of the subject string, it is never re-entered, even if
1457 it contains untried alternatives and there is a subsequent matching failure.
1458 </P>
1459 <P>
1460 This PCRE pattern solves the nested parentheses problem (assume the
1461 PCRE_EXTENDED option is set so that white space is ignored):
1462 <pre>
1463 \( ( (?&#62;[^()]+) | (?R) )* \)
1464 </pre>
1465 First it matches an opening parenthesis. Then it matches any number of
1466 substrings which can either be a sequence of non-parentheses, or a recursive
1467 match of the pattern itself (that is, a correctly parenthesized substring).
1468 Finally there is a closing parenthesis.
1469 </P>
1470 <P>
1471 If this were part of a larger pattern, you would not want to recurse the entire
1472 pattern, so instead you could use this:
1473 <pre>
1474 ( \( ( (?&#62;[^()]+) | (?1) )* \) )
1475 </pre>
1476 We have put the pattern into parentheses, and caused the recursion to refer to
1477 them instead of the whole pattern. In a larger pattern, keeping track of
1478 parenthesis numbers can be tricky. It may be more convenient to use named
1479 parentheses instead. For this, PCRE uses (?P&#62;name), which is an extension to
1480 the Python syntax that PCRE uses for named parentheses (Perl does not provide
1481 named parentheses). We could rewrite the above example as follows:
1482 <pre>
1483 (?P&#60;pn&#62; \( ( (?&#62;[^()]+) | (?P&#62;pn) )* \) )
1484 </pre>
1485 This particular example pattern contains nested unlimited repeats, and so the
1486 use of atomic grouping for matching strings of non-parentheses is important
1487 when applying the pattern to strings that do not match. For example, when this
1488 pattern is applied to
1489 <pre>
1490 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
1491 </pre>
1492 it yields "no match" quickly. However, if atomic grouping is not used,
1493 the match runs for a very long time indeed because there are so many different
1494 ways the + and * repeats can carve up the subject, and all have to be tested
1495 before failure can be reported.
1496 </P>
1497 <P>
1498 At the end of a match, the values set for any capturing subpatterns are those
1499 from the outermost level of the recursion at which the subpattern value is set.
1500 If you want to obtain intermediate values, a callout function can be used (see
1501 the next section and the
1502 <a href="pcrecallout.html"><b>pcrecallout</b></a>
1503 documentation). If the pattern above is matched against
1504 <pre>
1505 (ab(cd)ef)
1506 </pre>
1507 the value for the capturing parentheses is "ef", which is the last value taken
1508 on at the top level. If additional parentheses are added, giving
1509 <pre>
1510 \( ( ( (?&#62;[^()]+) | (?R) )* ) \)
1511 ^ ^
1512 ^ ^
1513 </pre>
1514 the string they capture is "ab(cd)ef", the contents of the top level
1515 parentheses. If there are more than 15 capturing parentheses in a pattern, PCRE
1516 has to obtain extra memory to store data during a recursion, which it does by
1517 using <b>pcre_malloc</b>, freeing it via <b>pcre_free</b> afterwards. If no
1518 memory can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
1519 </P>
1520 <P>
1521 Do not confuse the (?R) item with the condition (R), which tests for recursion.
1522 Consider this pattern, which matches text in angle brackets, allowing for
1523 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
1524 recursing), whereas any characters are permitted at the outer level.
1525 <pre>
1526 &#60; (?: (?(R) \d++ | [^&#60;&#62;]*+) | (?R)) * &#62;
1527 </pre>
1528 In this pattern, (?(R) is the start of a conditional subpattern, with two
1529 different alternatives for the recursive and non-recursive cases. The (?R) item
1530 is the actual recursive call.
1531 <a name="subpatternsassubroutines"></a></P>
1532 <br><a name="SEC19" href="#TOC1">SUBPATTERNS AS SUBROUTINES</a><br>
1533 <P>
1534 If the syntax for a recursive subpattern reference (either by number or by
1535 name) is used outside the parentheses to which it refers, it operates like a
1536 subroutine in a programming language. An earlier example pointed out that the
1537 pattern
1538 <pre>
1539 (sens|respons)e and \1ibility
1540 </pre>
1541 matches "sense and sensibility" and "response and responsibility", but not
1542 "sense and responsibility". If instead the pattern
1543 <pre>
1544 (sens|respons)e and (?1)ibility
1545 </pre>
1546 is used, it does match "sense and responsibility" as well as the other two
1547 strings. Such references must, however, follow the subpattern to which they
1548 refer.
1549 </P>
1550 <P>
1551 Like recursive subpatterns, a "subroutine" call is always treated as an atomic
1552 group. That is, once it has matched some of the subject string, it is never
1553 re-entered, even if it contains untried alternatives and there is a subsequent
1554 matching failure.
1555 </P>
1556 <br><a name="SEC20" href="#TOC1">CALLOUTS</a><br>
1557 <P>
1558 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
1559 code to be obeyed in the middle of matching a regular expression. This makes it
1560 possible, amongst other things, to extract different substrings that match the
1561 same pair of parentheses when there is a repetition.
1562 </P>
1563 <P>
1564 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
1565 code. The feature is called "callout". The caller of PCRE provides an external
1566 function by putting its entry point in the global variable <i>pcre_callout</i>.
1567 By default, this variable contains NULL, which disables all calling out.
1568 </P>
1569 <P>
1570 Within a regular expression, (?C) indicates the points at which the external
1571 function is to be called. If you want to identify different callout points, you
1572 can put a number less than 256 after the letter C. The default value is zero.
1573 For example, this pattern has two callout points:
1574 <pre>
1575 (?C1)\dabc(?C2)def
1576 </pre>
1577 If the PCRE_AUTO_CALLOUT flag is passed to <b>pcre_compile()</b>, callouts are
1578 automatically installed before each item in the pattern. They are all numbered
1579 255.
1580 </P>
1581 <P>
1582 During matching, when PCRE reaches a callout point (and <i>pcre_callout</i> is
1583 set), the external function is called. It is provided with the number of the
1584 callout, the position in the pattern, and, optionally, one item of data
1585 originally supplied by the caller of <b>pcre_exec()</b>. The callout function
1586 may cause matching to proceed, to backtrack, or to fail altogether. A complete
1587 description of the interface to the callout function is given in the
1588 <a href="pcrecallout.html"><b>pcrecallout</b></a>
1589 documentation.
1590 </P>
1591 <P>
1592 Last updated: 24 January 2006
1593 <br>
1594 Copyright &copy; 1997-2006 University of Cambridge.
1595 <p>
1596 Return to the <a href="index.html">PCRE index page</a>.
1597 </p>

  ViewVC Help
Powered by ViewVC 1.1.5