ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 572 - (show annotations)
Wed Nov 17 17:55:57 2010 UTC (9 years, 2 months ago) by ph10
File MIME type: text/html
File size: 115642 byte(s)
Documentation updates and tidies.
1 <html>
2 <head>
3 <title>pcrepattern specification</title>
4 </head>
5 <body bgcolor="#FFFFFF" text="#00005A" link="#0066FF" alink="#3399FF" vlink="#2222BB">
6 <h1>pcrepattern man page</h1>
7 <p>
8 Return to the <a href="index.html">PCRE index page</a>.
9 </p>
10 <p>
11 This page is part of the PCRE HTML documentation. It was generated automatically
12 from the original man page. If there is any nonsense in it, please consult the
13 man page, in case the conversion went wrong.
14 <br>
15 <ul>
16 <li><a name="TOC1" href="#SEC1">PCRE REGULAR EXPRESSION DETAILS</a>
17 <li><a name="TOC2" href="#SEC2">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>
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>
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>
47 The syntax and semantics of the regular expressions that are supported by PCRE
48 are described in detail below. There is a quick-reference syntax summary in the
49 <a href="pcresyntax.html"><b>pcresyntax</b></a>
50 page. PCRE tries to match Perl syntax and semantics as closely as it can. PCRE
51 also supports some alternative regular expression syntax (which does not
52 conflict with the Perl syntax) in order to provide some compatibility with
53 regular expressions in Python, .NET, and Oniguruma.
54 </P>
55 <P>
56 Perl's regular expressions are described in its own documentation, and
57 regular expressions in general are covered in a number of books, some of which
58 have copious examples. Jeffrey Friedl's "Mastering Regular Expressions",
59 published by O'Reilly, covers regular expressions in great detail. This
60 description of PCRE's regular expressions is intended as reference material.
61 </P>
62 <P>
63 The original operation of PCRE was on strings of one-byte characters. However,
64 there is now also support for UTF-8 character strings. To use this,
65 PCRE 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
67 is also a special sequence that can be given at the start of a pattern:
68 <pre>
69 (*UTF8)
70 </pre>
71 Starting a pattern with this sequence is equivalent to setting the PCRE_UTF8
72 option. This feature is not Perl-compatible. How setting UTF-8 mode affects
73 pattern matching is mentioned in several places below. There is also a summary
74 of UTF-8 features in the
75 <a href="pcre.html#utf8support">section on UTF-8 support</a>
76 in the main
77 <a href="pcre.html"><b>pcre</b></a>
78 page.
79 </P>
80 <P>
81 Another special sequence that may appear at the start of a pattern or in
82 combination with (*UTF8) is:
83 <pre>
84 (*UCP)
85 </pre>
86 This has the same effect as setting the PCRE_UCP option: it causes sequences
87 such as \d and \w to use Unicode properties to determine character types,
88 instead of recognizing only characters with codes less than 128 via a lookup
89 table.
90 </P>
91 <P>
92 The remainder of this document discusses the patterns that are supported by
93 PCRE when its main matching function, <b>pcre_exec()</b>, is used.
94 From release 6.0, PCRE offers a second matching function,
95 <b>pcre_dfa_exec()</b>, which matches using a different algorithm that is not
96 Perl-compatible. Some of the features discussed below are not available when
97 <b>pcre_dfa_exec()</b> is used. The advantages and disadvantages of the
98 alternative function, and how it differs from the normal function, are
99 discussed in the
100 <a href="pcrematching.html"><b>pcrematching</b></a>
101 page.
102 <a name="newlines"></a></P>
103 <br><a name="SEC2" href="#TOC1">NEWLINE CONVENTIONS</a><br>
104 <P>
105 PCRE supports five different conventions for indicating line breaks in
106 strings: a single CR (carriage return) character, a single LF (linefeed)
107 character, the two-character sequence CRLF, any of the three preceding, or any
108 Unicode newline sequence. The
109 <a href="pcreapi.html"><b>pcreapi</b></a>
110 page has
111 <a href="pcreapi.html#newlines">further discussion</a>
112 about newlines, and shows how to set the newline convention in the
113 <i>options</i> arguments for the compiling and matching functions.
114 </P>
115 <P>
116 It is also possible to specify a newline convention by starting a pattern
117 string with one of the following five sequences:
118 <pre>
119 (*CR) carriage return
120 (*LF) linefeed
121 (*CRLF) carriage return, followed by linefeed
122 (*ANYCRLF) any of the three above
123 (*ANY) all Unicode newline sequences
124 </pre>
125 These override the default and the options given to <b>pcre_compile()</b> or
126 <b>pcre_compile2()</b>. For example, on a Unix system where LF is the default
127 newline sequence, the pattern
128 <pre>
129 (*CR)a.b
130 </pre>
131 changes the convention to CR. That pattern matches "a\nb" because LF is no
132 longer a newline. Note that these special settings, which are not
133 Perl-compatible, are recognized only at the very start of a pattern, and that
134 they must be in upper case. If more than one of them is present, the last one
135 is used.
136 </P>
137 <P>
138 The newline convention affects the interpretation of the dot metacharacter when
139 PCRE_DOTALL is not set, and also the behaviour of \N. However, it does not
140 affect what the \R escape sequence matches. By default, this is any Unicode
141 newline sequence, for Perl compatibility. However, this can be changed; see the
142 description of \R in the section entitled
143 <a href="#newlineseq">"Newline sequences"</a>
144 below. A change of \R setting can be combined with a change of newline
145 convention.
146 </P>
147 <br><a name="SEC3" href="#TOC1">CHARACTERS AND METACHARACTERS</a><br>
148 <P>
149 A regular expression is a pattern that is matched against a subject string from
150 left to right. Most characters stand for themselves in a pattern, and match the
151 corresponding characters in the subject. As a trivial example, the pattern
152 <pre>
153 The quick brown fox
154 </pre>
155 matches a portion of a subject string that is identical to itself. When
156 caseless matching is specified (the PCRE_CASELESS option), letters are matched
157 independently of case. In UTF-8 mode, PCRE always understands the concept of
158 case for characters whose values are less than 128, so caseless matching is
159 always possible. For characters with higher values, the concept of case is
160 supported if PCRE is compiled with Unicode property support, but not otherwise.
161 If you want to use caseless matching for characters 128 and above, you must
162 ensure that PCRE is compiled with Unicode property support as well as with
163 UTF-8 support.
164 </P>
165 <P>
166 The power of regular expressions comes from the ability to include alternatives
167 and repetitions in the pattern. These are encoded in the pattern by the use of
168 <i>metacharacters</i>, which do not stand for themselves but instead are
169 interpreted in some special way.
170 </P>
171 <P>
172 There are two different sets of metacharacters: those that are recognized
173 anywhere in the pattern except within square brackets, and those that are
174 recognized within square brackets. Outside square brackets, the metacharacters
175 are as follows:
176 <pre>
177 \ general escape character with several uses
178 ^ assert start of string (or line, in multiline mode)
179 $ assert end of string (or line, in multiline mode)
180 . match any character except newline (by default)
181 [ start character class definition
182 | start of alternative branch
183 ( start subpattern
184 ) end subpattern
185 ? extends the meaning of (
186 also 0 or 1 quantifier
187 also quantifier minimizer
188 * 0 or more quantifier
189 + 1 or more quantifier
190 also "possessive quantifier"
191 { start min/max quantifier
192 </pre>
193 Part of a pattern that is in square brackets is called a "character class". In
194 a character class the only metacharacters are:
195 <pre>
196 \ general escape character
197 ^ negate the class, but only if the first character
198 - indicates character range
199 [ POSIX character class (only if followed by POSIX syntax)
200 ] terminates the character class
201 </pre>
202 The following sections describe the use of each of the metacharacters.
203 </P>
204 <br><a name="SEC4" href="#TOC1">BACKSLASH</a><br>
205 <P>
206 The backslash character has several uses. Firstly, if it is followed by a
207 non-alphanumeric character, it takes away any special meaning that character
208 may have. This use of backslash as an escape character applies both inside and
209 outside character classes.
210 </P>
211 <P>
212 For example, if you want to match a * character, you write \* in the pattern.
213 This escaping action applies whether or not the following character would
214 otherwise be interpreted as a metacharacter, so it is always safe to precede a
215 non-alphanumeric with backslash to specify that it stands for itself. In
216 particular, if you want to match a backslash, you write \\.
217 </P>
218 <P>
219 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
220 pattern (other than in a character class) and characters between a # outside
221 a character class and the next newline are ignored. An escaping backslash can
222 be used to include a whitespace or # character as part of the pattern.
223 </P>
224 <P>
225 If you want to remove the special meaning from a sequence of characters, you
226 can do so by putting them between \Q and \E. This is different from Perl in
227 that $ and @ are handled as literals in \Q...\E sequences in PCRE, whereas in
228 Perl, $ and @ cause variable interpolation. Note the following examples:
229 <pre>
230 Pattern PCRE matches Perl matches
232 \Qabc$xyz\E abc$xyz abc followed by the contents of $xyz
233 \Qabc\$xyz\E abc\$xyz abc\$xyz
234 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
235 </pre>
236 The \Q...\E sequence is recognized both inside and outside character classes.
237 An isolated \E that is not preceded by \Q is ignored.
238 <a name="digitsafterbackslash"></a></P>
239 <br><b>
240 Non-printing characters
241 </b><br>
242 <P>
243 A second use of backslash provides a way of encoding non-printing characters
244 in patterns in a visible manner. There is no restriction on the appearance of
245 non-printing characters, apart from the binary zero that terminates a pattern,
246 but when a pattern is being prepared by text editing, it is often easier to use
247 one of the following escape sequences than the binary character it represents:
248 <pre>
249 \a alarm, that is, the BEL character (hex 07)
250 \cx "control-x", where x is any character
251 \e escape (hex 1B)
252 \f formfeed (hex 0C)
253 \n linefeed (hex 0A)
254 \r carriage return (hex 0D)
255 \t tab (hex 09)
256 \ddd character with octal code ddd, or back reference
257 \xhh character with hex code hh
258 \x{hhh..} character with hex code hhh..
259 </pre>
260 The precise effect of \cx is as follows: if x is a lower case letter, it
261 is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
262 Thus \cz becomes hex 1A, but \c{ becomes hex 3B, while \c; becomes hex
263 7B.
264 </P>
265 <P>
266 After \x, from zero to two hexadecimal digits are read (letters can be in
267 upper or lower case). Any number of hexadecimal digits may appear between \x{
268 and }, but the value of the character code must be less than 256 in non-UTF-8
269 mode, and less than 2**31 in UTF-8 mode. That is, the maximum value in
270 hexadecimal is 7FFFFFFF. Note that this is bigger than the largest Unicode code
271 point, which is 10FFFF.
272 </P>
273 <P>
274 If characters other than hexadecimal digits appear between \x{ and }, or if
275 there is no terminating }, this form of escape is not recognized. Instead, the
276 initial \x will be interpreted as a basic hexadecimal escape, with no
277 following digits, giving a character whose value is zero.
278 </P>
279 <P>
280 Characters whose value is less than 256 can be defined by either of the two
281 syntaxes for \x. There is no difference in the way they are handled. For
282 example, \xdc is exactly the same as \x{dc}.
283 </P>
284 <P>
285 After \0 up to two further octal digits are read. If there are fewer than two
286 digits, just those that are present are used. Thus the sequence \0\x\07
287 specifies two binary zeros followed by a BEL character (code value 7). Make
288 sure you supply two digits after the initial zero if the pattern character that
289 follows is itself an octal digit.
290 </P>
291 <P>
292 The handling of a backslash followed by a digit other than 0 is complicated.
293 Outside a character class, PCRE reads it and any following digits as a decimal
294 number. If the number is less than 10, or if there have been at least that many
295 previous capturing left parentheses in the expression, the entire sequence is
296 taken as a <i>back reference</i>. A description of how this works is given
297 <a href="#backreferences">later,</a>
298 following the discussion of
299 <a href="#subpattern">parenthesized subpatterns.</a>
300 </P>
301 <P>
302 Inside a character class, or if the decimal number is greater than 9 and there
303 have not been that many capturing subpatterns, PCRE re-reads up to three octal
304 digits following the backslash, and uses them to generate a data character. Any
305 subsequent digits stand for themselves. In non-UTF-8 mode, the value of a
306 character specified in octal must be less than \400. In UTF-8 mode, values up
307 to \777 are permitted. For example:
308 <pre>
309 \040 is another way of writing a space
310 \40 is the same, provided there are fewer than 40 previous capturing subpatterns
311 \7 is always a back reference
312 \11 might be a back reference, or another way of writing a tab
313 \011 is always a tab
314 \0113 is a tab followed by the character "3"
315 \113 might be a back reference, otherwise the character with octal code 113
316 \377 might be a back reference, otherwise the byte consisting entirely of 1 bits
317 \81 is either a back reference, or a binary zero followed by the two characters "8" and "1"
318 </pre>
319 Note that octal values of 100 or greater must not be introduced by a leading
320 zero, because no more than three octal digits are ever read.
321 </P>
322 <P>
323 All the sequences that define a single character value can be used both inside
324 and outside character classes. In addition, inside a character class, the
325 sequence \b is interpreted as the backspace character (hex 08). The sequences
326 \B, \N, \R, and \X are not special inside a character class. Like any other
327 unrecognized escape sequences, they are treated as the literal characters "B",
328 "N", "R", and "X" by default, but cause an error if the PCRE_EXTRA option is
329 set. Outside a character class, these sequences have different meanings.
330 </P>
331 <br><b>
332 Absolute and relative back references
333 </b><br>
334 <P>
335 The sequence \g followed by an unsigned or a negative number, optionally
336 enclosed in braces, is an absolute or relative back reference. A named back
337 reference can be coded as \g{name}. Back references are discussed
338 <a href="#backreferences">later,</a>
339 following the discussion of
340 <a href="#subpattern">parenthesized subpatterns.</a>
341 </P>
342 <br><b>
343 Absolute and relative subroutine calls
344 </b><br>
345 <P>
346 For compatibility with Oniguruma, the non-Perl syntax \g followed by a name or
347 a number enclosed either in angle brackets or single quotes, is an alternative
348 syntax for referencing a subpattern as a "subroutine". Details are discussed
349 <a href="#onigurumasubroutines">later.</a>
350 Note that \g{...} (Perl syntax) and \g&#60;...&#62; (Oniguruma syntax) are <i>not</i>
351 synonymous. The former is a back reference; the latter is a
352 <a href="#subpatternsassubroutines">subroutine</a>
353 call.
