• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1<html>
2<head>
3<title>pcrepattern specification</title>
4</head>
5<body bgcolor="#FFFFFF" text="#00005A" link="#0066FF" alink="#3399FF" vlink="#2222BB">
6<h1>pcrepattern man page</h1>
7<p>
8Return to the <a href="index.html">PCRE index page</a>.
9</p>
10<p>
11This page is part of the PCRE HTML documentation. It was generated automatically
12from the original man page. If there is any nonsense in it, please consult the
13man page, in case the conversion went wrong.
14<br>
15<ul>
16<li><a name="TOC1" href="#SEC1">PCRE REGULAR EXPRESSION DETAILS</a>
17<li><a name="TOC2" href="#SEC2">NEWLINE CONVENTIONS</a>
18<li><a name="TOC3" href="#SEC3">CHARACTERS AND METACHARACTERS</a>
19<li><a name="TOC4" href="#SEC4">BACKSLASH</a>
20<li><a name="TOC5" href="#SEC5">CIRCUMFLEX AND DOLLAR</a>
21<li><a name="TOC6" href="#SEC6">FULL STOP (PERIOD, DOT) AND \N</a>
22<li><a name="TOC7" href="#SEC7">MATCHING A SINGLE BYTE</a>
23<li><a name="TOC8" href="#SEC8">SQUARE BRACKETS AND CHARACTER CLASSES</a>
24<li><a name="TOC9" href="#SEC9">POSIX CHARACTER CLASSES</a>
25<li><a name="TOC10" href="#SEC10">VERTICAL BAR</a>
26<li><a name="TOC11" href="#SEC11">INTERNAL OPTION SETTING</a>
27<li><a name="TOC12" href="#SEC12">SUBPATTERNS</a>
28<li><a name="TOC13" href="#SEC13">DUPLICATE SUBPATTERN NUMBERS</a>
29<li><a name="TOC14" href="#SEC14">NAMED SUBPATTERNS</a>
30<li><a name="TOC15" href="#SEC15">REPETITION</a>
31<li><a name="TOC16" href="#SEC16">ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS</a>
32<li><a name="TOC17" href="#SEC17">BACK REFERENCES</a>
33<li><a name="TOC18" href="#SEC18">ASSERTIONS</a>
34<li><a name="TOC19" href="#SEC19">CONDITIONAL SUBPATTERNS</a>
35<li><a name="TOC20" href="#SEC20">COMMENTS</a>
36<li><a name="TOC21" href="#SEC21">RECURSIVE PATTERNS</a>
37<li><a name="TOC22" href="#SEC22">SUBPATTERNS AS SUBROUTINES</a>
38<li><a name="TOC23" href="#SEC23">ONIGURUMA SUBROUTINE SYNTAX</a>
39<li><a name="TOC24" href="#SEC24">CALLOUTS</a>
40<li><a name="TOC25" href="#SEC25">BACKTRACKING CONTROL</a>
41<li><a name="TOC26" href="#SEC26">SEE ALSO</a>
42<li><a name="TOC27" href="#SEC27">AUTHOR</a>
43<li><a name="TOC28" href="#SEC28">REVISION</a>
44</ul>
45<br><a name="SEC1" href="#TOC1">PCRE REGULAR EXPRESSION DETAILS</a><br>
46<P>
47The syntax and semantics of the regular expressions that are supported by PCRE
48are described in detail below. There is a quick-reference syntax summary in the
49<a href="pcresyntax.html"><b>pcresyntax</b></a>
50page. PCRE tries to match Perl syntax and semantics as closely as it can. PCRE
51also supports some alternative regular expression syntax (which does not
52conflict with the Perl syntax) in order to provide some compatibility with
53regular expressions in Python, .NET, and Oniguruma.
54</P>
55<P>
56Perl's regular expressions are described in its own documentation, and
57regular expressions in general are covered in a number of books, some of which
58have copious examples. Jeffrey Friedl's "Mastering Regular Expressions",
59published by O'Reilly, covers regular expressions in great detail. This
60description of PCRE's regular expressions is intended as reference material.
61</P>
62<P>
63The original operation of PCRE was on strings of one-byte characters. However,
64there is now also support for UTF-8 character strings. To use this,
65PCRE must be built to include UTF-8 support, and you must call
66<b>pcre_compile()</b> or <b>pcre_compile2()</b> with the PCRE_UTF8 option. There
67is also a special sequence that can be given at the start of a pattern:
68<pre>
69  (*UTF8)
70</pre>
71Starting a pattern with this sequence is equivalent to setting the PCRE_UTF8
72option. This feature is not Perl-compatible. How setting UTF-8 mode affects
73pattern matching is mentioned in several places below. There is also a summary
74of UTF-8 features in the
75<a href="pcre.html#utf8support">section on UTF-8 support</a>
76in the main
77<a href="pcre.html"><b>pcre</b></a>
78page.
79</P>
80<P>
81Another special sequence that may appear at the start of a pattern or in
82combination with (*UTF8) is:
83<pre>
84  (*UCP)
85</pre>
86This has the same effect as setting the PCRE_UCP option: it causes sequences
87such as \d and \w to use Unicode properties to determine character types,
88instead of recognizing only characters with codes less than 128 via a lookup
89table.
90</P>
91<P>
92If a pattern starts with (*NO_START_OPT), it has the same effect as setting the
93PCRE_NO_START_OPTIMIZE option either at compile or matching time. There are
94also some more of these special sequences that are concerned with the handling
95of newlines; they are described below.
96</P>
97<P>
98The remainder of this document discusses the patterns that are supported by
99PCRE when its main matching function, <b>pcre_exec()</b>, is used.
100From release 6.0, PCRE offers a second matching function,
101<b>pcre_dfa_exec()</b>, which matches using a different algorithm that is not
102Perl-compatible. Some of the features discussed below are not available when
103<b>pcre_dfa_exec()</b> is used. The advantages and disadvantages of the
104alternative function, and how it differs from the normal function, are
105discussed in the
106<a href="pcrematching.html"><b>pcrematching</b></a>
107page.
108<a name="newlines"></a></P>
109<br><a name="SEC2" href="#TOC1">NEWLINE CONVENTIONS</a><br>
110<P>
111PCRE supports five different conventions for indicating line breaks in
112strings: a single CR (carriage return) character, a single LF (linefeed)
113character, the two-character sequence CRLF, any of the three preceding, or any
114Unicode newline sequence. The
115<a href="pcreapi.html"><b>pcreapi</b></a>
116page has
117<a href="pcreapi.html#newlines">further discussion</a>
118about newlines, and shows how to set the newline convention in the
119<i>options</i> arguments for the compiling and matching functions.
120</P>
121<P>
122It is also possible to specify a newline convention by starting a pattern
123string with one of the following five sequences:
124<pre>
125  (*CR)        carriage return
126  (*LF)        linefeed
127  (*CRLF)      carriage return, followed by linefeed
128  (*ANYCRLF)   any of the three above
129  (*ANY)       all Unicode newline sequences
130</pre>
131These override the default and the options given to <b>pcre_compile()</b> or
132<b>pcre_compile2()</b>. For example, on a Unix system where LF is the default
133newline sequence, the pattern
134<pre>
135  (*CR)a.b
136</pre>
137changes the convention to CR. That pattern matches "a\nb" because LF is no
138longer a newline. Note that these special settings, which are not
139Perl-compatible, are recognized only at the very start of a pattern, and that
140they must be in upper case. If more than one of them is present, the last one
141is used.
142</P>
143<P>
144The newline convention affects the interpretation of the dot metacharacter when
145PCRE_DOTALL is not set, and also the behaviour of \N. However, it does not
146affect what the \R escape sequence matches. By default, this is any Unicode
147newline sequence, for Perl compatibility. However, this can be changed; see the
148description of \R in the section entitled
149<a href="#newlineseq">"Newline sequences"</a>
150below. A change of \R setting can be combined with a change of newline
151convention.
152</P>
153<br><a name="SEC3" href="#TOC1">CHARACTERS AND METACHARACTERS</a><br>
154<P>
155A regular expression is a pattern that is matched against a subject string from
156left to right. Most characters stand for themselves in a pattern, and match the
157corresponding characters in the subject. As a trivial example, the pattern
158<pre>
159  The quick brown fox
160</pre>
161matches a portion of a subject string that is identical to itself. When
162caseless matching is specified (the PCRE_CASELESS option), letters are matched
163independently of case. In UTF-8 mode, PCRE always understands the concept of
164case for characters whose values are less than 128, so caseless matching is
165always possible. For characters with higher values, the concept of case is
166supported if PCRE is compiled with Unicode property support, but not otherwise.
167If you want to use caseless matching for characters 128 and above, you must
168ensure that PCRE is compiled with Unicode property support as well as with
169UTF-8 support.
170</P>
171<P>
172The power of regular expressions comes from the ability to include alternatives
173and repetitions in the pattern. These are encoded in the pattern by the use of
174<i>metacharacters</i>, which do not stand for themselves but instead are
175interpreted in some special way.
176</P>
177<P>
178There are two different sets of metacharacters: those that are recognized
179anywhere in the pattern except within square brackets, and those that are
180recognized within square brackets. Outside square brackets, the metacharacters
181are as follows:
182<pre>
183  \      general escape character with several uses
184  ^      assert start of string (or line, in multiline mode)
185  $      assert end of string (or line, in multiline mode)
186  .      match any character except newline (by default)
187  [      start character class definition
188  |      start of alternative branch
189  (      start subpattern
190  )      end subpattern
191  ?      extends the meaning of (
192         also 0 or 1 quantifier
193         also quantifier minimizer
194  *      0 or more quantifier
195  +      1 or more quantifier
196         also "possessive quantifier"
197  {      start min/max quantifier
198</pre>
199Part of a pattern that is in square brackets is called a "character class". In
200a character class the only metacharacters are:
201<pre>
202  \      general escape character
203  ^      negate the class, but only if the first character
204  -      indicates character range
205  [      POSIX character class (only if followed by POSIX syntax)
206  ]      terminates the character class
207</pre>
208The following sections describe the use of each of the metacharacters.
209</P>
210<br><a name="SEC4" href="#TOC1">BACKSLASH</a><br>
211<P>
212The backslash character has several uses. Firstly, if it is followed by a
213character that is not a number or a letter, it takes away any special meaning
214that character may have. This use of backslash as an escape character applies
215both inside and outside character classes.
216</P>
217<P>
218For example, if you want to match a * character, you write \* in the pattern.
219This escaping action applies whether or not the following character would
220otherwise be interpreted as a metacharacter, so it is always safe to precede a
221non-alphanumeric with backslash to specify that it stands for itself. In
222particular, if you want to match a backslash, you write \\.
223</P>
224<P>
225In UTF-8 mode, only ASCII numbers and letters have any special meaning after a
226backslash. All other characters (in particular, those whose codepoints are
227greater than 127) are treated as literals.
228</P>
229<P>
230If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
231pattern (other than in a character class) and characters between a # outside
232a character class and the next newline are ignored. An escaping backslash can
233be used to include a whitespace or # character as part of the pattern.
234</P>
235<P>
236If you want to remove the special meaning from a sequence of characters, you
237can do so by putting them between \Q and \E. This is different from Perl in
238that $ and @ are handled as literals in \Q...\E sequences in PCRE, whereas in
239Perl, $ and @ cause variable interpolation. Note the following examples:
240<pre>
241  Pattern            PCRE matches   Perl matches
242
243  \Qabc$xyz\E        abc$xyz        abc followed by the contents of $xyz
244  \Qabc\$xyz\E       abc\$xyz       abc\$xyz
245  \Qabc\E\$\Qxyz\E   abc$xyz        abc$xyz
246</pre>
247The \Q...\E sequence is recognized both inside and outside character classes.
248An isolated \E that is not preceded by \Q is ignored.
249<a name="digitsafterbackslash"></a></P>
250<br><b>
251Non-printing characters
252</b><br>
253<P>
254A second use of backslash provides a way of encoding non-printing characters
255in patterns in a visible manner. There is no restriction on the appearance of
256non-printing characters, apart from the binary zero that terminates a pattern,
257but when a pattern is being prepared by text editing, it is often easier to use
258one of the following escape sequences than the binary character it represents:
259<pre>
260  \a        alarm, that is, the BEL character (hex 07)
261  \cx       "control-x", where x is any ASCII character
262  \e        escape (hex 1B)
263  \f        formfeed (hex 0C)
264  \n        linefeed (hex 0A)
265  \r        carriage return (hex 0D)
266  \t        tab (hex 09)
267  \ddd      character with octal code ddd, or back reference
268  \xhh      character with hex code hh
269  \x{hhh..} character with hex code hhh..
270</pre>
271The precise effect of \cx is as follows: if x is a lower case letter, it
272is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
273Thus \cz becomes hex 1A (z is 7A), but \c{ becomes hex 3B ({ is 7B), while
274\c; becomes hex 7B (; is 3B). If the byte following \c has a value greater
275than 127, a compile-time error occurs. This locks out non-ASCII characters in
276both byte mode and UTF-8 mode. (When PCRE is compiled in EBCDIC mode, all byte
277values are valid. A lower case letter is converted to upper case, and then the
2780xc0 bits are flipped.)
279</P>
280<P>
281After \x, from zero to two hexadecimal digits are read (letters can be in
282upper or lower case). Any number of hexadecimal digits may appear between \x{
283and }, but the value of the character code must be less than 256 in non-UTF-8
284mode, and less than 2**31 in UTF-8 mode. That is, the maximum value in
285hexadecimal is 7FFFFFFF. Note that this is bigger than the largest Unicode code
286point, which is 10FFFF.
287</P>
288<P>
289If characters other than hexadecimal digits appear between \x{ and }, or if
290there is no terminating }, this form of escape is not recognized. Instead, the
291initial \x will be interpreted as a basic hexadecimal escape, with no
292following digits, giving a character whose value is zero.
293</P>
294<P>
295Characters whose value is less than 256 can be defined by either of the two
296syntaxes for \x. There is no difference in the way they are handled. For
297example, \xdc is exactly the same as \x{dc}.
