3 <title>pcrepattern specification</title>
5 <body bgcolor="#FFFFFF" text="#00005A" link="#0066FF" alink="#3399FF" vlink="#2222BB">
6 <h1>pcrepattern man page</h1>
8 Return to the <a href="index.html">PCRE index page</a>.
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.
16 <li><a name="TOC1" href="#SEC1">PCRE REGULAR EXPRESSION DETAILS</a>
17 <li><a name="TOC2" href="#SEC2">SPECIAL START-OF-PATTERN ITEMS</a>
18 <li><a name="TOC3" href="#SEC3">EBCDIC CHARACTER CODES</a>
19 <li><a name="TOC4" href="#SEC4">CHARACTERS AND METACHARACTERS</a>
20 <li><a name="TOC5" href="#SEC5">BACKSLASH</a>
21 <li><a name="TOC6" href="#SEC6">CIRCUMFLEX AND DOLLAR</a>
22 <li><a name="TOC7" href="#SEC7">FULL STOP (PERIOD, DOT) AND \N</a>
23 <li><a name="TOC8" href="#SEC8">MATCHING A SINGLE DATA UNIT</a>
24 <li><a name="TOC9" href="#SEC9">SQUARE BRACKETS AND CHARACTER CLASSES</a>
25 <li><a name="TOC10" href="#SEC10">POSIX CHARACTER CLASSES</a>
26 <li><a name="TOC11" href="#SEC11">COMPATIBILITY FEATURE FOR WORD BOUNDARIES</a>
27 <li><a name="TOC12" href="#SEC12">VERTICAL BAR</a>
28 <li><a name="TOC13" href="#SEC13">INTERNAL OPTION SETTING</a>
29 <li><a name="TOC14" href="#SEC14">SUBPATTERNS</a>
30 <li><a name="TOC15" href="#SEC15">DUPLICATE SUBPATTERN NUMBERS</a>
31 <li><a name="TOC16" href="#SEC16">NAMED SUBPATTERNS</a>
32 <li><a name="TOC17" href="#SEC17">REPETITION</a>
33 <li><a name="TOC18" href="#SEC18">ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS</a>
34 <li><a name="TOC19" href="#SEC19">BACK REFERENCES</a>
35 <li><a name="TOC20" href="#SEC20">ASSERTIONS</a>
36 <li><a name="TOC21" href="#SEC21">CONDITIONAL SUBPATTERNS</a>
37 <li><a name="TOC22" href="#SEC22">COMMENTS</a>
38 <li><a name="TOC23" href="#SEC23">RECURSIVE PATTERNS</a>
39 <li><a name="TOC24" href="#SEC24">SUBPATTERNS AS SUBROUTINES</a>
40 <li><a name="TOC25" href="#SEC25">ONIGURUMA SUBROUTINE SYNTAX</a>
41 <li><a name="TOC26" href="#SEC26">CALLOUTS</a>
42 <li><a name="TOC27" href="#SEC27">BACKTRACKING CONTROL</a>
43 <li><a name="TOC28" href="#SEC28">SEE ALSO</a>
44 <li><a name="TOC29" href="#SEC29">AUTHOR</a>
45 <li><a name="TOC30" href="#SEC30">REVISION</a>
47 <br><a name="SEC1" href="#TOC1">PCRE REGULAR EXPRESSION DETAILS</a><br>
49 The syntax and semantics of the regular expressions that are supported by PCRE
50 are described in detail below. There is a quick-reference syntax summary in the
51 <a href="pcresyntax.html"><b>pcresyntax</b></a>
52 page. PCRE tries to match Perl syntax and semantics as closely as it can. PCRE
53 also supports some alternative regular expression syntax (which does not
54 conflict with the Perl syntax) in order to provide some compatibility with
55 regular expressions in Python, .NET, and Oniguruma.
58 Perl's regular expressions are described in its own documentation, and
59 regular expressions in general are covered in a number of books, some of which
60 have copious examples. Jeffrey Friedl's "Mastering Regular Expressions",
61 published by O'Reilly, covers regular expressions in great detail. This
62 description of PCRE's regular expressions is intended as reference material.
65 This document discusses the patterns that are supported by PCRE when one its
66 main matching functions, <b>pcre_exec()</b> (8-bit) or <b>pcre[16|32]_exec()</b>
67 (16- or 32-bit), is used. PCRE also has alternative matching functions,
68 <b>pcre_dfa_exec()</b> and <b>pcre[16|32_dfa_exec()</b>, which match using a
69 different algorithm that is not Perl-compatible. Some of the features discussed
70 below are not available when DFA matching is used. The advantages and
71 disadvantages of the alternative functions, and how they differ from the normal
72 functions, are discussed in the
73 <a href="pcrematching.html"><b>pcrematching</b></a>
76 <br><a name="SEC2" href="#TOC1">SPECIAL START-OF-PATTERN ITEMS</a><br>
78 A number of options that can be passed to <b>pcre_compile()</b> can also be set
79 by special items at the start of a pattern. These are not Perl-compatible, but
80 are provided to make these options accessible to pattern writers who are not
81 able to change the program that processes the pattern. Any number of these
82 items may appear, but they must all be together right at the start of the
83 pattern string, and the letters must be in upper case.
89 The original operation of PCRE was on strings of one-byte characters. However,
90 there is now also support for UTF-8 strings in the original library, an
91 extra library that supports 16-bit and UTF-16 character strings, and a
92 third library that supports 32-bit and UTF-32 character strings. To use these
93 features, PCRE must be built to include appropriate support. When using UTF
94 strings you must either call the compiling function with the PCRE_UTF8,
95 PCRE_UTF16, or PCRE_UTF32 option, or the pattern must start with one of
96 these special sequences:
103 (*UTF) is a generic sequence that can be used with any of the libraries.
104 Starting a pattern with such a sequence is equivalent to setting the relevant
105 option. How setting a UTF mode affects pattern matching is mentioned in several
106 places below. There is also a summary of features in the
107 <a href="pcreunicode.html"><b>pcreunicode</b></a>
111 Some applications that allow their users to supply patterns may wish to
112 restrict them to non-UTF data for security reasons. If the PCRE_NEVER_UTF
113 option is set at compile time, (*UTF) etc. are not allowed, and their
114 appearance causes an error.
117 Unicode property support
120 Another special sequence that may appear at the start of a pattern is (*UCP).
121 This has the same effect as setting the PCRE_UCP option: it causes sequences
122 such as \d and \w to use Unicode properties to determine character types,
123 instead of recognizing only characters with codes less than 128 via a lookup
127 Disabling auto-possessification
130 If a pattern starts with (*NO_AUTO_POSSESS), it has the same effect as setting
131 the PCRE_NO_AUTO_POSSESS option at compile time. This stops PCRE from making
132 quantifiers possessive when what follows cannot match the repeated item. For
133 example, by default a+b is treated as a++b. For more details, see the
134 <a href="pcreapi.html"><b>pcreapi</b></a>
138 Disabling start-up optimizations
141 If a pattern starts with (*NO_START_OPT), it has the same effect as setting the
142 PCRE_NO_START_OPTIMIZE option either at compile or matching time. This disables
143 several optimizations for quickly reaching "no match" results. For more
145 <a href="pcreapi.html"><b>pcreapi</b></a>
147 <a name="newlines"></a></P>
152 PCRE supports five different conventions for indicating line breaks in
153 strings: a single CR (carriage return) character, a single LF (linefeed)
154 character, the two-character sequence CRLF, any of the three preceding, or any
155 Unicode newline sequence. The
156 <a href="pcreapi.html"><b>pcreapi</b></a>
158 <a href="pcreapi.html#newlines">further discussion</a>
159 about newlines, and shows how to set the newline convention in the
160 <i>options</i> arguments for the compiling and matching functions.
163 It is also possible to specify a newline convention by starting a pattern
164 string with one of the following five sequences:
166 (*CR) carriage return
168 (*CRLF) carriage return, followed by linefeed
169 (*ANYCRLF) any of the three above
170 (*ANY) all Unicode newline sequences
172 These override the default and the options given to the compiling function. For
173 example, on a Unix system where LF is the default newline sequence, the pattern
177 changes the convention to CR. That pattern matches "a\nb" because LF is no
178 longer a newline. If more than one of these settings is present, the last one
182 The newline convention affects where the circumflex and dollar assertions are
183 true. It also affects the interpretation of the dot metacharacter when
184 PCRE_DOTALL is not set, and the behaviour of \N. However, it does not affect
185 what the \R escape sequence matches. By default, this is any Unicode newline
186 sequence, for Perl compatibility. However, this can be changed; see the
187 description of \R in the section entitled
188 <a href="#newlineseq">"Newline sequences"</a>
189 below. A change of \R setting can be combined with a change of newline
193 Setting match and recursion limits
196 The caller of <b>pcre_exec()</b> can set a limit on the number of times the
197 internal <b>match()</b> function is called and on the maximum depth of
198 recursive calls. These facilities are provided to catch runaway matches that
199 are provoked by patterns with huge matching trees (a typical example is a
200 pattern with nested unlimited repeats) and to avoid running out of system stack
201 by too much recursion. When one of these limits is reached, <b>pcre_exec()</b>
202 gives an error return. The limits can also be set by items at the start of the
208 where d is any number of decimal digits. However, the value of the setting must
209 be less than the value set (or defaulted) by the caller of <b>pcre_exec()</b>
210 for it to have any effect. In other words, the pattern writer can lower the
211 limits set by the programmer, but not raise them. If there is more than one
212 setting of one of these limits, the lower value is used.
214 <br><a name="SEC3" href="#TOC1">EBCDIC CHARACTER CODES</a><br>
216 PCRE can be compiled to run in an environment that uses EBCDIC as its character
217 code rather than ASCII or Unicode (typically a mainframe system). In the
218 sections below, character code values are ASCII or Unicode; in an EBCDIC
219 environment these characters may have different code values, and there are no
220 code points greater than 255.
222 <br><a name="SEC4" href="#TOC1">CHARACTERS AND METACHARACTERS</a><br>
224 A regular expression is a pattern that is matched against a subject string from
225 left to right. Most characters stand for themselves in a pattern, and match the
226 corresponding characters in the subject. As a trivial example, the pattern
230 matches a portion of a subject string that is identical to itself. When
231 caseless matching is specified (the PCRE_CASELESS option), letters are matched
232 independently of case. In a UTF mode, PCRE always understands the concept of
233 case for characters whose values are less than 128, so caseless matching is
234 always possible. For characters with higher values, the concept of case is
235 supported if PCRE is compiled with Unicode property support, but not otherwise.
236 If you want to use caseless matching for characters 128 and above, you must
237 ensure that PCRE is compiled with Unicode property support as well as with
241 The power of regular expressions comes from the ability to include alternatives
242 and repetitions in the pattern. These are encoded in the pattern by the use of
243 <i>metacharacters</i>, which do not stand for themselves but instead are
244 interpreted in some special way.
247 There are two different sets of metacharacters: those that are recognized
248 anywhere in the pattern except within square brackets, and those that are
249 recognized within square brackets. Outside square brackets, the metacharacters
252 \ general escape character with several uses
253 ^ assert start of string (or line, in multiline mode)
254 $ assert end of string (or line, in multiline mode)
255 . match any character except newline (by default)
256 [ start character class definition
257 | start of alternative branch
260 ? extends the meaning of (
261 also 0 or 1 quantifier
262 also quantifier minimizer
263 * 0 or more quantifier
264 + 1 or more quantifier
265 also "possessive quantifier"
266 { start min/max quantifier
268 Part of a pattern that is in square brackets is called a "character class". In
269 a character class the only metacharacters are:
271 \ general escape character
272 ^ negate the class, but only if the first character
273 - indicates character range
274 [ POSIX character class (only if followed by POSIX syntax)
275 ] terminates the character class
277 The following sections describe the use of each of the metacharacters.
279 <br><a name="SEC5" href="#TOC1">BACKSLASH</a><br>
281 The backslash character has several uses. Firstly, if it is followed by a
282 character that is not a number or a letter, it takes away any special meaning
283 that character may have. This use of backslash as an escape character applies
284 both inside and outside character classes.
287 For example, if you want to match a * character, you write \* in the pattern.
288 This escaping action applies whether or not the following character would
289 otherwise be interpreted as a metacharacter, so it is always safe to precede a
290 non-alphanumeric with backslash to specify that it stands for itself. In
291 particular, if you want to match a backslash, you write \\.
294 In a UTF mode, only ASCII numbers and letters have any special meaning after a
295 backslash. All other characters (in particular, those whose codepoints are
296 greater than 127) are treated as literals.
299 If a pattern is compiled with the PCRE_EXTENDED option, most white space in the
300 pattern (other than in a character class), and characters between a # outside a
301 character class and the next newline, inclusive, are ignored. An escaping
302 backslash can be used to include a white space or # character as part of the
306 If you want to remove the special meaning from a sequence of characters, you
307 can do so by putting them between \Q and \E. This is different from Perl in
308 that $ and @ are handled as literals in \Q...\E sequences in PCRE, whereas in
309 Perl, $ and @ cause variable interpolation. Note the following examples:
311 Pattern PCRE matches Perl matches
313 \Qabc$xyz\E abc$xyz abc followed by the contents of $xyz
314 \Qabc\$xyz\E abc\$xyz abc\$xyz
315 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
317 The \Q...\E sequence is recognized both inside and outside character classes.
318 An isolated \E that is not preceded by \Q is ignored. If \Q is not followed
319 by \E later in the pattern, the literal interpretation continues to the end of
320 the pattern (that is, \E is assumed at the end). If the isolated \Q is inside
321 a character class, this causes an error, because the character class is not
323 <a name="digitsafterbackslash"></a></P>
325 Non-printing characters
328 A second use of backslash provides a way of encoding non-printing characters
329 in patterns in a visible manner. There is no restriction on the appearance of
330 non-printing characters, apart from the binary zero that terminates a pattern,
331 but when a pattern is being prepared by text editing, it is often easier to use
332 one of the following escape sequences than the binary character it represents.
333 In an ASCII or Unicode environment, these escapes are as follows:
335 \a alarm, that is, the BEL character (hex 07)
336 \cx "control-x", where x is any ASCII character
338 \f form feed (hex 0C)
340 \r carriage return (hex 0D)
342 \0dd character with octal code 0dd
343 \ddd character with octal code ddd, or back reference
344 \o{ddd..} character with octal code ddd..