354 <a name="genericchartypes"></a></P>
355 <br><b>
356 Generic character types
357 </b><br>
358 <P>
359 Another use of backslash is for specifying generic character types:
360 <pre>
361 \d any decimal digit
362 \D any character that is not a decimal digit
363 \h any horizontal whitespace character
364 \H any character that is not a horizontal whitespace character
365 \s any whitespace character
366 \S any character that is not a whitespace character
367 \v any vertical whitespace character
368 \V any character that is not a vertical whitespace character
369 \w any "word" character
370 \W any "non-word" character
371 </pre>
372 There is also the single sequence \N, which matches a non-newline character.
373 This is the same as
374 <a href="#fullstopdot">the "." metacharacter</a>
375 when PCRE_DOTALL is not set.
376 </P>
377 <P>
378 Each pair of lower and upper case escape sequences partitions the complete set
379 of characters into two disjoint sets. Any given character matches one, and only
380 one, of each pair. The sequences can appear both inside and outside character
381 classes. They each match one character of the appropriate type. If the current
382 matching point is at the end of the subject string, all of them fail, because
383 there is no character to match.
384 </P>
385 <P>
386 For compatibility with Perl, \s does not match the VT character (code 11).
387 This makes it different from the the POSIX "space" class. The \s characters
388 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
389 included in a Perl script, \s may match the VT character. In PCRE, it never
390 does.
391 </P>
392 <P>
393 A "word" character is an underscore or any character that is a letter or digit.
394 By default, the definition of letters and digits is controlled by PCRE's
395 low-valued character tables, and may vary if locale-specific matching is taking
396 place (see
397 <a href="pcreapi.html#localesupport">"Locale support"</a>
398 in the
399 <a href="pcreapi.html"><b>pcreapi</b></a>
400 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
401 or "french" in Windows, some character codes greater than 128 are used for
402 accented letters, and these are then matched by \w. The use of locales with
403 Unicode is discouraged.
404 </P>
405 <P>
406 By default, in UTF-8 mode, characters with values greater than 128 never match
407 \d, \s, or \w, and always match \D, \S, and \W. These sequences retain
408 their original meanings from before UTF-8 support was available, mainly for
409 efficiency reasons. However, if PCRE is compiled with Unicode property support,
410 and the PCRE_UCP option is set, the behaviour is changed so that Unicode
411 properties are used to determine character types, as follows:
412 <pre>
413 \d any character that \p{Nd} matches (decimal digit)
414 \s any character that \p{Z} matches, plus HT, LF, FF, CR
415 \w any character that \p{L} or \p{N} matches, plus underscore
416 </pre>
417 The upper case escapes match the inverse sets of characters. Note that \d
418 matches only decimal digits, whereas \w matches any Unicode digit, as well as
419 any Unicode letter, and underscore. Note also that PCRE_UCP affects \b, and
420 \B because they are defined in terms of \w and \W. Matching these sequences
421 is noticeably slower when PCRE_UCP is set.
422 </P>
423 <P>
424 The sequences \h, \H, \v, and \V are features that were added to Perl at
425 release 5.10. In contrast to the other sequences, which match only ASCII
426 characters by default, these always match certain high-valued codepoints in
427 UTF-8 mode, whether or not PCRE_UCP is set. The horizontal space characters
428 are:
429 <pre>
430 U+0009 Horizontal tab
431 U+0020 Space
432 U+00A0 Non-break space
433 U+1680 Ogham space mark
434 U+180E Mongolian vowel separator
435 U+2000 En quad
436 U+2001 Em quad
437 U+2002 En space
438 U+2003 Em space
439 U+2004 Three-per-em space
440 U+2005 Four-per-em space
441 U+2006 Six-per-em space
442 U+2007 Figure space
443 U+2008 Punctuation space
444 U+2009 Thin space
445 U+200A Hair space
446 U+202F Narrow no-break space
447 U+205F Medium mathematical space
448 U+3000 Ideographic space
449 </pre>
450 The vertical space characters are:
451 <pre>
452 U+000A Linefeed
453 U+000B Vertical tab
454 U+000C Formfeed
455 U+000D Carriage return
456 U+0085 Next line
457 U+2028 Line separator
458 U+2029 Paragraph separator
459 <a name="newlineseq"></a></PRE>
460 </P>
461 <br><b>
462 Newline sequences
463 </b><br>
464 <P>
465 Outside a character class, by default, the escape sequence \R matches any
466 Unicode newline sequence. In non-UTF-8 mode \R is equivalent to the following:
467 <pre>
468 (?&#62;\r\n|\n|\x0b|\f|\r|\x85)
469 </pre>
470 This is an example of an "atomic group", details of which are given
471 <a href="#atomicgroup">below.</a>
472 This particular group matches either the two-character sequence CR followed by
473 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
474 U+000B), FF (formfeed, U+000C), CR (carriage return, U+000D), or NEL (next
475 line, U+0085). The two-character sequence is treated as a single unit that
476 cannot be split.
477 </P>
478 <P>
479 In UTF-8 mode, two additional characters whose codepoints are greater than 255
480 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
481 Unicode character property support is not needed for these characters to be
482 recognized.
483 </P>
484 <P>
485 It is possible to restrict \R to match only CR, LF, or CRLF (instead of the
486 complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
487 either at compile time or when the pattern is matched. (BSR is an abbrevation
488 for "backslash R".) This can be made the default when PCRE is built; if this is
489 the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
490 It is also possible to specify these settings by starting a pattern string with
491 one of the following sequences:
492 <pre>
493 (*BSR_ANYCRLF) CR, LF, or CRLF only
494 (*BSR_UNICODE) any Unicode newline sequence
495 </pre>
496 These override the default and the options given to <b>pcre_compile()</b> or
497 <b>pcre_compile2()</b>, but they can be overridden by options given to
498 <b>pcre_exec()</b> or <b>pcre_dfa_exec()</b>. Note that these special settings,
499 which are not Perl-compatible, are recognized only at the very start of a
500 pattern, and that they must be in upper case. If more than one of them is
501 present, the last one is used. They can be combined with a change of newline
502 convention; for example, a pattern can start with:
503 <pre>
505 </pre>
506 They can also be combined with the (*UTF8) or (*UCP) special sequences. Inside
507 a character class, \R is treated as an unrecognized escape sequence, and so
508 matches the letter "R" by default, but causes an error if PCRE_EXTRA is set.
509 <a name="uniextseq"></a></P>
510 <br><b>
511 Unicode character properties
512 </b><br>
513 <P>
514 When PCRE is built with Unicode character property support, three additional
515 escape sequences that match characters with specific properties are available.
516 When not in UTF-8 mode, these sequences are of course limited to testing
517 characters whose codepoints are less than 256, but they do work in this mode.
518 The extra escape sequences are:
519 <pre>
520 \p{<i>xx</i>} a character with the <i>xx</i> property
521 \P{<i>xx</i>} a character without the <i>xx</i> property
522 \X an extended Unicode sequence
523 </pre>
524 The property names represented by <i>xx</i> above are limited to the Unicode
525 script names, the general category properties, "Any", which matches any
526 character (including newline), and some special PCRE properties (described
527 in the
528 <a href="#extraprops">next section).</a>
529 Other Perl properties such as "InMusicalSymbols" are not currently supported by
530 PCRE. Note that \P{Any} does not match any characters, so always causes a
531 match failure.
532 </P>
533 <P>
534 Sets of Unicode characters are defined as belonging to certain scripts. A
535 character from one of these sets can be matched using a script name. For
536 example:
537 <pre>
538 \p{Greek}
539 \P{Han}
540 </pre>
541 Those that are not part of an identified script are lumped together as
542 "Common". The current list of scripts is:
543 </P>
544 <P>
545 Arabic,
546 Armenian,
547 Avestan,
548 Balinese,
549 Bamum,
550 Bengali,
551 Bopomofo,
552 Braille,
553 Buginese,
554 Buhid,
555 Canadian_Aboriginal,
556 Carian,
557 Cham,
558 Cherokee,
559 Common,
560 Coptic,
561 Cuneiform,
562 Cypriot,
563 Cyrillic,
564 Deseret,
565 Devanagari,
566 Egyptian_Hieroglyphs,
567 Ethiopic,
568 Georgian,
569 Glagolitic,
570 Gothic,
571 Greek,
572 Gujarati,
573 Gurmukhi,
574 Han,
575 Hangul,
576 Hanunoo,
577 Hebrew,
578 Hiragana,
579 Imperial_Aramaic,
580 Inherited,
581 Inscriptional_Pahlavi,
582 Inscriptional_Parthian,
583 Javanese,
584 Kaithi,
585 Kannada,
586 Katakana,
587 Kayah_Li,
588 Kharoshthi,
589 Khmer,
590 Lao,
591 Latin,
592 Lepcha,
593 Limbu,
594 Linear_B,
595 Lisu,
596 Lycian,
597 Lydian,
598 Malayalam,
599 Meetei_Mayek,
600 Mongolian,
601 Myanmar,
602 New_Tai_Lue,
603 Nko,
604 Ogham,
605 Old_Italic,
606 Old_Persian,
607 Old_South_Arabian,
608 Old_Turkic,
609 Ol_Chiki,
610 Oriya,
611 Osmanya,
612 Phags_Pa,
613 Phoenician,
614 Rejang,
615 Runic,
616 Samaritan,
617 Saurashtra,
618 Shavian,
619 Sinhala,
620 Sundanese,
621 Syloti_Nagri,
622 Syriac,
623 Tagalog,
624 Tagbanwa,
625 Tai_Le,
626 Tai_Tham,
627 Tai_Viet,
628 Tamil,
629 Telugu,
630 Thaana,
631 Thai,
632 Tibetan,
633 Tifinagh,
634 Ugaritic,
635 Vai,
636 Yi.
637 </P>
638 <P>
639 Each character has exactly one Unicode general category property, specified by
640 a two-letter abbreviation. For compatibility with Perl, negation can be
641 specified by including a circumflex between the opening brace and the property
642 name. For example, \p{^Lu} is the same as \P{Lu}.
643 </P>
644 <P>
645 If only one letter is specified with \p or \P, it includes all the general
646 category properties that start with that letter. In this case, in the absence
647 of negation, the curly brackets in the escape sequence are optional; these two
648 examples have the same effect:
649 <pre>
650 \p{L}
651 \pL
652 </pre>
653 The following general category property codes are supported:
654 <pre>
655 C Other
656 Cc Control
657 Cf Format
658 Cn Unassigned
659 Co Private use
660 Cs Surrogate
662 L Letter
663 Ll Lower case letter
664 Lm Modifier letter
665 Lo Other letter
666 Lt Title case letter
667 Lu Upper case letter
669 M Mark
670 Mc Spacing mark
671 Me Enclosing mark
672 Mn Non-spacing mark
674 N Number
675 Nd Decimal number
676 Nl Letter number
677 No Other number
679 P Punctuation
680 Pc Connector punctuation
681 Pd Dash punctuation
682 Pe Close punctuation
683 Pf Final punctuation
684 Pi Initial punctuation
685 Po Other punctuation
686 Ps Open punctuation
688 S Symbol
689 Sc Currency symbol
690 Sk Modifier symbol
691 Sm Mathematical symbol
692 So Other symbol
694 Z Separator
695 Zl Line separator
696 Zp Paragraph separator
697 Zs Space separator
698 </pre>
699 The special property L& is also supported: it matches a character that has
700 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
701 a modifier or "other".
702 </P>
703 <P>
704 The Cs (Surrogate) property applies only to characters in the range U+D800 to
705 U+DFFF. Such characters are not valid in UTF-8 strings (see RFC 3629) and so
706 cannot be tested by PCRE, unless UTF-8 validity checking has been turned off
707 (see the discussion of PCRE_NO_UTF8_CHECK in the
708 <a href="pcreapi.html"><b>pcreapi</b></a>
709 page). Perl does not support the Cs property.
710 </P>
711 <P>
712 The long synonyms for property names that Perl supports (such as \p{Letter})
713 are not supported by PCRE, nor is it permitted to prefix any of these
714 properties with "Is".
715 </P>
716 <P>
717 No character that is in the Unicode table has the Cn (unassigned) property.
718 Instead, this property is assumed for any code point that is not in the
719 Unicode table.
720 </P>
721 <P>
722 Specifying caseless matching does not affect these escape sequences. For
723 example, \p{Lu} always matches only upper case letters.
724 </P>
725 <P>
726 The \X escape matches any number of Unicode characters that form an extended
727 Unicode sequence. \X is equivalent to
728 <pre>
729 (?&#62;\PM\pM*)
730 </pre>
731 That is, it matches a character without the "mark" property, followed by zero
732 or more characters with the "mark" property, and treats the sequence as an
733 atomic group
734 <a href="#atomicgroup">(see below).</a>
735 Characters with the "mark" property are typically accents that affect the
736 preceding character. None of them have codepoints less than 256, so in
737 non-UTF-8 mode \X matches any one character.
738 </P>
739 <P>
740 Matching characters by Unicode property is not fast, because PCRE has to search
741 a structure that contains data for over fifteen thousand characters. That is
742 why the traditional escape sequences such as \d and \w do not use Unicode
743 properties in PCRE by default, though you can make them do so by setting the
744 PCRE_UCP option for <b>pcre_compile()</b> or by starting the pattern with
745 (*UCP).
746 <a name="extraprops"></a></P>
747 <br><b>
748 PCRE's additional properties
749 </b><br>
750 <P>
751 As well as the standard Unicode properties described in the previous
752 section, PCRE supports four more that make it possible to convert traditional
753 escape sequences such as \w and \s and POSIX character classes to use Unicode
754 properties. PCRE uses these non-standard, non-Perl properties internally when
755 PCRE_UCP is set. They are:
756 <pre>
757 Xan Any alphanumeric character
758 Xps Any POSIX space character
759 Xsp Any Perl space character
760 Xwd Any Perl "word" character
761 </pre>
762 Xan matches characters that have either the L (letter) or the N (number)
763 property. Xps matches the characters tab, linefeed, vertical tab, formfeed, or
764 carriage return, and any other character that has the Z (separator) property.
