ViewVC logotype

Contents of /code/trunk/doc/pcrepattern.3

Parent Directory Parent Directory | Revision Log Revision Log

Revision 607 - (show annotations)
Sun Jun 12 15:09:49 2011 UTC (9 years, 8 months ago) by ph10
File size: 114129 byte(s)
Error occurred while calculating annotation data.
Document that isolated \Q continues till the end of the pattern.
3 PCRE - Perl-compatible regular expressions
5 .rs
6 .sp
7 The syntax and semantics of the regular expressions that are supported by PCRE
8 are described in detail below. There is a quick-reference syntax summary in the
9 .\" HREF
10 \fBpcresyntax\fP
11 .\"
12 page. PCRE tries to match Perl syntax and semantics as closely as it can. PCRE
13 also supports some alternative regular expression syntax (which does not
14 conflict with the Perl syntax) in order to provide some compatibility with
15 regular expressions in Python, .NET, and Oniguruma.
16 .P
17 Perl's regular expressions are described in its own documentation, and
18 regular expressions in general are covered in a number of books, some of which
19 have copious examples. Jeffrey Friedl's "Mastering Regular Expressions",
20 published by O'Reilly, covers regular expressions in great detail. This
21 description of PCRE's regular expressions is intended as reference material.
22 .P
23 The original operation of PCRE was on strings of one-byte characters. However,
24 there is now also support for UTF-8 character strings. To use this,
25 PCRE must be built to include UTF-8 support, and you must call
26 \fBpcre_compile()\fP or \fBpcre_compile2()\fP with the PCRE_UTF8 option. There
27 is also a special sequence that can be given at the start of a pattern:
28 .sp
29 (*UTF8)
30 .sp
31 Starting a pattern with this sequence is equivalent to setting the PCRE_UTF8
32 option. This feature is not Perl-compatible. How setting UTF-8 mode affects
33 pattern matching is mentioned in several places below. There is also a summary
34 of UTF-8 features in the
35 .\" HTML <a href="pcre.html#utf8support">
36 .\" </a>
37 section on UTF-8 support
38 .\"
39 in the main
40 .\" HREF
41 \fBpcre\fP
42 .\"
43 page.
44 .P
45 Another special sequence that may appear at the start of a pattern or in
46 combination with (*UTF8) is:
47 .sp
48 (*UCP)
49 .sp
50 This has the same effect as setting the PCRE_UCP option: it causes sequences
51 such as \ed and \ew to use Unicode properties to determine character types,
52 instead of recognizing only characters with codes less than 128 via a lookup
53 table.
54 .P
55 If a pattern starts with (*NO_START_OPT), it has the same effect as setting the
56 PCRE_NO_START_OPTIMIZE option either at compile or matching time. There are
57 also some more of these special sequences that are concerned with the handling
58 of newlines; they are described below.
59 .P
60 The remainder of this document discusses the patterns that are supported by
61 PCRE when its main matching function, \fBpcre_exec()\fP, is used.
62 From release 6.0, PCRE offers a second matching function,
63 \fBpcre_dfa_exec()\fP, which matches using a different algorithm that is not
64 Perl-compatible. Some of the features discussed below are not available when
65 \fBpcre_dfa_exec()\fP is used. The advantages and disadvantages of the
66 alternative function, and how it differs from the normal function, are
67 discussed in the
68 .\" HREF
69 \fBpcrematching\fP
70 .\"
71 page.
72 .
73 .
74 .\" HTML <a name="newlines"></a>
76 .rs
77 .sp
78 PCRE supports five different conventions for indicating line breaks in
79 strings: a single CR (carriage return) character, a single LF (linefeed)
80 character, the two-character sequence CRLF, any of the three preceding, or any
81 Unicode newline sequence. The
82 .\" HREF
83 \fBpcreapi\fP
84 .\"
85 page has
86 .\" HTML <a href="pcreapi.html#newlines">
87 .\" </a>
88 further discussion
89 .\"
90 about newlines, and shows how to set the newline convention in the
91 \fIoptions\fP arguments for the compiling and matching functions.