298</P>
299<P>
300After \0 up to two further octal digits are read. If there are fewer than two
301digits, just those that are present are used. Thus the sequence \0\x\07
302specifies two binary zeros followed by a BEL character (code value 7). Make
303sure you supply two digits after the initial zero if the pattern character that
304follows is itself an octal digit.
305</P>
306<P>
307The handling of a backslash followed by a digit other than 0 is complicated.
308Outside a character class, PCRE reads it and any following digits as a decimal
309number. If the number is less than 10, or if there have been at least that many
310previous capturing left parentheses in the expression, the entire sequence is
311taken as a <i>back reference</i>. A description of how this works is given
312<a href="#backreferences">later,</a>
313following the discussion of
314<a href="#subpattern">parenthesized subpatterns.</a>
315</P>
316<P>
317Inside a character class, or if the decimal number is greater than 9 and there
318have not been that many capturing subpatterns, PCRE re-reads up to three octal
319digits following the backslash, and uses them to generate a data character. Any
320subsequent digits stand for themselves. In non-UTF-8 mode, the value of a
321character specified in octal must be less than \400. In UTF-8 mode, values up
322to \777 are permitted. For example:
323<pre>
324  \040   is another way of writing a space
325  \40    is the same, provided there are fewer than 40 previous capturing subpatterns
326  \7     is always a back reference
327  \11    might be a back reference, or another way of writing a tab
328  \011   is always a tab
329  \0113  is a tab followed by the character "3"
330  \113   might be a back reference, otherwise the character with octal code 113
331  \377   might be a back reference, otherwise the byte consisting entirely of 1 bits
332  \81    is either a back reference, or a binary zero followed by the two characters "8" and "1"
333</pre>
334Note that octal values of 100 or greater must not be introduced by a leading
335zero, because no more than three octal digits are ever read.
336</P>
337<P>
338All the sequences that define a single character value can be used both inside
339and outside character classes. In addition, inside a character class, the
340sequence \b is interpreted as the backspace character (hex 08). The sequences
341\B, \N, \R, and \X are not special inside a character class. Like any other
342unrecognized escape sequences, they are treated as the literal characters "B",
343"N", "R", and "X" by default, but cause an error if the PCRE_EXTRA option is
344set. Outside a character class, these sequences have different meanings.
345</P>
346<br><b>
347Absolute and relative back references
348</b><br>
349<P>
350The sequence \g followed by an unsigned or a negative number, optionally
351enclosed in braces, is an absolute or relative back reference. A named back
352reference can be coded as \g{name}. Back references are discussed
353<a href="#backreferences">later,</a>
354following the discussion of
355<a href="#subpattern">parenthesized subpatterns.</a>
356</P>
357<br><b>
358Absolute and relative subroutine calls
359</b><br>
360<P>
361For compatibility with Oniguruma, the non-Perl syntax \g followed by a name or
362a number enclosed either in angle brackets or single quotes, is an alternative
363syntax for referencing a subpattern as a "subroutine". Details are discussed
364<a href="#onigurumasubroutines">later.</a>
365Note that \g{...} (Perl syntax) and \g&#60;...&#62; (Oniguruma syntax) are <i>not</i>
366synonymous. The former is a back reference; the latter is a
367<a href="#subpatternsassubroutines">subroutine</a>
368call.
369<a name="genericchartypes"></a></P>
370<br><b>
371Generic character types
372</b><br>
373<P>
374Another use of backslash is for specifying generic character types:
375<pre>
376  \d     any decimal digit
377  \D     any character that is not a decimal digit
378  \h     any horizontal whitespace character
379  \H     any character that is not a horizontal whitespace character
380  \s     any whitespace character
381  \S     any character that is not a whitespace character
382  \v     any vertical whitespace character
383  \V     any character that is not a vertical whitespace character
384  \w     any "word" character
385  \W     any "non-word" character
386</pre>
387There is also the single sequence \N, which matches a non-newline character.
388This is the same as
389<a href="#fullstopdot">the "." metacharacter</a>
390when PCRE_DOTALL is not set.
391</P>
392<P>
393Each pair of lower and upper case escape sequences partitions the complete set
394of characters into two disjoint sets. Any given character matches one, and only
395one, of each pair. The sequences can appear both inside and outside character
396classes. They each match one character of the appropriate type. If the current
397matching point is at the end of the subject string, all of them fail, because
398there is no character to match.
399</P>
400<P>
401For compatibility with Perl, \s does not match the VT character (code 11).
402This makes it different from the the POSIX "space" class. The \s characters
403are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
404included in a Perl script, \s may match the VT character. In PCRE, it never
405does.
406</P>
407<P>
408A "word" character is an underscore or any character that is a letter or digit.
409By default, the definition of letters and digits is controlled by PCRE's
410low-valued character tables, and may vary if locale-specific matching is taking
411place (see
412<a href="pcreapi.html#localesupport">"Locale support"</a>
413in the
414<a href="pcreapi.html"><b>pcreapi</b></a>
415page). For example, in a French locale such as "fr_FR" in Unix-like systems,
416or "french" in Windows, some character codes greater than 128 are used for
417accented letters, and these are then matched by \w. The use of locales with
418Unicode is discouraged.
419</P>
420<P>
421By default, in UTF-8 mode, characters with values greater than 128 never match
422\d, \s, or \w, and always match \D, \S, and \W. These sequences retain
423their original meanings from before UTF-8 support was available, mainly for
424efficiency reasons. However, if PCRE is compiled with Unicode property support,
425and the PCRE_UCP option is set, the behaviour is changed so that Unicode
426properties are used to determine character types, as follows:
427<pre>
428  \d  any character that \p{Nd} matches (decimal digit)
429  \s  any character that \p{Z} matches, plus HT, LF, FF, CR
430  \w  any character that \p{L} or \p{N} matches, plus underscore
431</pre>
432The upper case escapes match the inverse sets of characters. Note that \d
433matches only decimal digits, whereas \w matches any Unicode digit, as well as
434any Unicode letter, and underscore. Note also that PCRE_UCP affects \b, and
435\B because they are defined in terms of \w and \W. Matching these sequences
436is noticeably slower when PCRE_UCP is set.
437</P>
438<P>
439The sequences \h, \H, \v, and \V are features that were added to Perl at
440release 5.10. In contrast to the other sequences, which match only ASCII
441characters by default, these always match certain high-valued codepoints in
442UTF-8 mode, whether or not PCRE_UCP is set. The horizontal space characters
443are:
444<pre>
445  U+0009     Horizontal tab
446  U+0020     Space
447  U+00A0     Non-break space
448  U+1680     Ogham space mark
449  U+180E     Mongolian vowel separator
450  U+2000     En quad
451  U+2001     Em quad
452  U+2002     En space
453  U+2003     Em space
454  U+2004     Three-per-em space
455  U+2005     Four-per-em space
456  U+2006     Six-per-em space
457  U+2007     Figure space
458  U+2008     Punctuation space
459  U+2009     Thin space
460  U+200A     Hair space
461  U+202F     Narrow no-break space
462  U+205F     Medium mathematical space
463  U+3000     Ideographic space
464</pre>
465The vertical space characters are:
466<pre>
467  U+000A     Linefeed
468  U+000B     Vertical tab
469  U+000C     Formfeed
470  U+000D     Carriage return
471  U+0085     Next line
472  U+2028     Line separator
473  U+2029     Paragraph separator
474<a name="newlineseq"></a></PRE>
475</P>
476<br><b>
477Newline sequences
478</b><br>
479<P>
480Outside a character class, by default, the escape sequence \R matches any
481Unicode newline sequence. In non-UTF-8 mode \R is equivalent to the following:
482<pre>
483  (?&#62;\r\n|\n|\x0b|\f|\r|\x85)
484</pre>
485This is an example of an "atomic group", details of which are given
486<a href="#atomicgroup">below.</a>
487This particular group matches either the two-character sequence CR followed by
488LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
489U+000B), FF (formfeed, U+000C), CR (carriage return, U+000D), or NEL (next
490line, U+0085). The two-character sequence is treated as a single unit that
491cannot be split.
492</P>
493<P>
494In UTF-8 mode, two additional characters whose codepoints are greater than 255
495are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
496Unicode character property support is not needed for these characters to be
497recognized.
498</P>
499<P>
500It is possible to restrict \R to match only CR, LF, or CRLF (instead of the
501complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
502either at compile time or when the pattern is matched. (BSR is an abbrevation
503for "backslash R".) This can be made the default when PCRE is built; if this is
504the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
505It is also possible to specify these settings by starting a pattern string with
506one of the following sequences:
507<pre>
508  (*BSR_ANYCRLF)   CR, LF, or CRLF only
509  (*BSR_UNICODE)   any Unicode newline sequence
510</pre>
511These override the default and the options given to <b>pcre_compile()</b> or
512<b>pcre_compile2()</b>, but they can be overridden by options given to
513<b>pcre_exec()</b> or <b>pcre_dfa_exec()</b>. Note that these special settings,
514which are not Perl-compatible, are recognized only at the very start of a
515pattern, and that they must be in upper case. If more than one of them is
516present, the last one is used. They can be combined with a change of newline
517convention; for example, a pattern can start with:
518<pre>
519  (*ANY)(*BSR_ANYCRLF)
520</pre>
521They can also be combined with the (*UTF8) or (*UCP) special sequences. Inside
522a character class, \R is treated as an unrecognized escape sequence, and so
523matches the letter "R" by default, but causes an error if PCRE_EXTRA is set.
524<a name="uniextseq"></a></P>
525<br><b>
526Unicode character properties
527</b><br>
528<P>
529When PCRE is built with Unicode character property support, three additional
530escape sequences that match characters with specific properties are available.
531When not in UTF-8 mode, these sequences are of course limited to testing
532characters whose codepoints are less than 256, but they do work in this mode.
533The extra escape sequences are:
534<pre>
535  \p{<i>xx</i>}   a character with the <i>xx</i> property
536  \P{<i>xx</i>}   a character without the <i>xx</i> property
537  \X       an extended Unicode sequence
538</pre>
539The property names represented by <i>xx</i> above are limited to the Unicode
540script names, the general category properties, "Any", which matches any
541character (including newline), and some special PCRE properties (described
542in the
543<a href="#extraprops">next section).</a>
544Other Perl properties such as "InMusicalSymbols" are not currently supported by
545PCRE. Note that \P{Any} does not match any characters, so always causes a
546match failure.
547</P>
548<P>
549Sets of Unicode characters are defined as belonging to certain scripts. A
550character from one of these sets can be matched using a script name. For
551example:
552<pre>
553  \p{Greek}
554  \P{Han}
555</pre>
556Those that are not part of an identified script are lumped together as
557"Common". The current list of scripts is:
558</P>
559<P>
560Arabic,
561Armenian,
562Avestan,
563Balinese,
564Bamum,
565Bengali,
566Bopomofo,
567Braille,
568Buginese,
569Buhid,
570Canadian_Aboriginal,
571Carian,
572Cham,
573Cherokee,
574Common,
575Coptic,
576Cuneiform,
577Cypriot,
578Cyrillic,
579Deseret,
580Devanagari,
581Egyptian_Hieroglyphs,
582Ethiopic,
583Georgian,
584Glagolitic,
585Gothic,
586Greek,
587Gujarati,
588Gurmukhi,
589Han,
590Hangul,
591Hanunoo,
592Hebrew,
593Hiragana,
594Imperial_Aramaic,
595Inherited,
596Inscriptional_Pahlavi,
597Inscriptional_Parthian,
598Javanese,
599Kaithi,
600Kannada,
601Katakana,
602Kayah_Li,
603Kharoshthi,
604Khmer,
605Lao,
606Latin,
607Lepcha,
608Limbu,
609Linear_B,
610Lisu,
611Lycian,
612Lydian,
613Malayalam,
614Meetei_Mayek,
615Mongolian,
616Myanmar,
617New_Tai_Lue,
618Nko,
619Ogham,
620Old_Italic,
621Old_Persian,
622Old_South_Arabian,
623Old_Turkic,
624Ol_Chiki,
625Oriya,
626Osmanya,
627Phags_Pa,
628Phoenician,
629Rejang,
630Runic,
631Samaritan,
632Saurashtra,
633Shavian,
634Sinhala,
635Sundanese,
636Syloti_Nagri,
637Syriac,
638Tagalog,
639Tagbanwa,
640Tai_Le,
641Tai_Tham,
642Tai_Viet,
643Tamil,
644Telugu,
645Thaana,
646Thai,
647Tibetan,
648Tifinagh,
649Ugaritic,
650Vai,
651Yi.
652</P>
653<P>
654Each character has exactly one Unicode general category property, specified by
655a two-letter abbreviation. For compatibility with Perl, negation can be
656specified by including a circumflex between the opening brace and the property
657name. For example, \p{^Lu} is the same as \P{Lu}.
658</P>
659<P>
660If only one letter is specified with \p or \P, it includes all the general
661category properties that start with that letter. In this case, in the absence
662of negation, the curly brackets in the escape sequence are optional; these two
663examples have the same effect:
664<pre>
665  \p{L}
666  \pL
667</pre>
668The following general category property codes are supported:
669<pre>
670  C     Other
671  Cc    Control
672  Cf    Format
673  Cn    Unassigned
674  Co    Private use
675  Cs    Surrogate
676
677  L     Letter
678  Ll    Lower case letter
679  Lm    Modifier letter
680  Lo    Other letter
681  Lt    Title case letter
682  Lu    Upper case letter
683
684  M     Mark
685  Mc    Spacing mark
686  Me    Enclosing mark
687  Mn    Non-spacing mark
688
689  N     Number
690  Nd    Decimal number
691  Nl    Letter number
692  No    Other number
693
694  P     Punctuation
695  Pc    Connector punctuation
696  Pd    Dash punctuation
697  Pe    Close punctuation
698  Pf    Final punctuation
699  Pi    Initial punctuation
700  Po    Other punctuation
701  Ps    Open punctuation
702
703  S     Symbol
704  Sc    Currency symbol
705  Sk    Modifier symbol
706  Sm    Mathematical symbol
707  So    Other symbol
708
709  Z     Separator
710  Zl    Line separator
711  Zp    Paragraph separator
712  Zs    Space separator
713</pre>
714The special property L& is also supported: it matches a character that has
715the Lu, Ll, or Lt property, in other words, a letter that is not classified as
716a modifier or "other".