345 \xhh character with hex code hh
346 \x{hhh..} character with hex code hhh.. (non-JavaScript mode)
347 \uhhhh character with hex code hhhh (JavaScript mode only)
349 The precise effect of \cx on ASCII characters is as follows: if x is a lower
350 case letter, it is converted to upper case. Then bit 6 of the character (hex
351 40) is inverted. Thus \cA to \cZ become hex 01 to hex 1A (A is 41, Z is 5A),
352 but \c{ becomes hex 3B ({ is 7B), and \c; becomes hex 7B (; is 3B). If the
353 data item (byte or 16-bit value) following \c has a value greater than 127, a
354 compile-time error occurs. This locks out non-ASCII characters in all modes.
357 When PCRE is compiled in EBCDIC mode, \a, \e, \f, \n, \r, and \t
358 generate the appropriate EBCDIC code values. The \c escape is processed
359 as specified for Perl in the <b>perlebcdic</b> document. The only characters
360 that are allowed after \c are A-Z, a-z, or one of @, [, \, ], ^, _, or ?. Any
361 other character provokes a compile-time error. The sequence \@ encodes
362 character code 0; the letters (in either case) encode characters 1-26 (hex 01
363 to hex 1A); [, \, ], ^, and _ encode characters 27-31 (hex 1B to hex 1F), and
364 \? becomes either 255 (hex FF) or 95 (hex 5F).
367 Thus, apart from \?, these escapes generate the same character code values as
368 they do in an ASCII environment, though the meanings of the values mostly
369 differ. For example, \G always generates code value 7, which is BEL in ASCII
373 The sequence \? generates DEL (127, hex 7F) in an ASCII environment, but
374 because 127 is not a control character in EBCDIC, Perl makes it generate the
375 APC character. Unfortunately, there are several variants of EBCDIC. In most of
376 them the APC character has the value 255 (hex FF), but in the one Perl calls
377 POSIX-BC its value is 95 (hex 5F). If certain other characters have POSIX-BC
378 values, PCRE makes \? generate 95; otherwise it generates 255.
381 After \0 up to two further octal digits are read. If there are fewer than two
382 digits, just those that are present are used. Thus the sequence \0\x\015
383 specifies two binary zeros followed by a CR character (code value 13). Make
384 sure you supply two digits after the initial zero if the pattern character that
385 follows is itself an octal digit.
388 The escape \o must be followed by a sequence of octal digits, enclosed in
389 braces. An error occurs if this is not the case. This escape is a recent
390 addition to Perl; it provides way of specifying character code points as octal
391 numbers greater than 0777, and it also allows octal numbers and back references
392 to be unambiguously specified.
395 For greater clarity and unambiguity, it is best to avoid following \ by a
396 digit greater than zero. Instead, use \o{} or \x{} to specify character
397 numbers, and \g{} to specify back references. The following paragraphs
398 describe the old, ambiguous syntax.
401 The handling of a backslash followed by a digit other than 0 is complicated,
402 and Perl has changed in recent releases, causing PCRE also to change. Outside a
403 character class, PCRE reads the digit and any following digits as a decimal
404 number. If the number is less than 8, or if there have been at least that many
405 previous capturing left parentheses in the expression, the entire sequence is
406 taken as a <i>back reference</i>. A description of how this works is given
407 <a href="#backreferences">later,</a>
408 following the discussion of
409 <a href="#subpattern">parenthesized subpatterns.</a>
412 Inside a character class, or if the decimal number following \ is greater than
413 7 and there have not been that many capturing subpatterns, PCRE handles \8 and
414 \9 as the literal characters "8" and "9", and otherwise re-reads up to three
415 octal digits following the backslash, using them to generate a data character.
416 Any subsequent digits stand for themselves. For example:
418 \040 is another way of writing an ASCII space
419 \40 is the same, provided there are fewer than 40 previous capturing subpatterns
420 \7 is always a back reference
421 \11 might be a back reference, or another way of writing a tab
423 \0113 is a tab followed by the character "3"
424 \113 might be a back reference, otherwise the character with octal code 113
425 \377 might be a back reference, otherwise the value 255 (decimal)
426 \81 is either a back reference, or the two characters "8" and "1"
428 Note that octal values of 100 or greater that are specified using this syntax
429 must not be introduced by a leading zero, because no more than three octal
430 digits are ever read.
433 By default, after \x that is not followed by {, from zero to two hexadecimal
434 digits are read (letters can be in upper or lower case). Any number of
435 hexadecimal digits may appear between \x{ and }. If a character other than
436 a hexadecimal digit appears between \x{ and }, or if there is no terminating
440 If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \x is
441 as just described only when it is followed by two hexadecimal digits.
442 Otherwise, it matches a literal "x" character. In JavaScript mode, support for
443 code points greater than 256 is provided by \u, which must be followed by
444 four hexadecimal digits; otherwise it matches a literal "u" character.
447 Characters whose value is less than 256 can be defined by either of the two
448 syntaxes for \x (or by \u in JavaScript mode). There is no difference in the
449 way they are handled. For example, \xdc is exactly the same as \x{dc} (or
450 \u00dc in JavaScript mode).
453 Constraints on character values
456 Characters that are specified using octal or hexadecimal numbers are
457 limited to certain values, as follows:
459 8-bit non-UTF mode less than 0x100
460 8-bit UTF-8 mode less than 0x10ffff and a valid codepoint
461 16-bit non-UTF mode less than 0x10000
462 16-bit UTF-16 mode less than 0x10ffff and a valid codepoint
463 32-bit non-UTF mode less than 0x100000000
464 32-bit UTF-32 mode less than 0x10ffff and a valid codepoint
466 Invalid Unicode codepoints are the range 0xd800 to 0xdfff (the so-called
467 "surrogate" codepoints), and 0xffef.
470 Escape sequences in character classes
473 All the sequences that define a single character value can be used both inside
474 and outside character classes. In addition, inside a character class, \b is
475 interpreted as the backspace character (hex 08).
478 \N is not allowed in a character class. \B, \R, and \X are not special
479 inside a character class. Like other unrecognized escape sequences, they are
480 treated as the literal characters "B", "R", and "X" by default, but cause an
481 error if the PCRE_EXTRA option is set. Outside a character class, these
482 sequences have different meanings.
485 Unsupported escape sequences
488 In Perl, the sequences \l, \L, \u, and \U are recognized by its string
489 handler and used to modify the case of following characters. By default, PCRE
490 does not support these escape sequences. However, if the PCRE_JAVASCRIPT_COMPAT
491 option is set, \U matches a "U" character, and \u can be used to define a
492 character by code point, as described in the previous section.
495 Absolute and relative back references
498 The sequence \g followed by an unsigned or a negative number, optionally
499 enclosed in braces, is an absolute or relative back reference. A named back
500 reference can be coded as \g{name}. Back references are discussed
501 <a href="#backreferences">later,</a>
502 following the discussion of
503 <a href="#subpattern">parenthesized subpatterns.</a>
506 Absolute and relative subroutine calls
509 For compatibility with Oniguruma, the non-Perl syntax \g followed by a name or
510 a number enclosed either in angle brackets or single quotes, is an alternative
511 syntax for referencing a subpattern as a "subroutine". Details are discussed
512 <a href="#onigurumasubroutines">later.</a>
513 Note that \g{...} (Perl syntax) and \g<...> (Oniguruma syntax) are <i>not</i>
514 synonymous. The former is a back reference; the latter is a
515 <a href="#subpatternsassubroutines">subroutine</a>
517 <a name="genericchartypes"></a></P>
519 Generic character types
522 Another use of backslash is for specifying generic character types:
525 \D any character that is not a decimal digit
526 \h any horizontal white space character
527 \H any character that is not a horizontal white space character
528 \s any white space character
529 \S any character that is not a white space character
530 \v any vertical white space character
531 \V any character that is not a vertical white space character
532 \w any "word" character
533 \W any "non-word" character
535 There is also the single sequence \N, which matches a non-newline character.
537 <a href="#fullstopdot">the "." metacharacter</a>
538 when PCRE_DOTALL is not set. Perl also uses \N to match characters by name;
539 PCRE does not support this.
542 Each pair of lower and upper case escape sequences partitions the complete set
543 of characters into two disjoint sets. Any given character matches one, and only
544 one, of each pair. The sequences can appear both inside and outside character
545 classes. They each match one character of the appropriate type. If the current
546 matching point is at the end of the subject string, all of them fail, because
547 there is no character to match.
550 For compatibility with Perl, \s did not used to match the VT character (code
551 11), which made it different from the the POSIX "space" class. However, Perl
552 added VT at release 5.18, and PCRE followed suit at release 8.34. The default
553 \s characters are now HT (9), LF (10), VT (11), FF (12), CR (13), and space
554 (32), which are defined as white space in the "C" locale. This list may vary if
555 locale-specific matching is taking place. For example, in some locales the
556 "non-breaking space" character (\xA0) is recognized as white space, and in
557 others the VT character is not.
560 A "word" character is an underscore or any character that is a letter or digit.
561 By default, the definition of letters and digits is controlled by PCRE's
562 low-valued character tables, and may vary if locale-specific matching is taking
564 <a href="pcreapi.html#localesupport">"Locale support"</a>
566 <a href="pcreapi.html"><b>pcreapi</b></a>
567 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
568 or "french" in Windows, some character codes greater than 127 are used for
569 accented letters, and these are then matched by \w. The use of locales with
570 Unicode is discouraged.
573 By default, characters whose code points are greater than 127 never match \d,
574 \s, or \w, and always match \D, \S, and \W, although this may vary for
575 characters in the range 128-255 when locale-specific matching is happening.
576 These escape sequences retain their original meanings from before Unicode
577 support was available, mainly for efficiency reasons. If PCRE is compiled with
578 Unicode property support, and the PCRE_UCP option is set, the behaviour is
579 changed so that Unicode properties are used to determine character types, as
582 \d any character that matches \p{Nd} (decimal digit)
583 \s any character that matches \p{Z} or \h or \v
584 \w any character that matches \p{L} or \p{N}, plus underscore
586 The upper case escapes match the inverse sets of characters. Note that \d
587 matches only decimal digits, whereas \w matches any Unicode digit, as well as
588 any Unicode letter, and underscore. Note also that PCRE_UCP affects \b, and
589 \B because they are defined in terms of \w and \W. Matching these sequences
590 is noticeably slower when PCRE_UCP is set.
593 The sequences \h, \H, \v, and \V are features that were added to Perl at
594 release 5.10. In contrast to the other sequences, which match only ASCII
595 characters by default, these always match certain high-valued code points,
596 whether or not PCRE_UCP is set. The horizontal space characters are:
598 U+0009 Horizontal tab (HT)
600 U+00A0 Non-break space
601 U+1680 Ogham space mark
602 U+180E Mongolian vowel separator
607 U+2004 Three-per-em space
608 U+2005 Four-per-em space
609 U+2006 Six-per-em space
611 U+2008 Punctuation space
614 U+202F Narrow no-break space
615 U+205F Medium mathematical space
616 U+3000 Ideographic space
618 The vertical space characters are:
621 U+000B Vertical tab (VT)
622 U+000C Form feed (FF)
623 U+000D Carriage return (CR)
624 U+0085 Next line (NEL)
625 U+2028 Line separator
626 U+2029 Paragraph separator
628 In 8-bit, non-UTF-8 mode, only the characters with codepoints less than 256 are
630 <a name="newlineseq"></a></P>
635 Outside a character class, by default, the escape sequence \R matches any
636 Unicode newline sequence. In 8-bit non-UTF-8 mode \R is equivalent to the
639 (?>\r\n|\n|\x0b|\f|\r|\x85)
641 This is an example of an "atomic group", details of which are given
642 <a href="#atomicgroup">below.</a>
643 This particular group matches either the two-character sequence CR followed by
644 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
645 U+000B), FF (form feed, U+000C), CR (carriage return, U+000D), or NEL (next
646 line, U+0085). The two-character sequence is treated as a single unit that
650 In other modes, two additional characters whose codepoints are greater than 255
651 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
652 Unicode character property support is not needed for these characters to be
656 It is possible to restrict \R to match only CR, LF, or CRLF (instead of the
657 complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
658 either at compile time or when the pattern is matched. (BSR is an abbrevation
659 for "backslash R".) This can be made the default when PCRE is built; if this is
660 the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
661 It is also possible to specify these settings by starting a pattern string with
662 one of the following sequences:
664 (*BSR_ANYCRLF) CR, LF, or CRLF only
665 (*BSR_UNICODE) any Unicode newline sequence
667 These override the default and the options given to the compiling function, but
668 they can themselves be overridden by options given to a matching function. Note
669 that these special settings, which are not Perl-compatible, are recognized only
670 at the very start of a pattern, and that they must be in upper case. If more
671 than one of them is present, the last one is used. They can be combined with a
672 change of newline convention; for example, a pattern can start with:
676 They can also be combined with the (*UTF8), (*UTF16), (*UTF32), (*UTF) or
677 (*UCP) special sequences. Inside a character class, \R is treated as an
678 unrecognized escape sequence, and so matches the letter "R" by default, but
679 causes an error if PCRE_EXTRA is set.
680 <a name="uniextseq"></a></P>
682 Unicode character properties
685 When PCRE is built with Unicode character property support, three additional
686 escape sequences that match characters with specific properties are available.
687 When in 8-bit non-UTF-8 mode, these sequences are of course limited to testing
688 characters whose codepoints are less than 256, but they do work in this mode.
689 The extra escape sequences are:
691 \p{<i>xx</i>} a character with the <i>xx</i> property
692 \P{<i>xx</i>} a character without the <i>xx</i> property
693 \X a Unicode extended grapheme cluster
695 The property names represented by <i>xx</i> above are limited to the Unicode
696 script names, the general category properties, "Any", which matches any
697 character (including newline), and some special PCRE properties (described
699 <a href="#extraprops">next section).</a>
700 Other Perl properties such as "InMusicalSymbols" are not currently supported by
701 PCRE. Note that \P{Any} does not match any characters, so always causes a
705 Sets of Unicode characters are defined as belonging to certain scripts. A
706 character from one of these sets can be matched using a script name. For
712 Those that are not part of an identified script are lumped together as
713 "Common". The current list of scripts is:
743 Egyptian_Hieroglyphs,
760 Inscriptional_Pahlavi,
761 Inscriptional_Parthian,
787 Meroitic_Hieroglyphs,
843 Each character has exactly one Unicode general category property, specified by
844 a two-letter abbreviation. For compatibility with Perl, negation can be
845 specified by including a circumflex between the opening brace and the property
846 name. For example, \p{^Lu} is the same as \P{Lu}.
849 If only one letter is specified with \p or \P, it includes all the general
850 category properties that start with that letter. In this case, in the absence
851 of negation, the curly brackets in the escape sequence are optional; these two
852 examples have the same effect:
857 The following general category property codes are supported:
884 Pc Connector punctuation
888 Pi Initial punctuation
895 Sm Mathematical symbol
900 Zp Paragraph separator
903 The special property L& is also supported: it matches a character that has
904 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
905 a modifier or "other".