92 .P
93 It is also possible to specify a newline convention by starting a pattern
94 string with one of the following five sequences:
95 .sp
96 (*CR) carriage return
97 (*LF) linefeed
98 (*CRLF) carriage return, followed by linefeed
99 (*ANYCRLF) any of the three above
100 (*ANY) all Unicode newline sequences
101 .sp
102 These override the default and the options given to \fBpcre_compile()\fP or
103 \fBpcre_compile2()\fP. For example, on a Unix system where LF is the default
104 newline sequence, the pattern
105 .sp
106 (*CR)a.b
107 .sp
108 changes the convention to CR. That pattern matches "a\enb" because LF is no
109 longer a newline. Note that these special settings, which are not
110 Perl-compatible, are recognized only at the very start of a pattern, and that
111 they must be in upper case. If more than one of them is present, the last one
112 is used.
113 .P
114 The newline convention affects the interpretation of the dot metacharacter when
115 PCRE_DOTALL is not set, and also the behaviour of \eN. However, it does not
116 affect what the \eR escape sequence matches. By default, this is any Unicode
117 newline sequence, for Perl compatibility. However, this can be changed; see the
118 description of \eR in the section entitled
119 .\" HTML <a href="#newlineseq">
120 .\" </a>
121 "Newline sequences"
122 .\"
123 below. A change of \eR setting can be combined with a change of newline
124 convention.
125 .
126 .
128 .rs
129 .sp
130 A regular expression is a pattern that is matched against a subject string from
131 left to right. Most characters stand for themselves in a pattern, and match the
132 corresponding characters in the subject. As a trivial example, the pattern
133 .sp
134 The quick brown fox
135 .sp
136 matches a portion of a subject string that is identical to itself. When
137 caseless matching is specified (the PCRE_CASELESS option), letters are matched
138 independently of case. In UTF-8 mode, PCRE always understands the concept of
139 case for characters whose values are less than 128, so caseless matching is
140 always possible. For characters with higher values, the concept of case is
141 supported if PCRE is compiled with Unicode property support, but not otherwise.
142 If you want to use caseless matching for characters 128 and above, you must
143 ensure that PCRE is compiled with Unicode property support as well as with
144 UTF-8 support.
145 .P
146 The power of regular expressions comes from the ability to include alternatives
147 and repetitions in the pattern. These are encoded in the pattern by the use of
148 \fImetacharacters\fP, which do not stand for themselves but instead are
149 interpreted in some special way.
150 .P
151 There are two different sets of metacharacters: those that are recognized
152 anywhere in the pattern except within square brackets, and those that are
153 recognized within square brackets. Outside square brackets, the metacharacters
154 are as follows:
155 .sp
156 \e general escape character with several uses
157 ^ assert start of string (or line, in multiline mode)
158 $ assert end of string (or line, in multiline mode)
159 . match any character except newline (by default)
160 [ start character class definition
161 | start of alternative branch
162 ( start subpattern
163 ) end subpattern
164 ? extends the meaning of (
165 also 0 or 1 quantifier
166 also quantifier minimizer
167 * 0 or more quantifier
168 + 1 or more quantifier
169 also "possessive quantifier"
170 { start min/max quantifier
171 .sp
172 Part of a pattern that is in square brackets is called a "character class". In
173 a character class the only metacharacters are:
174 .sp
175 \e general escape character
176 ^ negate the class, but only if the first character
177 - indicates character range
178 .\" JOIN
179 [ POSIX character class (only if followed by POSIX
180 syntax)
181 ] terminates the character class
182 .sp
183 The following sections describe the use of each of the metacharacters.
184 .
185 .
187 .rs
188 .sp
189 The backslash character has several uses. Firstly, if it is followed by a
190 character that is not a number or a letter, it takes away any special meaning
191 that character may have. This use of backslash as an escape character applies
192 both inside and outside character classes.