717</P>
718<P>
719The Cs (Surrogate) property applies only to characters in the range U+D800 to
720U+DFFF. Such characters are not valid in UTF-8 strings (see RFC 3629) and so
721cannot be tested by PCRE, unless UTF-8 validity checking has been turned off
722(see the discussion of PCRE_NO_UTF8_CHECK in the
723<a href="pcreapi.html"><b>pcreapi</b></a>
724page). Perl does not support the Cs property.
725</P>
726<P>
727The long synonyms for property names that Perl supports (such as \p{Letter})
728are not supported by PCRE, nor is it permitted to prefix any of these
729properties with "Is".
730</P>
731<P>
732No character that is in the Unicode table has the Cn (unassigned) property.
733Instead, this property is assumed for any code point that is not in the
734Unicode table.
735</P>
736<P>
737Specifying caseless matching does not affect these escape sequences. For
738example, \p{Lu} always matches only upper case letters.
739</P>
740<P>
741The \X escape matches any number of Unicode characters that form an extended
742Unicode sequence. \X is equivalent to
743<pre>
744  (?&#62;\PM\pM*)
745</pre>
746That is, it matches a character without the "mark" property, followed by zero
747or more characters with the "mark" property, and treats the sequence as an
748atomic group
749<a href="#atomicgroup">(see below).</a>
750Characters with the "mark" property are typically accents that affect the
751preceding character. None of them have codepoints less than 256, so in
752non-UTF-8 mode \X matches any one character.
753</P>
754<P>
755Matching characters by Unicode property is not fast, because PCRE has to search
756a structure that contains data for over fifteen thousand characters. That is
757why the traditional escape sequences such as \d and \w do not use Unicode
758properties in PCRE by default, though you can make them do so by setting the
759PCRE_UCP option for <b>pcre_compile()</b> or by starting the pattern with
760(*UCP).
761<a name="extraprops"></a></P>
762<br><b>
763PCRE's additional properties
764</b><br>
765<P>
766As well as the standard Unicode properties described in the previous
767section, PCRE supports four more that make it possible to convert traditional
768escape sequences such as \w and \s and POSIX character classes to use Unicode
769properties. PCRE uses these non-standard, non-Perl properties internally when
770PCRE_UCP is set. They are:
771<pre>
772  Xan   Any alphanumeric character
773  Xps   Any POSIX space character
774  Xsp   Any Perl space character
775  Xwd   Any Perl "word" character
776</pre>
777Xan matches characters that have either the L (letter) or the N (number)
778property. Xps matches the characters tab, linefeed, vertical tab, formfeed, or
779carriage return, and any other character that has the Z (separator) property.
780Xsp is the same as Xps, except that vertical tab is excluded. Xwd matches the
781same characters as Xan, plus underscore.
782<a name="resetmatchstart"></a></P>
783<br><b>
784Resetting the match start
785</b><br>
786<P>
787The escape sequence \K causes any previously matched characters not to be
788included in the final matched sequence. For example, the pattern:
789<pre>
790  foo\Kbar
791</pre>
792matches "foobar", but reports that it has matched "bar". This feature is
793similar to a lookbehind assertion
794<a href="#lookbehind">(described below).</a>
795However, in this case, the part of the subject before the real match does not
796have to be of fixed length, as lookbehind assertions do. The use of \K does
797not interfere with the setting of
798<a href="#subpattern">captured substrings.</a>
799For example, when the pattern
800<pre>
801  (foo)\Kbar
802</pre>
803matches "foobar", the first substring is still set to "foo".
804</P>
805<P>
806Perl documents that the use of \K within assertions is "not well defined". In
807PCRE, \K is acted upon when it occurs inside positive assertions, but is
808ignored in negative assertions.
809<a name="smallassertions"></a></P>
810<br><b>
811Simple assertions
812</b><br>
813<P>
814The final use of backslash is for certain simple assertions. An assertion
815specifies a condition that has to be met at a particular point in a match,
816without consuming any characters from the subject string. The use of
817subpatterns for more complicated assertions is described
818<a href="#bigassertions">below.</a>
819The backslashed assertions are:
820<pre>
821  \b     matches at a word boundary
822  \B     matches when not at a word boundary
823  \A     matches at the start of the subject
824  \Z     matches at the end of the subject
825          also matches before a newline at the end of the subject
826  \z     matches only at the end of the subject
827  \G     matches at the first matching position in the subject
828</pre>
829Inside a character class, \b has a different meaning; it matches the backspace
830character. If any other of these assertions appears in a character class, by
831default it matches the corresponding literal character (for example, \B
832matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
833escape sequence" error is generated instead.
834</P>
835<P>
836A word boundary is a position in the subject string where the current character
837and the previous character do not both match \w or \W (i.e. one matches
838\w and the other matches \W), or the start or end of the string if the
839first or last character matches \w, respectively. In UTF-8 mode, the meanings
840of \w and \W can be changed by setting the PCRE_UCP option. When this is
841done, it also affects \b and \B. Neither PCRE nor Perl has a separate "start
842of word" or "end of word" metasequence. However, whatever follows \b normally
843determines which it is. For example, the fragment \ba matches "a" at the start
844of a word.
845</P>
846<P>
847The \A, \Z, and \z assertions differ from the traditional circumflex and
848dollar (described in the next section) in that they only ever match at the very
849start and end of the subject string, whatever options are set. Thus, they are
850independent of multiline mode. These three assertions are not affected by the
851PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
852circumflex and dollar metacharacters. However, if the <i>startoffset</i>
853argument of <b>pcre_exec()</b> is non-zero, indicating that matching is to start
854at a point other than the beginning of the subject, \A can never match. The
855difference between \Z and \z is that \Z matches before a newline at the end
856of the string as well as at the very end, whereas \z matches only at the end.
857</P>
858<P>
859The \G assertion is true only when the current matching position is at the
860start point of the match, as specified by the <i>startoffset</i> argument of
861<b>pcre_exec()</b>. It differs from \A when the value of <i>startoffset</i> is
862non-zero. By calling <b>pcre_exec()</b> multiple times with appropriate
863arguments, you can mimic Perl's /g option, and it is in this kind of
864implementation where \G can be useful.
865</P>
866<P>
867Note, however, that PCRE's interpretation of \G, as the start of the current
868match, is subtly different from Perl's, which defines it as the end of the
869previous match. In Perl, these can be different when the previously matched
870string was empty. Because PCRE does just one match at a time, it cannot
871reproduce this behaviour.
872</P>
873<P>
874If all the alternatives of a pattern begin with \G, the expression is anchored
875to the starting match position, and the "anchored" flag is set in the compiled
876regular expression.
877</P>
878<br><a name="SEC5" href="#TOC1">CIRCUMFLEX AND DOLLAR</a><br>
879<P>
880Outside a character class, in the default matching mode, the circumflex
881character is an assertion that is true only if the current matching point is
882at the start of the subject string. If the <i>startoffset</i> argument of
883<b>pcre_exec()</b> is non-zero, circumflex can never match if the PCRE_MULTILINE
884option is unset. Inside a character class, circumflex has an entirely different
885meaning
886<a href="#characterclass">(see below).</a>
887</P>
888<P>
889Circumflex need not be the first character of the pattern if a number of
890alternatives are involved, but it should be the first thing in each alternative
891in which it appears if the pattern is ever to match that branch. If all
892possible alternatives start with a circumflex, that is, if the pattern is
893constrained to match only at the start of the subject, it is said to be an
894"anchored" pattern. (There are also other constructs that can cause a pattern
895to be anchored.)
896</P>
897<P>
898A dollar character is an assertion that is true only if the current matching
899point is at the end of the subject string, or immediately before a newline
900at the end of the string (by default). Dollar need not be the last character of
901the pattern if a number of alternatives are involved, but it should be the last
902item in any branch in which it appears. Dollar has no special meaning in a
903character class.
904</P>
905<P>
906The meaning of dollar can be changed so that it matches only at the very end of
907the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
908does not affect the \Z assertion.
909</P>
910<P>
911The meanings of the circumflex and dollar characters are changed if the
912PCRE_MULTILINE option is set. When this is the case, a circumflex matches
913immediately after internal newlines as well as at the start of the subject
914string. It does not match after a newline that ends the string. A dollar
915matches before any newlines in the string, as well as at the very end, when
916PCRE_MULTILINE is set. When newline is specified as the two-character
917sequence CRLF, isolated CR and LF characters do not indicate newlines.
918</P>
919<P>
920For example, the pattern /^abc$/ matches the subject string "def\nabc" (where
921\n represents a newline) in multiline mode, but not otherwise. Consequently,
922patterns that are anchored in single line mode because all branches start with
923^ are not anchored in multiline mode, and a match for circumflex is possible
924when the <i>startoffset</i> argument of <b>pcre_exec()</b> is non-zero. The
925PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
926</P>
927<P>
928Note that the sequences \A, \Z, and \z can be used to match the start and
929end of the subject in both modes, and if all branches of a pattern start with
930\A it is always anchored, whether or not PCRE_MULTILINE is set.
931<a name="fullstopdot"></a></P>
932<br><a name="SEC6" href="#TOC1">FULL STOP (PERIOD, DOT) AND \N</a><br>
933<P>
934Outside a character class, a dot in the pattern matches any one character in
935the subject string except (by default) a character that signifies the end of a
936line. In UTF-8 mode, the matched character may be more than one byte long.
937</P>
938<P>
939When a line ending is defined as a single character, dot never matches that
940character; when the two-character sequence CRLF is used, dot does not match CR
941if it is immediately followed by LF, but otherwise it matches all characters
942(including isolated CRs and LFs). When any Unicode line endings are being
943recognized, dot does not match CR or LF or any of the other line ending
944characters.
945</P>
946<P>
947The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
948option is set, a dot matches any one character, without exception. If the
949two-character sequence CRLF is present in the subject string, it takes two dots
950to match it.
951</P>
952<P>
953The handling of dot is entirely independent of the handling of circumflex and
954dollar, the only relationship being that they both involve newlines. Dot has no
955special meaning in a character class.
956</P>
957<P>
958The escape sequence \N behaves like a dot, except that it is not affected by
959the PCRE_DOTALL option. In other words, it matches any character except one
960that signifies the end of a line.
961</P>
962<br><a name="SEC7" href="#TOC1">MATCHING A SINGLE BYTE</a><br>
963<P>
964Outside a character class, the escape sequence \C matches any one byte, both
965in and out of UTF-8 mode. Unlike a dot, it always matches any line-ending
966characters. The feature is provided in Perl in order to match individual bytes
967in UTF-8 mode. Because it breaks up UTF-8 characters into individual bytes, the
968rest of the string may start with a malformed UTF-8 character. For this reason,
969the \C escape sequence is best avoided.
970</P>
971<P>
972PCRE does not allow \C to appear in lookbehind assertions
973<a href="#lookbehind">(described below),</a>
974because in UTF-8 mode this would make it impossible to calculate the length of
975the lookbehind.
976<a name="characterclass"></a></P>
977<br><a name="SEC8" href="#TOC1">SQUARE BRACKETS AND CHARACTER CLASSES</a><br>
978<P>
979An opening square bracket introduces a character class, terminated by a closing
980square bracket. A closing square bracket on its own is not special by default.
981However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
982bracket causes a compile-time error. If a closing square bracket is required as
983a member of the class, it should be the first data character in the class
984(after an initial circumflex, if present) or escaped with a backslash.
985</P>
986<P>
987A character class matches a single character in the subject. In UTF-8 mode, the
988character may be more than one byte long. A matched character must be in the
989set of characters defined by the class, unless the first character in the class
990definition is a circumflex, in which case the subject character must not be in
991the set defined by the class. If a circumflex is actually required as a member
992of the class, ensure it is not the first character, or escape it with a
993backslash.
994</P>
995<P>
996For example, the character class [aeiou] matches any lower case vowel, while
997[^aeiou] matches any character that is not a lower case vowel. Note that a
998circumflex is just a convenient notation for specifying the characters that
999are in the class by enumerating those that are not. A class that starts with a
1000circumflex is not an assertion; it still consumes a character from the subject
1001string, and therefore it fails if the current pointer is at the end of the
1002string.
1003</P>
1004<P>
1005In UTF-8 mode, characters with values greater than 255 can be included in a
1006class as a literal string of bytes, or by using the \x{ escaping mechanism.
1007</P>
1008<P>
1009When caseless matching is set, any letters in a class represent both their
1010upper case and lower case versions, so for example, a caseless [aeiou] matches
1011"A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1012caseful version would. In UTF-8 mode, PCRE always understands the concept of
1013case for characters whose values are less than 128, so caseless matching is
1014always possible. For characters with higher values, the concept of case is
1015supported if PCRE is compiled with Unicode property support, but not otherwise.
1016If you want to use caseless matching in UTF8-mode for characters 128 and above,
1017you must ensure that PCRE is compiled with Unicode property support as well as
1018with UTF-8 support.
1019</P>
1020<P>
1021Characters that might indicate line breaks are never treated in any special way
1022when matching character classes, whatever line-ending sequence is in use, and
1023whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1024such as [^a] always matches one of these characters.
1025</P>
1026<P>
1027The minus (hyphen) character can be used to specify a range of characters in a
1028character class. For example, [d-m] matches any letter between d and m,
1029inclusive. If a minus character is required in a class, it must be escaped with
1030a backslash or appear in a position where it cannot be interpreted as
1031indicating a range, typically as the first or last character in the class.
1032</P>
1033<P>
1034It is not possible to have the literal character "]" as the end character of a
1035range. A pattern such as [W-]46] is interpreted as a class of two characters
1036("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1037"-46]". However, if the "]" is escaped with a backslash it is interpreted as
1038the end of range, so [W-\]46] is interpreted as a class containing a range
1039followed by two other characters. The octal or hexadecimal representation of
1040"]" can also be used to end a range.
1041</P>
1042<P>
1043Ranges operate in the collating sequence of character values. They can also be
1044used for characters specified numerically, for example [\000-\037]. In UTF-8
1045mode, ranges can include characters whose values are greater than 255, for
1046example [\x{100}-\x{2ff}].