908 The Cs (Surrogate) property applies only to characters in the range U+D800 to
909 U+DFFF. Such characters are not valid in Unicode strings and so
910 cannot be tested by PCRE, unless UTF validity checking has been turned off
911 (see the discussion of PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK and
912 PCRE_NO_UTF32_CHECK in the
913 <a href="pcreapi.html"><b>pcreapi</b></a>
914 page). Perl does not support the Cs property.
917 The long synonyms for property names that Perl supports (such as \p{Letter})
918 are not supported by PCRE, nor is it permitted to prefix any of these
919 properties with "Is".
922 No character that is in the Unicode table has the Cn (unassigned) property.
923 Instead, this property is assumed for any code point that is not in the
927 Specifying caseless matching does not affect these escape sequences. For
928 example, \p{Lu} always matches only upper case letters. This is different from
929 the behaviour of current versions of Perl.
932 Matching characters by Unicode property is not fast, because PCRE has to do a
933 multistage table lookup in order to find a character's property. That is why
934 the traditional escape sequences such as \d and \w do not use Unicode
935 properties in PCRE by default, though you can make them do so by setting the
936 PCRE_UCP option or by starting the pattern with (*UCP).
939 Extended grapheme clusters
942 The \X escape matches any number of Unicode characters that form an "extended
943 grapheme cluster", and treats the sequence as an atomic group
944 <a href="#atomicgroup">(see below).</a>
945 Up to and including release 8.31, PCRE matched an earlier, simpler definition
946 that was equivalent to
950 That is, it matched a character without the "mark" property, followed by zero
951 or more characters with the "mark" property. Characters with the "mark"
952 property are typically non-spacing accents that affect the preceding character.
955 This simple definition was extended in Unicode to include more complicated
956 kinds of composite character by giving each character a grapheme breaking
957 property, and creating rules that use these properties to define the boundaries
958 of extended grapheme clusters. In releases of PCRE later than 8.31, \X matches
959 one of these clusters.
962 \X always matches at least one character. Then it decides whether to add
963 additional characters according to the following rules for ending a cluster:
966 1. End at the end of the subject string.
969 2. Do not end between CR and LF; otherwise end after any control character.
972 3. Do not break Hangul (a Korean script) syllable sequences. Hangul characters
973 are of five types: L, V, T, LV, and LVT. An L character may be followed by an
974 L, V, LV, or LVT character; an LV or V character may be followed by a V or T
975 character; an LVT or T character may be follwed only by a T character.
978 4. Do not end before extending characters or spacing marks. Characters with
979 the "mark" property always have the "extend" grapheme breaking property.
982 5. Do not end after prepend characters.
985 6. Otherwise, end the cluster.
986 <a name="extraprops"></a></P>
988 PCRE's additional properties
991 As well as the standard Unicode properties described above, PCRE supports four
992 more that make it possible to convert traditional escape sequences such as \w
993 and \s to use Unicode properties. PCRE uses these non-standard, non-Perl
994 properties internally when PCRE_UCP is set. However, they may also be used
995 explicitly. These properties are:
997 Xan Any alphanumeric character
998 Xps Any POSIX space character
999 Xsp Any Perl space character
1000 Xwd Any Perl "word" character
1002 Xan matches characters that have either the L (letter) or the N (number)
1003 property. Xps matches the characters tab, linefeed, vertical tab, form feed, or
1004 carriage return, and any other character that has the Z (separator) property.
1005 Xsp is the same as Xps; it used to exclude vertical tab, for Perl
1006 compatibility, but Perl changed, and so PCRE followed at release 8.34. Xwd
1007 matches the same characters as Xan, plus underscore.
1010 There is another non-standard property, Xuc, which matches any character that
1011 can be represented by a Universal Character Name in C++ and other programming
1012 languages. These are the characters $, @, ` (grave accent), and all characters
1013 with Unicode code points greater than or equal to U+00A0, except for the
1014 surrogates U+D800 to U+DFFF. Note that most base (ASCII) characters are
1015 excluded. (Universal Character Names are of the form \uHHHH or \UHHHHHHHH
1016 where H is a hexadecimal digit. Note that the Xuc property does not match these
1017 sequences but the characters that they represent.)
1018 <a name="resetmatchstart"></a></P>
1020 Resetting the match start
1023 The escape sequence \K causes any previously matched characters not to be
1024 included in the final matched sequence. For example, the pattern:
1028 matches "foobar", but reports that it has matched "bar". This feature is
1029 similar to a lookbehind assertion
1030 <a href="#lookbehind">(described below).</a>
1031 However, in this case, the part of the subject before the real match does not
1032 have to be of fixed length, as lookbehind assertions do. The use of \K does
1033 not interfere with the setting of
1034 <a href="#subpattern">captured substrings.</a>
1035 For example, when the pattern
1039 matches "foobar", the first substring is still set to "foo".
1042 Perl documents that the use of \K within assertions is "not well defined". In
1043 PCRE, \K is acted upon when it occurs inside positive assertions, but is
1044 ignored in negative assertions. Note that when a pattern such as (?=ab\K)
1045 matches, the reported start of the match can be greater than the end of the
1047 <a name="smallassertions"></a></P>
1052 The final use of backslash is for certain simple assertions. An assertion
1053 specifies a condition that has to be met at a particular point in a match,
1054 without consuming any characters from the subject string. The use of
1055 subpatterns for more complicated assertions is described
1056 <a href="#bigassertions">below.</a>
1057 The backslashed assertions are:
1059 \b matches at a word boundary
1060 \B matches when not at a word boundary
1061 \A matches at the start of the subject
1062 \Z matches at the end of the subject
1063 also matches before a newline at the end of the subject
1064 \z matches only at the end of the subject
1065 \G matches at the first matching position in the subject
1067 Inside a character class, \b has a different meaning; it matches the backspace
1068 character. If any other of these assertions appears in a character class, by
1069 default it matches the corresponding literal character (for example, \B
1070 matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
1071 escape sequence" error is generated instead.
1074 A word boundary is a position in the subject string where the current character
1075 and the previous character do not both match \w or \W (i.e. one matches
1076 \w and the other matches \W), or the start or end of the string if the
1077 first or last character matches \w, respectively. In a UTF mode, the meanings
1078 of \w and \W can be changed by setting the PCRE_UCP option. When this is
1079 done, it also affects \b and \B. Neither PCRE nor Perl has a separate "start
1080 of word" or "end of word" metasequence. However, whatever follows \b normally
1081 determines which it is. For example, the fragment \ba matches "a" at the start
1085 The \A, \Z, and \z assertions differ from the traditional circumflex and
1086 dollar (described in the next section) in that they only ever match at the very
1087 start and end of the subject string, whatever options are set. Thus, they are
1088 independent of multiline mode. These three assertions are not affected by the
1089 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
1090 circumflex and dollar metacharacters. However, if the <i>startoffset</i>
1091 argument of <b>pcre_exec()</b> is non-zero, indicating that matching is to start
1092 at a point other than the beginning of the subject, \A can never match. The
1093 difference between \Z and \z is that \Z matches before a newline at the end
1094 of the string as well as at the very end, whereas \z matches only at the end.
1097 The \G assertion is true only when the current matching position is at the
1098 start point of the match, as specified by the <i>startoffset</i> argument of
1099 <b>pcre_exec()</b>. It differs from \A when the value of <i>startoffset</i> is
1100 non-zero. By calling <b>pcre_exec()</b> multiple times with appropriate
1101 arguments, you can mimic Perl's /g option, and it is in this kind of
1102 implementation where \G can be useful.
1105 Note, however, that PCRE's interpretation of \G, as the start of the current
1106 match, is subtly different from Perl's, which defines it as the end of the
1107 previous match. In Perl, these can be different when the previously matched
1108 string was empty. Because PCRE does just one match at a time, it cannot
1109 reproduce this behaviour.
1112 If all the alternatives of a pattern begin with \G, the expression is anchored
1113 to the starting match position, and the "anchored" flag is set in the compiled
1116 <br><a name="SEC6" href="#TOC1">CIRCUMFLEX AND DOLLAR</a><br>
1118 The circumflex and dollar metacharacters are zero-width assertions. That is,
1119 they test for a particular condition being true without consuming any
1120 characters from the subject string.
1123 Outside a character class, in the default matching mode, the circumflex
1124 character is an assertion that is true only if the current matching point is at
1125 the start of the subject string. If the <i>startoffset</i> argument of
1126 <b>pcre_exec()</b> is non-zero, circumflex can never match if the PCRE_MULTILINE
1127 option is unset. Inside a character class, circumflex has an entirely different
1129 <a href="#characterclass">(see below).</a>
1132 Circumflex need not be the first character of the pattern if a number of
1133 alternatives are involved, but it should be the first thing in each alternative
1134 in which it appears if the pattern is ever to match that branch. If all
1135 possible alternatives start with a circumflex, that is, if the pattern is
1136 constrained to match only at the start of the subject, it is said to be an
1137 "anchored" pattern. (There are also other constructs that can cause a pattern
1141 The dollar character is an assertion that is true only if the current matching
1142 point is at the end of the subject string, or immediately before a newline at
1143 the end of the string (by default). Note, however, that it does not actually
1144 match the newline. Dollar need not be the last character of the pattern if a
1145 number of alternatives are involved, but it should be the last item in any
1146 branch in which it appears. Dollar has no special meaning in a character class.
1149 The meaning of dollar can be changed so that it matches only at the very end of
1150 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
1151 does not affect the \Z assertion.
1154 The meanings of the circumflex and dollar characters are changed if the
1155 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
1156 immediately after internal newlines as well as at the start of the subject
1157 string. It does not match after a newline that ends the string. A dollar
1158 matches before any newlines in the string, as well as at the very end, when
1159 PCRE_MULTILINE is set. When newline is specified as the two-character
1160 sequence CRLF, isolated CR and LF characters do not indicate newlines.
1163 For example, the pattern /^abc$/ matches the subject string "def\nabc" (where
1164 \n represents a newline) in multiline mode, but not otherwise. Consequently,
1165 patterns that are anchored in single line mode because all branches start with
1166 ^ are not anchored in multiline mode, and a match for circumflex is possible
1167 when the <i>startoffset</i> argument of <b>pcre_exec()</b> is non-zero. The
1168 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
1171 Note that the sequences \A, \Z, and \z can be used to match the start and
1172 end of the subject in both modes, and if all branches of a pattern start with
1173 \A it is always anchored, whether or not PCRE_MULTILINE is set.
1174 <a name="fullstopdot"></a></P>
1175 <br><a name="SEC7" href="#TOC1">FULL STOP (PERIOD, DOT) AND \N</a><br>
1177 Outside a character class, a dot in the pattern matches any one character in
1178 the subject string except (by default) a character that signifies the end of a
1182 When a line ending is defined as a single character, dot never matches that
1183 character; when the two-character sequence CRLF is used, dot does not match CR
1184 if it is immediately followed by LF, but otherwise it matches all characters
1185 (including isolated CRs and LFs). When any Unicode line endings are being
1186 recognized, dot does not match CR or LF or any of the other line ending
1190 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
1191 option is set, a dot matches any one character, without exception. If the
1192 two-character sequence CRLF is present in the subject string, it takes two dots
1196 The handling of dot is entirely independent of the handling of circumflex and
1197 dollar, the only relationship being that they both involve newlines. Dot has no
1198 special meaning in a character class.
1201 The escape sequence \N behaves like a dot, except that it is not affected by
1202 the PCRE_DOTALL option. In other words, it matches any character except one
1203 that signifies the end of a line. Perl also uses \N to match characters by
1204 name; PCRE does not support this.
1206 <br><a name="SEC8" href="#TOC1">MATCHING A SINGLE DATA UNIT</a><br>
1208 Outside a character class, the escape sequence \C matches any one data unit,
1209 whether or not a UTF mode is set. In the 8-bit library, one data unit is one
1210 byte; in the 16-bit library it is a 16-bit unit; in the 32-bit library it is
1211 a 32-bit unit. Unlike a dot, \C always
1212 matches line-ending characters. The feature is provided in Perl in order to
1213 match individual bytes in UTF-8 mode, but it is unclear how it can usefully be
1214 used. Because \C breaks up characters into individual data units, matching one
1215 unit with \C in a UTF mode means that the rest of the string may start with a
1216 malformed UTF character. This has undefined results, because PCRE assumes that
1217 it is dealing with valid UTF strings (and by default it checks this at the
1218 start of processing unless the PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK or
1219 PCRE_NO_UTF32_CHECK option is used).
1222 PCRE does not allow \C to appear in lookbehind assertions
1223 <a href="#lookbehind">(described below)</a>
1224 in a UTF mode, because this would make it impossible to calculate the length of
1228 In general, the \C escape sequence is best avoided. However, one
1229 way of using it that avoids the problem of malformed UTF characters is to use a
1230 lookahead to check the length of the next character, as in this pattern, which
1231 could be used with a UTF-8 string (ignore white space and line breaks):
1233 (?| (?=[\x00-\x7f])(\C) |
1234 (?=[\x80-\x{7ff}])(\C)(\C) |
1235 (?=[\x{800}-\x{ffff}])(\C)(\C)(\C) |
1236 (?=[\x{10000}-\x{1fffff}])(\C)(\C)(\C)(\C))
1238 A group that starts with (?| resets the capturing parentheses numbers in each
1240 <a href="#dupsubpatternnumber">"Duplicate Subpattern Numbers"</a>
1241 below). The assertions at the start of each branch check the next UTF-8
1242 character for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The
1243 character's individual bytes are then captured by the appropriate number of
1245 <a name="characterclass"></a></P>
1246 <br><a name="SEC9" href="#TOC1">SQUARE BRACKETS AND CHARACTER CLASSES</a><br>
1248 An opening square bracket introduces a character class, terminated by a closing
1249 square bracket. A closing square bracket on its own is not special by default.
1250 However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
1251 bracket causes a compile-time error. If a closing square bracket is required as
1252 a member of the class, it should be the first data character in the class
1253 (after an initial circumflex, if present) or escaped with a backslash.