765 Xsp is the same as Xps, except that vertical tab is excluded. Xwd matches the
766 same characters as Xan, plus underscore.
767 <a name="resetmatchstart"></a></P>
768 <br><b>
769 Resetting the match start
770 </b><br>
771 <P>
772 The escape sequence \K causes any previously matched characters not to be
773 included in the final matched sequence. For example, the pattern:
774 <pre>
775 foo\Kbar
776 </pre>
777 matches "foobar", but reports that it has matched "bar". This feature is
778 similar to a lookbehind assertion
779 <a href="#lookbehind">(described below).</a>
780 However, in this case, the part of the subject before the real match does not
781 have to be of fixed length, as lookbehind assertions do. The use of \K does
782 not interfere with the setting of
783 <a href="#subpattern">captured substrings.</a>
784 For example, when the pattern
785 <pre>
786 (foo)\Kbar
787 </pre>
788 matches "foobar", the first substring is still set to "foo".
789 </P>
790 <P>
791 Perl documents that the use of \K within assertions is "not well defined". In
792 PCRE, \K is acted upon when it occurs inside positive assertions, but is
793 ignored in negative assertions.
794 <a name="smallassertions"></a></P>
795 <br><b>
796 Simple assertions
797 </b><br>
798 <P>
799 The final use of backslash is for certain simple assertions. An assertion
800 specifies a condition that has to be met at a particular point in a match,
801 without consuming any characters from the subject string. The use of
802 subpatterns for more complicated assertions is described
803 <a href="#bigassertions">below.</a>
804 The backslashed assertions are:
805 <pre>
806 \b matches at a word boundary
807 \B matches when not at a word boundary
808 \A matches at the start of the subject
809 \Z matches at the end of the subject
810 also matches before a newline at the end of the subject
811 \z matches only at the end of the subject
812 \G matches at the first matching position in the subject
813 </pre>
814 Inside a character class, \b has a different meaning; it matches the backspace
815 character. If any other of these assertions appears in a character class, by
816 default it matches the corresponding literal character (for example, \B
817 matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
818 escape sequence" error is generated instead.
819 </P>
820 <P>
821 A word boundary is a position in the subject string where the current character
822 and the previous character do not both match \w or \W (i.e. one matches
823 \w and the other matches \W), or the start or end of the string if the
824 first or last character matches \w, respectively. In UTF-8 mode, the meanings
825 of \w and \W can be changed by setting the PCRE_UCP option. When this is
826 done, it also affects \b and \B. Neither PCRE nor Perl has a separate "start
827 of word" or "end of word" metasequence. However, whatever follows \b normally
828 determines which it is. For example, the fragment \ba matches "a" at the start
829 of a word.
830 </P>
831 <P>
832 The \A, \Z, and \z assertions differ from the traditional circumflex and
833 dollar (described in the next section) in that they only ever match at the very
834 start and end of the subject string, whatever options are set. Thus, they are
835 independent of multiline mode. These three assertions are not affected by the
836 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
837 circumflex and dollar metacharacters. However, if the <i>startoffset</i>
838 argument of <b>pcre_exec()</b> is non-zero, indicating that matching is to start
839 at a point other than the beginning of the subject, \A can never match. The
840 difference between \Z and \z is that \Z matches before a newline at the end
841 of the string as well as at the very end, whereas \z matches only at the end.
842 </P>
843 <P>
844 The \G assertion is true only when the current matching position is at the
845 start point of the match, as specified by the <i>startoffset</i> argument of
846 <b>pcre_exec()</b>. It differs from \A when the value of <i>startoffset</i> is
847 non-zero. By calling <b>pcre_exec()</b> multiple times with appropriate
848 arguments, you can mimic Perl's /g option, and it is in this kind of
849 implementation where \G can be useful.
850 </P>
851 <P>
852 Note, however, that PCRE's interpretation of \G, as the start of the current
853 match, is subtly different from Perl's, which defines it as the end of the
854 previous match. In Perl, these can be different when the previously matched
855 string was empty. Because PCRE does just one match at a time, it cannot
856 reproduce this behaviour.
857 </P>
858 <P>
859 If all the alternatives of a pattern begin with \G, the expression is anchored
860 to the starting match position, and the "anchored" flag is set in the compiled
861 regular expression.
862 </P>
863 <br><a name="SEC5" href="#TOC1">CIRCUMFLEX AND DOLLAR</a><br>
864 <P>
865 Outside a character class, in the default matching mode, the circumflex
866 character is an assertion that is true only if the current matching point is
867 at the start of the subject string. If the <i>startoffset</i> argument of
868 <b>pcre_exec()</b> is non-zero, circumflex can never match if the PCRE_MULTILINE
869 option is unset. Inside a character class, circumflex has an entirely different
870 meaning
871 <a href="#characterclass">(see below).</a>
872 </P>
873 <P>
874 Circumflex need not be the first character of the pattern if a number of
875 alternatives are involved, but it should be the first thing in each alternative
876 in which it appears if the pattern is ever to match that branch. If all
877 possible alternatives start with a circumflex, that is, if the pattern is
878 constrained to match only at the start of the subject, it is said to be an
879 "anchored" pattern. (There are also other constructs that can cause a pattern
880 to be anchored.)
881 </P>
882 <P>
883 A dollar character is an assertion that is true only if the current matching
884 point is at the end of the subject string, or immediately before a newline
885 at the end of the string (by default). Dollar need not be the last character of
886 the pattern if a number of alternatives are involved, but it should be the last
887 item in any branch in which it appears. Dollar has no special meaning in a
888 character class.
889 </P>
890 <P>
891 The meaning of dollar can be changed so that it matches only at the very end of
892 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
893 does not affect the \Z assertion.
894 </P>
895 <P>
896 The meanings of the circumflex and dollar characters are changed if the
897 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
898 immediately after internal newlines as well as at the start of the subject
899 string. It does not match after a newline that ends the string. A dollar
900 matches before any newlines in the string, as well as at the very end, when
901 PCRE_MULTILINE is set. When newline is specified as the two-character
902 sequence CRLF, isolated CR and LF characters do not indicate newlines.
903 </P>
904 <P>
905 For example, the pattern /^abc$/ matches the subject string "def\nabc" (where
906 \n represents a newline) in multiline mode, but not otherwise. Consequently,
907 patterns that are anchored in single line mode because all branches start with
908 ^ are not anchored in multiline mode, and a match for circumflex is possible
909 when the <i>startoffset</i> argument of <b>pcre_exec()</b> is non-zero. The
910 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
911 </P>
912 <P>
913 Note that the sequences \A, \Z, and \z can be used to match the start and
914 end of the subject in both modes, and if all branches of a pattern start with
915 \A it is always anchored, whether or not PCRE_MULTILINE is set.
916 <a name="fullstopdot"></a></P>
917 <br><a name="SEC6" href="#TOC1">FULL STOP (PERIOD, DOT) AND \N</a><br>
918 <P>
919 Outside a character class, a dot in the pattern matches any one character in
920 the subject string except (by default) a character that signifies the end of a
921 line. In UTF-8 mode, the matched character may be more than one byte long.
922 </P>
923 <P>
924 When a line ending is defined as a single character, dot never matches that
925 character; when the two-character sequence CRLF is used, dot does not match CR
926 if it is immediately followed by LF, but otherwise it matches all characters
927 (including isolated CRs and LFs). When any Unicode line endings are being
928 recognized, dot does not match CR or LF or any of the other line ending
929 characters.
930 </P>
931 <P>
932 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
933 option is set, a dot matches any one character, without exception. If the
934 two-character sequence CRLF is present in the subject string, it takes two dots
935 to match it.
936 </P>
937 <P>
938 The handling of dot is entirely independent of the handling of circumflex and
939 dollar, the only relationship being that they both involve newlines. Dot has no
940 special meaning in a character class.
941 </P>
942 <P>
943 The escape sequence \N behaves like a dot, except that it is not affected by
944 the PCRE_DOTALL option. In other words, it matches any character except one
945 that signifies the end of a line.
946 </P>
947 <br><a name="SEC7" href="#TOC1">MATCHING A SINGLE BYTE</a><br>
948 <P>
949 Outside a character class, the escape sequence \C matches any one byte, both
950 in and out of UTF-8 mode. Unlike a dot, it always matches any line-ending
951 characters. The feature is provided in Perl in order to match individual bytes
952 in UTF-8 mode. Because it breaks up UTF-8 characters into individual bytes, the
953 rest of the string may start with a malformed UTF-8 character. For this reason,
954 the \C escape sequence is best avoided.
955 </P>
956 <P>
957 PCRE does not allow \C to appear in lookbehind assertions
958 <a href="#lookbehind">(described below),</a>
959 because in UTF-8 mode this would make it impossible to calculate the length of
960 the lookbehind.
961 <a name="characterclass"></a></P>
962 <br><a name="SEC8" href="#TOC1">SQUARE BRACKETS AND CHARACTER CLASSES</a><br>
963 <P>
964 An opening square bracket introduces a character class, terminated by a closing
965 square bracket. A closing square bracket on its own is not special by default.
966 However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
967 bracket causes a compile-time error. If a closing square bracket is required as
968 a member of the class, it should be the first data character in the class
969 (after an initial circumflex, if present) or escaped with a backslash.
970 </P>
971 <P>
972 A character class matches a single character in the subject. In UTF-8 mode, the
973 character may be more than one byte long. A matched character must be in the
974 set of characters defined by the class, unless the first character in the class
975 definition is a circumflex, in which case the subject character must not be in
976 the set defined by the class. If a circumflex is actually required as a member
977 of the class, ensure it is not the first character, or escape it with a
978 backslash.
979 </P>
980 <P>
981 For example, the character class [aeiou] matches any lower case vowel, while
982 [^aeiou] matches any character that is not a lower case vowel. Note that a
983 circumflex is just a convenient notation for specifying the characters that
984 are in the class by enumerating those that are not. A class that starts with a
985 circumflex is not an assertion; it still consumes a character from the subject
986 string, and therefore it fails if the current pointer is at the end of the
987 string.
988 </P>
989 <P>
990 In UTF-8 mode, characters with values greater than 255 can be included in a
991 class as a literal string of bytes, or by using the \x{ escaping mechanism.
992 </P>
993 <P>
994 When caseless matching is set, any letters in a class represent both their
995 upper case and lower case versions, so for example, a caseless [aeiou] matches
996 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
997 caseful version would. In UTF-8 mode, PCRE always understands the concept of
998 case for characters whose values are less than 128, so caseless matching is
999 always possible. For characters with higher values, the concept of case is
1000 supported if PCRE is compiled with Unicode property support, but not otherwise.
1001 If you want to use caseless matching in UTF8-mode for characters 128 and above,
1002 you must ensure that PCRE is compiled with Unicode property support as well as
1003 with UTF-8 support.
1004 </P>
1005 <P>
1006 Characters that might indicate line breaks are never treated in any special way
1007 when matching character classes, whatever line-ending sequence is in use, and
1008 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1009 such as [^a] always matches one of these characters.
1010 </P>
1011 <P>
1012 The minus (hyphen) character can be used to specify a range of characters in a
1013 character class. For example, [d-m] matches any letter between d and m,
1014 inclusive. If a minus character is required in a class, it must be escaped with
1015 a backslash or appear in a position where it cannot be interpreted as
1016 indicating a range, typically as the first or last character in the class.
1017 </P>
1018 <P>
1019 It is not possible to have the literal character "]" as the end character of a
1020 range. A pattern such as [W-]46] is interpreted as a class of two characters
1021 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1022 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1023 the end of range, so [W-\]46] is interpreted as a class containing a range
1024 followed by two other characters. The octal or hexadecimal representation of
1025 "]" can also be used to end a range.
1026 </P>
1027 <P>
1028 Ranges operate in the collating sequence of character values. They can also be
1029 used for characters specified numerically, for example [\000-\037]. In UTF-8
1030 mode, ranges can include characters whose values are greater than 255, for
1031 example [\x{100}-\x{2ff}].
1032 </P>
1033 <P>
1034 If a range that includes letters is used when caseless matching is set, it
1035 matches the letters in either case. For example, [W-c] is equivalent to
1036 [][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
1037 tables for a French locale are in use, [\xc8-\xcb] matches accented E
1038 characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
1039 characters with values greater than 128 only when it is compiled with Unicode
1040 property support.
1041 </P>
1042 <P>
1043 The character types \d, \D, \h, \H, \p, \P, \s, \S, \v, \V, \w, and
1044 \W may also appear in a character class, and add the characters that they
1045 match to the class. For example, [\dABCDEF] matches any hexadecimal digit. A
1046 circumflex can conveniently be used with the upper case character types to
1047 specify a more restricted set of characters than the matching lower case type.
1048 For example, the class [^\W_] matches any letter or digit, but not underscore.
1049 </P>
1050 <P>
1051 The only metacharacters that are recognized in character classes are backslash,
1052 hyphen (only where it can be interpreted as specifying a range), circumflex
1053 (only at the start), opening square bracket (only when it can be interpreted as
1054 introducing a POSIX class name - see the next section), and the terminating
1055 closing square bracket. However, escaping other non-alphanumeric characters
1056 does no harm.
1057 </P>
1058 <br><a name="SEC9" href="#TOC1">POSIX CHARACTER CLASSES</a><br>
1059 <P>
1060 Perl supports the POSIX notation for character classes. This uses names
1061 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1062 this notation. For example,
1063 <pre>
1064 [01[:alpha:]%]
1065 </pre>
1066 matches "0", "1", any alphabetic character, or "%". The supported class names
1067 are:
1068 <pre>
1069 alnum letters and digits
1070 alpha letters
1071 ascii character codes 0 - 127
1072 blank space or tab only
1073 cntrl control characters
1074 digit decimal digits (same as \d)
1075 graph printing characters, excluding space
1076 lower lower case letters
1077 print printing characters, including space
1078 punct printing characters, excluding letters and digits and space
1079 space white space (not quite the same as \s)
1080 upper upper case letters
1081 word "word" characters (same as \w)
1082 xdigit hexadecimal digits
1083 </pre>
1084 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
1085 space (32). Notice that this list includes the VT character (code 11). This
1086 makes "space" different to \s, which does not include VT (for Perl
1087 compatibility).