1047</P>
1048<P>
1049If a range that includes letters is used when caseless matching is set, it
1050matches the letters in either case. For example, [W-c] is equivalent to
1051[][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
1052tables for a French locale are in use, [\xc8-\xcb] matches accented E
1053characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
1054characters with values greater than 128 only when it is compiled with Unicode
1055property support.
1056</P>
1057<P>
1058The character escape sequences \d, \D, \h, \H, \p, \P, \s, \S, \v,
1059\V, \w, and \W may appear in a character class, and add the characters that
1060they match to the class. For example, [\dABCDEF] matches any hexadecimal
1061digit. In UTF-8 mode, the PCRE_UCP option affects the meanings of \d, \s, \w
1062and their upper case partners, just as it does when they appear outside a
1063character class, as described in the section entitled
1064<a href="#genericchartypes">"Generic character types"</a>
1065above. The escape sequence \b has a different meaning inside a character
1066class; it matches the backspace character. The sequences \B, \N, \R, and \X
1067are not special inside a character class. Like any other unrecognized escape
1068sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1069default, but cause an error if the PCRE_EXTRA option is set.
1070</P>
1071<P>
1072A circumflex can conveniently be used with the upper case character types to
1073specify a more restricted set of characters than the matching lower case type.
1074For example, the class [^\W_] matches any letter or digit, but not underscore,
1075whereas [\w] includes underscore. A positive character class should be read as
1076"something OR something OR ..." and a negative class as "NOT something AND NOT
1077something AND NOT ...".
1078</P>
1079<P>
1080The only metacharacters that are recognized in character classes are backslash,
1081hyphen (only where it can be interpreted as specifying a range), circumflex
1082(only at the start), opening square bracket (only when it can be interpreted as
1083introducing a POSIX class name - see the next section), and the terminating
1084closing square bracket. However, escaping other non-alphanumeric characters
1085does no harm.
1086</P>
1087<br><a name="SEC9" href="#TOC1">POSIX CHARACTER CLASSES</a><br>
1088<P>
1089Perl supports the POSIX notation for character classes. This uses names
1090enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1091this notation. For example,
1092<pre>
1093  [01[:alpha:]%]
1094</pre>
1095matches "0", "1", any alphabetic character, or "%". The supported class names
1096are:
1097<pre>
1098  alnum    letters and digits
1099  alpha    letters
1100  ascii    character codes 0 - 127
1101  blank    space or tab only
1102  cntrl    control characters
1103  digit    decimal digits (same as \d)
1104  graph    printing characters, excluding space
1105  lower    lower case letters
1106  print    printing characters, including space
1107  punct    printing characters, excluding letters and digits and space
1108  space    white space (not quite the same as \s)
1109  upper    upper case letters
1110  word     "word" characters (same as \w)
1111  xdigit   hexadecimal digits
1112</pre>
1113The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
1114space (32). Notice that this list includes the VT character (code 11). This
1115makes "space" different to \s, which does not include VT (for Perl
1116compatibility).
1117</P>
1118<P>
1119The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
11205.8. Another Perl extension is negation, which is indicated by a ^ character
1121after the colon. For example,
1122<pre>
1123  [12[:^digit:]]
1124</pre>
1125matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1126syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1127supported, and an error is given if they are encountered.
1128</P>
1129<P>
1130By default, in UTF-8 mode, characters with values greater than 128 do not match
1131any of the POSIX character classes. However, if the PCRE_UCP option is passed
1132to <b>pcre_compile()</b>, some of the classes are changed so that Unicode
1133character properties are used. This is achieved by replacing the POSIX classes
1134by other sequences, as follows:
1135<pre>
1136  [:alnum:]  becomes  \p{Xan}
1137  [:alpha:]  becomes  \p{L}
1138  [:blank:]  becomes  \h
1139  [:digit:]  becomes  \p{Nd}
1140  [:lower:]  becomes  \p{Ll}
1141  [:space:]  becomes  \p{Xps}
1142  [:upper:]  becomes  \p{Lu}
1143  [:word:]   becomes  \p{Xwd}
1144</pre>
1145Negated versions, such as [:^alpha:] use \P instead of \p. The other POSIX
1146classes are unchanged, and match only characters with code points less than
1147128.
1148</P>
1149<br><a name="SEC10" href="#TOC1">VERTICAL BAR</a><br>
1150<P>
1151Vertical bar characters are used to separate alternative patterns. For example,
1152the pattern
1153<pre>
1154  gilbert|sullivan
1155</pre>
1156matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1157and an empty alternative is permitted (matching the empty string). The matching
1158process tries each alternative in turn, from left to right, and the first one
1159that succeeds is used. If the alternatives are within a subpattern
1160<a href="#subpattern">(defined below),</a>
1161"succeeds" means matching the rest of the main pattern as well as the
1162alternative in the subpattern.
1163</P>
1164<br><a name="SEC11" href="#TOC1">INTERNAL OPTION SETTING</a><br>
1165<P>
1166The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1167PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1168the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1169The option letters are
1170<pre>
1171  i  for PCRE_CASELESS
1172  m  for PCRE_MULTILINE
1173  s  for PCRE_DOTALL
1174  x  for PCRE_EXTENDED
1175</pre>
1176For example, (?im) sets caseless, multiline matching. It is also possible to
1177unset these options by preceding the letter with a hyphen, and a combined
1178setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1179PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1180permitted. If a letter appears both before and after the hyphen, the option is
1181unset.
1182</P>
1183<P>
1184The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1185changed in the same way as the Perl-compatible options by using the characters
1186J, U and X respectively.
1187</P>
1188<P>
1189When one of these option changes occurs at top level (that is, not inside
1190subpattern parentheses), the change applies to the remainder of the pattern
1191that follows. If the change is placed right at the start of a pattern, PCRE
1192extracts it into the global options (and it will therefore show up in data
1193extracted by the <b>pcre_fullinfo()</b> function).
1194</P>
1195<P>
1196An option change within a subpattern (see below for a description of
1197subpatterns) affects only that part of the subpattern that follows it, so
1198<pre>
1199  (a(?i)b)c
1200</pre>
1201matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1202By this means, options can be made to have different settings in different
1203parts of the pattern. Any changes made in one alternative do carry on
1204into subsequent branches within the same subpattern. For example,
1205<pre>
1206  (a(?i)b|c)
1207</pre>
1208matches "ab", "aB", "c", and "C", even though when matching "C" the first
1209branch is abandoned before the option setting. This is because the effects of
1210option settings happen at compile time. There would be some very weird
1211behaviour otherwise.
1212</P>
1213<P>
1214<b>Note:</b> There are other PCRE-specific options that can be set by the
1215application when the compile or match functions are called. In some cases the
1216pattern can contain special leading sequences such as (*CRLF) to override what
1217the application has set or what has been defaulted. Details are given in the
1218section entitled
1219<a href="#newlineseq">"Newline sequences"</a>
1220above. There are also the (*UTF8) and (*UCP) leading sequences that can be used
1221to set UTF-8 and Unicode property modes; they are equivalent to setting the
1222PCRE_UTF8 and the PCRE_UCP options, respectively.
1223<a name="subpattern"></a></P>
1224<br><a name="SEC12" href="#TOC1">SUBPATTERNS</a><br>
1225<P>
1226Subpatterns are delimited by parentheses (round brackets), which can be nested.
1227Turning part of a pattern into a subpattern does two things:
1228<br>
1229<br>
12301. It localizes a set of alternatives. For example, the pattern
1231<pre>
1232  cat(aract|erpillar|)
1233</pre>
1234matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1235match "cataract", "erpillar" or an empty string.
1236<br>
1237<br>
12382. It sets up the subpattern as a capturing subpattern. This means that, when
1239the whole pattern matches, that portion of the subject string that matched the
1240subpattern is passed back to the caller via the <i>ovector</i> argument of
1241<b>pcre_exec()</b>. Opening parentheses are counted from left to right (starting
1242from 1) to obtain numbers for the capturing subpatterns. For example, if the
1243string "the red king" is matched against the pattern
1244<pre>
1245  the ((red|white) (king|queen))
1246</pre>
1247the captured substrings are "red king", "red", and "king", and are numbered 1,
12482, and 3, respectively.
1249</P>
1250<P>
1251The fact that plain parentheses fulfil two functions is not always helpful.
1252There are often times when a grouping subpattern is required without a
1253capturing requirement. If an opening parenthesis is followed by a question mark
1254and a colon, the subpattern does not do any capturing, and is not counted when
1255computing the number of any subsequent capturing subpatterns. For example, if
1256the string "the white queen" is matched against the pattern
1257<pre>
1258  the ((?:red|white) (king|queen))
1259</pre>
1260the captured substrings are "white queen" and "queen", and are numbered 1 and
12612. The maximum number of capturing subpatterns is 65535.
1262</P>
1263<P>
1264As a convenient shorthand, if any option settings are required at the start of
1265a non-capturing subpattern, the option letters may appear between the "?" and
1266the ":". Thus the two patterns
1267<pre>
1268  (?i:saturday|sunday)
1269  (?:(?i)saturday|sunday)
1270</pre>
1271match exactly the same set of strings. Because alternative branches are tried
1272from left to right, and options are not reset until the end of the subpattern
1273is reached, an option setting in one branch does affect subsequent branches, so
1274the above patterns match "SUNDAY" as well as "Saturday".
1275<a name="dupsubpatternnumber"></a></P>
1276<br><a name="SEC13" href="#TOC1">DUPLICATE SUBPATTERN NUMBERS</a><br>
1277<P>
1278Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1279the same numbers for its capturing parentheses. Such a subpattern starts with
1280(?| and is itself a non-capturing subpattern. For example, consider this
1281pattern:
1282<pre>
1283  (?|(Sat)ur|(Sun))day
1284</pre>
1285Because the two alternatives are inside a (?| group, both sets of capturing
1286parentheses are numbered one. Thus, when the pattern matches, you can look
1287at captured substring number one, whichever alternative matched. This construct
1288is useful when you want to capture part, but not all, of one of a number of
1289alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1290number is reset at the start of each branch. The numbers of any capturing
1291parentheses that follow the subpattern start after the highest number used in
1292any branch. The following example is taken from the Perl documentation. The
1293numbers underneath show in which buffer the captured content will be stored.
1294<pre>
1295  # before  ---------------branch-reset----------- after
1296  / ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1297  # 1            2         2  3        2     3     4
1298</pre>
1299A back reference to a numbered subpattern uses the most recent value that is
1300set for that number by any subpattern. The following pattern matches "abcabc"
1301or "defdef":
1302<pre>
1303  /(?|(abc)|(def))\1/
1304</pre>
1305In contrast, a recursive or "subroutine" call to a numbered subpattern always
1306refers to the first one in the pattern with the given number. The following
1307pattern matches "abcabc" or "defabc":
1308<pre>
1309  /(?|(abc)|(def))(?1)/
1310</pre>
1311If a
1312<a href="#conditions">condition test</a>
1313for a subpattern's having matched refers to a non-unique number, the test is
1314true if any of the subpatterns of that number have matched.
1315</P>
1316<P>
1317An alternative approach to using this "branch reset" feature is to use
1318duplicate named subpatterns, as described in the next section.
1319</P>
1320<br><a name="SEC14" href="#TOC1">NAMED SUBPATTERNS</a><br>
1321<P>
1322Identifying capturing parentheses by number is simple, but it can be very hard
1323to keep track of the numbers in complicated regular expressions. Furthermore,
1324if an expression is modified, the numbers may change. To help with this
1325difficulty, PCRE supports the naming of subpatterns. This feature was not
1326added to Perl until release 5.10. Python had the feature earlier, and PCRE
1327introduced it at release 4.0, using the Python syntax. PCRE now supports both
1328the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1329have different names, but PCRE does not.
1330</P>
1331<P>
1332In PCRE, a subpattern can be named in one of three ways: (?&#60;name&#62;...) or
1333(?'name'...) as in Perl, or (?P&#60;name&#62;...) as in Python. References to capturing
1334parentheses from other parts of the pattern, such as
1335<a href="#backreferences">back references,</a>
1336<a href="#recursion">recursion,</a>
1337and
1338<a href="#conditions">conditions,</a>
1339can be made by name as well as by number.
1340</P>
1341<P>
1342Names consist of up to 32 alphanumeric characters and underscores. Named
1343capturing parentheses are still allocated numbers as well as names, exactly as
1344if the names were not present. The PCRE API provides function calls for
1345extracting the name-to-number translation table from a compiled pattern. There
1346is also a convenience function for extracting a captured substring by name.
1347</P>
1348<P>
1349By default, a name must be unique within a pattern, but it is possible to relax
1350this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1351names are also always permitted for subpatterns with the same number, set up as
1352described in the previous section.) Duplicate names can be useful for patterns
1353where only one instance of the named parentheses can match. Suppose you want to
1354match the name of a weekday, either as a 3-letter abbreviation or as the full
1355name, and in both cases you want to extract the abbreviation. This pattern
1356(ignoring the line breaks) does the job:
1357<pre>
1358  (?&#60;DN&#62;Mon|Fri|Sun)(?:day)?|
1359  (?&#60;DN&#62;Tue)(?:sday)?|
1360  (?&#60;DN&#62;Wed)(?:nesday)?|
1361  (?&#60;DN&#62;Thu)(?:rsday)?|
1362  (?&#60;DN&#62;Sat)(?:urday)?
1363</pre>
1364There are five capturing substrings, but only one is ever set after a match.
1365(An alternative way of solving this problem is to use a "branch reset"
1366subpattern, as described in the previous section.)
1367</P>
1368<P>
1369The convenience function for extracting the data by name returns the substring
1370for the first (and in this example, the only) subpattern of that name that
1371matched. This saves searching to find which numbered subpattern it was.
1372</P>
1373<P>
1374If you make a back reference to a non-unique named subpattern from elsewhere in
1375the pattern, the one that corresponds to the first occurrence of the name is
1376used. In the absence of duplicate numbers (see the previous section) this is
1377the one with the lowest number. If you use a named reference in a condition
1378test (see the
1379<a href="#conditions">section about conditions</a>
1380below), either to check whether a subpattern has matched, or to check for
1381recursion, all subpatterns with the same name are tested. If the condition is
1382true for any one of them, the overall condition is true. This is the same
1383behaviour as testing by number. For further details of the interfaces for
1384handling named subpatterns, see the
1385<a href="pcreapi.html"><b>pcreapi</b></a>
1386documentation.