1256 A character class matches a single character in the subject. In a UTF mode, the
1257 character may be more than one data unit long. A matched character must be in
1258 the set of characters defined by the class, unless the first character in the
1259 class definition is a circumflex, in which case the subject character must not
1260 be in the set defined by the class. If a circumflex is actually required as a
1261 member of the class, ensure it is not the first character, or escape it with a
1265 For example, the character class [aeiou] matches any lower case vowel, while
1266 [^aeiou] matches any character that is not a lower case vowel. Note that a
1267 circumflex is just a convenient notation for specifying the characters that
1268 are in the class by enumerating those that are not. A class that starts with a
1269 circumflex is not an assertion; it still consumes a character from the subject
1270 string, and therefore it fails if the current pointer is at the end of the
1274 In UTF-8 (UTF-16, UTF-32) mode, characters with values greater than 255 (0xffff)
1275 can be included in a class as a literal string of data units, or by using the
1276 \x{ escaping mechanism.
1279 When caseless matching is set, any letters in a class represent both their
1280 upper case and lower case versions, so for example, a caseless [aeiou] matches
1281 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1282 caseful version would. In a UTF mode, PCRE always understands the concept of
1283 case for characters whose values are less than 128, so caseless matching is
1284 always possible. For characters with higher values, the concept of case is
1285 supported if PCRE is compiled with Unicode property support, but not otherwise.
1286 If you want to use caseless matching in a UTF mode for characters 128 and
1287 above, you must ensure that PCRE is compiled with Unicode property support as
1288 well as with UTF support.
1291 Characters that might indicate line breaks are never treated in any special way
1292 when matching character classes, whatever line-ending sequence is in use, and
1293 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1294 such as [^a] always matches one of these characters.
1297 The minus (hyphen) character can be used to specify a range of characters in a
1298 character class. For example, [d-m] matches any letter between d and m,
1299 inclusive. If a minus character is required in a class, it must be escaped with
1300 a backslash or appear in a position where it cannot be interpreted as
1301 indicating a range, typically as the first or last character in the class, or
1302 immediately after a range. For example, [b-d-z] matches letters in the range b
1303 to d, a hyphen character, or z.
1306 It is not possible to have the literal character "]" as the end character of a
1307 range. A pattern such as [W-]46] is interpreted as a class of two characters
1308 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1309 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1310 the end of range, so [W-\]46] is interpreted as a class containing a range
1311 followed by two other characters. The octal or hexadecimal representation of
1312 "]" can also be used to end a range.
1315 An error is generated if a POSIX character class (see below) or an escape
1316 sequence other than one that defines a single character appears at a point
1317 where a range ending character is expected. For example, [z-\xff] is valid,
1318 but [A-\d] and [A-[:digit:]] are not.
1321 Ranges operate in the collating sequence of character values. They can also be
1322 used for characters specified numerically, for example [\000-\037]. Ranges
1323 can include any characters that are valid for the current mode.
1326 If a range that includes letters is used when caseless matching is set, it
1327 matches the letters in either case. For example, [W-c] is equivalent to
1328 [][\\^_`wxyzabc], matched caselessly, and in a non-UTF mode, if character
1329 tables for a French locale are in use, [\xc8-\xcb] matches accented E
1330 characters in both cases. In UTF modes, PCRE supports the concept of case for
1331 characters with values greater than 128 only when it is compiled with Unicode
1335 The character escape sequences \d, \D, \h, \H, \p, \P, \s, \S, \v,
1336 \V, \w, and \W may appear in a character class, and add the characters that
1337 they match to the class. For example, [\dABCDEF] matches any hexadecimal
1338 digit. In UTF modes, the PCRE_UCP option affects the meanings of \d, \s, \w
1339 and their upper case partners, just as it does when they appear outside a
1340 character class, as described in the section entitled
1341 <a href="#genericchartypes">"Generic character types"</a>
1342 above. The escape sequence \b has a different meaning inside a character
1343 class; it matches the backspace character. The sequences \B, \N, \R, and \X
1344 are not special inside a character class. Like any other unrecognized escape
1345 sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1346 default, but cause an error if the PCRE_EXTRA option is set.
1349 A circumflex can conveniently be used with the upper case character types to
1350 specify a more restricted set of characters than the matching lower case type.
1351 For example, the class [^\W_] matches any letter or digit, but not underscore,
1352 whereas [\w] includes underscore. A positive character class should be read as
1353 "something OR something OR ..." and a negative class as "NOT something AND NOT
1354 something AND NOT ...".
1357 The only metacharacters that are recognized in character classes are backslash,
1358 hyphen (only where it can be interpreted as specifying a range), circumflex
1359 (only at the start), opening square bracket (only when it can be interpreted as
1360 introducing a POSIX class name, or for a special compatibility feature - see
1361 the next two sections), and the terminating closing square bracket. However,
1362 escaping other non-alphanumeric characters does no harm.
1364 <br><a name="SEC10" href="#TOC1">POSIX CHARACTER CLASSES</a><br>
1366 Perl supports the POSIX notation for character classes. This uses names
1367 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1368 this notation. For example,
1372 matches "0", "1", any alphabetic character, or "%". The supported class names
1375 alnum letters and digits
1377 ascii character codes 0 - 127
1378 blank space or tab only
1379 cntrl control characters
1380 digit decimal digits (same as \d)
1381 graph printing characters, excluding space
1382 lower lower case letters
1383 print printing characters, including space
1384 punct printing characters, excluding letters and digits and space
1385 space white space (the same as \s from PCRE 8.34)
1386 upper upper case letters
1387 word "word" characters (same as \w)
1388 xdigit hexadecimal digits
1390 The default "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13),
1391 and space (32). If locale-specific matching is taking place, the list of space
1392 characters may be different; there may be fewer or more of them. "Space" used
1393 to be different to \s, which did not include VT, for Perl compatibility.
1394 However, Perl changed at release 5.18, and PCRE followed at release 8.34.
1395 "Space" and \s now match the same set of characters.
1398 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1399 5.8. Another Perl extension is negation, which is indicated by a ^ character
1400 after the colon. For example,
1404 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1405 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1406 supported, and an error is given if they are encountered.
1409 By default, characters with values greater than 128 do not match any of the
1410 POSIX character classes. However, if the PCRE_UCP option is passed to
1411 <b>pcre_compile()</b>, some of the classes are changed so that Unicode character
1412 properties are used. This is achieved by replacing certain POSIX classes by
1413 other sequences, as follows:
1415 [:alnum:] becomes \p{Xan}
1416 [:alpha:] becomes \p{L}
1417 [:blank:] becomes \h
1418 [:digit:] becomes \p{Nd}
1419 [:lower:] becomes \p{Ll}
1420 [:space:] becomes \p{Xps}
1421 [:upper:] becomes \p{Lu}
1422 [:word:] becomes \p{Xwd}
1424 Negated versions, such as [:^alpha:] use \P instead of \p. Three other POSIX
1425 classes are handled specially in UCP mode:
1429 This matches characters that have glyphs that mark the page when printed. In
1430 Unicode property terms, it matches all characters with the L, M, N, P, S, or Cf
1431 properties, except for:
1433 U+061C Arabic Letter Mark
1434 U+180E Mongolian Vowel Separator
1435 U+2066 - U+2069 Various "isolate"s
1441 This matches the same characters as [:graph:] plus space characters that are
1442 not controls, that is, characters with the Zs property.
1446 This matches all characters that have the Unicode P (punctuation) property,
1447 plus those characters whose code points are less than 128 that have the S
1451 The other POSIX classes are unchanged, and match only characters with code
1452 points less than 128.
1454 <br><a name="SEC11" href="#TOC1">COMPATIBILITY FEATURE FOR WORD BOUNDARIES</a><br>
1456 In the POSIX.2 compliant library that was included in 4.4BSD Unix, the ugly
1457 syntax [[:<:]] and [[:>:]] is used for matching "start of word" and "end of
1458 word". PCRE treats these items as follows:
1460 [[:<:]] is converted to \b(?=\w)
1461 [[:>:]] is converted to \b(?<=\w)
1463 Only these exact character sequences are recognized. A sequence such as
1464 [a[:<:]b] provokes error for an unrecognized POSIX class name. This support is
1465 not compatible with Perl. It is provided to help migrations from other
1466 environments, and is best not used in any new patterns. Note that \b matches
1467 at the start and the end of a word (see
1468 <a href="#smallassertions">"Simple assertions"</a>
1469 above), and in a Perl-style pattern the preceding or following character
1470 normally shows which is wanted, without the need for the assertions that are
1471 used above in order to give exactly the POSIX behaviour.
1473 <br><a name="SEC12" href="#TOC1">VERTICAL BAR</a><br>
1475 Vertical bar characters are used to separate alternative patterns. For example,
1480 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1481 and an empty alternative is permitted (matching the empty string). The matching
1482 process tries each alternative in turn, from left to right, and the first one
1483 that succeeds is used. If the alternatives are within a subpattern
1484 <a href="#subpattern">(defined below),</a>
1485 "succeeds" means matching the rest of the main pattern as well as the
1486 alternative in the subpattern.
1488 <br><a name="SEC13" href="#TOC1">INTERNAL OPTION SETTING</a><br>
1490 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1491 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1492 the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1493 The option letters are
1496 m for PCRE_MULTILINE
1500 For example, (?im) sets caseless, multiline matching. It is also possible to
1501 unset these options by preceding the letter with a hyphen, and a combined
1502 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1503 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1504 permitted. If a letter appears both before and after the hyphen, the option is
1508 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1509 changed in the same way as the Perl-compatible options by using the characters
1510 J, U and X respectively.
1513 When one of these option changes occurs at top level (that is, not inside
1514 subpattern parentheses), the change applies to the remainder of the pattern
1515 that follows. If the change is placed right at the start of a pattern, PCRE
1516 extracts it into the global options (and it will therefore show up in data
1517 extracted by the <b>pcre_fullinfo()</b> function).
1520 An option change within a subpattern (see below for a description of
1521 subpatterns) affects only that part of the subpattern that follows it, so
1525 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1526 By this means, options can be made to have different settings in different
1527 parts of the pattern. Any changes made in one alternative do carry on
1528 into subsequent branches within the same subpattern. For example,
1532 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1533 branch is abandoned before the option setting. This is because the effects of
1534 option settings happen at compile time. There would be some very weird
1535 behaviour otherwise.
1538 <b>Note:</b> There are other PCRE-specific options that can be set by the
1539 application when the compiling or matching functions are called. In some cases
1540 the pattern can contain special leading sequences such as (*CRLF) to override
1541 what the application has set or what has been defaulted. Details are given in
1542 the section entitled
1543 <a href="#newlineseq">"Newline sequences"</a>
1544 above. There are also the (*UTF8), (*UTF16),(*UTF32), and (*UCP) leading
1545 sequences that can be used to set UTF and Unicode property modes; they are
1546 equivalent to setting the PCRE_UTF8, PCRE_UTF16, PCRE_UTF32 and the PCRE_UCP
1547 options, respectively. The (*UTF) sequence is a generic version that can be
1548 used with any of the libraries. However, the application can set the
1549 PCRE_NEVER_UTF option, which locks out the use of the (*UTF) sequences.
1550 <a name="subpattern"></a></P>
1551 <br><a name="SEC14" href="#TOC1">SUBPATTERNS</a><br>
1553 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1554 Turning part of a pattern into a subpattern does two things:
1557 1. It localizes a set of alternatives. For example, the pattern
1559 cat(aract|erpillar|)
1561 matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1562 match "cataract", "erpillar" or an empty string.
1565 2. It sets up the subpattern as a capturing subpattern. This means that, when
1566 the whole pattern matches, that portion of the subject string that matched the
1567 subpattern is passed back to the caller via the <i>ovector</i> argument of the
1568 matching function. (This applies only to the traditional matching functions;
1569 the DFA matching functions do not support capturing.)
1572 Opening parentheses are counted from left to right (starting from 1) to obtain
1573 numbers for the capturing subpatterns. For example, if the string "the red
1574 king" is matched against the pattern
1576 the ((red|white) (king|queen))
1578 the captured substrings are "red king", "red", and "king", and are numbered 1,
1579 2, and 3, respectively.
1582 The fact that plain parentheses fulfil two functions is not always helpful.
1583 There are often times when a grouping subpattern is required without a
1584 capturing requirement. If an opening parenthesis is followed by a question mark
1585 and a colon, the subpattern does not do any capturing, and is not counted when
1586 computing the number of any subsequent capturing subpatterns. For example, if
1587 the string "the white queen" is matched against the pattern
1589 the ((?:red|white) (king|queen))
1591 the captured substrings are "white queen" and "queen", and are numbered 1 and
1592 2. The maximum number of capturing subpatterns is 65535.
1595 As a convenient shorthand, if any option settings are required at the start of
1596 a non-capturing subpattern, the option letters may appear between the "?" and
1597 the ":". Thus the two patterns
1599 (?i:saturday|sunday)
1600 (?:(?i)saturday|sunday)
1602 match exactly the same set of strings. Because alternative branches are tried
1603 from left to right, and options are not reset until the end of the subpattern
1604 is reached, an option setting in one branch does affect subsequent branches, so
1605 the above patterns match "SUNDAY" as well as "Saturday".
1606 <a name="dupsubpatternnumber"></a></P>
1607 <br><a name="SEC15" href="#TOC1">DUPLICATE SUBPATTERN NUMBERS</a><br>
1609 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1610 the same numbers for its capturing parentheses. Such a subpattern starts with
1611 (?| and is itself a non-capturing subpattern. For example, consider this
1614 (?|(Sat)ur|(Sun))day
1616 Because the two alternatives are inside a (?| group, both sets of capturing
1617 parentheses are numbered one. Thus, when the pattern matches, you can look
1618 at captured substring number one, whichever alternative matched. This construct
1619 is useful when you want to capture part, but not all, of one of a number of
1620 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1621 number is reset at the start of each branch. The numbers of any capturing
1622 parentheses that follow the subpattern start after the highest number used in
1623 any branch. The following example is taken from the Perl documentation. The
1624 numbers underneath show in which buffer the captured content will be stored.
1626 # before ---------------branch-reset----------- after
1627 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1630 A back reference to a numbered subpattern uses the most recent value that is
1631 set for that number by any subpattern. The following pattern matches "abcabc"
1636 In contrast, a subroutine call to a numbered subpattern always refers to the
1637 first one in the pattern with the given number. The following pattern matches
1638 "abcabc" or "defabc":
1640 /(?|(abc)|(def))(?1)/
1643 <a href="#conditions">condition test</a>
1644 for a subpattern's having matched refers to a non-unique number, the test is
1645 true if any of the subpatterns of that number have matched.
1648 An alternative approach to using this "branch reset" feature is to use
1649 duplicate named subpatterns, as described in the next section.
1651 <br><a name="SEC16" href="#TOC1">NAMED SUBPATTERNS</a><br>
1653 Identifying capturing parentheses by number is simple, but it can be very hard
1654 to keep track of the numbers in complicated regular expressions. Furthermore,
1655 if an expression is modified, the numbers may change. To help with this
1656 difficulty, PCRE supports the naming of subpatterns. This feature was not
1657 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1658 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1659 the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1660 have different names, but PCRE does not.