193 .P
194 For example, if you want to match a * character, you write \e* in the pattern.
195 This escaping action applies whether or not the following character would
196 otherwise be interpreted as a metacharacter, so it is always safe to precede a
197 non-alphanumeric with backslash to specify that it stands for itself. In
198 particular, if you want to match a backslash, you write \e\e.
199 .P
200 In UTF-8 mode, only ASCII numbers and letters have any special meaning after a
201 backslash. All other characters (in particular, those whose codepoints are
202 greater than 127) are treated as literals.
203 .P
204 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
205 pattern (other than in a character class) and characters between a # outside
206 a character class and the next newline are ignored. An escaping backslash can
207 be used to include a whitespace or # character as part of the pattern.
208 .P
209 If you want to remove the special meaning from a sequence of characters, you
210 can do so by putting them between \eQ and \eE. This is different from Perl in
211 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
212 Perl, $ and @ cause variable interpolation. Note the following examples:
213 .sp
214 Pattern PCRE matches Perl matches
215 .sp
216 .\" JOIN
217 \eQabc$xyz\eE abc$xyz abc followed by the
218 contents of $xyz
219 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
220 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
221 .sp
222 The \eQ...\eE sequence is recognized both inside and outside character classes.
223 An isolated \eE that is not preceded by \eQ is ignored. If \eQ is not followed
224 by \eE later in the pattern, the literal interpretation continues to the end of
225 the pattern (that is, \eE is assumed at the end). If the isolated \eQ is inside
226 a character class, this causes an error, because the character class is not
227 terminated.
228 .
229 .
230 .\" HTML <a name="digitsafterbackslash"></a>
231 .SS "Non-printing characters"
232 .rs
233 .sp
234 A second use of backslash provides a way of encoding non-printing characters
235 in patterns in a visible manner. There is no restriction on the appearance of
236 non-printing characters, apart from the binary zero that terminates a pattern,
237 but when a pattern is being prepared by text editing, it is often easier to use
238 one of the following escape sequences than the binary character it represents:
239 .sp
240 \ea alarm, that is, the BEL character (hex 07)
241 \ecx "control-x", where x is any ASCII character
242 \ee escape (hex 1B)
243 \ef formfeed (hex 0C)
244 \en linefeed (hex 0A)
245 \er carriage return (hex 0D)
246 \et tab (hex 09)
247 \eddd character with octal code ddd, or back reference
248 \exhh character with hex code hh
249 \ex{hhh..} character with hex code hhh..
250 .sp
251 The precise effect of \ecx is as follows: if x is a lower case letter, it
252 is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
253 Thus \ecz becomes hex 1A (z is 7A), but \ec{ becomes hex 3B ({ is 7B), while
254 \ec; becomes hex 7B (; is 3B). If the byte following \ec has a value greater
255 than 127, a compile-time error occurs. This locks out non-ASCII characters in
256 both byte mode and UTF-8 mode. (When PCRE is compiled in EBCDIC mode, all byte
257 values are valid. A lower case letter is converted to upper case, and then the
258 0xc0 bits are flipped.)
259 .P
260 After \ex, from zero to two hexadecimal digits are read (letters can be in
261 upper or lower case). Any number of hexadecimal digits may appear between \ex{
262 and }, but the value of the character code must be less than 256 in non-UTF-8
263 mode, and less than 2**31 in UTF-8 mode. That is, the maximum value in
264 hexadecimal is 7FFFFFFF. Note that this is bigger than the largest Unicode code
265 point, which is 10FFFF.
266 .P
267 If characters other than hexadecimal digits appear between \ex{ and }, or if
268 there is no terminating }, this form of escape is not recognized. Instead, the
269 initial \ex will be interpreted as a basic hexadecimal escape, with no
270 following digits, giving a character whose value is zero.
271 .P
272 Characters whose value is less than 256 can be defined by either of the two
273 syntaxes for \ex. There is no difference in the way they are handled. For
274 example, \exdc is exactly the same as \ex{dc}.