1387</P>
1388<P>
1389<b>Warning:</b> You cannot use different names to distinguish between two
1390subpatterns with the same number because PCRE uses only the numbers when
1391matching. For this reason, an error is given at compile time if different names
1392are given to subpatterns with the same number. However, you can give the same
1393name to subpatterns with the same number, even when PCRE_DUPNAMES is not set.
1394</P>
1395<br><a name="SEC15" href="#TOC1">REPETITION</a><br>
1396<P>
1397Repetition is specified by quantifiers, which can follow any of the following
1398items:
1399<pre>
1400  a literal data character
1401  the dot metacharacter
1402  the \C escape sequence
1403  the \X escape sequence (in UTF-8 mode with Unicode properties)
1404  the \R escape sequence
1405  an escape such as \d or \pL that matches a single character
1406  a character class
1407  a back reference (see next section)
1408  a parenthesized subpattern (unless it is an assertion)
1409  a recursive or "subroutine" call to a subpattern
1410</pre>
1411The general repetition quantifier specifies a minimum and maximum number of
1412permitted matches, by giving the two numbers in curly brackets (braces),
1413separated by a comma. The numbers must be less than 65536, and the first must
1414be less than or equal to the second. For example:
1415<pre>
1416  z{2,4}
1417</pre>
1418matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1419character. If the second number is omitted, but the comma is present, there is
1420no upper limit; if the second number and the comma are both omitted, the
1421quantifier specifies an exact number of required matches. Thus
1422<pre>
1423  [aeiou]{3,}
1424</pre>
1425matches at least 3 successive vowels, but may match many more, while
1426<pre>
1427  \d{8}
1428</pre>
1429matches exactly 8 digits. An opening curly bracket that appears in a position
1430where a quantifier is not allowed, or one that does not match the syntax of a
1431quantifier, is taken as a literal character. For example, {,6} is not a
1432quantifier, but a literal string of four characters.
1433</P>
1434<P>
1435In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
1436bytes. Thus, for example, \x{100}{2} matches two UTF-8 characters, each of
1437which is represented by a two-byte sequence. Similarly, when Unicode property
1438support is available, \X{3} matches three Unicode extended sequences, each of
1439which may be several bytes long (and they may be of different lengths).
1440</P>
1441<P>
1442The quantifier {0} is permitted, causing the expression to behave as if the
1443previous item and the quantifier were not present. This may be useful for
1444subpatterns that are referenced as
1445<a href="#subpatternsassubroutines">subroutines</a>
1446from elsewhere in the pattern (but see also the section entitled
1447<a href="#subdefine">"Defining subpatterns for use by reference only"</a>
1448below). Items other than subpatterns that have a {0} quantifier are omitted
1449from the compiled pattern.
1450</P>
1451<P>
1452For convenience, the three most common quantifiers have single-character
1453abbreviations:
1454<pre>
1455  *    is equivalent to {0,}
1456  +    is equivalent to {1,}
1457  ?    is equivalent to {0,1}
1458</pre>
1459It is possible to construct infinite loops by following a subpattern that can
1460match no characters with a quantifier that has no upper limit, for example:
1461<pre>
1462  (a?)*
1463</pre>
1464Earlier versions of Perl and PCRE used to give an error at compile time for
1465such patterns. However, because there are cases where this can be useful, such
1466patterns are now accepted, but if any repetition of the subpattern does in fact
1467match no characters, the loop is forcibly broken.
1468</P>
1469<P>
1470By default, the quantifiers are "greedy", that is, they match as much as
1471possible (up to the maximum number of permitted times), without causing the
1472rest of the pattern to fail. The classic example of where this gives problems
1473is in trying to match comments in C programs. These appear between /* and */
1474and within the comment, individual * and / characters may appear. An attempt to
1475match C comments by applying the pattern
1476<pre>
1477  /\*.*\*/
1478</pre>
1479to the string
1480<pre>
1481  /* first comment */  not comment  /* second comment */
1482</pre>
1483fails, because it matches the entire string owing to the greediness of the .*
1484item.
1485</P>
1486<P>
1487However, if a quantifier is followed by a question mark, it ceases to be
1488greedy, and instead matches the minimum number of times possible, so the
1489pattern
1490<pre>
1491  /\*.*?\*/
1492</pre>
1493does the right thing with the C comments. The meaning of the various
1494quantifiers is not otherwise changed, just the preferred number of matches.
1495Do not confuse this use of question mark with its use as a quantifier in its
1496own right. Because it has two uses, it can sometimes appear doubled, as in
1497<pre>
1498  \d??\d
1499</pre>
1500which matches one digit by preference, but can match two if that is the only
1501way the rest of the pattern matches.
1502</P>
1503<P>
1504If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1505the quantifiers are not greedy by default, but individual ones can be made
1506greedy by following them with a question mark. In other words, it inverts the
1507default behaviour.
1508</P>
1509<P>
1510When a parenthesized subpattern is quantified with a minimum repeat count that
1511is greater than 1 or with a limited maximum, more memory is required for the
1512compiled pattern, in proportion to the size of the minimum or maximum.
1513</P>
1514<P>
1515If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1516to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1517implicitly anchored, because whatever follows will be tried against every
1518character position in the subject string, so there is no point in retrying the
1519overall match at any position after the first. PCRE normally treats such a
1520pattern as though it were preceded by \A.
1521</P>
1522<P>
1523In cases where it is known that the subject string contains no newlines, it is
1524worth setting PCRE_DOTALL in order to obtain this optimization, or
1525alternatively using ^ to indicate anchoring explicitly.
1526</P>
1527<P>
1528However, there is one situation where the optimization cannot be used. When .*
1529is inside capturing parentheses that are the subject of a back reference
1530elsewhere in the pattern, a match at the start may fail where a later one
1531succeeds. Consider, for example:
1532<pre>
1533  (.*)abc\1
1534</pre>
1535If the subject is "xyz123abc123" the match point is the fourth character. For
1536this reason, such a pattern is not implicitly anchored.
1537</P>
1538<P>
1539When a capturing subpattern is repeated, the value captured is the substring
1540that matched the final iteration. For example, after
1541<pre>
1542  (tweedle[dume]{3}\s*)+
1543</pre>
1544has matched "tweedledum tweedledee" the value of the captured substring is
1545"tweedledee". However, if there are nested capturing subpatterns, the
1546corresponding captured values may have been set in previous iterations. For
1547example, after
1548<pre>
1549  /(a|(b))+/
1550</pre>
1551matches "aba" the value of the second captured substring is "b".
1552<a name="atomicgroup"></a></P>
1553<br><a name="SEC16" href="#TOC1">ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS</a><br>
1554<P>
1555With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1556repetition, failure of what follows normally causes the repeated item to be
1557re-evaluated to see if a different number of repeats allows the rest of the
1558pattern to match. Sometimes it is useful to prevent this, either to change the
1559nature of the match, or to cause it fail earlier than it otherwise might, when
1560the author of the pattern knows there is no point in carrying on.
1561</P>
1562<P>
1563Consider, for example, the pattern \d+foo when applied to the subject line
1564<pre>
1565  123456bar
1566</pre>
1567After matching all 6 digits and then failing to match "foo", the normal
1568action of the matcher is to try again with only 5 digits matching the \d+
1569item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1570(a term taken from Jeffrey Friedl's book) provides the means for specifying
1571that once a subpattern has matched, it is not to be re-evaluated in this way.
1572</P>
1573<P>
1574If we use atomic grouping for the previous example, the matcher gives up
1575immediately on failing to match "foo" the first time. The notation is a kind of
1576special parenthesis, starting with (?&#62; as in this example:
1577<pre>
1578  (?&#62;\d+)foo
1579</pre>
1580This kind of parenthesis "locks up" the  part of the pattern it contains once
1581it has matched, and a failure further into the pattern is prevented from
1582backtracking into it. Backtracking past it to previous items, however, works as
1583normal.
1584</P>
1585<P>
1586An alternative description is that a subpattern of this type matches the string
1587of characters that an identical standalone pattern would match, if anchored at
1588the current point in the subject string.
1589</P>
1590<P>
1591Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1592the above example can be thought of as a maximizing repeat that must swallow
1593everything it can. So, while both \d+ and \d+? are prepared to adjust the
1594number of digits they match in order to make the rest of the pattern match,
1595(?&#62;\d+) can only match an entire sequence of digits.
1596</P>
1597<P>
1598Atomic groups in general can of course contain arbitrarily complicated
1599subpatterns, and can be nested. However, when the subpattern for an atomic
1600group is just a single repeated item, as in the example above, a simpler
1601notation, called a "possessive quantifier" can be used. This consists of an
1602additional + character following a quantifier. Using this notation, the
1603previous example can be rewritten as
1604<pre>
1605  \d++foo
1606</pre>
1607Note that a possessive quantifier can be used with an entire group, for
1608example:
1609<pre>
1610  (abc|xyz){2,3}+
1611</pre>
1612Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1613option is ignored. They are a convenient notation for the simpler forms of
1614atomic group. However, there is no difference in the meaning of a possessive
1615quantifier and the equivalent atomic group, though there may be a performance
1616difference; possessive quantifiers should be slightly faster.
1617</P>
1618<P>
1619The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1620Jeffrey Friedl originated the idea (and the name) in the first edition of his
1621book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1622package, and PCRE copied it from there. It ultimately found its way into Perl
1623at release 5.10.
1624</P>
1625<P>
1626PCRE has an optimization that automatically "possessifies" certain simple
1627pattern constructs. For example, the sequence A+B is treated as A++B because
1628there is no point in backtracking into a sequence of A's when B must follow.
1629</P>
1630<P>
1631When a pattern contains an unlimited repeat inside a subpattern that can itself
1632be repeated an unlimited number of times, the use of an atomic group is the
1633only way to avoid some failing matches taking a very long time indeed. The
1634pattern
1635<pre>
1636  (\D+|&#60;\d+&#62;)*[!?]
1637</pre>
1638matches an unlimited number of substrings that either consist of non-digits, or
1639digits enclosed in &#60;&#62;, followed by either ! or ?. When it matches, it runs
1640quickly. However, if it is applied to
1641<pre>
1642  aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1643</pre>
1644it takes a long time before reporting failure. This is because the string can
1645be divided between the internal \D+ repeat and the external * repeat in a
1646large number of ways, and all have to be tried. (The example uses [!?] rather
1647than a single character at the end, because both PCRE and Perl have an
1648optimization that allows for fast failure when a single character is used. They
1649remember the last single character that is required for a match, and fail early
1650if it is not present in the string.) If the pattern is changed so that it uses
1651an atomic group, like this:
1652<pre>
1653  ((?&#62;\D+)|&#60;\d+&#62;)*[!?]
1654</pre>
1655sequences of non-digits cannot be broken, and failure happens quickly.
1656<a name="backreferences"></a></P>
1657<br><a name="SEC17" href="#TOC1">BACK REFERENCES</a><br>
1658<P>
1659Outside a character class, a backslash followed by a digit greater than 0 (and
1660possibly further digits) is a back reference to a capturing subpattern earlier
1661(that is, to its left) in the pattern, provided there have been that many
1662previous capturing left parentheses.
1663</P>
1664<P>
1665However, if the decimal number following the backslash is less than 10, it is
1666always taken as a back reference, and causes an error only if there are not
1667that many capturing left parentheses in the entire pattern. In other words, the
1668parentheses that are referenced need not be to the left of the reference for
1669numbers less than 10. A "forward back reference" of this type can make sense
1670when a repetition is involved and the subpattern to the right has participated
1671in an earlier iteration.
1672</P>
1673<P>
1674It is not possible to have a numerical "forward back reference" to a subpattern
1675whose number is 10 or more using this syntax because a sequence such as \50 is
1676interpreted as a character defined in octal. See the subsection entitled
1677"Non-printing characters"
1678<a href="#digitsafterbackslash">above</a>
1679for further details of the handling of digits following a backslash. There is
1680no such problem when named parentheses are used. A back reference to any
1681subpattern is possible using named parentheses (see below).
1682</P>
1683<P>
1684Another way of avoiding the ambiguity inherent in the use of digits following a
1685backslash is to use the \g escape sequence. This escape must be followed by an
1686unsigned number or a negative number, optionally enclosed in braces. These
1687examples are all identical:
1688<pre>
1689  (ring), \1
1690  (ring), \g1
1691  (ring), \g{1}
1692</pre>
1693An unsigned number specifies an absolute reference without the ambiguity that
1694is present in the older syntax. It is also useful when literal digits follow
1695the reference. A negative number is a relative reference. Consider this
1696example:
1697<pre>
1698  (abc(def)ghi)\g{-1}
1699</pre>
1700The sequence \g{-1} is a reference to the most recently started capturing
1701subpattern before \g, that is, is it equivalent to \2 in this example.
1702Similarly, \g{-2} would be equivalent to \1. The use of relative references
1703can be helpful in long patterns, and also in patterns that are created by
1704joining together fragments that contain references within themselves.
1705</P>
1706<P>
1707A back reference matches whatever actually matched the capturing subpattern in
1708the current subject string, rather than anything matching the subpattern
1709itself (see
1710<a href="#subpatternsassubroutines">"Subpatterns as subroutines"</a>
1711below for a way of doing that). So the pattern
1712<pre>
1713  (sens|respons)e and \1ibility
1714</pre>
1715matches "sense and sensibility" and "response and responsibility", but not
1716"sense and responsibility". If caseful matching is in force at the time of the
1717back reference, the case of letters is relevant. For example,
1718<pre>
1719  ((?i)rah)\s+\1
1720</pre>
1721matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1722capturing subpattern is matched caselessly.