1663 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1664 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1665 parentheses from other parts of the pattern, such as
1666 <a href="#backreferences">back references,</a>
1667 <a href="#recursion">recursion,</a>
1669 <a href="#conditions">conditions,</a>
1670 can be made by name as well as by number.
1673 Names consist of up to 32 alphanumeric characters and underscores, but must
1674 start with a non-digit. Named capturing parentheses are still allocated numbers
1675 as well as names, exactly as if the names were not present. The PCRE API
1676 provides function calls for extracting the name-to-number translation table
1677 from a compiled pattern. There is also a convenience function for extracting a
1678 captured substring by name.
1681 By default, a name must be unique within a pattern, but it is possible to relax
1682 this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1683 names are also always permitted for subpatterns with the same number, set up as
1684 described in the previous section.) Duplicate names can be useful for patterns
1685 where only one instance of the named parentheses can match. Suppose you want to
1686 match the name of a weekday, either as a 3-letter abbreviation or as the full
1687 name, and in both cases you want to extract the abbreviation. This pattern
1688 (ignoring the line breaks) does the job:
1690 (?<DN>Mon|Fri|Sun)(?:day)?|
1691 (?<DN>Tue)(?:sday)?|
1692 (?<DN>Wed)(?:nesday)?|
1693 (?<DN>Thu)(?:rsday)?|
1694 (?<DN>Sat)(?:urday)?
1696 There are five capturing substrings, but only one is ever set after a match.
1697 (An alternative way of solving this problem is to use a "branch reset"
1698 subpattern, as described in the previous section.)
1701 The convenience function for extracting the data by name returns the substring
1702 for the first (and in this example, the only) subpattern of that name that
1703 matched. This saves searching to find which numbered subpattern it was.
1706 If you make a back reference to a non-unique named subpattern from elsewhere in
1707 the pattern, the subpatterns to which the name refers are checked in the order
1708 in which they appear in the overall pattern. The first one that is set is used
1709 for the reference. For example, this pattern matches both "foofoo" and
1710 "barbar" but not "foobar" or "barfoo":
1712 (?:(?<n>foo)|(?<n>bar))\k<n>
1717 If you make a subroutine call to a non-unique named subpattern, the one that
1718 corresponds to the first occurrence of the name is used. In the absence of
1719 duplicate numbers (see the previous section) this is the one with the lowest
1723 If you use a named reference in a condition
1725 <a href="#conditions">section about conditions</a>
1726 below), either to check whether a subpattern has matched, or to check for
1727 recursion, all subpatterns with the same name are tested. If the condition is
1728 true for any one of them, the overall condition is true. This is the same
1729 behaviour as testing by number. For further details of the interfaces for
1730 handling named subpatterns, see the
1731 <a href="pcreapi.html"><b>pcreapi</b></a>
1735 <b>Warning:</b> You cannot use different names to distinguish between two
1736 subpatterns with the same number because PCRE uses only the numbers when
1737 matching. For this reason, an error is given at compile time if different names
1738 are given to subpatterns with the same number. However, you can always give the
1739 same name to subpatterns with the same number, even when PCRE_DUPNAMES is not
1742 <br><a name="SEC17" href="#TOC1">REPETITION</a><br>
1744 Repetition is specified by quantifiers, which can follow any of the following
1747 a literal data character
1748 the dot metacharacter
1749 the \C escape sequence
1750 the \X escape sequence
1751 the \R escape sequence
1752 an escape such as \d or \pL that matches a single character
1754 a back reference (see next section)
1755 a parenthesized subpattern (including assertions)
1756 a subroutine call to a subpattern (recursive or otherwise)
1758 The general repetition quantifier specifies a minimum and maximum number of
1759 permitted matches, by giving the two numbers in curly brackets (braces),
1760 separated by a comma. The numbers must be less than 65536, and the first must
1761 be less than or equal to the second. For example:
1765 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1766 character. If the second number is omitted, but the comma is present, there is
1767 no upper limit; if the second number and the comma are both omitted, the
1768 quantifier specifies an exact number of required matches. Thus
1772 matches at least 3 successive vowels, but may match many more, while
1776 matches exactly 8 digits. An opening curly bracket that appears in a position
1777 where a quantifier is not allowed, or one that does not match the syntax of a
1778 quantifier, is taken as a literal character. For example, {,6} is not a
1779 quantifier, but a literal string of four characters.
1782 In UTF modes, quantifiers apply to characters rather than to individual data
1783 units. Thus, for example, \x{100}{2} matches two characters, each of
1784 which is represented by a two-byte sequence in a UTF-8 string. Similarly,
1785 \X{3} matches three Unicode extended grapheme clusters, each of which may be
1786 several data units long (and they may be of different lengths).
1789 The quantifier {0} is permitted, causing the expression to behave as if the
1790 previous item and the quantifier were not present. This may be useful for
1791 subpatterns that are referenced as
1792 <a href="#subpatternsassubroutines">subroutines</a>
1793 from elsewhere in the pattern (but see also the section entitled
1794 <a href="#subdefine">"Defining subpatterns for use by reference only"</a>
1795 below). Items other than subpatterns that have a {0} quantifier are omitted
1796 from the compiled pattern.
1799 For convenience, the three most common quantifiers have single-character
1802 * is equivalent to {0,}
1803 + is equivalent to {1,}
1804 ? is equivalent to {0,1}
1806 It is possible to construct infinite loops by following a subpattern that can
1807 match no characters with a quantifier that has no upper limit, for example:
1811 Earlier versions of Perl and PCRE used to give an error at compile time for
1812 such patterns. However, because there are cases where this can be useful, such
1813 patterns are now accepted, but if any repetition of the subpattern does in fact
1814 match no characters, the loop is forcibly broken.
1817 By default, the quantifiers are "greedy", that is, they match as much as
1818 possible (up to the maximum number of permitted times), without causing the
1819 rest of the pattern to fail. The classic example of where this gives problems
1820 is in trying to match comments in C programs. These appear between /* and */
1821 and within the comment, individual * and / characters may appear. An attempt to
1822 match C comments by applying the pattern
1828 /* first comment */ not comment /* second comment */
1830 fails, because it matches the entire string owing to the greediness of the .*
1834 However, if a quantifier is followed by a question mark, it ceases to be
1835 greedy, and instead matches the minimum number of times possible, so the
1840 does the right thing with the C comments. The meaning of the various
1841 quantifiers is not otherwise changed, just the preferred number of matches.
1842 Do not confuse this use of question mark with its use as a quantifier in its
1843 own right. Because it has two uses, it can sometimes appear doubled, as in
1847 which matches one digit by preference, but can match two if that is the only
1848 way the rest of the pattern matches.
1851 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1852 the quantifiers are not greedy by default, but individual ones can be made
1853 greedy by following them with a question mark. In other words, it inverts the
1857 When a parenthesized subpattern is quantified with a minimum repeat count that
1858 is greater than 1 or with a limited maximum, more memory is required for the
1859 compiled pattern, in proportion to the size of the minimum or maximum.
1862 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1863 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1864 implicitly anchored, because whatever follows will be tried against every
1865 character position in the subject string, so there is no point in retrying the
1866 overall match at any position after the first. PCRE normally treats such a
1867 pattern as though it were preceded by \A.
1870 In cases where it is known that the subject string contains no newlines, it is
1871 worth setting PCRE_DOTALL in order to obtain this optimization, or
1872 alternatively using ^ to indicate anchoring explicitly.
1875 However, there are some cases where the optimization cannot be used. When .*
1876 is inside capturing parentheses that are the subject of a back reference
1877 elsewhere in the pattern, a match at the start may fail where a later one
1878 succeeds. Consider, for example:
1882 If the subject is "xyz123abc123" the match point is the fourth character. For
1883 this reason, such a pattern is not implicitly anchored.
1886 Another case where implicit anchoring is not applied is when the leading .* is
1887 inside an atomic group. Once again, a match at the start may fail where a later
1888 one succeeds. Consider this pattern:
1892 It matches "ab" in the subject "aab". The use of the backtracking control verbs
1893 (*PRUNE) and (*SKIP) also disable this optimization.
1896 When a capturing subpattern is repeated, the value captured is the substring
1897 that matched the final iteration. For example, after
1899 (tweedle[dume]{3}\s*)+
1901 has matched "tweedledum tweedledee" the value of the captured substring is
1902 "tweedledee". However, if there are nested capturing subpatterns, the
1903 corresponding captured values may have been set in previous iterations. For
1908 matches "aba" the value of the second captured substring is "b".
1909 <a name="atomicgroup"></a></P>
1910 <br><a name="SEC18" href="#TOC1">ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS</a><br>
1912 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1913 repetition, failure of what follows normally causes the repeated item to be
1914 re-evaluated to see if a different number of repeats allows the rest of the
1915 pattern to match. Sometimes it is useful to prevent this, either to change the
1916 nature of the match, or to cause it fail earlier than it otherwise might, when
1917 the author of the pattern knows there is no point in carrying on.
1920 Consider, for example, the pattern \d+foo when applied to the subject line
1924 After matching all 6 digits and then failing to match "foo", the normal
1925 action of the matcher is to try again with only 5 digits matching the \d+
1926 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1927 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1928 that once a subpattern has matched, it is not to be re-evaluated in this way.
1931 If we use atomic grouping for the previous example, the matcher gives up
1932 immediately on failing to match "foo" the first time. The notation is a kind of
1933 special parenthesis, starting with (?> as in this example:
1937 This kind of parenthesis "locks up" the part of the pattern it contains once
1938 it has matched, and a failure further into the pattern is prevented from
1939 backtracking into it. Backtracking past it to previous items, however, works as
1943 An alternative description is that a subpattern of this type matches the string
1944 of characters that an identical standalone pattern would match, if anchored at
1945 the current point in the subject string.
1948 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1949 the above example can be thought of as a maximizing repeat that must swallow
1950 everything it can. So, while both \d+ and \d+? are prepared to adjust the
1951 number of digits they match in order to make the rest of the pattern match,
1952 (?>\d+) can only match an entire sequence of digits.
1955 Atomic groups in general can of course contain arbitrarily complicated
1956 subpatterns, and can be nested. However, when the subpattern for an atomic
1957 group is just a single repeated item, as in the example above, a simpler
1958 notation, called a "possessive quantifier" can be used. This consists of an
1959 additional + character following a quantifier. Using this notation, the
1960 previous example can be rewritten as
1964 Note that a possessive quantifier can be used with an entire group, for
1969 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1970 option is ignored. They are a convenient notation for the simpler forms of
1971 atomic group. However, there is no difference in the meaning of a possessive
1972 quantifier and the equivalent atomic group, though there may be a performance
1973 difference; possessive quantifiers should be slightly faster.
1976 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1977 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1978 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1979 package, and PCRE copied it from there. It ultimately found its way into Perl
1983 PCRE has an optimization that automatically "possessifies" certain simple
1984 pattern constructs. For example, the sequence A+B is treated as A++B because
1985 there is no point in backtracking into a sequence of A's when B must follow.
1988 When a pattern contains an unlimited repeat inside a subpattern that can itself
1989 be repeated an unlimited number of times, the use of an atomic group is the
1990 only way to avoid some failing matches taking a very long time indeed. The
1993 (\D+|<\d+>)*[!?]
1995 matches an unlimited number of substrings that either consist of non-digits, or
1996 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1997 quickly. However, if it is applied to
1999 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
2001 it takes a long time before reporting failure. This is because the string can
2002 be divided between the internal \D+ repeat and the external * repeat in a
2003 large number of ways, and all have to be tried. (The example uses [!?] rather
2004 than a single character at the end, because both PCRE and Perl have an
2005 optimization that allows for fast failure when a single character is used. They
2006 remember the last single character that is required for a match, and fail early
2007 if it is not present in the string.) If the pattern is changed so that it uses
2008 an atomic group, like this:
2010 ((?>\D+)|<\d+>)*[!?]
2012 sequences of non-digits cannot be broken, and failure happens quickly.
2013 <a name="backreferences"></a></P>
2014 <br><a name="SEC19" href="#TOC1">BACK REFERENCES</a><br>
2016 Outside a character class, a backslash followed by a digit greater than 0 (and
2017 possibly further digits) is a back reference to a capturing subpattern earlier
2018 (that is, to its left) in the pattern, provided there have been that many
2019 previous capturing left parentheses.
2022 However, if the decimal number following the backslash is less than 10, it is
2023 always taken as a back reference, and causes an error only if there are not
2024 that many capturing left parentheses in the entire pattern. In other words, the
2025 parentheses that are referenced need not be to the left of the reference for
2026 numbers less than 10. A "forward back reference" of this type can make sense
2027 when a repetition is involved and the subpattern to the right has participated
2028 in an earlier iteration.
2031 It is not possible to have a numerical "forward back reference" to a subpattern
2032 whose number is 10 or more using this syntax because a sequence such as \50 is
2033 interpreted as a character defined in octal. See the subsection entitled
2034 "Non-printing characters"
2035 <a href="#digitsafterbackslash">above</a>
2036 for further details of the handling of digits following a backslash. There is
2037 no such problem when named parentheses are used. A back reference to any
2038 subpattern is possible using named parentheses (see below).
2041 Another way of avoiding the ambiguity inherent in the use of digits following a
2042 backslash is to use the \g escape sequence. This escape must be followed by an
2043 unsigned number or a negative number, optionally enclosed in braces. These
2044 examples are all identical:
2050 An unsigned number specifies an absolute reference without the ambiguity that
2051 is present in the older syntax. It is also useful when literal digits follow
2052 the reference. A negative number is a relative reference. Consider this
2057 The sequence \g{-1} is a reference to the most recently started capturing
2058 subpattern before \g, that is, is it equivalent to \2 in this example.
2059 Similarly, \g{-2} would be equivalent to \1. The use of relative references
2060 can be helpful in long patterns, and also in patterns that are created by
2061 joining together fragments that contain references within themselves.
2064 A back reference matches whatever actually matched the capturing subpattern in
2065 the current subject string, rather than anything matching the subpattern
2067 <a href="#subpatternsassubroutines">"Subpatterns as subroutines"</a>
2068 below for a way of doing that). So the pattern
2070 (sens|respons)e and \1ibility
2072 matches "sense and sensibility" and "response and responsibility", but not
2073 "sense and responsibility". If caseful matching is in force at the time of the
2074 back reference, the case of letters is relevant. For example,
2078 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
2079 capturing subpattern is matched caselessly.