1088 </P>
1089 <P>
1090 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1091 5.8. Another Perl extension is negation, which is indicated by a ^ character
1092 after the colon. For example,
1093 <pre>
1094 [12[:^digit:]]
1095 </pre>
1096 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1097 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1098 supported, and an error is given if they are encountered.
1099 </P>
1100 <P>
1101 By default, in UTF-8 mode, characters with values greater than 128 do not match
1102 any of the POSIX character classes. However, if the PCRE_UCP option is passed
1103 to <b>pcre_compile()</b>, some of the classes are changed so that Unicode
1104 character properties are used. This is achieved by replacing the POSIX classes
1105 by other sequences, as follows:
1106 <pre>
1107 [:alnum:] becomes \p{Xan}
1108 [:alpha:] becomes \p{L}
1109 [:blank:] becomes \h
1110 [:digit:] becomes \p{Nd}
1111 [:lower:] becomes \p{Ll}
1112 [:space:] becomes \p{Xps}
1113 [:upper:] becomes \p{Lu}
1114 [:word:] becomes \p{Xwd}
1115 </pre>
1116 Negated versions, such as [:^alpha:] use \P instead of \p. The other POSIX
1117 classes are unchanged, and match only characters with code points less than
1118 128.
1119 </P>
1120 <br><a name="SEC10" href="#TOC1">VERTICAL BAR</a><br>
1121 <P>
1122 Vertical bar characters are used to separate alternative patterns. For example,
1123 the pattern
1124 <pre>
1125 gilbert|sullivan
1126 </pre>
1127 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1128 and an empty alternative is permitted (matching the empty string). The matching
1129 process tries each alternative in turn, from left to right, and the first one
1130 that succeeds is used. If the alternatives are within a subpattern
1131 <a href="#subpattern">(defined below),</a>
1132 "succeeds" means matching the rest of the main pattern as well as the
1133 alternative in the subpattern.
1134 </P>
1135 <br><a name="SEC11" href="#TOC1">INTERNAL OPTION SETTING</a><br>
1136 <P>
1137 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1138 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1139 the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1140 The option letters are
1141 <pre>
1142 i for PCRE_CASELESS
1144 s for PCRE_DOTALL
1145 x for PCRE_EXTENDED
1146 </pre>
1147 For example, (?im) sets caseless, multiline matching. It is also possible to
1148 unset these options by preceding the letter with a hyphen, and a combined
1149 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1150 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1151 permitted. If a letter appears both before and after the hyphen, the option is
1152 unset.
1153 </P>
1154 <P>
1155 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1156 changed in the same way as the Perl-compatible options by using the characters
1157 J, U and X respectively.
1158 </P>
1159 <P>
1160 When one of these option changes occurs at top level (that is, not inside
1161 subpattern parentheses), the change applies to the remainder of the pattern
1162 that follows. If the change is placed right at the start of a pattern, PCRE
1163 extracts it into the global options (and it will therefore show up in data
1164 extracted by the <b>pcre_fullinfo()</b> function).
1165 </P>
1166 <P>
1167 An option change within a subpattern (see below for a description of
1168 subpatterns) affects only that part of the subpattern that follows it, so
1169 <pre>
1170 (a(?i)b)c
1171 </pre>
1172 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1173 By this means, options can be made to have different settings in different
1174 parts of the pattern. Any changes made in one alternative do carry on
1175 into subsequent branches within the same subpattern. For example,
1176 <pre>
1177 (a(?i)b|c)
1178 </pre>
1179 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1180 branch is abandoned before the option setting. This is because the effects of
1181 option settings happen at compile time. There would be some very weird
1182 behaviour otherwise.
1183 </P>
1184 <P>
1185 <b>Note:</b> There are other PCRE-specific options that can be set by the
1186 application when the compile or match functions are called. In some cases the
1187 pattern can contain special leading sequences such as (*CRLF) to override what
1188 the application has set or what has been defaulted. Details are given in the
1189 section entitled
1190 <a href="#newlineseq">"Newline sequences"</a>
1191 above. There are also the (*UTF8) and (*UCP) leading sequences that can be used
1192 to set UTF-8 and Unicode property modes; they are equivalent to setting the
1193 PCRE_UTF8 and the PCRE_UCP options, respectively.
1194 <a name="subpattern"></a></P>
1195 <br><a name="SEC12" href="#TOC1">SUBPATTERNS</a><br>
1196 <P>
1197 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1198 Turning part of a pattern into a subpattern does two things:
1199 <br>
1200 <br>
1201 1. It localizes a set of alternatives. For example, the pattern
1202 <pre>
1203 cat(aract|erpillar|)
1204 </pre>
1205 matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1206 match "cataract", "erpillar" or an empty string.
1207 <br>
1208 <br>
1209 2. It sets up the subpattern as a capturing subpattern. This means that, when
1210 the whole pattern matches, that portion of the subject string that matched the
1211 subpattern is passed back to the caller via the <i>ovector</i> argument of
1212 <b>pcre_exec()</b>. Opening parentheses are counted from left to right (starting
1213 from 1) to obtain numbers for the capturing subpatterns. For example, if the
1214 string "the red king" is matched against the pattern
1215 <pre>
1216 the ((red|white) (king|queen))
1217 </pre>
1218 the captured substrings are "red king", "red", and "king", and are numbered 1,
1219 2, and 3, respectively.
1220 </P>
1221 <P>
1222 The fact that plain parentheses fulfil two functions is not always helpful.
1223 There are often times when a grouping subpattern is required without a
1224 capturing requirement. If an opening parenthesis is followed by a question mark
1225 and a colon, the subpattern does not do any capturing, and is not counted when
1226 computing the number of any subsequent capturing subpatterns. For example, if
1227 the string "the white queen" is matched against the pattern
1228 <pre>
1229 the ((?:red|white) (king|queen))
1230 </pre>
1231 the captured substrings are "white queen" and "queen", and are numbered 1 and
1232 2. The maximum number of capturing subpatterns is 65535.
1233 </P>
1234 <P>
1235 As a convenient shorthand, if any option settings are required at the start of
1236 a non-capturing subpattern, the option letters may appear between the "?" and
1237 the ":". Thus the two patterns
1238 <pre>
1239 (?i:saturday|sunday)
1240 (?:(?i)saturday|sunday)
1241 </pre>
1242 match exactly the same set of strings. Because alternative branches are tried
1243 from left to right, and options are not reset until the end of the subpattern
1244 is reached, an option setting in one branch does affect subsequent branches, so
1245 the above patterns match "SUNDAY" as well as "Saturday".
1246 <a name="dupsubpatternnumber"></a></P>
1247 <br><a name="SEC13" href="#TOC1">DUPLICATE SUBPATTERN NUMBERS</a><br>
1248 <P>
1249 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1250 the same numbers for its capturing parentheses. Such a subpattern starts with
1251 (?| and is itself a non-capturing subpattern. For example, consider this
1252 pattern:
1253 <pre>
1254 (?|(Sat)ur|(Sun))day
1255 </pre>
1256 Because the two alternatives are inside a (?| group, both sets of capturing
1257 parentheses are numbered one. Thus, when the pattern matches, you can look
1258 at captured substring number one, whichever alternative matched. This construct
1259 is useful when you want to capture part, but not all, of one of a number of
1260 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1261 number is reset at the start of each branch. The numbers of any capturing
1262 parentheses that follow the subpattern start after the highest number used in
1263 any branch. The following example is taken from the Perl documentation. The
1264 numbers underneath show in which buffer the captured content will be stored.
1265 <pre>
1266 # before ---------------branch-reset----------- after
1267 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1268 # 1 2 2 3 2 3 4
1269 </pre>
1270 A back reference to a numbered subpattern uses the most recent value that is
1271 set for that number by any subpattern. The following pattern matches "abcabc"
1272 or "defdef":
1273 <pre>
1274 /(?|(abc)|(def))\1/
1275 </pre>
1276 In contrast, a recursive or "subroutine" call to a numbered subpattern always
1277 refers to the first one in the pattern with the given number. The following
1278 pattern matches "abcabc" or "defabc":
1279 <pre>
1280 /(?|(abc)|(def))(?1)/
1281 </pre>
1282 If a
1283 <a href="#conditions">condition test</a>
1284 for a subpattern's having matched refers to a non-unique number, the test is
1285 true if any of the subpatterns of that number have matched.
1286 </P>
1287 <P>
1288 An alternative approach to using this "branch reset" feature is to use
1289 duplicate named subpatterns, as described in the next section.
1290 </P>
1291 <br><a name="SEC14" href="#TOC1">NAMED SUBPATTERNS</a><br>
1292 <P>
1293 Identifying capturing parentheses by number is simple, but it can be very hard
1294 to keep track of the numbers in complicated regular expressions. Furthermore,
1295 if an expression is modified, the numbers may change. To help with this
1296 difficulty, PCRE supports the naming of subpatterns. This feature was not
1297 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1298 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1299 the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1300 have different names, but PCRE does not.
1301 </P>
1302 <P>
1303 In PCRE, a subpattern can be named in one of three ways: (?&#60;name&#62;...) or
1304 (?'name'...) as in Perl, or (?P&#60;name&#62;...) as in Python. References to capturing
1305 parentheses from other parts of the pattern, such as
1306 <a href="#backreferences">back references,</a>
1307 <a href="#recursion">recursion,</a>
1308 and
1309 <a href="#conditions">conditions,</a>
1310 can be made by name as well as by number.
1311 </P>
1312 <P>
1313 Names consist of up to 32 alphanumeric characters and underscores. Named
1314 capturing parentheses are still allocated numbers as well as names, exactly as
1315 if the names were not present. The PCRE API provides function calls for
1316 extracting the name-to-number translation table from a compiled pattern. There
1317 is also a convenience function for extracting a captured substring by name.
1318 </P>
1319 <P>
1320 By default, a name must be unique within a pattern, but it is possible to relax
1321 this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1322 names are also always permitted for subpatterns with the same number, set up as
1323 described in the previous section.) Duplicate names can be useful for patterns
1324 where only one instance of the named parentheses can match. Suppose you want to
1325 match the name of a weekday, either as a 3-letter abbreviation or as the full
1326 name, and in both cases you want to extract the abbreviation. This pattern
1327 (ignoring the line breaks) does the job:
1328 <pre>
1329 (?&#60;DN&#62;Mon|Fri|Sun)(?:day)?|
1330 (?&#60;DN&#62;Tue)(?:sday)?|
1331 (?&#60;DN&#62;Wed)(?:nesday)?|
1332 (?&#60;DN&#62;Thu)(?:rsday)?|
1333 (?&#60;DN&#62;Sat)(?:urday)?
1334 </pre>
1335 There are five capturing substrings, but only one is ever set after a match.
1336 (An alternative way of solving this problem is to use a "branch reset"
1337 subpattern, as described in the previous section.)
1338 </P>
1339 <P>
1340 The convenience function for extracting the data by name returns the substring
1341 for the first (and in this example, the only) subpattern of that name that
1342 matched. This saves searching to find which numbered subpattern it was.
1343 </P>
1344 <P>
1345 If you make a back reference to a non-unique named subpattern from elsewhere in
1346 the pattern, the one that corresponds to the first occurrence of the name is
1347 used. In the absence of duplicate numbers (see the previous section) this is
1348 the one with the lowest number. If you use a named reference in a condition
1349 test (see the
1350 <a href="#conditions">section about conditions</a>
1351 below), either to check whether a subpattern has matched, or to check for
1352 recursion, all subpatterns with the same name are tested. If the condition is
1353 true for any one of them, the overall condition is true. This is the same
1354 behaviour as testing by number. For further details of the interfaces for
1355 handling named subpatterns, see the
1356 <a href="pcreapi.html"><b>pcreapi</b></a>
1357 documentation.
1358 </P>
1359 <P>
1360 <b>Warning:</b> You cannot use different names to distinguish between two
1361 subpatterns with the same number because PCRE uses only the numbers when
1362 matching. For this reason, an error is given at compile time if different names
1363 are given to subpatterns with the same number. However, you can give the same
1364 name to subpatterns with the same number, even when PCRE_DUPNAMES is not set.
1365 </P>
1366 <br><a name="SEC15" href="#TOC1">REPETITION</a><br>
1367 <P>
1368 Repetition is specified by quantifiers, which can follow any of the following
1369 items:
1370 <pre>
1371 a literal data character
1372 the dot metacharacter
1373 the \C escape sequence
1374 the \X escape sequence (in UTF-8 mode with Unicode properties)
1375 the \R escape sequence
1376 an escape such as \d or \pL that matches a single character
1377 a character class
1378 a back reference (see next section)
1379 a parenthesized subpattern (unless it is an assertion)
1380 a recursive or "subroutine" call to a subpattern
1381 </pre>
1382 The general repetition quantifier specifies a minimum and maximum number of
1383 permitted matches, by giving the two numbers in curly brackets (braces),
1384 separated by a comma. The numbers must be less than 65536, and the first must
1385 be less than or equal to the second. For example:
1386 <pre>
1387 z{2,4}
1388 </pre>
1389 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1390 character. If the second number is omitted, but the comma is present, there is
1391 no upper limit; if the second number and the comma are both omitted, the
1392 quantifier specifies an exact number of required matches. Thus
1393 <pre>
1394 [aeiou]{3,}
1395 </pre>
1396 matches at least 3 successive vowels, but may match many more, while
1397 <pre>
1398 \d{8}
1399 </pre>
1400 matches exactly 8 digits. An opening curly bracket that appears in a position
1401 where a quantifier is not allowed, or one that does not match the syntax of a
1402 quantifier, is taken as a literal character. For example, {,6} is not a
1403 quantifier, but a literal string of four characters.