1723</P>
1724<P>
1725There are several different ways of writing back references to named
1726subpatterns. The .NET syntax \k{name} and the Perl syntax \k&#60;name&#62; or
1727\k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1728back reference syntax, in which \g can be used for both numeric and named
1729references, is also supported. We could rewrite the above example in any of
1730the following ways:
1731<pre>
1732  (?&#60;p1&#62;(?i)rah)\s+\k&#60;p1&#62;
1733  (?'p1'(?i)rah)\s+\k{p1}
1734  (?P&#60;p1&#62;(?i)rah)\s+(?P=p1)
1735  (?&#60;p1&#62;(?i)rah)\s+\g{p1}
1736</pre>
1737A subpattern that is referenced by name may appear in the pattern before or
1738after the reference.
1739</P>
1740<P>
1741There may be more than one back reference to the same subpattern. If a
1742subpattern has not actually been used in a particular match, any back
1743references to it always fail by default. For example, the pattern
1744<pre>
1745  (a|(bc))\2
1746</pre>
1747always fails if it starts to match "a" rather than "bc". However, if the
1748PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
1749unset value matches an empty string.
1750</P>
1751<P>
1752Because there may be many capturing parentheses in a pattern, all digits
1753following a backslash are taken as part of a potential back reference number.
1754If the pattern continues with a digit character, some delimiter must be used to
1755terminate the back reference. If the PCRE_EXTENDED option is set, this can be
1756whitespace. Otherwise, the \g{ syntax or an empty comment (see
1757<a href="#comments">"Comments"</a>
1758below) can be used.
1759</P>
1760<br><b>
1761Recursive back references
1762</b><br>
1763<P>
1764A back reference that occurs inside the parentheses to which it refers fails
1765when the subpattern is first used, so, for example, (a\1) never matches.
1766However, such references can be useful inside repeated subpatterns. For
1767example, the pattern
1768<pre>
1769  (a|b\1)+
1770</pre>
1771matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1772the subpattern, the back reference matches the character string corresponding
1773to the previous iteration. In order for this to work, the pattern must be such
1774that the first iteration does not need to match the back reference. This can be
1775done using alternation, as in the example above, or by a quantifier with a
1776minimum of zero.
1777</P>
1778<P>
1779Back references of this type cause the group that they reference to be treated
1780as an
1781<a href="#atomicgroup">atomic group.</a>
1782Once the whole group has been matched, a subsequent matching failure cannot
1783cause backtracking into the middle of the group.
1784<a name="bigassertions"></a></P>
1785<br><a name="SEC18" href="#TOC1">ASSERTIONS</a><br>
1786<P>
1787An assertion is a test on the characters following or preceding the current
1788matching point that does not actually consume any characters. The simple
1789assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are described
1790<a href="#smallassertions">above.</a>
1791</P>
1792<P>
1793More complicated assertions are coded as subpatterns. There are two kinds:
1794those that look ahead of the current position in the subject string, and those
1795that look behind it. An assertion subpattern is matched in the normal way,
1796except that it does not cause the current matching position to be changed.
1797</P>
1798<P>
1799Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1800because it makes no sense to assert the same thing several times. If any kind
1801of assertion contains capturing subpatterns within it, these are counted for
1802the purposes of numbering the capturing subpatterns in the whole pattern.
1803However, substring capturing is carried out only for positive assertions,
1804because it does not make sense for negative assertions.
1805</P>
1806<br><b>
1807Lookahead assertions
1808</b><br>
1809<P>
1810Lookahead assertions start with (?= for positive assertions and (?! for
1811negative assertions. For example,
1812<pre>
1813  \w+(?=;)
1814</pre>
1815matches a word followed by a semicolon, but does not include the semicolon in
1816the match, and
1817<pre>
1818  foo(?!bar)
1819</pre>
1820matches any occurrence of "foo" that is not followed by "bar". Note that the
1821apparently similar pattern
1822<pre>
1823  (?!foo)bar
1824</pre>
1825does not find an occurrence of "bar" that is preceded by something other than
1826"foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1827(?!foo) is always true when the next three characters are "bar". A
1828lookbehind assertion is needed to achieve the other effect.
1829</P>
1830<P>
1831If you want to force a matching failure at some point in a pattern, the most
1832convenient way to do it is with (?!) because an empty string always matches, so
1833an assertion that requires there not to be an empty string must always fail.
1834The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
1835<a name="lookbehind"></a></P>
1836<br><b>
1837Lookbehind assertions
1838</b><br>
1839<P>
1840Lookbehind assertions start with (?&#60;= for positive assertions and (?&#60;! for
1841negative assertions. For example,
1842<pre>
1843  (?&#60;!foo)bar
1844</pre>
1845does find an occurrence of "bar" that is not preceded by "foo". The contents of
1846a lookbehind assertion are restricted such that all the strings it matches must
1847have a fixed length. However, if there are several top-level alternatives, they
1848do not all have to have the same fixed length. Thus
1849<pre>
1850  (?&#60;=bullock|donkey)
1851</pre>
1852is permitted, but
1853<pre>
1854  (?&#60;!dogs?|cats?)
1855</pre>
1856causes an error at compile time. Branches that match different length strings
1857are permitted only at the top level of a lookbehind assertion. This is an
1858extension compared with Perl, which requires all branches to match the same
1859length of string. An assertion such as
1860<pre>
1861  (?&#60;=ab(c|de))
1862</pre>
1863is not permitted, because its single top-level branch can match two different
1864lengths, but it is acceptable to PCRE if rewritten to use two top-level
1865branches:
1866<pre>
1867  (?&#60;=abc|abde)
1868</pre>
1869In some cases, the escape sequence \K
1870<a href="#resetmatchstart">(see above)</a>
1871can be used instead of a lookbehind assertion to get round the fixed-length
1872restriction.
1873</P>
1874<P>
1875The implementation of lookbehind assertions is, for each alternative, to
1876temporarily move the current position back by the fixed length and then try to
1877match. If there are insufficient characters before the current position, the
1878assertion fails.
1879</P>
1880<P>
1881PCRE does not allow the \C escape (which matches a single byte in UTF-8 mode)
1882to appear in lookbehind assertions, because it makes it impossible to calculate
1883the length of the lookbehind. The \X and \R escapes, which can match
1884different numbers of bytes, are also not permitted.
1885</P>
1886<P>
1887<a href="#subpatternsassubroutines">"Subroutine"</a>
1888calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
1889as the subpattern matches a fixed-length string.
1890<a href="#recursion">Recursion,</a>
1891however, is not supported.
1892</P>
1893<P>
1894Possessive quantifiers can be used in conjunction with lookbehind assertions to
1895specify efficient matching of fixed-length strings at the end of subject
1896strings. Consider a simple pattern such as
1897<pre>
1898  abcd$
1899</pre>
1900when applied to a long string that does not match. Because matching proceeds
1901from left to right, PCRE will look for each "a" in the subject and then see if
1902what follows matches the rest of the pattern. If the pattern is specified as
1903<pre>
1904  ^.*abcd$
1905</pre>
1906the initial .* matches the entire string at first, but when this fails (because
1907there is no following "a"), it backtracks to match all but the last character,
1908then all but the last two characters, and so on. Once again the search for "a"
1909covers the entire string, from right to left, so we are no better off. However,
1910if the pattern is written as
1911<pre>
1912  ^.*+(?&#60;=abcd)
1913</pre>
1914there can be no backtracking for the .*+ item; it can match only the entire
1915string. The subsequent lookbehind assertion does a single test on the last four
1916characters. If it fails, the match fails immediately. For long strings, this
1917approach makes a significant difference to the processing time.
1918</P>
1919<br><b>
1920Using multiple assertions
1921</b><br>
1922<P>
1923Several assertions (of any sort) may occur in succession. For example,
1924<pre>
1925  (?&#60;=\d{3})(?&#60;!999)foo
1926</pre>
1927matches "foo" preceded by three digits that are not "999". Notice that each of
1928the assertions is applied independently at the same point in the subject
1929string. First there is a check that the previous three characters are all
1930digits, and then there is a check that the same three characters are not "999".
1931This pattern does <i>not</i> match "foo" preceded by six characters, the first
1932of which are digits and the last three of which are not "999". For example, it
1933doesn't match "123abcfoo". A pattern to do that is
1934<pre>
1935  (?&#60;=\d{3}...)(?&#60;!999)foo
1936</pre>
1937This time the first assertion looks at the preceding six characters, checking
1938that the first three are digits, and then the second assertion checks that the
1939preceding three characters are not "999".
1940</P>
1941<P>
1942Assertions can be nested in any combination. For example,
1943<pre>
1944  (?&#60;=(?&#60;!foo)bar)baz
1945</pre>
1946matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1947preceded by "foo", while
1948<pre>
1949  (?&#60;=\d{3}(?!999)...)foo
1950</pre>
1951is another pattern that matches "foo" preceded by three digits and any three
1952characters that are not "999".
1953<a name="conditions"></a></P>
1954<br><a name="SEC19" href="#TOC1">CONDITIONAL SUBPATTERNS</a><br>
1955<P>
1956It is possible to cause the matching process to obey a subpattern
1957conditionally or to choose between two alternative subpatterns, depending on
1958the result of an assertion, or whether a specific capturing subpattern has
1959already been matched. The two possible forms of conditional subpattern are:
1960<pre>
1961  (?(condition)yes-pattern)
1962  (?(condition)yes-pattern|no-pattern)
1963</pre>
1964If the condition is satisfied, the yes-pattern is used; otherwise the
1965no-pattern (if present) is used. If there are more than two alternatives in the
1966subpattern, a compile-time error occurs. Each of the two alternatives may
1967itself contain nested subpatterns of any form, including conditional
1968subpatterns; the restriction to two alternatives applies only at the level of
1969the condition. This pattern fragment is an example where the alternatives are
1970complex:
1971<pre>
1972  (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
1973
1974</PRE>
1975</P>
1976<P>
1977There are four kinds of condition: references to subpatterns, references to
1978recursion, a pseudo-condition called DEFINE, and assertions.
1979</P>
1980<br><b>
1981Checking for a used subpattern by number
1982</b><br>
1983<P>
1984If the text between the parentheses consists of a sequence of digits, the
1985condition is true if a capturing subpattern of that number has previously
1986matched. If there is more than one capturing subpattern with the same number
1987(see the earlier
1988<a href="#recursion">section about duplicate subpattern numbers),</a>
1989the condition is true if any of them have matched. An alternative notation is
1990to precede the digits with a plus or minus sign. In this case, the subpattern
1991number is relative rather than absolute. The most recently opened parentheses
1992can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
1993loops it can also make sense to refer to subsequent groups. The next
1994parentheses to be opened can be referenced as (?(+1), and so on. (The value
1995zero in any of these forms is not used; it provokes a compile-time error.)
1996</P>
1997<P>
1998Consider the following pattern, which contains non-significant white space to
1999make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2000three parts for ease of discussion:
2001<pre>
2002  ( \( )?    [^()]+    (?(1) \) )
2003</pre>
2004The first part matches an optional opening parenthesis, and if that
2005character is present, sets it as the first captured substring. The second part
2006matches one or more characters that are not parentheses. The third part is a
2007conditional subpattern that tests whether or not the first set of parentheses
2008matched. If they did, that is, if subject started with an opening parenthesis,
2009the condition is true, and so the yes-pattern is executed and a closing
2010parenthesis is required. Otherwise, since no-pattern is not present, the
2011subpattern matches nothing. In other words, this pattern matches a sequence of
2012non-parentheses, optionally enclosed in parentheses.
2013</P>
2014<P>
2015If you were embedding this pattern in a larger one, you could use a relative
2016reference:
2017<pre>
2018  ...other stuff... ( \( )?    [^()]+    (?(-1) \) ) ...
2019</pre>
2020This makes the fragment independent of the parentheses in the larger pattern.
2021</P>
2022<br><b>
2023Checking for a used subpattern by name
2024</b><br>
2025<P>
2026Perl uses the syntax (?(&#60;name&#62;)...) or (?('name')...) to test for a used
2027subpattern by name. For compatibility with earlier versions of PCRE, which had
2028this facility before Perl, the syntax (?(name)...) is also recognized. However,
2029there is a possible ambiguity with this syntax, because subpattern names may
2030consist entirely of digits. PCRE looks first for a named subpattern; if it
2031cannot find one and the name consists entirely of digits, PCRE looks for a
2032subpattern of that number, which must be greater than zero. Using subpattern
2033names that consist entirely of digits is not recommended.
2034</P>
2035<P>
2036Rewriting the above example to use a named subpattern gives this:
2037<pre>
2038  (?&#60;OPEN&#62; \( )?    [^()]+    (?(&#60;OPEN&#62;) \) )
2039</pre>
2040If the name used in a condition of this kind is a duplicate, the test is
2041applied to all subpatterns of the same name, and is true if any one of them has
2042matched.
2043</P>
2044<br><b>
2045Checking for pattern recursion
2046</b><br>
2047<P>
2048If the condition is the string (R), and there is no subpattern with the name R,
2049the condition is true if a recursive call to the whole pattern or any
2050subpattern has been made. If digits or a name preceded by ampersand follow the
2051letter R, for example:
2052<pre>
2053  (?(R3)...) or (?(R&name)...)
2054</pre>
2055the condition is true if the most recent recursion is into a subpattern whose
2056number or name is given. This condition does not check the entire recursion
2057stack. If the name used in a condition of this kind is a duplicate, the test is
2058applied to all subpatterns of the same name, and is true if any one of them is
2059the most recent recursion.
2060</P>
2061<P>
2062At "top level", all these recursion test conditions are false.
2063<a href="#recursion">The syntax for recursive patterns</a>
2064is described below.
2065<a name="subdefine"></a></P>
2066<br><b>
2067Defining subpatterns for use by reference only
2068</b><br>
2069<P>
2070If the condition is the string (DEFINE), and there is no subpattern with the
2071name DEFINE, the condition is always false. In this case, there may be only one
2072alternative in the subpattern. It is always skipped if control reaches this
2073point in the pattern; the idea of DEFINE is that it can be used to define
2074"subroutines" that can be referenced from elsewhere. (The use of
2075<a href="#subpatternsassubroutines">"subroutines"</a>
2076is described below.) For example, a pattern to match an IPv4 address such as
2077"192.168.23.245" could be written like this (ignore whitespace and line
2078breaks):
2079<pre>
2080  (?(DEFINE) (?&#60;byte&#62; 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) )
2081  \b (?&byte) (\.(?&byte)){3} \b
2082</pre>
2083The first part of the pattern is a DEFINE group inside which a another group
2084named "byte" is defined. This matches an individual component of an IPv4
2085address (a number less than 256). When matching takes place, this part of the
2086pattern is skipped because DEFINE acts like a false condition. The rest of the
2087pattern uses references to the named group to match the four dot-separated
2088components of an IPv4 address, insisting on a word boundary at each end.