2082 There are several different ways of writing back references to named
2083 subpatterns. The .NET syntax \k{name} and the Perl syntax \k<name> or
2084 \k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
2085 back reference syntax, in which \g can be used for both numeric and named
2086 references, is also supported. We could rewrite the above example in any of
2089 (?<p1>(?i)rah)\s+\k<p1>
2090 (?'p1'(?i)rah)\s+\k{p1}
2091 (?P<p1>(?i)rah)\s+(?P=p1)
2092 (?<p1>(?i)rah)\s+\g{p1}
2094 A subpattern that is referenced by name may appear in the pattern before or
2095 after the reference.
2098 There may be more than one back reference to the same subpattern. If a
2099 subpattern has not actually been used in a particular match, any back
2100 references to it always fail by default. For example, the pattern
2104 always fails if it starts to match "a" rather than "bc". However, if the
2105 PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
2106 unset value matches an empty string.
2109 Because there may be many capturing parentheses in a pattern, all digits
2110 following a backslash are taken as part of a potential back reference number.
2111 If the pattern continues with a digit character, some delimiter must be used to
2112 terminate the back reference. If the PCRE_EXTENDED option is set, this can be
2113 white space. Otherwise, the \g{ syntax or an empty comment (see
2114 <a href="#comments">"Comments"</a>
2118 Recursive back references
2121 A back reference that occurs inside the parentheses to which it refers fails
2122 when the subpattern is first used, so, for example, (a\1) never matches.
2123 However, such references can be useful inside repeated subpatterns. For
2124 example, the pattern
2128 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
2129 the subpattern, the back reference matches the character string corresponding
2130 to the previous iteration. In order for this to work, the pattern must be such
2131 that the first iteration does not need to match the back reference. This can be
2132 done using alternation, as in the example above, or by a quantifier with a
2136 Back references of this type cause the group that they reference to be treated
2138 <a href="#atomicgroup">atomic group.</a>
2139 Once the whole group has been matched, a subsequent matching failure cannot
2140 cause backtracking into the middle of the group.
2141 <a name="bigassertions"></a></P>
2142 <br><a name="SEC20" href="#TOC1">ASSERTIONS</a><br>
2144 An assertion is a test on the characters following or preceding the current
2145 matching point that does not actually consume any characters. The simple
2146 assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are described
2147 <a href="#smallassertions">above.</a>
2150 More complicated assertions are coded as subpatterns. There are two kinds:
2151 those that look ahead of the current position in the subject string, and those
2152 that look behind it. An assertion subpattern is matched in the normal way,
2153 except that it does not cause the current matching position to be changed.
2156 Assertion subpatterns are not capturing subpatterns. If such an assertion
2157 contains capturing subpatterns within it, these are counted for the purposes of
2158 numbering the capturing subpatterns in the whole pattern. However, substring
2159 capturing is carried out only for positive assertions. (Perl sometimes, but not
2160 always, does do capturing in negative assertions.)
2163 For compatibility with Perl, assertion subpatterns may be repeated; though
2164 it makes no sense to assert the same thing several times, the side effect of
2165 capturing parentheses may occasionally be useful. In practice, there only three
2169 (1) If the quantifier is {0}, the assertion is never obeyed during matching.
2170 However, it may contain internal capturing parenthesized groups that are called
2171 from elsewhere via the
2172 <a href="#subpatternsassubroutines">subroutine mechanism.</a>
2175 (2) If quantifier is {0,n} where n is greater than zero, it is treated as if it
2176 were {0,1}. At run time, the rest of the pattern match is tried with and
2177 without the assertion, the order depending on the greediness of the quantifier.
2180 (3) If the minimum repetition is greater than zero, the quantifier is ignored.
2181 The assertion is obeyed just once when encountered during matching.
2184 Lookahead assertions
2187 Lookahead assertions start with (?= for positive assertions and (?! for
2188 negative assertions. For example,
2192 matches a word followed by a semicolon, but does not include the semicolon in
2197 matches any occurrence of "foo" that is not followed by "bar". Note that the
2198 apparently similar pattern
2202 does not find an occurrence of "bar" that is preceded by something other than
2203 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
2204 (?!foo) is always true when the next three characters are "bar". A
2205 lookbehind assertion is needed to achieve the other effect.
2208 If you want to force a matching failure at some point in a pattern, the most
2209 convenient way to do it is with (?!) because an empty string always matches, so
2210 an assertion that requires there not to be an empty string must always fail.
2211 The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
2212 <a name="lookbehind"></a></P>
2214 Lookbehind assertions
2217 Lookbehind assertions start with (?<= for positive assertions and (?<! for
2218 negative assertions. For example,
2222 does find an occurrence of "bar" that is not preceded by "foo". The contents of
2223 a lookbehind assertion are restricted such that all the strings it matches must
2224 have a fixed length. However, if there are several top-level alternatives, they
2225 do not all have to have the same fixed length. Thus
2227 (?<=bullock|donkey)
2231 (?<!dogs?|cats?)
2233 causes an error at compile time. Branches that match different length strings
2234 are permitted only at the top level of a lookbehind assertion. This is an
2235 extension compared with Perl, which requires all branches to match the same
2236 length of string. An assertion such as
2240 is not permitted, because its single top-level branch can match two different
2241 lengths, but it is acceptable to PCRE if rewritten to use two top-level
2246 In some cases, the escape sequence \K
2247 <a href="#resetmatchstart">(see above)</a>
2248 can be used instead of a lookbehind assertion to get round the fixed-length
2252 The implementation of lookbehind assertions is, for each alternative, to
2253 temporarily move the current position back by the fixed length and then try to
2254 match. If there are insufficient characters before the current position, the
2258 In a UTF mode, PCRE does not allow the \C escape (which matches a single data
2259 unit even in a UTF mode) to appear in lookbehind assertions, because it makes
2260 it impossible to calculate the length of the lookbehind. The \X and \R
2261 escapes, which can match different numbers of data units, are also not
2265 <a href="#subpatternsassubroutines">"Subroutine"</a>
2266 calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
2267 as the subpattern matches a fixed-length string.
2268 <a href="#recursion">Recursion,</a>
2269 however, is not supported.
2272 Possessive quantifiers can be used in conjunction with lookbehind assertions to
2273 specify efficient matching of fixed-length strings at the end of subject
2274 strings. Consider a simple pattern such as
2278 when applied to a long string that does not match. Because matching proceeds
2279 from left to right, PCRE will look for each "a" in the subject and then see if
2280 what follows matches the rest of the pattern. If the pattern is specified as
2284 the initial .* matches the entire string at first, but when this fails (because
2285 there is no following "a"), it backtracks to match all but the last character,
2286 then all but the last two characters, and so on. Once again the search for "a"
2287 covers the entire string, from right to left, so we are no better off. However,
2288 if the pattern is written as
2292 there can be no backtracking for the .*+ item; it can match only the entire
2293 string. The subsequent lookbehind assertion does a single test on the last four
2294 characters. If it fails, the match fails immediately. For long strings, this
2295 approach makes a significant difference to the processing time.
2298 Using multiple assertions
2301 Several assertions (of any sort) may occur in succession. For example,
2303 (?<=\d{3})(?<!999)foo
2305 matches "foo" preceded by three digits that are not "999". Notice that each of
2306 the assertions is applied independently at the same point in the subject
2307 string. First there is a check that the previous three characters are all
2308 digits, and then there is a check that the same three characters are not "999".
2309 This pattern does <i>not</i> match "foo" preceded by six characters, the first
2310 of which are digits and the last three of which are not "999". For example, it
2311 doesn't match "123abcfoo". A pattern to do that is
2313 (?<=\d{3}...)(?<!999)foo
2315 This time the first assertion looks at the preceding six characters, checking
2316 that the first three are digits, and then the second assertion checks that the
2317 preceding three characters are not "999".
2320 Assertions can be nested in any combination. For example,
2322 (?<=(?<!foo)bar)baz
2324 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
2325 preceded by "foo", while
2327 (?<=\d{3}(?!999)...)foo
2329 is another pattern that matches "foo" preceded by three digits and any three
2330 characters that are not "999".
2331 <a name="conditions"></a></P>
2332 <br><a name="SEC21" href="#TOC1">CONDITIONAL SUBPATTERNS</a><br>
2334 It is possible to cause the matching process to obey a subpattern
2335 conditionally or to choose between two alternative subpatterns, depending on
2336 the result of an assertion, or whether a specific capturing subpattern has
2337 already been matched. The two possible forms of conditional subpattern are:
2339 (?(condition)yes-pattern)
2340 (?(condition)yes-pattern|no-pattern)
2342 If the condition is satisfied, the yes-pattern is used; otherwise the
2343 no-pattern (if present) is used. If there are more than two alternatives in the
2344 subpattern, a compile-time error occurs. Each of the two alternatives may
2345 itself contain nested subpatterns of any form, including conditional
2346 subpatterns; the restriction to two alternatives applies only at the level of
2347 the condition. This pattern fragment is an example where the alternatives are
2350 (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
2355 There are four kinds of condition: references to subpatterns, references to
2356 recursion, a pseudo-condition called DEFINE, and assertions.
2359 Checking for a used subpattern by number
2362 If the text between the parentheses consists of a sequence of digits, the
2363 condition is true if a capturing subpattern of that number has previously
2364 matched. If there is more than one capturing subpattern with the same number
2366 <a href="#recursion">section about duplicate subpattern numbers),</a>
2367 the condition is true if any of them have matched. An alternative notation is
2368 to precede the digits with a plus or minus sign. In this case, the subpattern
2369 number is relative rather than absolute. The most recently opened parentheses
2370 can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
2371 loops it can also make sense to refer to subsequent groups. The next
2372 parentheses to be opened can be referenced as (?(+1), and so on. (The value
2373 zero in any of these forms is not used; it provokes a compile-time error.)
2376 Consider the following pattern, which contains non-significant white space to
2377 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2378 three parts for ease of discussion:
2380 ( \( )? [^()]+ (?(1) \) )
2382 The first part matches an optional opening parenthesis, and if that
2383 character is present, sets it as the first captured substring. The second part
2384 matches one or more characters that are not parentheses. The third part is a
2385 conditional subpattern that tests whether or not the first set of parentheses
2386 matched. If they did, that is, if subject started with an opening parenthesis,
2387 the condition is true, and so the yes-pattern is executed and a closing
2388 parenthesis is required. Otherwise, since no-pattern is not present, the
2389 subpattern matches nothing. In other words, this pattern matches a sequence of
2390 non-parentheses, optionally enclosed in parentheses.
2393 If you were embedding this pattern in a larger one, you could use a relative
2396 ...other stuff... ( \( )? [^()]+ (?(-1) \) ) ...
2398 This makes the fragment independent of the parentheses in the larger pattern.
2401 Checking for a used subpattern by name
2404 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
2405 subpattern by name. For compatibility with earlier versions of PCRE, which had
2406 this facility before Perl, the syntax (?(name)...) is also recognized.
2409 Rewriting the above example to use a named subpattern gives this:
2411 (?<OPEN> \( )? [^()]+ (?(<OPEN>) \) )
2413 If the name used in a condition of this kind is a duplicate, the test is
2414 applied to all subpatterns of the same name, and is true if any one of them has
2418 Checking for pattern recursion
2421 If the condition is the string (R), and there is no subpattern with the name R,
2422 the condition is true if a recursive call to the whole pattern or any
2423 subpattern has been made. If digits or a name preceded by ampersand follow the
2424 letter R, for example:
2426 (?(R3)...) or (?(R&name)...)
2428 the condition is true if the most recent recursion is into a subpattern whose
2429 number or name is given. This condition does not check the entire recursion
2430 stack. If the name used in a condition of this kind is a duplicate, the test is
2431 applied to all subpatterns of the same name, and is true if any one of them is
2432 the most recent recursion.
2435 At "top level", all these recursion test conditions are false.
2436 <a href="#recursion">The syntax for recursive patterns</a>
2438 <a name="subdefine"></a></P>
2440 Defining subpatterns for use by reference only
2443 If the condition is the string (DEFINE), and there is no subpattern with the
2444 name DEFINE, the condition is always false. In this case, there may be only one
2445 alternative in the subpattern. It is always skipped if control reaches this
2446 point in the pattern; the idea of DEFINE is that it can be used to define
2447 subroutines that can be referenced from elsewhere. (The use of
2448 <a href="#subpatternsassubroutines">subroutines</a>
2449 is described below.) For example, a pattern to match an IPv4 address such as
2450 "192.168.23.245" could be written like this (ignore white space and line
2453 (?(DEFINE) (?<byte> 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) )
2454 \b (?&byte) (\.(?&byte)){3} \b
2456 The first part of the pattern is a DEFINE group inside which a another group
2457 named "byte" is defined. This matches an individual component of an IPv4
2458 address (a number less than 256). When matching takes place, this part of the
2459 pattern is skipped because DEFINE acts like a false condition. The rest of the
2460 pattern uses references to the named group to match the four dot-separated
2461 components of an IPv4 address, insisting on a word boundary at each end.
2464 Assertion conditions
2467 If the condition is not in any of the above formats, it must be an assertion.
2468 This may be a positive or negative lookahead or lookbehind assertion. Consider
2469 this pattern, again containing non-significant white space, and with the two
2470 alternatives on the second line:
2473 \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )
2475 The condition is a positive lookahead assertion that matches an optional
2476 sequence of non-letters followed by a letter. In other words, it tests for the
2477 presence of at least one letter in the subject. If a letter is found, the
2478 subject is matched against the first alternative; otherwise it is matched
2479 against the second. This pattern matches strings in one of the two forms
2480 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2481 <a name="comments"></a></P>
2482 <br><a name="SEC22" href="#TOC1">COMMENTS</a><br>
2484 There are two ways of including comments in patterns that are processed by
2485 PCRE. In both cases, the start of the comment must not be in a character class,
2486 nor in the middle of any other sequence of related characters such as (?: or a
2487 subpattern name or number. The characters that make up a comment play no part
2488 in the pattern matching.
2491 The sequence (?# marks the start of a comment that continues up to the next
2492 closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2493 option is set, an unescaped # character also introduces a comment, which in
2494 this case continues to immediately after the next newline character or
2495 character sequence in the pattern. Which characters are interpreted as newlines
2496 is controlled by the options passed to a compiling function or by a special
2497 sequence at the start of the pattern, as described in the section entitled
2498 <a href="#newlines">"Newline conventions"</a>
2499 above. Note that the end of this type of comment is a literal newline sequence
2500 in the pattern; escape sequences that happen to represent a newline do not
2501 count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2502 default newline convention is in force:
2504 abc #comment \n still comment
2506 On encountering the # character, <b>pcre_compile()</b> skips along, looking for
2507 a newline in the pattern. The sequence \n is still literal at this stage, so
2508 it does not terminate the comment. Only an actual character with the code value
2509 0x0a (the default newline) does so.