1404 </P>
1405 <P>
1406 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
1407 bytes. Thus, for example, \x{100}{2} matches two UTF-8 characters, each of
1408 which is represented by a two-byte sequence. Similarly, when Unicode property
1409 support is available, \X{3} matches three Unicode extended sequences, each of
1410 which may be several bytes long (and they may be of different lengths).
1411 </P>
1412 <P>
1413 The quantifier {0} is permitted, causing the expression to behave as if the
1414 previous item and the quantifier were not present. This may be useful for
1415 subpatterns that are referenced as
1416 <a href="#subpatternsassubroutines">subroutines</a>
1417 from elsewhere in the pattern (but see also the section entitled
1418 <a href="#subdefine">"Defining subpatterns for use by reference only"</a>
1419 below). Items other than subpatterns that have a {0} quantifier are omitted
1420 from the compiled pattern.
1421 </P>
1422 <P>
1423 For convenience, the three most common quantifiers have single-character
1424 abbreviations:
1425 <pre>
1426 * is equivalent to {0,}
1427 + is equivalent to {1,}
1428 ? is equivalent to {0,1}
1429 </pre>
1430 It is possible to construct infinite loops by following a subpattern that can
1431 match no characters with a quantifier that has no upper limit, for example:
1432 <pre>
1433 (a?)*
1434 </pre>
1435 Earlier versions of Perl and PCRE used to give an error at compile time for
1436 such patterns. However, because there are cases where this can be useful, such
1437 patterns are now accepted, but if any repetition of the subpattern does in fact
1438 match no characters, the loop is forcibly broken.
1439 </P>
1440 <P>
1441 By default, the quantifiers are "greedy", that is, they match as much as
1442 possible (up to the maximum number of permitted times), without causing the
1443 rest of the pattern to fail. The classic example of where this gives problems
1444 is in trying to match comments in C programs. These appear between /* and */
1445 and within the comment, individual * and / characters may appear. An attempt to
1446 match C comments by applying the pattern
1447 <pre>
1448 /\*.*\*/
1449 </pre>
1450 to the string
1451 <pre>
1452 /* first comment */ not comment /* second comment */
1453 </pre>
1454 fails, because it matches the entire string owing to the greediness of the .*
1455 item.
1456 </P>
1457 <P>
1458 However, if a quantifier is followed by a question mark, it ceases to be
1459 greedy, and instead matches the minimum number of times possible, so the
1460 pattern
1461 <pre>
1462 /\*.*?\*/
1463 </pre>
1464 does the right thing with the C comments. The meaning of the various
1465 quantifiers is not otherwise changed, just the preferred number of matches.
1466 Do not confuse this use of question mark with its use as a quantifier in its
1467 own right. Because it has two uses, it can sometimes appear doubled, as in
1468 <pre>
1469 \d??\d
1470 </pre>
1471 which matches one digit by preference, but can match two if that is the only
1472 way the rest of the pattern matches.
1473 </P>
1474 <P>
1475 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1476 the quantifiers are not greedy by default, but individual ones can be made
1477 greedy by following them with a question mark. In other words, it inverts the
1478 default behaviour.
1479 </P>
1480 <P>
1481 When a parenthesized subpattern is quantified with a minimum repeat count that
1482 is greater than 1 or with a limited maximum, more memory is required for the
1483 compiled pattern, in proportion to the size of the minimum or maximum.
1484 </P>
1485 <P>
1486 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1487 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1488 implicitly anchored, because whatever follows will be tried against every
1489 character position in the subject string, so there is no point in retrying the
1490 overall match at any position after the first. PCRE normally treats such a
1491 pattern as though it were preceded by \A.
1492 </P>
1493 <P>
1494 In cases where it is known that the subject string contains no newlines, it is
1495 worth setting PCRE_DOTALL in order to obtain this optimization, or
1496 alternatively using ^ to indicate anchoring explicitly.
1497 </P>
1498 <P>
1499 However, there is one situation where the optimization cannot be used. When .*
1500 is inside capturing parentheses that are the subject of a back reference
1501 elsewhere in the pattern, a match at the start may fail where a later one
1502 succeeds. Consider, for example:
1503 <pre>
1504 (.*)abc\1
1505 </pre>
1506 If the subject is "xyz123abc123" the match point is the fourth character. For
1507 this reason, such a pattern is not implicitly anchored.
1508 </P>
1509 <P>
1510 When a capturing subpattern is repeated, the value captured is the substring
1511 that matched the final iteration. For example, after
1512 <pre>
1513 (tweedle[dume]{3}\s*)+
1514 </pre>
1515 has matched "tweedledum tweedledee" the value of the captured substring is
1516 "tweedledee". However, if there are nested capturing subpatterns, the
1517 corresponding captured values may have been set in previous iterations. For
1518 example, after
1519 <pre>
1520 /(a|(b))+/
1521 </pre>
1522 matches "aba" the value of the second captured substring is "b".
1523 <a name="atomicgroup"></a></P>
1524 <br><a name="SEC16" href="#TOC1">ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS</a><br>
1525 <P>
1526 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1527 repetition, failure of what follows normally causes the repeated item to be
1528 re-evaluated to see if a different number of repeats allows the rest of the
1529 pattern to match. Sometimes it is useful to prevent this, either to change the
1530 nature of the match, or to cause it fail earlier than it otherwise might, when
1531 the author of the pattern knows there is no point in carrying on.
1532 </P>
1533 <P>
1534 Consider, for example, the pattern \d+foo when applied to the subject line
1535 <pre>
1536 123456bar
1537 </pre>
1538 After matching all 6 digits and then failing to match "foo", the normal
1539 action of the matcher is to try again with only 5 digits matching the \d+
1540 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1541 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1542 that once a subpattern has matched, it is not to be re-evaluated in this way.
1543 </P>
1544 <P>
1545 If we use atomic grouping for the previous example, the matcher gives up
1546 immediately on failing to match "foo" the first time. The notation is a kind of
1547 special parenthesis, starting with (?&#62; as in this example:
1548 <pre>
1549 (?&#62;\d+)foo
1550 </pre>
1551 This kind of parenthesis "locks up" the part of the pattern it contains once
1552 it has matched, and a failure further into the pattern is prevented from
1553 backtracking into it. Backtracking past it to previous items, however, works as
1554 normal.
1555 </P>
1556 <P>
1557 An alternative description is that a subpattern of this type matches the string
1558 of characters that an identical standalone pattern would match, if anchored at
1559 the current point in the subject string.
1560 </P>
1561 <P>
1562 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1563 the above example can be thought of as a maximizing repeat that must swallow
1564 everything it can. So, while both \d+ and \d+? are prepared to adjust the
1565 number of digits they match in order to make the rest of the pattern match,
1566 (?&#62;\d+) can only match an entire sequence of digits.
1567 </P>
1568 <P>
1569 Atomic groups in general can of course contain arbitrarily complicated
1570 subpatterns, and can be nested. However, when the subpattern for an atomic
1571 group is just a single repeated item, as in the example above, a simpler
1572 notation, called a "possessive quantifier" can be used. This consists of an
1573 additional + character following a quantifier. Using this notation, the
1574 previous example can be rewritten as
1575 <pre>
1576 \d++foo
1577 </pre>
1578 Note that a possessive quantifier can be used with an entire group, for
1579 example:
1580 <pre>
1581 (abc|xyz){2,3}+
1582 </pre>
1583 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1584 option is ignored. They are a convenient notation for the simpler forms of
1585 atomic group. However, there is no difference in the meaning of a possessive
1586 quantifier and the equivalent atomic group, though there may be a performance
1587 difference; possessive quantifiers should be slightly faster.
1588 </P>
1589 <P>
1590 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1591 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1592 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1593 package, and PCRE copied it from there. It ultimately found its way into Perl
1594 at release 5.10.
1595 </P>
1596 <P>
1597 PCRE has an optimization that automatically "possessifies" certain simple
1598 pattern constructs. For example, the sequence A+B is treated as A++B because
1599 there is no point in backtracking into a sequence of A's when B must follow.
1600 </P>
1601 <P>
1602 When a pattern contains an unlimited repeat inside a subpattern that can itself
1603 be repeated an unlimited number of times, the use of an atomic group is the
1604 only way to avoid some failing matches taking a very long time indeed. The
1605 pattern
1606 <pre>
1607 (\D+|&#60;\d+&#62;)*[!?]
1608 </pre>
1609 matches an unlimited number of substrings that either consist of non-digits, or
1610 digits enclosed in &#60;&#62;, followed by either ! or ?. When it matches, it runs
1611 quickly. However, if it is applied to
1612 <pre>
1613 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1614 </pre>
1615 it takes a long time before reporting failure. This is because the string can
1616 be divided between the internal \D+ repeat and the external * repeat in a
1617 large number of ways, and all have to be tried. (The example uses [!?] rather
1618 than a single character at the end, because both PCRE and Perl have an
1619 optimization that allows for fast failure when a single character is used. They
1620 remember the last single character that is required for a match, and fail early
1621 if it is not present in the string.) If the pattern is changed so that it uses
1622 an atomic group, like this:
1623 <pre>
1624 ((?&#62;\D+)|&#60;\d+&#62;)*[!?]
1625 </pre>
1626 sequences of non-digits cannot be broken, and failure happens quickly.
1627 <a name="backreferences"></a></P>
1628 <br><a name="SEC17" href="#TOC1">BACK REFERENCES</a><br>
1629 <P>
1630 Outside a character class, a backslash followed by a digit greater than 0 (and
1631 possibly further digits) is a back reference to a capturing subpattern earlier
1632 (that is, to its left) in the pattern, provided there have been that many
1633 previous capturing left parentheses.
1634 </P>
1635 <P>
1636 However, if the decimal number following the backslash is less than 10, it is
1637 always taken as a back reference, and causes an error only if there are not
1638 that many capturing left parentheses in the entire pattern. In other words, the
1639 parentheses that are referenced need not be to the left of the reference for
1640 numbers less than 10. A "forward back reference" of this type can make sense
1641 when a repetition is involved and the subpattern to the right has participated
1642 in an earlier iteration.
1643 </P>
1644 <P>
1645 It is not possible to have a numerical "forward back reference" to a subpattern
1646 whose number is 10 or more using this syntax because a sequence such as \50 is
1647 interpreted as a character defined in octal. See the subsection entitled
1648 "Non-printing characters"
1649 <a href="#digitsafterbackslash">above</a>
1650 for further details of the handling of digits following a backslash. There is
1651 no such problem when named parentheses are used. A back reference to any
1652 subpattern is possible using named parentheses (see below).
1653 </P>
1654 <P>
1655 Another way of avoiding the ambiguity inherent in the use of digits following a
1656 backslash is to use the \g escape sequence. This escape must be followed by an
1657 unsigned number or a negative number, optionally enclosed in braces. These
1658 examples are all identical:
1659 <pre>
1660 (ring), \1
1661 (ring), \g1
1662 (ring), \g{1}
1663 </pre>
1664 An unsigned number specifies an absolute reference without the ambiguity that
1665 is present in the older syntax. It is also useful when literal digits follow
1666 the reference. A negative number is a relative reference. Consider this
1667 example:
1668 <pre>
1669 (abc(def)ghi)\g{-1}
1670 </pre>
1671 The sequence \g{-1} is a reference to the most recently started capturing
1672 subpattern before \g, that is, is it equivalent to \2. Similarly, \g{-2}
1673 would be equivalent to \1. The use of relative references can be helpful in
1674 long patterns, and also in patterns that are created by joining together
1675 fragments that contain references within themselves.
1676 </P>
1677 <P>
1678 A back reference matches whatever actually matched the capturing subpattern in
1679 the current subject string, rather than anything matching the subpattern
1680 itself (see
1681 <a href="#subpatternsassubroutines">"Subpatterns as subroutines"</a>
1682 below for a way of doing that). So the pattern
1683 <pre>
1684 (sens|respons)e and \1ibility
1685 </pre>
1686 matches "sense and sensibility" and "response and responsibility", but not
1687 "sense and responsibility". If caseful matching is in force at the time of the
1688 back reference, the case of letters is relevant. For example,
1689 <pre>
1690 ((?i)rah)\s+\1
1691 </pre>
1692 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1693 capturing subpattern is matched caselessly.
1694 </P>
1695 <P>
1696 There are several different ways of writing back references to named
1697 subpatterns. The .NET syntax \k{name} and the Perl syntax \k&#60;name&#62; or
1698 \k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1699 back reference syntax, in which \g can be used for both numeric and named
1700 references, is also supported. We could rewrite the above example in any of
1701 the following ways:
1702 <pre>
1703 (?&#60;p1&#62;(?i)rah)\s+\k&#60;p1&#62;
1704 (?'p1'(?i)rah)\s+\k{p1}
1705 (?P&#60;p1&#62;(?i)rah)\s+(?P=p1)
1706 (?&#60;p1&#62;(?i)rah)\s+\g{p1}
1707 </pre>
1708 A subpattern that is referenced by name may appear in the pattern before or
1709 after the reference.
1710 </P>
1711 <P>
1712 There may be more than one back reference to the same subpattern. If a
1713 subpattern has not actually been used in a particular match, any back
1714 references to it always fail by default. For example, the pattern
1715 <pre>
1716 (a|(bc))\2
1717 </pre>
1718 always fails if it starts to match "a" rather than "bc". However, if the
1719 PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
1720 unset value matches an empty string.
1721 </P>
1722 <P>
1723 Because there may be many capturing parentheses in a pattern, all digits
1724 following a backslash are taken as part of a potential back reference number.
1725 If the pattern continues with a digit character, some delimiter must be used to
1726 terminate the back reference. If the PCRE_EXTENDED option is set, this can be
1727 whitespace. Otherwise, the \g{ syntax or an empty comment (see
1728 <a href="#comments">"Comments"</a>
1729 below) can be used.
1730 </P>
1731 <br><b>
1732 Recursive back references
1733 </b><br>
1734 <P>
1735 A back reference that occurs inside the parentheses to which it refers fails
1736 when the subpattern is first used, so, for example, (a\1) never matches.