275 .P
276 After \e0 up to two further octal digits are read. If there are fewer than two
277 digits, just those that are present are used. Thus the sequence \e0\ex\e07
278 specifies two binary zeros followed by a BEL character (code value 7). Make
279 sure you supply two digits after the initial zero if the pattern character that
280 follows is itself an octal digit.
281 .P
282 The handling of a backslash followed by a digit other than 0 is complicated.
283 Outside a character class, PCRE reads it and any following digits as a decimal
284 number. If the number is less than 10, or if there have been at least that many
285 previous capturing left parentheses in the expression, the entire sequence is
286 taken as a \fIback reference\fP. A description of how this works is given
287 .\" HTML <a href="#backreferences">
288 .\" </a>
289 later,
290 .\"
291 following the discussion of
292 .\" HTML <a href="#subpattern">
293 .\" </a>
294 parenthesized subpatterns.
295 .\"
296 .P
297 Inside a character class, or if the decimal number is greater than 9 and there
298 have not been that many capturing subpatterns, PCRE re-reads up to three octal
299 digits following the backslash, and uses them to generate a data character. Any
300 subsequent digits stand for themselves. In non-UTF-8 mode, the value of a
301 character specified in octal must be less than \e400. In UTF-8 mode, values up
302 to \e777 are permitted. For example:
303 .sp
304 \e040 is another way of writing a space
305 .\" JOIN
306 \e40 is the same, provided there are fewer than 40
307 previous capturing subpatterns
308 \e7 is always a back reference
309 .\" JOIN
310 \e11 might be a back reference, or another way of
311 writing a tab
312 \e011 is always a tab
313 \e0113 is a tab followed by the character "3"
314 .\" JOIN
315 \e113 might be a back reference, otherwise the
316 character with octal code 113
317 .\" JOIN
318 \e377 might be a back reference, otherwise
319 the byte consisting entirely of 1 bits
320 .\" JOIN
321 \e81 is either a back reference, or a binary zero
322 followed by the two characters "8" and "1"
323 .sp
324 Note that octal values of 100 or greater must not be introduced by a leading
325 zero, because no more than three octal digits are ever read.
326 .P
327 All the sequences that define a single character value can be used both inside
328 and outside character classes. In addition, inside a character class, the
329 sequence \eb is interpreted as the backspace character (hex 08). The sequences
330 \eB, \eN, \eR, and \eX are not special inside a character class. Like any other
331 unrecognized escape sequences, they are treated as the literal characters "B",
332 "N", "R", and "X" by default, but cause an error if the PCRE_EXTRA option is
333 set. Outside a character class, these sequences have different meanings.
334 .
335 .
336 .SS "Absolute and relative back references"
337 .rs
338 .sp
339 The sequence \eg followed by an unsigned or a negative number, optionally
340 enclosed in braces, is an absolute or relative back reference. A named back
341 reference can be coded as \eg{name}. Back references are discussed
342 .\" HTML <a href="#backreferences">
343 .\" </a>
344 later,
345 .\"
346 following the discussion of
347 .\" HTML <a href="#subpattern">
348 .\" </a>
349 parenthesized subpatterns.
350 .\"
351 .
352 .
353 .SS "Absolute and relative subroutine calls"
354 .rs
355 .sp
356 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
357 a number enclosed either in angle brackets or single quotes, is an alternative
358 syntax for referencing a subpattern as a "subroutine". Details are discussed
359 .\" HTML <a href="#onigurumasubroutines">
360 .\" </a>
361 later.
362 .\"
363 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
364 synonymous. The former is a back reference; the latter is a
365 .\" HTML <a href="#subpatternsassubroutines">
366 .\" </a>
367 subroutine
368 .\"
369 call.
370 .
371 .