2510 <a name="recursion"></a></P>
2511 <br><a name="SEC23" href="#TOC1">RECURSIVE PATTERNS</a><br>
2513 Consider the problem of matching a string in parentheses, allowing for
2514 unlimited nested parentheses. Without the use of recursion, the best that can
2515 be done is to use a pattern that matches up to some fixed depth of nesting. It
2516 is not possible to handle an arbitrary nesting depth.
2519 For some time, Perl has provided a facility that allows regular expressions to
2520 recurse (amongst other things). It does this by interpolating Perl code in the
2521 expression at run time, and the code can refer to the expression itself. A Perl
2522 pattern using code interpolation to solve the parentheses problem can be
2525 $re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x;
2527 The (?p{...}) item interpolates Perl code at run time, and in this case refers
2528 recursively to the pattern in which it appears.
2531 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2532 supports special syntax for recursion of the entire pattern, and also for
2533 individual subpattern recursion. After its introduction in PCRE and Python,
2534 this kind of recursion was subsequently introduced into Perl at release 5.10.
2537 A special item that consists of (? followed by a number greater than zero and a
2538 closing parenthesis is a recursive subroutine call of the subpattern of the
2539 given number, provided that it occurs inside that subpattern. (If not, it is a
2540 <a href="#subpatternsassubroutines">non-recursive subroutine</a>
2541 call, which is described in the next section.) The special item (?R) or (?0) is
2542 a recursive call of the entire regular expression.
2545 This PCRE pattern solves the nested parentheses problem (assume the
2546 PCRE_EXTENDED option is set so that white space is ignored):
2548 \( ( [^()]++ | (?R) )* \)
2550 First it matches an opening parenthesis. Then it matches any number of
2551 substrings which can either be a sequence of non-parentheses, or a recursive
2552 match of the pattern itself (that is, a correctly parenthesized substring).
2553 Finally there is a closing parenthesis. Note the use of a possessive quantifier
2554 to avoid backtracking into sequences of non-parentheses.
2557 If this were part of a larger pattern, you would not want to recurse the entire
2558 pattern, so instead you could use this:
2560 ( \( ( [^()]++ | (?1) )* \) )
2562 We have put the pattern into parentheses, and caused the recursion to refer to
2563 them instead of the whole pattern.
2566 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2567 is made easier by the use of relative references. Instead of (?1) in the
2568 pattern above you can write (?-2) to refer to the second most recently opened
2569 parentheses preceding the recursion. In other words, a negative number counts
2570 capturing parentheses leftwards from the point at which it is encountered.
2573 It is also possible to refer to subsequently opened parentheses, by writing
2574 references such as (?+2). However, these cannot be recursive because the
2575 reference is not inside the parentheses that are referenced. They are always
2576 <a href="#subpatternsassubroutines">non-recursive subroutine</a>
2577 calls, as described in the next section.
2580 An alternative approach is to use named parentheses instead. The Perl syntax
2581 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
2582 could rewrite the above example as follows:
2584 (?<pn> \( ( [^()]++ | (?&pn) )* \) )
2586 If there is more than one subpattern with the same name, the earliest one is
2590 This particular example pattern that we have been looking at contains nested
2591 unlimited repeats, and so the use of a possessive quantifier for matching
2592 strings of non-parentheses is important when applying the pattern to strings
2593 that do not match. For example, when this pattern is applied to
2595 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2597 it yields "no match" quickly. However, if a possessive quantifier is not used,
2598 the match runs for a very long time indeed because there are so many different
2599 ways the + and * repeats can carve up the subject, and all have to be tested
2600 before failure can be reported.
2603 At the end of a match, the values of capturing parentheses are those from
2604 the outermost level. If you want to obtain intermediate values, a callout
2605 function can be used (see below and the
2606 <a href="pcrecallout.html"><b>pcrecallout</b></a>
2607 documentation). If the pattern above is matched against
2611 the value for the inner capturing parentheses (numbered 2) is "ef", which is
2612 the last value taken on at the top level. If a capturing subpattern is not
2613 matched at the top level, its final captured value is unset, even if it was
2614 (temporarily) set at a deeper level during the matching process.
2617 If there are more than 15 capturing parentheses in a pattern, PCRE has to
2618 obtain extra memory to store data during a recursion, which it does by using
2619 <b>pcre_malloc</b>, freeing it via <b>pcre_free</b> afterwards. If no memory can
2620 be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2623 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2624 Consider this pattern, which matches text in angle brackets, allowing for
2625 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2626 recursing), whereas any characters are permitted at the outer level.
2628 < (?: (?(R) \d++ | [^<>]*+) | (?R)) * >
2630 In this pattern, (?(R) is the start of a conditional subpattern, with two
2631 different alternatives for the recursive and non-recursive cases. The (?R) item
2632 is the actual recursive call.
2633 <a name="recursiondifference"></a></P>
2635 Differences in recursion processing between PCRE and Perl
2638 Recursion processing in PCRE differs from Perl in two important ways. In PCRE
2639 (like Python, but unlike Perl), a recursive subpattern call is always treated
2640 as an atomic group. That is, once it has matched some of the subject string, it
2641 is never re-entered, even if it contains untried alternatives and there is a
2642 subsequent matching failure. This can be illustrated by the following pattern,
2643 which purports to match a palindromic string that contains an odd number of
2644 characters (for example, "a", "aba", "abcba", "abcdcba"):
2648 The idea is that it either matches a single character, or two identical
2649 characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2650 it does not if the pattern is longer than three characters. Consider the
2651 subject string "abcba":
2654 At the top level, the first character is matched, but as it is not at the end
2655 of the string, the first alternative fails; the second alternative is taken
2656 and the recursion kicks in. The recursive call to subpattern 1 successfully
2657 matches the next character ("b"). (Note that the beginning and end of line
2658 tests are not part of the recursion).
2661 Back at the top level, the next character ("c") is compared with what
2662 subpattern 2 matched, which was "a". This fails. Because the recursion is
2663 treated as an atomic group, there are now no backtracking points, and so the
2664 entire match fails. (Perl is able, at this point, to re-enter the recursion and
2665 try the second alternative.) However, if the pattern is written with the
2666 alternatives in the other order, things are different:
2670 This time, the recursing alternative is tried first, and continues to recurse
2671 until it runs out of characters, at which point the recursion fails. But this
2672 time we do have another alternative to try at the higher level. That is the big
2673 difference: in the previous case the remaining alternative is at a deeper
2674 recursion level, which PCRE cannot use.
2677 To change the pattern so that it matches all palindromic strings, not just
2678 those with an odd number of characters, it is tempting to change the pattern to
2683 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2684 deeper recursion has matched a single character, it cannot be entered again in
2685 order to match an empty string. The solution is to separate the two cases, and
2686 write out the odd and even cases as alternatives at the higher level:
2688 ^(?:((.)(?1)\2|)|((.)(?3)\4|.))
2690 If you want to match typical palindromic phrases, the pattern has to ignore all
2691 non-word characters, which can be done like this:
2693 ^\W*+(?:((.)\W*+(?1)\W*+\2|)|((.)\W*+(?3)\W*+\4|\W*+.\W*+))\W*+$
2695 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2696 man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2697 the use of the possessive quantifier *+ to avoid backtracking into sequences of
2698 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2699 more) to match typical phrases, and Perl takes so long that you think it has
2703 <b>WARNING</b>: The palindrome-matching patterns above work only if the subject
2704 string does not start with a palindrome that is shorter than the entire string.
2705 For example, although "abcba" is correctly matched, if the subject is "ababa",
2706 PCRE finds the palindrome "aba" at the start, then fails at top level because
2707 the end of the string does not follow. Once again, it cannot jump back into the
2708 recursion to try other alternatives, so the entire match fails.
2711 The second way in which PCRE and Perl differ in their recursion processing is
2712 in the handling of captured values. In Perl, when a subpattern is called
2713 recursively or as a subpattern (see the next section), it has no access to any
2714 values that were captured outside the recursion, whereas in PCRE these values
2715 can be referenced. Consider this pattern:
2719 In PCRE, this pattern matches "bab". The first capturing parentheses match "b",
2720 then in the second group, when the back reference \1 fails to match "b", the
2721 second alternative matches "a" and then recurses. In the recursion, \1 does
2722 now match "b" and so the whole match succeeds. In Perl, the pattern fails to
2723 match because inside the recursive call \1 cannot access the externally set
2725 <a name="subpatternsassubroutines"></a></P>
2726 <br><a name="SEC24" href="#TOC1">SUBPATTERNS AS SUBROUTINES</a><br>
2728 If the syntax for a recursive subpattern call (either by number or by
2729 name) is used outside the parentheses to which it refers, it operates like a
2730 subroutine in a programming language. The called subpattern may be defined
2731 before or after the reference. A numbered reference can be absolute or
2732 relative, as in these examples:
2734 (...(absolute)...)...(?2)...
2735 (...(relative)...)...(?-1)...
2736 (...(?+1)...(relative)...
2738 An earlier example pointed out that the pattern
2740 (sens|respons)e and \1ibility
2742 matches "sense and sensibility" and "response and responsibility", but not
2743 "sense and responsibility". If instead the pattern
2745 (sens|respons)e and (?1)ibility
2747 is used, it does match "sense and responsibility" as well as the other two
2748 strings. Another example is given in the discussion of DEFINE above.
2751 All subroutine calls, whether recursive or not, are always treated as atomic
2752 groups. That is, once a subroutine has matched some of the subject string, it
2753 is never re-entered, even if it contains untried alternatives and there is a
2754 subsequent matching failure. Any capturing parentheses that are set during the
2755 subroutine call revert to their previous values afterwards.
2758 Processing options such as case-independence are fixed when a subpattern is
2759 defined, so if it is used as a subroutine, such options cannot be changed for
2760 different calls. For example, consider this pattern:
2764 It matches "abcabc". It does not match "abcABC" because the change of
2765 processing option does not affect the called subpattern.
2766 <a name="onigurumasubroutines"></a></P>
2767 <br><a name="SEC25" href="#TOC1">ONIGURUMA SUBROUTINE SYNTAX</a><br>
2769 For compatibility with Oniguruma, the non-Perl syntax \g followed by a name or
2770 a number enclosed either in angle brackets or single quotes, is an alternative
2771 syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2772 are two of the examples used above, rewritten using this syntax:
2774 (?<pn> \( ( (?>[^()]+) | \g<pn> )* \) )
2775 (sens|respons)e and \g'1'ibility
2777 PCRE supports an extension to Oniguruma: if a number is preceded by a
2778 plus or a minus sign it is taken as a relative reference. For example:
2780 (abc)(?i:\g<-1>)
2782 Note that \g{...} (Perl syntax) and \g<...> (Oniguruma syntax) are <i>not</i>
2783 synonymous. The former is a back reference; the latter is a subroutine call.
2785 <br><a name="SEC26" href="#TOC1">CALLOUTS</a><br>
2787 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2788 code to be obeyed in the middle of matching a regular expression. This makes it
2789 possible, amongst other things, to extract different substrings that match the
2790 same pair of parentheses when there is a repetition.
2793 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2794 code. The feature is called "callout". The caller of PCRE provides an external
2795 function by putting its entry point in the global variable <i>pcre_callout</i>
2796 (8-bit library) or <i>pcre[16|32]_callout</i> (16-bit or 32-bit library).
2797 By default, this variable contains NULL, which disables all calling out.
2800 Within a regular expression, (?C) indicates the points at which the external
2801 function is to be called. If you want to identify different callout points, you
2802 can put a number less than 256 after the letter C. The default value is zero.
2803 For example, this pattern has two callout points:
2807 If the PCRE_AUTO_CALLOUT flag is passed to a compiling function, callouts are
2808 automatically installed before each item in the pattern. They are all numbered
2809 255. If there is a conditional group in the pattern whose condition is an
2810 assertion, an additional callout is inserted just before the condition. An
2811 explicit callout may also be set at this position, as in this example:
2813 (?(?C9)(?=a)abc|def)
2815 Note that this applies only to assertion conditions, not to other types of
2819 During matching, when PCRE reaches a callout point, the external function is
2820 called. It is provided with the number of the callout, the position in the
2821 pattern, and, optionally, one item of data originally supplied by the caller of
2822 the matching function. The callout function may cause matching to proceed, to
2823 backtrack, or to fail altogether.
2826 By default, PCRE implements a number of optimizations at compile time and
2827 matching time, and one side-effect is that sometimes callouts are skipped. If
2828 you need all possible callouts to happen, you need to set options that disable
2829 the relevant optimizations. More details, and a complete description of the
2830 interface to the callout function, are given in the
2831 <a href="pcrecallout.html"><b>pcrecallout</b></a>
2833 <a name="backtrackcontrol"></a></P>
2834 <br><a name="SEC27" href="#TOC1">BACKTRACKING CONTROL</a><br>
2836 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2837 are still described in the Perl documentation as "experimental and subject to
2838 change or removal in a future version of Perl". It goes on to say: "Their usage
2839 in production code should be noted to avoid problems during upgrades." The same
2840 remarks apply to the PCRE features described in this section.
2843 The new verbs make use of what was previously invalid syntax: an opening
2844 parenthesis followed by an asterisk. They are generally of the form
2845 (*VERB) or (*VERB:NAME). Some may take either form, possibly behaving
2846 differently depending on whether or not a name is present. A name is any
2847 sequence of characters that does not include a closing parenthesis. The maximum
2848 length of name is 255 in the 8-bit library and 65535 in the 16-bit and 32-bit
2849 libraries. If the name is empty, that is, if the closing parenthesis
2850 immediately follows the colon, the effect is as if the colon were not there.
2851 Any number of these verbs may occur in a pattern.
2854 Since these verbs are specifically related to backtracking, most of them can be
2855 used only when the pattern is to be matched using one of the traditional
2856 matching functions, because these use a backtracking algorithm. With the
2857 exception of (*FAIL), which behaves like a failing negative assertion, the
2858 backtracking control verbs cause an error if encountered by a DFA matching
2862 The behaviour of these verbs in
2863 <a href="#btrepeat">repeated groups,</a>
2864 <a href="#btassert">assertions,</a>
2866 <a href="#btsub">subpatterns called as subroutines</a>
2867 (whether or not recursively) is documented below.