1737 However, such references can be useful inside repeated subpatterns. For
1738 example, the pattern
1739 <pre>
1740 (a|b\1)+
1741 </pre>
1742 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1743 the subpattern, the back reference matches the character string corresponding
1744 to the previous iteration. In order for this to work, the pattern must be such
1745 that the first iteration does not need to match the back reference. This can be
1746 done using alternation, as in the example above, or by a quantifier with a
1747 minimum of zero.
1748 </P>
1749 <P>
1750 Back references of this type cause the group that they reference to be treated
1751 as an
1752 <a href="#atomicgroup">atomic group.</a>
1753 Once the whole group has been matched, a subsequent matching failure cannot
1754 cause backtracking into the middle of the group.
1755 <a name="bigassertions"></a></P>
1756 <br><a name="SEC18" href="#TOC1">ASSERTIONS</a><br>
1757 <P>
1758 An assertion is a test on the characters following or preceding the current
1759 matching point that does not actually consume any characters. The simple
1760 assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are described
1761 <a href="#smallassertions">above.</a>
1762 </P>
1763 <P>
1764 More complicated assertions are coded as subpatterns. There are two kinds:
1765 those that look ahead of the current position in the subject string, and those
1766 that look behind it. An assertion subpattern is matched in the normal way,
1767 except that it does not cause the current matching position to be changed.
1768 </P>
1769 <P>
1770 Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1771 because it makes no sense to assert the same thing several times. If any kind
1772 of assertion contains capturing subpatterns within it, these are counted for
1773 the purposes of numbering the capturing subpatterns in the whole pattern.
1774 However, substring capturing is carried out only for positive assertions,
1775 because it does not make sense for negative assertions.
1776 </P>
1777 <br><b>
1778 Lookahead assertions
1779 </b><br>
1780 <P>
1781 Lookahead assertions start with (?= for positive assertions and (?! for
1782 negative assertions. For example,
1783 <pre>
1784 \w+(?=;)
1785 </pre>
1786 matches a word followed by a semicolon, but does not include the semicolon in
1787 the match, and
1788 <pre>
1789 foo(?!bar)
1790 </pre>
1791 matches any occurrence of "foo" that is not followed by "bar". Note that the
1792 apparently similar pattern
1793 <pre>
1794 (?!foo)bar
1795 </pre>
1796 does not find an occurrence of "bar" that is preceded by something other than
1797 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1798 (?!foo) is always true when the next three characters are "bar". A
1799 lookbehind assertion is needed to achieve the other effect.
1800 </P>
1801 <P>
1802 If you want to force a matching failure at some point in a pattern, the most
1803 convenient way to do it is with (?!) because an empty string always matches, so
1804 an assertion that requires there not to be an empty string must always fail.
1805 The backtracking control verb (*FAIL) or (*F) is essentially a synonym for
1806 (?!).
1807 <a name="lookbehind"></a></P>
1808 <br><b>
1809 Lookbehind assertions
1810 </b><br>
1811 <P>
1812 Lookbehind assertions start with (?&#60;= for positive assertions and (?&#60;! for
1813 negative assertions. For example,
1814 <pre>
1815 (?&#60;!foo)bar
1816 </pre>
1817 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1818 a lookbehind assertion are restricted such that all the strings it matches must
1819 have a fixed length. However, if there are several top-level alternatives, they
1820 do not all have to have the same fixed length. Thus
1821 <pre>
1822 (?&#60;=bullock|donkey)
1823 </pre>
1824 is permitted, but
1825 <pre>
1826 (?&#60;!dogs?|cats?)
1827 </pre>
1828 causes an error at compile time. Branches that match different length strings
1829 are permitted only at the top level of a lookbehind assertion. This is an
1830 extension compared with Perl, which requires all branches to match the same
1831 length of string. An assertion such as
1832 <pre>
1833 (?&#60;=ab(c|de))
1834 </pre>
1835 is not permitted, because its single top-level branch can match two different
1836 lengths, but it is acceptable to PCRE if rewritten to use two top-level
1837 branches:
1838 <pre>
1839 (?&#60;=abc|abde)
1840 </pre>
1841 In some cases, the escape sequence \K
1842 <a href="#resetmatchstart">(see above)</a>
1843 can be used instead of a lookbehind assertion to get round the fixed-length
1844 restriction.
1845 </P>
1846 <P>
1847 The implementation of lookbehind assertions is, for each alternative, to
1848 temporarily move the current position back by the fixed length and then try to
1849 match. If there are insufficient characters before the current position, the
1850 assertion fails.
1851 </P>
1852 <P>
1853 PCRE does not allow the \C escape (which matches a single byte in UTF-8 mode)
1854 to appear in lookbehind assertions, because it makes it impossible to calculate
1855 the length of the lookbehind. The \X and \R escapes, which can match
1856 different numbers of bytes, are also not permitted.
1857 </P>
1858 <P>
1859 <a href="#subpatternsassubroutines">"Subroutine"</a>
1860 calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
1861 as the subpattern matches a fixed-length string.
1862 <a href="#recursion">Recursion,</a>
1863 however, is not supported.
1864 </P>
1865 <P>
1866 Possessive quantifiers can be used in conjunction with lookbehind assertions to
1867 specify efficient matching of fixed-length strings at the end of subject
1868 strings. Consider a simple pattern such as
1869 <pre>
1870 abcd$
1871 </pre>
1872 when applied to a long string that does not match. Because matching proceeds
1873 from left to right, PCRE will look for each "a" in the subject and then see if
1874 what follows matches the rest of the pattern. If the pattern is specified as
1875 <pre>
1876 ^.*abcd$
1877 </pre>
1878 the initial .* matches the entire string at first, but when this fails (because
1879 there is no following "a"), it backtracks to match all but the last character,
1880 then all but the last two characters, and so on. Once again the search for "a"
1881 covers the entire string, from right to left, so we are no better off. However,
1882 if the pattern is written as
1883 <pre>
1884 ^.*+(?&#60;=abcd)
1885 </pre>
1886 there can be no backtracking for the .*+ item; it can match only the entire
1887 string. The subsequent lookbehind assertion does a single test on the last four
1888 characters. If it fails, the match fails immediately. For long strings, this
1889 approach makes a significant difference to the processing time.
1890 </P>
1891 <br><b>
1892 Using multiple assertions
1893 </b><br>
1894 <P>
1895 Several assertions (of any sort) may occur in succession. For example,
1896 <pre>
1897 (?&#60;=\d{3})(?&#60;!999)foo
1898 </pre>
1899 matches "foo" preceded by three digits that are not "999". Notice that each of
1900 the assertions is applied independently at the same point in the subject
1901 string. First there is a check that the previous three characters are all
1902 digits, and then there is a check that the same three characters are not "999".
1903 This pattern does <i>not</i> match "foo" preceded by six characters, the first
1904 of which are digits and the last three of which are not "999". For example, it
1905 doesn't match "123abcfoo". A pattern to do that is
1906 <pre>
1907 (?&#60;=\d{3}...)(?&#60;!999)foo
1908 </pre>
1909 This time the first assertion looks at the preceding six characters, checking
1910 that the first three are digits, and then the second assertion checks that the
1911 preceding three characters are not "999".
1912 </P>
1913 <P>
1914 Assertions can be nested in any combination. For example,
1915 <pre>
1916 (?&#60;=(?&#60;!foo)bar)baz
1917 </pre>
1918 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1919 preceded by "foo", while
1920 <pre>
1921 (?&#60;=\d{3}(?!999)...)foo
1922 </pre>
1923 is another pattern that matches "foo" preceded by three digits and any three
1924 characters that are not "999".
1925 <a name="conditions"></a></P>
1926 <br><a name="SEC19" href="#TOC1">CONDITIONAL SUBPATTERNS</a><br>
1927 <P>
1928 It is possible to cause the matching process to obey a subpattern
1929 conditionally or to choose between two alternative subpatterns, depending on
1930 the result of an assertion, or whether a specific capturing subpattern has
1931 already been matched. The two possible forms of conditional subpattern are:
1932 <pre>
1933 (?(condition)yes-pattern)
1934 (?(condition)yes-pattern|no-pattern)
1935 </pre>
1936 If the condition is satisfied, the yes-pattern is used; otherwise the
1937 no-pattern (if present) is used. If there are more than two alternatives in the
1938 subpattern, a compile-time error occurs. Each of the two alternatives may
1939 itself contain nested subpatterns of any form, including conditional
1940 subpatterns; the restriction to two alternatives applies only at the level of
1941 the condition. This pattern fragment is an example where the alternatives are
1942 complex:
1943 <pre>
1944 (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
1946 </PRE>
1947 </P>
1948 <P>
1949 There are four kinds of condition: references to subpatterns, references to
1950 recursion, a pseudo-condition called DEFINE, and assertions.
1951 </P>
1952 <br><b>
1953 Checking for a used subpattern by number
1954 </b><br>
1955 <P>
1956 If the text between the parentheses consists of a sequence of digits, the
1957 condition is true if a capturing subpattern of that number has previously
1958 matched. If there is more than one capturing subpattern with the same number
1959 (see the earlier
1960 <a href="#recursion">section about duplicate subpattern numbers),</a>
1961 the condition is true if any of them have been set. An alternative notation is
1962 to precede the digits with a plus or minus sign. In this case, the subpattern
1963 number is relative rather than absolute. The most recently opened parentheses
1964 can be referenced by (?(-1), the next most recent by (?(-2), and so on. In
1965 looping constructs it can also make sense to refer to subsequent groups with
1966 constructs such as (?(+2).
1967 </P>
1968 <P>
1969 Consider the following pattern, which contains non-significant white space to
1970 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
1971 three parts for ease of discussion:
1972 <pre>
1973 ( \( )? [^()]+ (?(1) \) )
1974 </pre>
1975 The first part matches an optional opening parenthesis, and if that
1976 character is present, sets it as the first captured substring. The second part
1977 matches one or more characters that are not parentheses. The third part is a
1978 conditional subpattern that tests whether the first set of parentheses matched
1979 or not. If they did, that is, if subject started with an opening parenthesis,
1980 the condition is true, and so the yes-pattern is executed and a closing
1981 parenthesis is required. Otherwise, since no-pattern is not present, the
1982 subpattern matches nothing. In other words, this pattern matches a sequence of
1983 non-parentheses, optionally enclosed in parentheses.
1984 </P>
1985 <P>
1986 If you were embedding this pattern in a larger one, you could use a relative
1987 reference:
1988 <pre>
1989 ...other stuff... ( \( )? [^()]+ (?(-1) \) ) ...
1990 </pre>
1991 This makes the fragment independent of the parentheses in the larger pattern.
1992 </P>
1993 <br><b>
1994 Checking for a used subpattern by name
1995 </b><br>
1996 <P>
1997 Perl uses the syntax (?(&#60;name&#62;)...) or (?('name')...) to test for a used
1998 subpattern by name. For compatibility with earlier versions of PCRE, which had
1999 this facility before Perl, the syntax (?(name)...) is also recognized. However,
2000 there is a possible ambiguity with this syntax, because subpattern names may
2001 consist entirely of digits. PCRE looks first for a named subpattern; if it
2002 cannot find one and the name consists entirely of digits, PCRE looks for a
2003 subpattern of that number, which must be greater than zero. Using subpattern
2004 names that consist entirely of digits is not recommended.
2005 </P>
2006 <P>
2007 Rewriting the above example to use a named subpattern gives this:
2008 <pre>
2009 (?&#60;OPEN&#62; \( )? [^()]+ (?(&#60;OPEN&#62;) \) )
2010 </pre>
2011 If the name used in a condition of this kind is a duplicate, the test is
2012 applied to all subpatterns of the same name, and is true if any one of them has
2013 matched.
2014 </P>
2015 <br><b>
2016 Checking for pattern recursion
2017 </b><br>
2018 <P>
2019 If the condition is the string (R), and there is no subpattern with the name R,
2020 the condition is true if a recursive call to the whole pattern or any
2021 subpattern has been made. If digits or a name preceded by ampersand follow the
2022 letter R, for example:
2023 <pre>
2024 (?(R3)...) or (?(R&name)...)
2025 </pre>
2026 the condition is true if the most recent recursion is into a subpattern whose
2027 number or name is given. This condition does not check the entire recursion
2028 stack. If the name used in a condition of this kind is a duplicate, the test is
2029 applied to all subpatterns of the same name, and is true if any one of them is
2030 the most recent recursion.
2031 </P>
2032 <P>
2033 At "top level", all these recursion test conditions are false.
2034 <a href="#recursion">The syntax for recursive patterns</a>
2035 is described below.
2036 <a name="subdefine"></a></P>
2037 <br><b>
2038 Defining subpatterns for use by reference only
2039 </b><br>
2040 <P>
2041 If the condition is the string (DEFINE), and there is no subpattern with the
2042 name DEFINE, the condition is always false. In this case, there may be only one
2043 alternative in the subpattern. It is always skipped if control reaches this
2044 point in the pattern; the idea of DEFINE is that it can be used to define
2045 "subroutines" that can be referenced from elsewhere. (The use of
2046 <a href="#subpatternsassubroutines">"subroutines"</a>
2047 is described below.) For example, a pattern to match an IPv4 address could be
2048 written like this (ignore whitespace and line breaks):
2049 <pre>
2050 (?(DEFINE) (?&#60;byte&#62; 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) )
2051 \b (?&byte) (\.(?&byte)){3} \b
2052 </pre>
2053 The first part of the pattern is a DEFINE group inside which a another group
2054 named "byte" is defined. This matches an individual component of an IPv4
2055 address (a number less than 256). When matching takes place, this part of the
2056 pattern is skipped because DEFINE acts like a false condition. The rest of the
2057 pattern uses references to the named group to match the four dot-separated
2058 components of an IPv4 address, insisting on a word boundary at each end.