372 .\" HTML <a name="genericchartypes"></a>
373 .SS "Generic character types"
374 .rs
375 .sp
376 Another use of backslash is for specifying generic character types:
377 .sp
378 \ed any decimal digit
379 \eD any character that is not a decimal digit
380 \eh any horizontal whitespace character
381 \eH any character that is not a horizontal whitespace character
382 \es any whitespace character
383 \eS any character that is not a whitespace character
384 \ev any vertical whitespace character
385 \eV any character that is not a vertical whitespace character
386 \ew any "word" character
387 \eW any "non-word" character
388 .sp
389 There is also the single sequence \eN, which matches a non-newline character.
390 This is the same as
391 .\" HTML <a href="#fullstopdot">
392 .\" </a>
393 the "." metacharacter
394 .\"
395 when PCRE_DOTALL is not set.
396 .P
397 Each pair of lower and upper case escape sequences partitions the complete set
398 of characters into two disjoint sets. Any given character matches one, and only
399 one, of each pair. The sequences can appear both inside and outside character
400 classes. They each match one character of the appropriate type. If the current
401 matching point is at the end of the subject string, all of them fail, because
402 there is no character to match.
403 .P
404 For compatibility with Perl, \es does not match the VT character (code 11).
405 This makes it different from the the POSIX "space" class. The \es characters
406 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
407 included in a Perl script, \es may match the VT character. In PCRE, it never
408 does.
409 .P
410 A "word" character is an underscore or any character that is a letter or digit.
411 By default, the definition of letters and digits is controlled by PCRE's
412 low-valued character tables, and may vary if locale-specific matching is taking
413 place (see
414 .\" HTML <a href="pcreapi.html#localesupport">
415 .\" </a>
416 "Locale support"
417 .\"
418 in the
419 .\" HREF
420 \fBpcreapi\fP
421 .\"
422 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
423 or "french" in Windows, some character codes greater than 128 are used for
424 accented letters, and these are then matched by \ew. The use of locales with
425 Unicode is discouraged.
426 .P
427 By default, in UTF-8 mode, characters with values greater than 128 never match
428 \ed, \es, or \ew, and always match \eD, \eS, and \eW. These sequences retain
429 their original meanings from before UTF-8 support was available, mainly for
430 efficiency reasons. However, if PCRE is compiled with Unicode property support,
431 and the PCRE_UCP option is set, the behaviour is changed so that Unicode
432 properties are used to determine character types, as follows:
433 .sp
434 \ed any character that \ep{Nd} matches (decimal digit)
435 \es any character that \ep{Z} matches, plus HT, LF, FF, CR
436 \ew any character that \ep{L} or \ep{N} matches, plus underscore
437 .sp
438 The upper case escapes match the inverse sets of characters. Note that \ed
439 matches only decimal digits, whereas \ew matches any Unicode digit, as well as
440 any Unicode letter, and underscore. Note also that PCRE_UCP affects \eb, and
441 \eB because they are defined in terms of \ew and \eW. Matching these sequences
442 is noticeably slower when PCRE_UCP is set.
443 .P
444 The sequences \eh, \eH, \ev, and \eV are features that were added to Perl at
445 release 5.10. In contrast to the other sequences, which match only ASCII
446 characters by default, these always match certain high-valued codepoints in
447 UTF-8 mode, whether or not PCRE_UCP is set. The horizontal space characters
448 are:
449 .sp
450 U+0009 Horizontal tab
451 U+0020 Space
452 U+00A0 Non-break space
453 U+1680 Ogham space mark
454 U+180E Mongolian vowel separator
455 U+2000 En quad
456 U+2001 Em quad
457 U+2002 En space
458 U+2003 Em space
459 U+2004 Three-per-em space
460 U+2005 Four-per-em space
461 U+2006 Six-per-em space
462 U+2007 Figure space
463 U+2008 Punctuation space
464 U+2009 Thin space
465 U+200A Hair space
466 U+202F Narrow no-break space
467 U+205F Medium mathematical space
468 U+3000 Ideographic space
469 .sp
470 The vertical space characters are:
471 .sp
472 U+000A Linefeed
473 U+000B Vertical tab
474 U+000C Formfeed
475 U+000D Carriage return
476 U+0085 Next line
477 U+2028 Line separator
478 U+2029 Paragraph separator
479 .