2868 <a name="nooptimize"></a></P>
2870 Optimizations that affect backtracking verbs
2873 PCRE contains some optimizations that are used to speed up matching by running
2874 some checks at the start of each match attempt. For example, it may know the
2875 minimum length of matching subject, or that a particular character must be
2876 present. When one of these optimizations bypasses the running of a match, any
2877 included backtracking verbs will not, of course, be processed. You can suppress
2878 the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2879 when calling <b>pcre_compile()</b> or <b>pcre_exec()</b>, or by starting the
2880 pattern with (*NO_START_OPT). There is more discussion of this option in the
2882 <a href="pcreapi.html#execoptions">"Option bits for <b>pcre_exec()</b>"</a>
2884 <a href="pcreapi.html"><b>pcreapi</b></a>
2888 Experiments with Perl suggest that it too has similar optimizations, sometimes
2889 leading to anomalous results.
2892 Verbs that act immediately
2895 The following verbs act as soon as they are encountered. They may not be
2900 This verb causes the match to end successfully, skipping the remainder of the
2901 pattern. However, when it is inside a subpattern that is called as a
2902 subroutine, only that subpattern is ended successfully. Matching then continues
2903 at the outer level. If (*ACCEPT) in triggered in a positive assertion, the
2904 assertion succeeds; in a negative assertion, the assertion fails.
2907 If (*ACCEPT) is inside capturing parentheses, the data so far is captured. For
2910 A((?:A|B(*ACCEPT)|C)D)
2912 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2913 the outer parentheses.
2917 This verb causes a matching failure, forcing backtracking to occur. It is
2918 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2919 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2920 Perl features that are not present in PCRE. The nearest equivalent is the
2921 callout feature, as for example in this pattern:
2925 A match with the string "aaaa" always fails, but the callout is taken before
2926 each backtrack happens (in this example, 10 times).
2929 Recording which path was taken
2932 There is one verb whose main purpose is to track how a match was arrived at,
2933 though it also has a secondary use in conjunction with advancing the match
2934 starting point (see (*SKIP) below).
2936 (*MARK:NAME) or (*:NAME)
2938 A name is always required with this verb. There may be as many instances of
2939 (*MARK) as you like in a pattern, and their names do not have to be unique.
2942 When a match succeeds, the name of the last-encountered (*MARK:NAME),
2943 (*PRUNE:NAME), or (*THEN:NAME) on the matching path is passed back to the
2944 caller as described in the section entitled
2945 <a href="pcreapi.html#extradata">"Extra data for <b>pcre_exec()</b>"</a>
2947 <a href="pcreapi.html"><b>pcreapi</b></a>
2948 documentation. Here is an example of <b>pcretest</b> output, where the /K
2949 modifier requests the retrieval and outputting of (*MARK) data:
2951 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2959 The (*MARK) name is tagged with "MK:" in this output, and in this example it
2960 indicates which of the two alternatives matched. This is a more efficient way
2961 of obtaining this information than putting each alternative in its own
2962 capturing parentheses.
2965 If a verb with a name is encountered in a positive assertion that is true, the
2966 name is recorded and passed back if it is the last-encountered. This does not
2967 happen for negative assertions or failing positive assertions.
2970 After a partial match or a failed match, the last encountered name in the
2971 entire match process is returned. For example:
2973 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2977 Note that in this unanchored example the mark is retained from the match
2978 attempt that started at the letter "X" in the subject. Subsequent match
2979 attempts starting at "P" and then with an empty string do not get as far as the
2980 (*MARK) item, but nevertheless do not reset it.
2983 If you are interested in (*MARK) values after failed matches, you should
2984 probably set the PCRE_NO_START_OPTIMIZE option
2985 <a href="#nooptimize">(see above)</a>
2986 to ensure that the match is always attempted.
2989 Verbs that act after backtracking
2992 The following verbs do nothing when they are encountered. Matching continues
2993 with what follows, but if there is no subsequent match, causing a backtrack to
2994 the verb, a failure is forced. That is, backtracking cannot pass to the left of
2995 the verb. However, when one of these verbs appears inside an atomic group or an
2996 assertion that is true, its effect is confined to that group, because once the
2997 group has been matched, there is never any backtracking into it. In this
2998 situation, backtracking can "jump back" to the left of the entire atomic group
2999 or assertion. (Remember also, as stated above, that this localization also
3000 applies in subroutine calls.)
3003 These verbs differ in exactly what kind of failure occurs when backtracking
3004 reaches them. The behaviour described below is what happens when the verb is
3005 not in a subroutine or an assertion. Subsequent sections cover these special
3010 This verb, which may not be followed by a name, causes the whole match to fail
3011 outright if there is a later matching failure that causes backtracking to reach
3012 it. Even if the pattern is unanchored, no further attempts to find a match by
3013 advancing the starting point take place. If (*COMMIT) is the only backtracking
3014 verb that is encountered, once it has been passed <b>pcre_exec()</b> is
3015 committed to finding a match at the current starting point, or not at all. For
3020 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
3021 dynamic anchor, or "I've started, so I must finish." The name of the most
3022 recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
3026 If there is more than one backtracking verb in a pattern, a different one that
3027 follows (*COMMIT) may be triggered first, so merely passing (*COMMIT) during a
3028 match does not always guarantee that a match must be at this starting point.
3031 Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
3032 unless PCRE's start-of-match optimizations are turned off, as shown in this
3033 output from <b>pcretest</b>:
3035 re> /(*COMMIT)abc/
3041 For this pattern, PCRE knows that any match must start with "a", so the
3042 optimization skips along the subject to "a" before applying the pattern to the
3043 first set of data. The match attempt then succeeds. In the second set of data,
3044 the escape sequence \Y is interpreted by the <b>pcretest</b> program. It causes
3045 the PCRE_NO_START_OPTIMIZE option to be set when <b>pcre_exec()</b> is called.
3046 This disables the optimization that skips along to the first character. The
3047 pattern is now applied starting at "x", and so the (*COMMIT) causes the match
3048 to fail without trying any other starting points.
3050 (*PRUNE) or (*PRUNE:NAME)
3052 This verb causes the match to fail at the current starting position in the
3053 subject if there is a later matching failure that causes backtracking to reach
3054 it. If the pattern is unanchored, the normal "bumpalong" advance to the next
3055 starting character then happens. Backtracking can occur as usual to the left of
3056 (*PRUNE), before it is reached, or when matching to the right of (*PRUNE), but
3057 if there is no match to the right, backtracking cannot cross (*PRUNE). In
3058 simple cases, the use of (*PRUNE) is just an alternative to an atomic group or
3059 possessive quantifier, but there are some uses of (*PRUNE) that cannot be
3060 expressed in any other way. In an anchored pattern (*PRUNE) has the same effect
3064 The behaviour of (*PRUNE:NAME) is the not the same as (*MARK:NAME)(*PRUNE).
3065 It is like (*MARK:NAME) in that the name is remembered for passing back to the
3066 caller. However, (*SKIP:NAME) searches only for names set with (*MARK).
3070 This verb, when given without a name, is like (*PRUNE), except that if the
3071 pattern is unanchored, the "bumpalong" advance is not to the next character,
3072 but to the position in the subject where (*SKIP) was encountered. (*SKIP)
3073 signifies that whatever text was matched leading up to it cannot be part of a
3074 successful match. Consider:
3078 If the subject is "aaaac...", after the first match attempt fails (starting at
3079 the first character in the string), the starting point skips on to start the
3080 next attempt at "c". Note that a possessive quantifer does not have the same
3081 effect as this example; although it would suppress backtracking during the
3082 first match attempt, the second attempt would start at the second character
3083 instead of skipping on to "c".
3087 When (*SKIP) has an associated name, its behaviour is modified. When it is
3088 triggered, the previous path through the pattern is searched for the most
3089 recent (*MARK) that has the same name. If one is found, the "bumpalong" advance
3090 is to the subject position that corresponds to that (*MARK) instead of to where
3091 (*SKIP) was encountered. If no (*MARK) with a matching name is found, the
3095 Note that (*SKIP:NAME) searches only for names set by (*MARK:NAME). It ignores
3096 names that are set by (*PRUNE:NAME) or (*THEN:NAME).
3098 (*THEN) or (*THEN:NAME)
3100 This verb causes a skip to the next innermost alternative when backtracking
3101 reaches it. That is, it cancels any further backtracking within the current
3102 alternative. Its name comes from the observation that it can be used for a
3103 pattern-based if-then-else block:
3105 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
3107 If the COND1 pattern matches, FOO is tried (and possibly further items after
3108 the end of the group if FOO succeeds); on failure, the matcher skips to the
3109 second alternative and tries COND2, without backtracking into COND1. If that
3110 succeeds and BAR fails, COND3 is tried. If subsequently BAZ fails, there are no
3111 more alternatives, so there is a backtrack to whatever came before the entire
3112 group. If (*THEN) is not inside an alternation, it acts like (*PRUNE).
3115 The behaviour of (*THEN:NAME) is the not the same as (*MARK:NAME)(*THEN).
3116 It is like (*MARK:NAME) in that the name is remembered for passing back to the
3117 caller. However, (*SKIP:NAME) searches only for names set with (*MARK).
3120 A subpattern that does not contain a | character is just a part of the
3121 enclosing alternative; it is not a nested alternation with only one
3122 alternative. The effect of (*THEN) extends beyond such a subpattern to the
3123 enclosing alternative. Consider this pattern, where A, B, etc. are complex
3124 pattern fragments that do not contain any | characters at this level:
3128 If A and B are matched, but there is a failure in C, matching does not
3129 backtrack into A; instead it moves to the next alternative, that is, D.
3130 However, if the subpattern containing (*THEN) is given an alternative, it
3131 behaves differently:
3133 A (B(*THEN)C | (*FAIL)) | D
3135 The effect of (*THEN) is now confined to the inner subpattern. After a failure
3136 in C, matching moves to (*FAIL), which causes the whole subpattern to fail
3137 because there are no more alternatives to try. In this case, matching does now
3141 Note that a conditional subpattern is not considered as having two
3142 alternatives, because only one is ever used. In other words, the | character in
3143 a conditional subpattern has a different meaning. Ignoring white space,
3146 ^.*? (?(?=a) a | b(*THEN)c )
3148 If the subject is "ba", this pattern does not match. Because .*? is ungreedy,
3149 it initially matches zero characters. The condition (?=a) then fails, the
3150 character "b" is matched, but "c" is not. At this point, matching does not
3151 backtrack to .*? as might perhaps be expected from the presence of the |
3152 character. The conditional subpattern is part of the single alternative that
3153 comprises the whole pattern, and so the match fails. (If there was a backtrack
3154 into .*?, allowing it to match "b", the match would succeed.)
3157 The verbs just described provide four different "strengths" of control when
3158 subsequent matching fails. (*THEN) is the weakest, carrying on the match at the
3159 next alternative. (*PRUNE) comes next, failing the match at the current
3160 starting position, but allowing an advance to the next character (for an
3161 unanchored pattern). (*SKIP) is similar, except that the advance may be more
3162 than one character. (*COMMIT) is the strongest, causing the entire match to
3166 More than one backtracking verb
3169 If more than one backtracking verb is present in a pattern, the one that is
3170 backtracked onto first acts. For example, consider this pattern, where A, B,
3171 etc. are complex pattern fragments:
3173 (A(*COMMIT)B(*THEN)C|ABD)
3175 If A matches but B fails, the backtrack to (*COMMIT) causes the entire match to
3176 fail. However, if A and B match, but C fails, the backtrack to (*THEN) causes
3177 the next alternative (ABD) to be tried. This behaviour is consistent, but is
3178 not always the same as Perl's. It means that if two or more backtracking verbs
3179 appear in succession, all the the last of them has no effect. Consider this
3182 ...(*COMMIT)(*PRUNE)...
3184 If there is a matching failure to the right, backtracking onto (*PRUNE) causes
3185 it to be triggered, and its action is taken. There can never be a backtrack
3187 <a name="btrepeat"></a></P>
3189 Backtracking verbs in repeated groups
3192 PCRE differs from Perl in its handling of backtracking verbs in repeated
3193 groups. For example, consider:
3197 If the subject is "abac", Perl matches, but PCRE fails because the (*COMMIT) in
3198 the second repeat of the group acts.
3199 <a name="btassert"></a></P>
3201 Backtracking verbs in assertions
3204 (*FAIL) in an assertion has its normal effect: it forces an immediate backtrack.
3207 (*ACCEPT) in a positive assertion causes the assertion to succeed without any
3208 further processing. In a negative assertion, (*ACCEPT) causes the assertion to
3209 fail without any further processing.
3212 The other backtracking verbs are not treated specially if they appear in a
3213 positive assertion. In particular, (*THEN) skips to the next alternative in the
3214 innermost enclosing group that has alternations, whether or not this is within
3218 Negative assertions are, however, different, in order to ensure that changing a
3219 positive assertion into a negative assertion changes its result. Backtracking
3220 into (*COMMIT), (*SKIP), or (*PRUNE) causes a negative assertion to be true,
3221 without considering any further alternative branches in the assertion.
3222 Backtracking into (*THEN) causes it to skip to the next enclosing alternative
3223 within the assertion (the normal behaviour), but if the assertion does not have
3224 such an alternative, (*THEN) behaves like (*PRUNE).
3225 <a name="btsub"></a></P>
3227 Backtracking verbs in subroutines
3230 These behaviours occur whether or not the subpattern is called recursively.
3231 Perl's treatment of subroutines is different in some cases.
3234 (*FAIL) in a subpattern called as a subroutine has its normal effect: it forces
3235 an immediate backtrack.
3238 (*ACCEPT) in a subpattern called as a subroutine causes the subroutine match to
3239 succeed without any further processing. Matching then continues after the
3243 (*COMMIT), (*SKIP), and (*PRUNE) in a subpattern called as a subroutine cause
3244 the subroutine match to fail.
3247 (*THEN) skips to the next alternative in the innermost enclosing group within
3248 the subpattern that has alternatives. If there is no such group within the
3249 subpattern, (*THEN) causes the subroutine match to fail.
3251 <br><a name="SEC28" href="#TOC1">SEE ALSO</a><br>
3253 <b>pcreapi</b>(3), <b>pcrecallout</b>(3), <b>pcrematching</b>(3),
3254 <b>pcresyntax</b>(3), <b>pcre</b>(3), <b>pcre16(3)</b>, <b>pcre32(3)</b>.
3256 <br><a name="SEC29" href="#TOC1">AUTHOR</a><br>
3260 University Computing Service
3262 Cambridge CB2 3QH, England.
3265 <br><a name="SEC30" href="#TOC1">REVISION</a><br>
3267 Last updated: 14 June 2015
3269 Copyright © 1997-2015 University of Cambridge.
3272 Return to the <a href="index.html">PCRE index page</a>.
~ian / pcre3.git / blob