2059 </P>
2060 <br><b>
2061 Assertion conditions
2062 </b><br>
2063 <P>
2064 If the condition is not in any of the above formats, it must be an assertion.
2065 This may be a positive or negative lookahead or lookbehind assertion. Consider
2066 this pattern, again containing non-significant white space, and with the two
2067 alternatives on the second line:
2068 <pre>
2069 (?(?=[^a-z]*[a-z])
2070 \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )
2071 </pre>
2072 The condition is a positive lookahead assertion that matches an optional
2073 sequence of non-letters followed by a letter. In other words, it tests for the
2074 presence of at least one letter in the subject. If a letter is found, the
2075 subject is matched against the first alternative; otherwise it is matched
2076 against the second. This pattern matches strings in one of the two forms
2077 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2078 <a name="comments"></a></P>
2079 <br><a name="SEC20" href="#TOC1">COMMENTS</a><br>
2080 <P>
2081 There are two ways of including comments in patterns that are processed by
2082 PCRE. In both cases, the start of the comment must not be in a character class,
2083 nor in the middle of any other sequence of related characters such as (?: or a
2084 subpattern name or number. The characters that make up a comment play no part
2085 in the pattern matching.
2086 </P>
2087 <P>
2088 The sequence (?# marks the start of a comment that continues up to the next
2089 closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2090 option is set, an unescaped # character also introduces a comment, which in
2091 this case continues to immediately after the next newline character or
2092 character sequence in the pattern. Which characters are interpreted as newlines
2093 is controlled by the options passed to <b>pcre_compile()</b> or by a special
2094 sequence at the start of the pattern, as described in the section entitled
2095 <a href="#newlines">"Newline conventions"</a>
2096 above. Note that the end of this type of comment is a literal newline sequence
2097 in the pattern; escape sequences that happen to represent a newline do not
2098 count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2099 default newline convention is in force:
2100 <pre>
2101 abc #comment \n still comment
2102 </pre>
2103 On encountering the # character, <b>pcre_compile()</b> skips along, looking for
2104 a newline in the pattern. The sequence \n is still literal at this stage, so
2105 it does not terminate the comment. Only an actual character with the code value
2106 0x0a (the default newline) does so.
2107 <a name="recursion"></a></P>
2108 <br><a name="SEC21" href="#TOC1">RECURSIVE PATTERNS</a><br>
2109 <P>
2110 Consider the problem of matching a string in parentheses, allowing for
2111 unlimited nested parentheses. Without the use of recursion, the best that can
2112 be done is to use a pattern that matches up to some fixed depth of nesting. It
2113 is not possible to handle an arbitrary nesting depth.
2114 </P>
2115 <P>
2116 For some time, Perl has provided a facility that allows regular expressions to
2117 recurse (amongst other things). It does this by interpolating Perl code in the
2118 expression at run time, and the code can refer to the expression itself. A Perl
2119 pattern using code interpolation to solve the parentheses problem can be
2120 created like this:
2121 <pre>
2122 $re = qr{\( (?: (?&#62;[^()]+) | (?p{$re}) )* \)}x;
2123 </pre>
2124 The (?p{...}) item interpolates Perl code at run time, and in this case refers
2125 recursively to the pattern in which it appears.
2126 </P>
2127 <P>
2128 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2129 supports special syntax for recursion of the entire pattern, and also for
2130 individual subpattern recursion. After its introduction in PCRE and Python,
2131 this kind of recursion was subsequently introduced into Perl at release 5.10.
2132 </P>
2133 <P>
2134 A special item that consists of (? followed by a number greater than zero and a
2135 closing parenthesis is a recursive call of the subpattern of the given number,
2136 provided that it occurs inside that subpattern. (If not, it is a
2137 <a href="#subpatternsassubroutines">"subroutine"</a>
2138 call, which is described in the next section.) The special item (?R) or (?0) is
2139 a recursive call of the entire regular expression.
2140 </P>
2141 <P>
2142 This PCRE pattern solves the nested parentheses problem (assume the
2143 PCRE_EXTENDED option is set so that white space is ignored):
2144 <pre>
2145 \( ( [^()]++ | (?R) )* \)
2146 </pre>
2147 First it matches an opening parenthesis. Then it matches any number of
2148 substrings which can either be a sequence of non-parentheses, or a recursive
2149 match of the pattern itself (that is, a correctly parenthesized substring).
2150 Finally there is a closing parenthesis. Note the use of a possessive quantifier
2151 to avoid backtracking into sequences of non-parentheses.
2152 </P>
2153 <P>
2154 If this were part of a larger pattern, you would not want to recurse the entire
2155 pattern, so instead you could use this:
2156 <pre>
2157 ( \( ( [^()]++ | (?1) )* \) )
2158 </pre>
2159 We have put the pattern into parentheses, and caused the recursion to refer to
2160 them instead of the whole pattern.
2161 </P>
2162 <P>
2163 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2164 is made easier by the use of relative references. Instead of (?1) in the
2165 pattern above you can write (?-2) to refer to the second most recently opened
2166 parentheses preceding the recursion. In other words, a negative number counts
2167 capturing parentheses leftwards from the point at which it is encountered.
2168 </P>
2169 <P>
2170 It is also possible to refer to subsequently opened parentheses, by writing
2171 references such as (?+2). However, these cannot be recursive because the
2172 reference is not inside the parentheses that are referenced. They are always
2173 <a href="#subpatternsassubroutines">"subroutine"</a>
2174 calls, as described in the next section.
2175 </P>
2176 <P>
2177 An alternative approach is to use named parentheses instead. The Perl syntax
2178 for this is (?&name); PCRE's earlier syntax (?P&#62;name) is also supported. We
2179 could rewrite the above example as follows:
2180 <pre>
2181 (?&#60;pn&#62; \( ( [^()]++ | (?&pn) )* \) )
2182 </pre>
2183 If there is more than one subpattern with the same name, the earliest one is
2184 used.
2185 </P>
2186 <P>
2187 This particular example pattern that we have been looking at contains nested
2188 unlimited repeats, and so the use of a possessive quantifier for matching
2189 strings of non-parentheses is important when applying the pattern to strings
2190 that do not match. For example, when this pattern is applied to
2191 <pre>
2192 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2193 </pre>
2194 it yields "no match" quickly. However, if a possessive quantifier is not used,
2195 the match runs for a very long time indeed because there are so many different
2196 ways the + and * repeats can carve up the subject, and all have to be tested
2197 before failure can be reported.
2198 </P>
2199 <P>
2200 At the end of a match, the values of capturing parentheses are those from
2201 the outermost level. If you want to obtain intermediate values, a callout
2202 function can be used (see below and the
2203 <a href="pcrecallout.html"><b>pcrecallout</b></a>
2204 documentation). If the pattern above is matched against
2205 <pre>
2206 (ab(cd)ef)
2207 </pre>
2208 the value for the inner capturing parentheses (numbered 2) is "ef", which is
2209 the last value taken on at the top level. If a capturing subpattern is not
2210 matched at the top level, its final value is unset, even if it is (temporarily)
2211 set at a deeper level.
2212 </P>
2213 <P>
2214 If there are more than 15 capturing parentheses in a pattern, PCRE has to
2215 obtain extra memory to store data during a recursion, which it does by using
2216 <b>pcre_malloc</b>, freeing it via <b>pcre_free</b> afterwards. If no memory can
2217 be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2218 </P>
2219 <P>
2220 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2221 Consider this pattern, which matches text in angle brackets, allowing for
2222 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2223 recursing), whereas any characters are permitted at the outer level.
2224 <pre>
2225 &#60; (?: (?(R) \d++ | [^&#60;&#62;]*+) | (?R)) * &#62;
2226 </pre>
2227 In this pattern, (?(R) is the start of a conditional subpattern, with two
2228 different alternatives for the recursive and non-recursive cases. The (?R) item
2229 is the actual recursive call.
2230 <a name="recursiondifference"></a></P>
2231 <br><b>
2232 Recursion difference from Perl
2233 </b><br>
2234 <P>
2235 In PCRE (like Python, but unlike Perl), a recursive subpattern call is always
2236 treated as an atomic group. That is, once it has matched some of the subject
2237 string, it is never re-entered, even if it contains untried alternatives and
2238 there is a subsequent matching failure. This can be illustrated by the
2239 following pattern, which purports to match a palindromic string that contains
2240 an odd number of characters (for example, "a", "aba", "abcba", "abcdcba"):
2241 <pre>
2242 ^(.|(.)(?1)\2)$
2243 </pre>
2244 The idea is that it either matches a single character, or two identical
2245 characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2246 it does not if the pattern is longer than three characters. Consider the
2247 subject string "abcba":
2248 </P>
2249 <P>
2250 At the top level, the first character is matched, but as it is not at the end
2251 of the string, the first alternative fails; the second alternative is taken
2252 and the recursion kicks in. The recursive call to subpattern 1 successfully
2253 matches the next character ("b"). (Note that the beginning and end of line
2254 tests are not part of the recursion).
2255 </P>
2256 <P>
2257 Back at the top level, the next character ("c") is compared with what
2258 subpattern 2 matched, which was "a". This fails. Because the recursion is
2259 treated as an atomic group, there are now no backtracking points, and so the
2260 entire match fails. (Perl is able, at this point, to re-enter the recursion and
2261 try the second alternative.) However, if the pattern is written with the
2262 alternatives in the other order, things are different:
2263 <pre>
2264 ^((.)(?1)\2|.)$
2265 </pre>
2266 This time, the recursing alternative is tried first, and continues to recurse
2267 until it runs out of characters, at which point the recursion fails. But this
2268 time we do have another alternative to try at the higher level. That is the big
2269 difference: in the previous case the remaining alternative is at a deeper
2270 recursion level, which PCRE cannot use.
2271 </P>
2272 <P>
2273 To change the pattern so that matches all palindromic strings, not just those
2274 with an odd number of characters, it is tempting to change the pattern to this:
2275 <pre>
2276 ^((.)(?1)\2|.?)$
2277 </pre>
2278 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2279 deeper recursion has matched a single character, it cannot be entered again in
2280 order to match an empty string. The solution is to separate the two cases, and
2281 write out the odd and even cases as alternatives at the higher level:
2282 <pre>
2283 ^(?:((.)(?1)\2|)|((.)(?3)\4|.))
2284 </pre>
2285 If you want to match typical palindromic phrases, the pattern has to ignore all
2286 non-word characters, which can be done like this:
2287 <pre>
2288 ^\W*+(?:((.)\W*+(?1)\W*+\2|)|((.)\W*+(?3)\W*+\4|\W*+.\W*+))\W*+$
2289 </pre>
2290 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2291 man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2292 the use of the possessive quantifier *+ to avoid backtracking into sequences of
2293 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2294 more) to match typical phrases, and Perl takes so long that you think it has
2295 gone into a loop.
2296 </P>
2297 <P>
2298 <b>WARNING</b>: The palindrome-matching patterns above work only if the subject
2299 string does not start with a palindrome that is shorter than the entire string.
2300 For example, although "abcba" is correctly matched, if the subject is "ababa",
2301 PCRE finds the palindrome "aba" at the start, then fails at top level because
2302 the end of the string does not follow. Once again, it cannot jump back into the
2303 recursion to try other alternatives, so the entire match fails.
2304 <a name="subpatternsassubroutines"></a></P>
2305 <br><a name="SEC22" href="#TOC1">SUBPATTERNS AS SUBROUTINES</a><br>
2306 <P>
2307 If the syntax for a recursive subpattern reference (either by number or by
2308 name) is used outside the parentheses to which it refers, it operates like a
2309 subroutine in a programming language. The "called" subpattern may be defined
2310 before or after the reference. A numbered reference can be absolute or
2311 relative, as in these examples:
2312 <pre>
2313 (...(absolute)...)...(?2)...
2314 (...(relative)...)...(?-1)...
2315 (...(?+1)...(relative)...
2316 </pre>
2317 An earlier example pointed out that the pattern
2318 <pre>
2319 (sens|respons)e and \1ibility
2320 </pre>
2321 matches "sense and sensibility" and "response and responsibility", but not
2322 "sense and responsibility". If instead the pattern
2323 <pre>
2324 (sens|respons)e and (?1)ibility
2325 </pre>
2326 is used, it does match "sense and responsibility" as well as the other two
2327 strings. Another example is given in the discussion of DEFINE above.
2328 </P>
2329 <P>
2330 Like recursive subpatterns, a subroutine call is always treated as an atomic
2331 group. That is, once it has matched some of the subject string, it is never
2332 re-entered, even if it contains untried alternatives and there is a subsequent
2333 matching failure. Any capturing parentheses that are set during the subroutine
2334 call revert to their previous values afterwards.
2335 </P>
2336 <P>
2337 When a subpattern is used as a subroutine, processing options such as
2338 case-independence are fixed when the subpattern is defined. They cannot be
2339 changed for different calls. For example, consider this pattern:
2340 <pre>
2341 (abc)(?i:(?-1))
2342 </pre>
2343 It matches "abcabc". It does not match "abcABC" because the change of
2344 processing option does not affect the called subpattern.
2345 <a name="onigurumasubroutines"></a></P>
2346 <br><a name="SEC23" href="#TOC1">ONIGURUMA SUBROUTINE SYNTAX</a><br>
2347 <P>
2348 For compatibility with Oniguruma, the non-Perl syntax \g followed by a name or
2349 a number enclosed either in angle brackets or single quotes, is an alternative
2350 syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2351 are two of the examples used above, rewritten using this syntax:
2352 <pre>
2353 (?&#60;pn&#62; \( ( (?&#62;[^()]+) | \g&#60;pn&#62; )* \) )
2354 (sens|respons)e and \g'1'ibility
2355 </pre>
2356 PCRE supports an extension to Oniguruma: if a number is preceded by a
2357 plus or a minus sign it is taken as a relative reference. For example:
2358 <pre>
2359 (abc)(?i:\g&#60;-1&#62;)
2360 </pre>
2361 Note that \g{...} (Perl syntax) and \g&#60;...&#62; (Oniguruma syntax) are <i>not</i>
2362 synonymous. The former is a back reference; the latter is a subroutine call.
2363 </P>
2364 <br><a name="SEC24" href="#TOC1">CALLOUTS</a><br>
2365 <P>
2366 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2367 code to be obeyed in the middle of matching a regular expression. This makes it
2368 possible, amongst other things, to extract different substrings that match the
2369 same pair of parentheses when there is a repetition.
2370 </P>
2371 <P>
2372 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2373 code. The feature is called "callout". The caller of PCRE provides an external
2374 function by putting its entry point in the global variable <i>pcre_callout</i>.