480 .
481 .\" HTML <a name="newlineseq"></a>
482 .SS "Newline sequences"
483 .rs
484 .sp
485 Outside a character class, by default, the escape sequence \eR matches any
486 Unicode newline sequence. In non-UTF-8 mode \eR is equivalent to the following:
487 .sp
488 (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
489 .sp
490 This is an example of an "atomic group", details of which are given
491 .\" HTML <a href="#atomicgroup">
492 .\" </a>
493 below.
494 .\"
495 This particular group matches either the two-character sequence CR followed by
496 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
497 U+000B), FF (formfeed, U+000C), CR (carriage return, U+000D), or NEL (next
498 line, U+0085). The two-character sequence is treated as a single unit that
499 cannot be split.
500 .P
501 In UTF-8 mode, two additional characters whose codepoints are greater than 255
502 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
503 Unicode character property support is not needed for these characters to be
504 recognized.
505 .P
506 It is possible to restrict \eR to match only CR, LF, or CRLF (instead of the
507 complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
508 either at compile time or when the pattern is matched. (BSR is an abbrevation
509 for "backslash R".) This can be made the default when PCRE is built; if this is
510 the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
511 It is also possible to specify these settings by starting a pattern string with
512 one of the following sequences:
513 .sp
514 (*BSR_ANYCRLF) CR, LF, or CRLF only
515 (*BSR_UNICODE) any Unicode newline sequence
516 .sp
517 These override the default and the options given to \fBpcre_compile()\fP or
518 \fBpcre_compile2()\fP, but they can be overridden by options given to
519 \fBpcre_exec()\fP or \fBpcre_dfa_exec()\fP. Note that these special settings,
520 which are not Perl-compatible, are recognized only at the very start of a
521 pattern, and that they must be in upper case. If more than one of them is
522 present, the last one is used. They can be combined with a change of newline
523 convention; for example, a pattern can start with:
524 .sp
526 .sp
527 They can also be combined with the (*UTF8) or (*UCP) special sequences. Inside
528 a character class, \eR is treated as an unrecognized escape sequence, and so
529 matches the letter "R" by default, but causes an error if PCRE_EXTRA is set.
530 .
531 .
532 .\" HTML <a name="uniextseq"></a>
533 .SS Unicode character properties
534 .rs
535 .sp
536 When PCRE is built with Unicode character property support, three additional
537 escape sequences that match characters with specific properties are available.
538 When not in UTF-8 mode, these sequences are of course limited to testing
539 characters whose codepoints are less than 256, but they do work in this mode.
540 The extra escape sequences are:
541 .sp
542 \ep{\fIxx\fP} a character with the \fIxx\fP property
543 \eP{\fIxx\fP} a character without the \fIxx\fP property
544 \eX an extended Unicode sequence
545 .sp
546 The property names represented by \fIxx\fP above are limited to the Unicode
547 script names, the general category properties, "Any", which matches any
548 character (including newline), and some special PCRE properties (described
549 in the
550 .\" HTML <a href="#extraprops">
551 .\" </a>
552 next section).
553 .\"
554 Other Perl properties such as "InMusicalSymbols" are not currently supported by
555 PCRE. Note that \eP{Any} does not match any characters, so always causes a
556 match failure.