2089</P>
2090<br><b>
2091Assertion conditions
2092</b><br>
2093<P>
2094If the condition is not in any of the above formats, it must be an assertion.
2095This may be a positive or negative lookahead or lookbehind assertion. Consider
2096this pattern, again containing non-significant white space, and with the two
2097alternatives on the second line:
2098<pre>
2099  (?(?=[^a-z]*[a-z])
2100  \d{2}-[a-z]{3}-\d{2}  |  \d{2}-\d{2}-\d{2} )
2101</pre>
2102The condition is a positive lookahead assertion that matches an optional
2103sequence of non-letters followed by a letter. In other words, it tests for the
2104presence of at least one letter in the subject. If a letter is found, the
2105subject is matched against the first alternative; otherwise it is matched
2106against the second. This pattern matches strings in one of the two forms
2107dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2108<a name="comments"></a></P>
2109<br><a name="SEC20" href="#TOC1">COMMENTS</a><br>
2110<P>
2111There are two ways of including comments in patterns that are processed by
2112PCRE. In both cases, the start of the comment must not be in a character class,
2113nor in the middle of any other sequence of related characters such as (?: or a
2114subpattern name or number. The characters that make up a comment play no part
2115in the pattern matching.
2116</P>
2117<P>
2118The sequence (?# marks the start of a comment that continues up to the next
2119closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2120option is set, an unescaped # character also introduces a comment, which in
2121this case continues to immediately after the next newline character or
2122character sequence in the pattern. Which characters are interpreted as newlines
2123is controlled by the options passed to <b>pcre_compile()</b> or by a special
2124sequence at the start of the pattern, as described in the section entitled
2125<a href="#newlines">"Newline conventions"</a>
2126above. Note that the end of this type of comment is a literal newline sequence
2127in the pattern; escape sequences that happen to represent a newline do not
2128count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2129default newline convention is in force:
2130<pre>
2131  abc #comment \n still comment
2132</pre>
2133On encountering the # character, <b>pcre_compile()</b> skips along, looking for
2134a newline in the pattern. The sequence \n is still literal at this stage, so
2135it does not terminate the comment. Only an actual character with the code value
21360x0a (the default newline) does so.
2137<a name="recursion"></a></P>
2138<br><a name="SEC21" href="#TOC1">RECURSIVE PATTERNS</a><br>
2139<P>
2140Consider the problem of matching a string in parentheses, allowing for
2141unlimited nested parentheses. Without the use of recursion, the best that can
2142be done is to use a pattern that matches up to some fixed depth of nesting. It
2143is not possible to handle an arbitrary nesting depth.
2144</P>
2145<P>
2146For some time, Perl has provided a facility that allows regular expressions to
2147recurse (amongst other things). It does this by interpolating Perl code in the
2148expression at run time, and the code can refer to the expression itself. A Perl
2149pattern using code interpolation to solve the parentheses problem can be
2150created like this:
2151<pre>
2152  $re = qr{\( (?: (?&#62;[^()]+) | (?p{$re}) )* \)}x;
2153</pre>
2154The (?p{...}) item interpolates Perl code at run time, and in this case refers
2155recursively to the pattern in which it appears.
2156</P>
2157<P>
2158Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2159supports special syntax for recursion of the entire pattern, and also for
2160individual subpattern recursion. After its introduction in PCRE and Python,
2161this kind of recursion was subsequently introduced into Perl at release 5.10.
2162</P>
2163<P>
2164A special item that consists of (? followed by a number greater than zero and a
2165closing parenthesis is a recursive call of the subpattern of the given number,
2166provided that it occurs inside that subpattern. (If not, it is a
2167<a href="#subpatternsassubroutines">"subroutine"</a>
2168call, which is described in the next section.) The special item (?R) or (?0) is
2169a recursive call of the entire regular expression.
2170</P>
2171<P>
2172This PCRE pattern solves the nested parentheses problem (assume the
2173PCRE_EXTENDED option is set so that white space is ignored):
2174<pre>
2175  \( ( [^()]++ | (?R) )* \)
2176</pre>
2177First it matches an opening parenthesis. Then it matches any number of
2178substrings which can either be a sequence of non-parentheses, or a recursive
2179match of the pattern itself (that is, a correctly parenthesized substring).
2180Finally there is a closing parenthesis. Note the use of a possessive quantifier
2181to avoid backtracking into sequences of non-parentheses.
2182</P>
2183<P>
2184If this were part of a larger pattern, you would not want to recurse the entire
2185pattern, so instead you could use this:
2186<pre>
2187  ( \( ( [^()]++ | (?1) )* \) )
2188</pre>
2189We have put the pattern into parentheses, and caused the recursion to refer to
2190them instead of the whole pattern.
2191</P>
2192<P>
2193In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2194is made easier by the use of relative references. Instead of (?1) in the
2195pattern above you can write (?-2) to refer to the second most recently opened
2196parentheses preceding the recursion. In other words, a negative number counts
2197capturing parentheses leftwards from the point at which it is encountered.
2198</P>
2199<P>
2200It is also possible to refer to subsequently opened parentheses, by writing
2201references such as (?+2). However, these cannot be recursive because the
2202reference is not inside the parentheses that are referenced. They are always
2203<a href="#subpatternsassubroutines">"subroutine"</a>
2204calls, as described in the next section.
2205</P>
2206<P>
2207An alternative approach is to use named parentheses instead. The Perl syntax
2208for this is (?&name); PCRE's earlier syntax (?P&#62;name) is also supported. We
2209could rewrite the above example as follows:
2210<pre>
2211  (?&#60;pn&#62; \( ( [^()]++ | (?&pn) )* \) )
2212</pre>
2213If there is more than one subpattern with the same name, the earliest one is
2214used.
2215</P>
2216<P>
2217This particular example pattern that we have been looking at contains nested
2218unlimited repeats, and so the use of a possessive quantifier for matching
2219strings of non-parentheses is important when applying the pattern to strings
2220that do not match. For example, when this pattern is applied to
2221<pre>
2222  (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2223</pre>
2224it yields "no match" quickly. However, if a possessive quantifier is not used,
2225the match runs for a very long time indeed because there are so many different
2226ways the + and * repeats can carve up the subject, and all have to be tested
2227before failure can be reported.
2228</P>
2229<P>
2230At the end of a match, the values of capturing parentheses are those from
2231the outermost level. If you want to obtain intermediate values, a callout
2232function can be used (see below and the
2233<a href="pcrecallout.html"><b>pcrecallout</b></a>
2234documentation). If the pattern above is matched against
2235<pre>
2236  (ab(cd)ef)
2237</pre>
2238the value for the inner capturing parentheses (numbered 2) is "ef", which is
2239the last value taken on at the top level. If a capturing subpattern is not
2240matched at the top level, its final value is unset, even if it is (temporarily)
2241set at a deeper level.
2242</P>
2243<P>
2244If there are more than 15 capturing parentheses in a pattern, PCRE has to
2245obtain extra memory to store data during a recursion, which it does by using
2246<b>pcre_malloc</b>, freeing it via <b>pcre_free</b> afterwards. If no memory can
2247be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2248</P>
2249<P>
2250Do not confuse the (?R) item with the condition (R), which tests for recursion.
2251Consider this pattern, which matches text in angle brackets, allowing for
2252arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2253recursing), whereas any characters are permitted at the outer level.
2254<pre>
2255  &#60; (?: (?(R) \d++  | [^&#60;&#62;]*+) | (?R)) * &#62;
2256</pre>
2257In this pattern, (?(R) is the start of a conditional subpattern, with two
2258different alternatives for the recursive and non-recursive cases. The (?R) item
2259is the actual recursive call.
2260<a name="recursiondifference"></a></P>
2261<br><b>
2262Recursion difference from Perl
2263</b><br>
2264<P>
2265In PCRE (like Python, but unlike Perl), a recursive subpattern call is always
2266treated as an atomic group. That is, once it has matched some of the subject
2267string, it is never re-entered, even if it contains untried alternatives and
2268there is a subsequent matching failure. This can be illustrated by the
2269following pattern, which purports to match a palindromic string that contains
2270an odd number of characters (for example, "a", "aba", "abcba", "abcdcba"):
2271<pre>
2272  ^(.|(.)(?1)\2)$
2273</pre>
2274The idea is that it either matches a single character, or two identical
2275characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2276it does not if the pattern is longer than three characters. Consider the
2277subject string "abcba":
2278</P>
2279<P>
2280At the top level, the first character is matched, but as it is not at the end
2281of the string, the first alternative fails; the second alternative is taken
2282and the recursion kicks in. The recursive call to subpattern 1 successfully
2283matches the next character ("b"). (Note that the beginning and end of line
2284tests are not part of the recursion).
2285</P>
2286<P>
2287Back at the top level, the next character ("c") is compared with what
2288subpattern 2 matched, which was "a". This fails. Because the recursion is
2289treated as an atomic group, there are now no backtracking points, and so the
2290entire match fails. (Perl is able, at this point, to re-enter the recursion and
2291try the second alternative.) However, if the pattern is written with the
2292alternatives in the other order, things are different:
2293<pre>
2294  ^((.)(?1)\2|.)$
2295</pre>
2296This time, the recursing alternative is tried first, and continues to recurse
2297until it runs out of characters, at which point the recursion fails. But this
2298time we do have another alternative to try at the higher level. That is the big
2299difference: in the previous case the remaining alternative is at a deeper
2300recursion level, which PCRE cannot use.
2301</P>
2302<P>
2303To change the pattern so that it matches all palindromic strings, not just
2304those with an odd number of characters, it is tempting to change the pattern to
2305this:
2306<pre>
2307  ^((.)(?1)\2|.?)$
2308</pre>
2309Again, this works in Perl, but not in PCRE, and for the same reason. When a
2310deeper recursion has matched a single character, it cannot be entered again in
2311order to match an empty string. The solution is to separate the two cases, and
2312write out the odd and even cases as alternatives at the higher level:
2313<pre>
2314  ^(?:((.)(?1)\2|)|((.)(?3)\4|.))
2315</pre>
2316If you want to match typical palindromic phrases, the pattern has to ignore all
2317non-word characters, which can be done like this:
2318<pre>
2319  ^\W*+(?:((.)\W*+(?1)\W*+\2|)|((.)\W*+(?3)\W*+\4|\W*+.\W*+))\W*+$
2320</pre>
2321If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2322man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2323the use of the possessive quantifier *+ to avoid backtracking into sequences of
2324non-word characters. Without this, PCRE takes a great deal longer (ten times or
2325more) to match typical phrases, and Perl takes so long that you think it has
2326gone into a loop.
2327</P>
2328<P>
2329<b>WARNING</b>: The palindrome-matching patterns above work only if the subject
2330string does not start with a palindrome that is shorter than the entire string.
2331For example, although "abcba" is correctly matched, if the subject is "ababa",
2332PCRE finds the palindrome "aba" at the start, then fails at top level because
2333the end of the string does not follow. Once again, it cannot jump back into the
2334recursion to try other alternatives, so the entire match fails.
2335<a name="subpatternsassubroutines"></a></P>
2336<br><a name="SEC22" href="#TOC1">SUBPATTERNS AS SUBROUTINES</a><br>
2337<P>
2338If the syntax for a recursive subpattern reference (either by number or by
2339name) is used outside the parentheses to which it refers, it operates like a
2340subroutine in a programming language. The "called" subpattern may be defined
2341before or after the reference. A numbered reference can be absolute or
2342relative, as in these examples:
2343<pre>
2344  (...(absolute)...)...(?2)...
2345  (...(relative)...)...(?-1)...
2346  (...(?+1)...(relative)...
2347</pre>
2348An earlier example pointed out that the pattern
2349<pre>
2350  (sens|respons)e and \1ibility
2351</pre>
2352matches "sense and sensibility" and "response and responsibility", but not
2353"sense and responsibility". If instead the pattern
2354<pre>
2355  (sens|respons)e and (?1)ibility
2356</pre>
2357is used, it does match "sense and responsibility" as well as the other two
2358strings. Another example is given in the discussion of DEFINE above.
2359</P>
2360<P>
2361Like recursive subpatterns, a subroutine call is always treated as an atomic
2362group. That is, once it has matched some of the subject string, it is never
2363re-entered, even if it contains untried alternatives and there is a subsequent
2364matching failure. Any capturing parentheses that are set during the subroutine
2365call revert to their previous values afterwards.
2366</P>
2367<P>
2368When a subpattern is used as a subroutine, processing options such as
2369case-independence are fixed when the subpattern is defined. They cannot be
2370changed for different calls. For example, consider this pattern:
2371<pre>
2372  (abc)(?i:(?-1))
2373</pre>
2374It matches "abcabc". It does not match "abcABC" because the change of
2375processing option does not affect the called subpattern.
2376<a name="onigurumasubroutines"></a></P>
2377<br><a name="SEC23" href="#TOC1">ONIGURUMA SUBROUTINE SYNTAX</a><br>
2378<P>
2379For compatibility with Oniguruma, the non-Perl syntax \g followed by a name or
2380a number enclosed either in angle brackets or single quotes, is an alternative
2381syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2382are two of the examples used above, rewritten using this syntax:
2383<pre>
2384  (?&#60;pn&#62; \( ( (?&#62;[^()]+) | \g&#60;pn&#62; )* \) )
2385  (sens|respons)e and \g'1'ibility
2386</pre>
2387PCRE supports an extension to Oniguruma: if a number is preceded by a
2388plus or a minus sign it is taken as a relative reference. For example:
2389<pre>
2390  (abc)(?i:\g&#60;-1&#62;)
2391</pre>
2392Note that \g{...} (Perl syntax) and \g&#60;...&#62; (Oniguruma syntax) are <i>not</i>
2393synonymous. The former is a back reference; the latter is a subroutine call.