2375 By default, this variable contains NULL, which disables all calling out.
2376 </P>
2377 <P>
2378 Within a regular expression, (?C) indicates the points at which the external
2379 function is to be called. If you want to identify different callout points, you
2380 can put a number less than 256 after the letter C. The default value is zero.
2381 For example, this pattern has two callout points:
2382 <pre>
2383 (?C1)abc(?C2)def
2384 </pre>
2385 If the PCRE_AUTO_CALLOUT flag is passed to <b>pcre_compile()</b>, callouts are
2386 automatically installed before each item in the pattern. They are all numbered
2387 255.
2388 </P>
2389 <P>
2390 During matching, when PCRE reaches a callout point (and <i>pcre_callout</i> is
2391 set), the external function is called. It is provided with the number of the
2392 callout, the position in the pattern, and, optionally, one item of data
2393 originally supplied by the caller of <b>pcre_exec()</b>. The callout function
2394 may cause matching to proceed, to backtrack, or to fail altogether. A complete
2395 description of the interface to the callout function is given in the
2396 <a href="pcrecallout.html"><b>pcrecallout</b></a>
2397 documentation.
2398 <a name="backtrackcontrol"></a></P>
2399 <br><a name="SEC25" href="#TOC1">BACKTRACKING CONTROL</a><br>
2400 <P>
2401 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2402 are described in the Perl documentation as "experimental and subject to change
2403 or removal in a future version of Perl". It goes on to say: "Their usage in
2404 production code should be noted to avoid problems during upgrades." The same
2405 remarks apply to the PCRE features described in this section.
2406 </P>
2407 <P>
2408 Since these verbs are specifically related to backtracking, most of them can be
2409 used only when the pattern is to be matched using <b>pcre_exec()</b>, which uses
2410 a backtracking algorithm. With the exception of (*FAIL), which behaves like a
2411 failing negative assertion, they cause an error if encountered by
2412 <b>pcre_dfa_exec()</b>.
2413 </P>
2414 <P>
2415 If any of these verbs are used in an assertion or subroutine subpattern
2416 (including recursive subpatterns), their effect is confined to that subpattern;
2417 it does not extend to the surrounding pattern. Note that such subpatterns are
2418 processed as anchored at the point where they are tested.
2419 </P>
2420 <P>
2421 The new verbs make use of what was previously invalid syntax: an opening
2422 parenthesis followed by an asterisk. They are generally of the form
2423 (*VERB) or (*VERB:NAME). Some may take either form, with differing behaviour,
2424 depending on whether or not an argument is present. An name is a sequence of
2425 letters, digits, and underscores. If the name is empty, that is, if the closing
2426 parenthesis immediately follows the colon, the effect is as if the colon were
2427 not there. Any number of these verbs may occur in a pattern.
2428 </P>
2429 <P>
2430 PCRE contains some optimizations that are used to speed up matching by running
2431 some checks at the start of each match attempt. For example, it may know the
2432 minimum length of matching subject, or that a particular character must be
2433 present. When one of these optimizations suppresses the running of a match, any
2434 included backtracking verbs will not, of course, be processed. You can suppress
2435 the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2436 when calling <b>pcre_exec()</b>.
2437 </P>
2438 <br><b>
2439 Verbs that act immediately
2440 </b><br>
2441 <P>
2442 The following verbs act as soon as they are encountered. They may not be
2443 followed by a name.
2444 <pre>
2445 (*ACCEPT)
2446 </pre>
2447 This verb causes the match to end successfully, skipping the remainder of the
2448 pattern. When inside a recursion, only the innermost pattern is ended
2449 immediately. If (*ACCEPT) is inside capturing parentheses, the data so far is
2450 captured. (This feature was added to PCRE at release 8.00.) For example:
2451 <pre>
2452 A((?:A|B(*ACCEPT)|C)D)
2453 </pre>
2454 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2455 the outer parentheses.
2456 <pre>
2457 (*FAIL) or (*F)
2458 </pre>
2459 This verb causes the match to fail, forcing backtracking to occur. It is
2460 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2461 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2462 Perl features that are not present in PCRE. The nearest equivalent is the
2463 callout feature, as for example in this pattern:
2464 <pre>
2465 a+(?C)(*FAIL)
2466 </pre>
2467 A match with the string "aaaa" always fails, but the callout is taken before
2468 each backtrack happens (in this example, 10 times).
2469 </P>
2470 <br><b>
2471 Recording which path was taken
2472 </b><br>
2473 <P>
2474 There is one verb whose main purpose is to track how a match was arrived at,
2475 though it also has a secondary use in conjunction with advancing the match
2476 starting point (see (*SKIP) below).
2477 <pre>
2478 (*MARK:NAME) or (*:NAME)
2479 </pre>
2480 A name is always required with this verb. There may be as many instances of
2481 (*MARK) as you like in a pattern, and their names do not have to be unique.
2482 </P>
2483 <P>
2484 When a match succeeds, the name of the last-encountered (*MARK) is passed back
2485 to the caller via the <i>pcre_extra</i> data structure, as described in the
2486 <a href="pcreapi.html#extradata">section on <i>pcre_extra</i></a>
2487 in the
2488 <a href="pcreapi.html"><b>pcreapi</b></a>
2489 documentation. No data is returned for a partial match. Here is an example of
2490 <b>pcretest</b> output, where the /K modifier requests the retrieval and
2491 outputting of (*MARK) data:
2492 <pre>
2493 /X(*MARK:A)Y|X(*MARK:B)Z/K
2494 XY
2495 0: XY
2496 MK: A
2497 XZ
2498 0: XZ
2499 MK: B
2500 </pre>
2501 The (*MARK) name is tagged with "MK:" in this output, and in this example it
2502 indicates which of the two alternatives matched. This is a more efficient way
2503 of obtaining this information than putting each alternative in its own
2504 capturing parentheses.
2505 </P>
2506 <P>
2507 A name may also be returned after a failed match if the final path through the
2508 pattern involves (*MARK). However, unless (*MARK) used in conjunction with
2509 (*COMMIT), this is unlikely to happen for an unanchored pattern because, as the
2510 starting point for matching is advanced, the final check is often with an empty
2511 string, causing a failure before (*MARK) is reached. For example:
2512 <pre>
2513 /X(*MARK:A)Y|X(*MARK:B)Z/K
2514 XP
2515 No match
2516 </pre>
2517 There are three potential starting points for this match (starting with X,
2518 starting with P, and with an empty string). If the pattern is anchored, the
2519 result is different:
2520 <pre>
2521 /^X(*MARK:A)Y|^X(*MARK:B)Z/K
2522 XP
2523 No match, mark = B
2524 </pre>
2525 PCRE's start-of-match optimizations can also interfere with this. For example,
2526 if, as a result of a call to <b>pcre_study()</b>, it knows the minimum
2527 subject length for a match, a shorter subject will not be scanned at all.
2528 </P>
2529 <P>
2530 Note that similar anomalies (though different in detail) exist in Perl, no
2531 doubt for the same reasons. The use of (*MARK) data after a failed match of an
2532 unanchored pattern is not recommended, unless (*COMMIT) is involved.
2533 </P>
2534 <br><b>
2535 Verbs that act after backtracking
2536 </b><br>
2537 <P>
2538 The following verbs do nothing when they are encountered. Matching continues
2539 with what follows, but if there is no subsequent match, causing a backtrack to
2540 the verb, a failure is forced. That is, backtracking cannot pass to the left of
2541 the verb. However, when one of these verbs appears inside an atomic group, its
2542 effect is confined to that group, because once the group has been matched,
2543 there is never any backtracking into it. In this situation, backtracking can
2544 "jump back" to the left of the entire atomic group. (Remember also, as stated
2545 above, that this localization also applies in subroutine calls and assertions.)
2546 </P>
2547 <P>
2548 These verbs differ in exactly what kind of failure occurs when backtracking
2549 reaches them.
2550 <pre>
2551 (*COMMIT)
2552 </pre>
2553 This verb, which may not be followed by a name, causes the whole match to fail
2554 outright if the rest of the pattern does not match. Even if the pattern is
2555 unanchored, no further attempts to find a match by advancing the starting point
2556 take place. Once (*COMMIT) has been passed, <b>pcre_exec()</b> is committed to
2557 finding a match at the current starting point, or not at all. For example:
2558 <pre>
2559 a+(*COMMIT)b
2560 </pre>
2561 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2562 dynamic anchor, or "I've started, so I must finish." The name of the most
2563 recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
2564 match failure.
2565 </P>
2566 <P>
2567 Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
2568 unless PCRE's start-of-match optimizations are turned off, as shown in this
2569 <b>pcretest</b> example:
2570 <pre>
2571 /(*COMMIT)abc/
2572 xyzabc
2573 0: abc
2574 xyzabc\Y
2575 No match
2576 </pre>
2577 PCRE knows that any match must start with "a", so the optimization skips along
2578 the subject to "a" before running the first match attempt, which succeeds. When
2579 the optimization is disabled by the \Y escape in the second subject, the match
2580 starts at "x" and so the (*COMMIT) causes it to fail without trying any other
2581 starting points.
2582 <pre>
2583 (*PRUNE) or (*PRUNE:NAME)
2584 </pre>
2585 This verb causes the match to fail at the current starting position in the
2586 subject if the rest of the pattern does not match. If the pattern is
2587 unanchored, the normal "bumpalong" advance to the next starting character then
2588 happens. Backtracking can occur as usual to the left of (*PRUNE), before it is
2589 reached, or when matching to the right of (*PRUNE), but if there is no match to
2590 the right, backtracking cannot cross (*PRUNE). In simple cases, the use of
2591 (*PRUNE) is just an alternative to an atomic group or possessive quantifier,
2592 but there are some uses of (*PRUNE) that cannot be expressed in any other way.
2593 The behaviour of (*PRUNE:NAME) is the same as (*MARK:NAME)(*PRUNE) when the
2594 match fails completely; the name is passed back if this is the final attempt.
2595 (*PRUNE:NAME) does not pass back a name if the match succeeds. In an anchored
2596 pattern (*PRUNE) has the same effect as (*COMMIT).
2597 <pre>
2598 (*SKIP)
2599 </pre>
2600 This verb, when given without a name, is like (*PRUNE), except that if the
2601 pattern is unanchored, the "bumpalong" advance is not to the next character,
2602 but to the position in the subject where (*SKIP) was encountered. (*SKIP)
2603 signifies that whatever text was matched leading up to it cannot be part of a
2604 successful match. Consider:
2605 <pre>
2606 a+(*SKIP)b
2607 </pre>
2608 If the subject is "aaaac...", after the first match attempt fails (starting at
2609 the first character in the string), the starting point skips on to start the
2610 next attempt at "c". Note that a possessive quantifer does not have the same
2611 effect as this example; although it would suppress backtracking during the
2612 first match attempt, the second attempt would start at the second character
2613 instead of skipping on to "c".
2614 <pre>
2615 (*SKIP:NAME)
2616 </pre>
2617 When (*SKIP) has an associated name, its behaviour is modified. If the
2618 following pattern fails to match, the previous path through the pattern is
2619 searched for the most recent (*MARK) that has the same name. If one is found,
2620 the "bumpalong" advance is to the subject position that corresponds to that
2621 (*MARK) instead of to where (*SKIP) was encountered. If no (*MARK) with a
2622 matching name is found, normal "bumpalong" of one character happens (the
2623 (*SKIP) is ignored).
2624 <pre>
2625 (*THEN) or (*THEN:NAME)
2626 </pre>
2627 This verb causes a skip to the next alternation in the innermost enclosing
2628 group if the rest of the pattern does not match. That is, it cancels pending
2629 backtracking, but only within the current alternation. Its name comes from the
2630 observation that it can be used for a pattern-based if-then-else block:
2631 <pre>
2632 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2633 </pre>
2634 If the COND1 pattern matches, FOO is tried (and possibly further items after
2635 the end of the group if FOO succeeds); on failure the matcher skips to the
2636 second alternative and tries COND2, without backtracking into COND1. The
2637 behaviour of (*THEN:NAME) is exactly the same as (*MARK:NAME)(*THEN) if the
2638 overall match fails. If (*THEN) is not directly inside an alternation, it acts
2639 like (*PRUNE).
2640 </P>
2641 <P>
2642 The above verbs provide four different "strengths" of control when subsequent
2643 matching fails. (*THEN) is the weakest, carrying on the match at the next
2644 alternation. (*PRUNE) comes next, failing the match at the current starting
2645 position, but allowing an advance to the next character (for an unanchored
2646 pattern). (*SKIP) is similar, except that the advance may be more than one
2647 character. (*COMMIT) is the strongest, causing the entire match to fail.
2648 </P>
2649 <P>
2650 If more than one is present in a pattern, the "stongest" one wins. For example,
2651 consider this pattern, where A, B, etc. are complex pattern fragments:
2652 <pre>
2653 (A(*COMMIT)B(*THEN)C|D)
2654 </pre>
2655 Once A has matched, PCRE is committed to this match, at the current starting
2656 position. If subsequently B matches, but C does not, the normal (*THEN) action
2657 of trying the next alternation (that is, D) does not happen because (*COMMIT)
2658 overrides.
2659 </P>
2660 <br><a name="SEC26" href="#TOC1">SEE ALSO</a><br>
2661 <P>
2662 <b>pcreapi</b>(3), <b>pcrecallout</b>(3), <b>pcrematching</b>(3),
2663 <b>pcresyntax</b>(3), <b>pcre</b>(3).
2664 </P>
2665 <br><a name="SEC27" href="#TOC1">AUTHOR</a><br>
2666 <P>
2667 Philip Hazel
2668 <br>
2669 University Computing Service
2670 <br>
2671 Cambridge CB2 3QH, England.
2672 <br>
2673 </P>
2674 <br><a name="SEC28" href="#TOC1">REVISION</a><br>
2675 <P>
2676 Last updated: 17 November 2010
2677 <br>
2678 Copyright &copy; 1997-2010 University of Cambridge.
2679 <br>
2680 <p>
2681 Return to the <a href="index.html">PCRE index page</a>.
2682 </p>


Name Value
svn:eol-style native
svn:keywords "Author Date Id Revision Url"

  ViewVC Help
Powered by ViewVC 1.1.5