557 .P
558 Sets of Unicode characters are defined as belonging to certain scripts. A
559 character from one of these sets can be matched using a script name. For
560 example:
561 .sp
562 \ep{Greek}
563 \eP{Han}
564 .sp
565 Those that are not part of an identified script are lumped together as
566 "Common". The current list of scripts is:
567 .P
568 Arabic,
569 Armenian,
570 Avestan,
571 Balinese,
572 Bamum,
573 Bengali,
574 Bopomofo,
575 Braille,
576 Buginese,
577 Buhid,
578 Canadian_Aboriginal,
579 Carian,
580 Cham,
581 Cherokee,
582 Common,
583 Coptic,
584 Cuneiform,
585 Cypriot,
586 Cyrillic,
587 Deseret,
588 Devanagari,
589 Egyptian_Hieroglyphs,
590 Ethiopic,
591 Georgian,
592 Glagolitic,
593 Gothic,
594 Greek,
595 Gujarati,
596 Gurmukhi,
597 Han,
598 Hangul,
599 Hanunoo,
600 Hebrew,
601 Hiragana,
602 Imperial_Aramaic,
603 Inherited,
604 Inscriptional_Pahlavi,
605 Inscriptional_Parthian,
606 Javanese,
607 Kaithi,
608 Kannada,
609 Katakana,
610 Kayah_Li,
611 Kharoshthi,
612 Khmer,
613 Lao,
614 Latin,
615 Lepcha,
616 Limbu,
617 Linear_B,
618 Lisu,
619 Lycian,
620 Lydian,
621 Malayalam,
622 Meetei_Mayek,
623 Mongolian,
624 Myanmar,
625 New_Tai_Lue,
626 Nko,
627 Ogham,
628 Old_Italic,
629 Old_Persian,
630 Old_South_Arabian,
631 Old_Turkic,
632 Ol_Chiki,
633 Oriya,
634 Osmanya,
635 Phags_Pa,
636 Phoenician,
637 Rejang,
638 Runic,
639 Samaritan,
640 Saurashtra,
641 Shavian,
642 Sinhala,
643 Sundanese,
644 Syloti_Nagri,
645 Syriac,
646 Tagalog,
647 Tagbanwa,
648 Tai_Le,
649 Tai_Tham,
650 Tai_Viet,
651 Tamil,
652 Telugu,
653 Thaana,
654 Thai,
655 Tibetan,
656 Tifinagh,
657 Ugaritic,
658 Vai,
659 Yi.
660 .P
661 Each character has exactly one Unicode general category property, specified by
662 a two-letter abbreviation. For compatibility with Perl, negation can be
663 specified by including a circumflex between the opening brace and the property
664 name. For example, \ep{^Lu} is the same as \eP{Lu}.
665 .P
666 If only one letter is specified with \ep or \eP, it includes all the general
667 category properties that start with that letter. In this case, in the absence
668 of negation, the curly brackets in the escape sequence are optional; these two
669 examples have the same effect:
670 .sp
671 \ep{L}
672 \epL
673 .sp
674 The following general category property codes are supported:
675 .sp
676 C Other
677 Cc Control
678 Cf Format
679 Cn Unassigned
680 Co Private use
681 Cs Surrogate
682 .sp
683 L Letter
684 Ll Lower case letter
685 Lm Modifier letter
686 Lo Other letter
687 Lt Title case letter
688 Lu Upper case letter
689 .sp
690 M Mark
691 Mc Spacing mark
692 Me Enclosing mark
693 Mn Non-spacing mark
694 .sp
695 N Number
696 Nd Decimal number
697 Nl Letter number
698 No Other number
699 .sp
700 P Punctuation
701 Pc Connector punctuation
702 Pd Dash punctuation
703 Pe Close punctuation
704 Pf Final punctuation
705 Pi Initial punctuation
706 Po Other punctuation
707 Ps Open punctuation
708 .sp
709 S Symbol
710 Sc Currency symbol
711 Sk Modifier symbol
712 Sm Mathematical symbol
713 So Other symbol
714 .sp
715 Z Separator
716 Zl Line separator
717 Zp Paragraph separator
718 Zs Space separator
719 .sp
720 The special property L& is also supported: it matches a character that has
721 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
722 a modifier or "other".
723 .P
724 The Cs (Surrogate) property applies only to characters in the range U+D800 to
725 U+DFFF. Such characters are not valid in UTF-8 strings (see RFC 3629) and so
726 cannot be tested by PCRE, unless UTF-8 validity checking has been turned off
727 (see the discussion of PCRE_NO_UTF8_CHECK in the
728 .\" HREF
729 \fBpcreapi\fP
730 .\"
731 page). Perl does not support the Cs property.