2394</P>
2395<br><a name="SEC24" href="#TOC1">CALLOUTS</a><br>
2396<P>
2397Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2398code to be obeyed in the middle of matching a regular expression. This makes it
2399possible, amongst other things, to extract different substrings that match the
2400same pair of parentheses when there is a repetition.
2401</P>
2402<P>
2403PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2404code. The feature is called "callout". The caller of PCRE provides an external
2405function by putting its entry point in the global variable <i>pcre_callout</i>.
2406By default, this variable contains NULL, which disables all calling out.
2407</P>
2408<P>
2409Within a regular expression, (?C) indicates the points at which the external
2410function is to be called. If you want to identify different callout points, you
2411can put a number less than 256 after the letter C. The default value is zero.
2412For example, this pattern has two callout points:
2413<pre>
2414  (?C1)abc(?C2)def
2415</pre>
2416If the PCRE_AUTO_CALLOUT flag is passed to <b>pcre_compile()</b>, callouts are
2417automatically installed before each item in the pattern. They are all numbered
2418255.
2419</P>
2420<P>
2421During matching, when PCRE reaches a callout point (and <i>pcre_callout</i> is
2422set), the external function is called. It is provided with the number of the
2423callout, the position in the pattern, and, optionally, one item of data
2424originally supplied by the caller of <b>pcre_exec()</b>. The callout function
2425may cause matching to proceed, to backtrack, or to fail altogether. A complete
2426description of the interface to the callout function is given in the
2427<a href="pcrecallout.html"><b>pcrecallout</b></a>
2428documentation.
2429<a name="backtrackcontrol"></a></P>
2430<br><a name="SEC25" href="#TOC1">BACKTRACKING CONTROL</a><br>
2431<P>
2432Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2433are described in the Perl documentation as "experimental and subject to change
2434or removal in a future version of Perl". It goes on to say: "Their usage in
2435production code should be noted to avoid problems during upgrades." The same
2436remarks apply to the PCRE features described in this section.
2437</P>
2438<P>
2439Since these verbs are specifically related to backtracking, most of them can be
2440used only when the pattern is to be matched using <b>pcre_exec()</b>, which uses
2441a backtracking algorithm. With the exception of (*FAIL), which behaves like a
2442failing negative assertion, they cause an error if encountered by
2443<b>pcre_dfa_exec()</b>.
2444</P>
2445<P>
2446If any of these verbs are used in an assertion or subroutine subpattern
2447(including recursive subpatterns), their effect is confined to that subpattern;
2448it does not extend to the surrounding pattern. Note that such subpatterns are
2449processed as anchored at the point where they are tested.
2450</P>
2451<P>
2452The new verbs make use of what was previously invalid syntax: an opening
2453parenthesis followed by an asterisk. They are generally of the form
2454(*VERB) or (*VERB:NAME). Some may take either form, with differing behaviour,
2455depending on whether or not an argument is present. An name is a sequence of
2456letters, digits, and underscores. If the name is empty, that is, if the closing
2457parenthesis immediately follows the colon, the effect is as if the colon were
2458not there. Any number of these verbs may occur in a pattern.
2459</P>
2460<P>
2461PCRE contains some optimizations that are used to speed up matching by running
2462some checks at the start of each match attempt. For example, it may know the
2463minimum length of matching subject, or that a particular character must be
2464present. When one of these optimizations suppresses the running of a match, any
2465included backtracking verbs will not, of course, be processed. You can suppress
2466the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2467when calling <b>pcre_compile()</b> or <b>pcre_exec()</b>, or by starting the
2468pattern with (*NO_START_OPT).
2469</P>
2470<br><b>
2471Verbs that act immediately
2472</b><br>
2473<P>
2474The following verbs act as soon as they are encountered. They may not be
2475followed by a name.
2476<pre>
2477   (*ACCEPT)
2478</pre>
2479This verb causes the match to end successfully, skipping the remainder of the
2480pattern. When inside a recursion, only the innermost pattern is ended
2481immediately. If (*ACCEPT) is inside capturing parentheses, the data so far is
2482captured. (This feature was added to PCRE at release 8.00.) For example:
2483<pre>
2484  A((?:A|B(*ACCEPT)|C)D)
2485</pre>
2486This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2487the outer parentheses.
2488<pre>
2489  (*FAIL) or (*F)
2490</pre>
2491This verb causes the match to fail, forcing backtracking to occur. It is
2492equivalent to (?!) but easier to read. The Perl documentation notes that it is
2493probably useful only when combined with (?{}) or (??{}). Those are, of course,
2494Perl features that are not present in PCRE. The nearest equivalent is the
2495callout feature, as for example in this pattern:
2496<pre>
2497  a+(?C)(*FAIL)
2498</pre>
2499A match with the string "aaaa" always fails, but the callout is taken before
2500each backtrack happens (in this example, 10 times).
2501</P>
2502<br><b>
2503Recording which path was taken
2504</b><br>
2505<P>
2506There is one verb whose main purpose is to track how a match was arrived at,
2507though it also has a secondary use in conjunction with advancing the match
2508starting point (see (*SKIP) below).
2509<pre>
2510  (*MARK:NAME) or (*:NAME)
2511</pre>
2512A name is always required with this verb. There may be as many instances of
2513(*MARK) as you like in a pattern, and their names do not have to be unique.
2514</P>
2515<P>
2516When a match succeeds, the name of the last-encountered (*MARK) is passed back
2517to the caller via the <i>pcre_extra</i> data structure, as described in the
2518<a href="pcreapi.html#extradata">section on <i>pcre_extra</i></a>
2519in the
2520<a href="pcreapi.html"><b>pcreapi</b></a>
2521documentation. No data is returned for a partial match. Here is an example of
2522<b>pcretest</b> output, where the /K modifier requests the retrieval and
2523outputting of (*MARK) data:
2524<pre>
2525  /X(*MARK:A)Y|X(*MARK:B)Z/K
2526  XY
2527   0: XY
2528  MK: A
2529  XZ
2530   0: XZ
2531  MK: B
2532</pre>
2533The (*MARK) name is tagged with "MK:" in this output, and in this example it
2534indicates which of the two alternatives matched. This is a more efficient way
2535of obtaining this information than putting each alternative in its own
2536capturing parentheses.
2537</P>
2538<P>
2539A name may also be returned after a failed match if the final path through the
2540pattern involves (*MARK). However, unless (*MARK) used in conjunction with
2541(*COMMIT), this is unlikely to happen for an unanchored pattern because, as the
2542starting point for matching is advanced, the final check is often with an empty
2543string, causing a failure before (*MARK) is reached. For example:
2544<pre>
2545  /X(*MARK:A)Y|X(*MARK:B)Z/K
2546  XP
2547  No match
2548</pre>
2549There are three potential starting points for this match (starting with X,
2550starting with P, and with an empty string). If the pattern is anchored, the
2551result is different:
2552<pre>
2553  /^X(*MARK:A)Y|^X(*MARK:B)Z/K
2554  XP
2555  No match, mark = B
2556</pre>
2557PCRE's start-of-match optimizations can also interfere with this. For example,
2558if, as a result of a call to <b>pcre_study()</b>, it knows the minimum
2559subject length for a match, a shorter subject will not be scanned at all.
2560</P>
2561<P>
2562Note that similar anomalies (though different in detail) exist in Perl, no
2563doubt for the same reasons. The use of (*MARK) data after a failed match of an
2564unanchored pattern is not recommended, unless (*COMMIT) is involved.
2565</P>
2566<br><b>
2567Verbs that act after backtracking
2568</b><br>
2569<P>
2570The following verbs do nothing when they are encountered. Matching continues
2571with what follows, but if there is no subsequent match, causing a backtrack to
2572the verb, a failure is forced. That is, backtracking cannot pass to the left of
2573the verb. However, when one of these verbs appears inside an atomic group, its
2574effect is confined to that group, because once the group has been matched,
2575there is never any backtracking into it. In this situation, backtracking can
2576"jump back" to the left of the entire atomic group. (Remember also, as stated
2577above, that this localization also applies in subroutine calls and assertions.)
2578</P>
2579<P>
2580These verbs differ in exactly what kind of failure occurs when backtracking
2581reaches them.
2582<pre>
2583  (*COMMIT)
2584</pre>
2585This verb, which may not be followed by a name, causes the whole match to fail
2586outright if the rest of the pattern does not match. Even if the pattern is
2587unanchored, no further attempts to find a match by advancing the starting point
2588take place. Once (*COMMIT) has been passed, <b>pcre_exec()</b> is committed to
2589finding a match at the current starting point, or not at all. For example:
2590<pre>
2591  a+(*COMMIT)b
2592</pre>
2593This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2594dynamic anchor, or "I've started, so I must finish." The name of the most
2595recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
2596match failure.
2597</P>
2598<P>
2599Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
2600unless PCRE's start-of-match optimizations are turned off, as shown in this
2601<b>pcretest</b> example:
2602<pre>
2603  /(*COMMIT)abc/
2604  xyzabc
2605   0: abc
2606  xyzabc\Y
2607  No match
2608</pre>
2609PCRE knows that any match must start with "a", so the optimization skips along
2610the subject to "a" before running the first match attempt, which succeeds. When
2611the optimization is disabled by the \Y escape in the second subject, the match
2612starts at "x" and so the (*COMMIT) causes it to fail without trying any other
2613starting points.
2614<pre>
2615  (*PRUNE) or (*PRUNE:NAME)
2616</pre>
2617This verb causes the match to fail at the current starting position in the
2618subject if the rest of the pattern does not match. If the pattern is
2619unanchored, the normal "bumpalong" advance to the next starting character then
2620happens. Backtracking can occur as usual to the left of (*PRUNE), before it is
2621reached, or when matching to the right of (*PRUNE), but if there is no match to
2622the right, backtracking cannot cross (*PRUNE). In simple cases, the use of
2623(*PRUNE) is just an alternative to an atomic group or possessive quantifier,
2624but there are some uses of (*PRUNE) that cannot be expressed in any other way.
2625The behaviour of (*PRUNE:NAME) is the same as (*MARK:NAME)(*PRUNE) when the
2626match fails completely; the name is passed back if this is the final attempt.
2627(*PRUNE:NAME) does not pass back a name if the match succeeds. In an anchored
2628pattern (*PRUNE) has the same effect as (*COMMIT).
2629<pre>
2630  (*SKIP)
2631</pre>
2632This verb, when given without a name, is like (*PRUNE), except that if the
2633pattern is unanchored, the "bumpalong" advance is not to the next character,
2634but to the position in the subject where (*SKIP) was encountered. (*SKIP)
2635signifies that whatever text was matched leading up to it cannot be part of a
2636successful match. Consider:
2637<pre>
2638  a+(*SKIP)b
2639</pre>
2640If the subject is "aaaac...", after the first match attempt fails (starting at
2641the first character in the string), the starting point skips on to start the
2642next attempt at "c". Note that a possessive quantifer does not have the same
2643effect as this example; although it would suppress backtracking during the
2644first match attempt, the second attempt would start at the second character
2645instead of skipping on to "c".
2646<pre>
2647  (*SKIP:NAME)
2648</pre>
2649When (*SKIP) has an associated name, its behaviour is modified. If the
2650following pattern fails to match, the previous path through the pattern is
2651searched for the most recent (*MARK) that has the same name. If one is found,
2652the "bumpalong" advance is to the subject position that corresponds to that
2653(*MARK) instead of to where (*SKIP) was encountered. If no (*MARK) with a
2654matching name is found, normal "bumpalong" of one character happens (the
2655(*SKIP) is ignored).
2656<pre>
2657  (*THEN) or (*THEN:NAME)
2658</pre>
2659This verb causes a skip to the next alternation in the innermost enclosing
2660group if the rest of the pattern does not match. That is, it cancels pending
2661backtracking, but only within the current alternation. Its name comes from the
2662observation that it can be used for a pattern-based if-then-else block:
2663<pre>
2664  ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2665</pre>
2666If the COND1 pattern matches, FOO is tried (and possibly further items after
2667the end of the group if FOO succeeds); on failure the matcher skips to the
2668second alternative and tries COND2, without backtracking into COND1. The
2669behaviour of (*THEN:NAME) is exactly the same as (*MARK:NAME)(*THEN) if the
2670overall match fails. If (*THEN) is not directly inside an alternation, it acts
2671like (*PRUNE).
2672</P>
2673<P>
2674The above verbs provide four different "strengths" of control when subsequent
2675matching fails. (*THEN) is the weakest, carrying on the match at the next
2676alternation. (*PRUNE) comes next, failing the match at the current starting
2677position, but allowing an advance to the next character (for an unanchored
2678pattern). (*SKIP) is similar, except that the advance may be more than one
2679character. (*COMMIT) is the strongest, causing the entire match to fail.
2680</P>
2681<P>
2682If more than one is present in a pattern, the "stongest" one wins. For example,
2683consider this pattern, where A, B, etc. are complex pattern fragments:
2684<pre>
2685  (A(*COMMIT)B(*THEN)C|D)
2686</pre>
2687Once A has matched, PCRE is committed to this match, at the current starting
2688position. If subsequently B matches, but C does not, the normal (*THEN) action
2689of trying the next alternation (that is, D) does not happen because (*COMMIT)
2690overrides.
2691</P>
2692<br><a name="SEC26" href="#TOC1">SEE ALSO</a><br>
2693<P>
2694<b>pcreapi</b>(3), <b>pcrecallout</b>(3), <b>pcrematching</b>(3),
2695<b>pcresyntax</b>(3), <b>pcre</b>(3).
2696</P>
2697<br><a name="SEC27" href="#TOC1">AUTHOR</a><br>
2698<P>
2699Philip Hazel
2700<br>
2701University Computing Service
2702<br>
2703Cambridge CB2 3QH, England.
2704<br>
2705</P>
2706<br><a name="SEC28" href="#TOC1">REVISION</a><br>
2707<P>
2708Last updated: 21 November 2010
2709<br>
2710Copyright &copy; 1997-2010 University of Cambridge.
2711<br>
2712<p>
2713Return to the <a href="index.html">PCRE index page</a>.
2714</p>
2715