732 .P
733 The long synonyms for property names that Perl supports (such as \ep{Letter})
734 are not supported by PCRE, nor is it permitted to prefix any of these
735 properties with "Is".
736 .P
737 No character that is in the Unicode table has the Cn (unassigned) property.
738 Instead, this property is assumed for any code point that is not in the
739 Unicode table.
740 .P
741 Specifying caseless matching does not affect these escape sequences. For
742 example, \ep{Lu} always matches only upper case letters.
743 .P
744 The \eX escape matches any number of Unicode characters that form an extended
745 Unicode sequence. \eX is equivalent to
746 .sp
747 (?>\ePM\epM*)
748 .sp
749 That is, it matches a character without the "mark" property, followed by zero
750 or more characters with the "mark" property, and treats the sequence as an
751 atomic group
752 .\" HTML <a href="#atomicgroup">
753 .\" </a>
754 (see below).
755 .\"
756 Characters with the "mark" property are typically accents that affect the
757 preceding character. None of them have codepoints less than 256, so in
758 non-UTF-8 mode \eX matches any one character.
759 .P
760 Matching characters by Unicode property is not fast, because PCRE has to search
761 a structure that contains data for over fifteen thousand characters. That is
762 why the traditional escape sequences such as \ed and \ew do not use Unicode
763 properties in PCRE by default, though you can make them do so by setting the
764 PCRE_UCP option for \fBpcre_compile()\fP or by starting the pattern with
765 (*UCP).
766 .
767 .
768 .\" HTML <a name="extraprops"></a>
769 .SS PCRE's additional properties
770 .rs
771 .sp
772 As well as the standard Unicode properties described in the previous
773 section, PCRE supports four more that make it possible to convert traditional
774 escape sequences such as \ew and \es and POSIX character classes to use Unicode
775 properties. PCRE uses these non-standard, non-Perl properties internally when
776 PCRE_UCP is set. They are:
777 .sp
778 Xan Any alphanumeric character
779 Xps Any POSIX space character
780 Xsp Any Perl space character
781 Xwd Any Perl "word" character
782 .sp
783 Xan matches characters that have either the L (letter) or the N (number)
784 property. Xps matches the characters tab, linefeed, vertical tab, formfeed, or
785 carriage return, and any other character that has the Z (separator) property.
786 Xsp is the same as Xps, except that vertical tab is excluded. Xwd matches the
787 same characters as Xan, plus underscore.
788 .
789 .
790 .\" HTML <a name="resetmatchstart"></a>
791 .SS "Resetting the match start"
792 .rs
793 .sp
794 The escape sequence \eK causes any previously matched characters not to be
795 included in the final matched sequence. For example, the pattern:
796 .sp
797 foo\eKbar
798 .sp
799 matches "foobar", but reports that it has matched "bar". This feature is
800 similar to a lookbehind assertion
801 .\" HTML <a href="#lookbehind">
802 .\" </a>
803 (described below).
804 .\"
805 However, in this case, the part of the subject before the real match does not
806 have to be of fixed length, as lookbehind assertions do. The use of \eK does
807 not interfere with the setting of
808 .\" HTML <a href="#subpattern">
809 .\" </a>
810 captured substrings.
811 .\"
812 For example, when the pattern
813 .sp
814 (foo)\eKbar
815 .sp
816 matches "foobar", the first substring is still set to "foo".
817 .P
818 Perl documents that the use of \eK within assertions is "not well defined". In
819 PCRE, \eK is acted upon when it occurs inside positive assertions, but is
820 ignored in negative assertions.
821 .
822 .
823 .\" HTML <a name="smallassertions"></a>
824 .SS "Simple assertions"
825 .rs
826 .sp
827 The final use of backslash is for certain simple assertions. An assertion
828 specifies a condition that has to be met at a particular point in a match,
829 without consuming any characters from the subject string. The use of
830 subpatterns for more complicated assertions is described
831 .\" HTML <a href="#bigassertions">
832 .\" </a>
833 below.
834 .\"
835 The backslashed assertions are:
836 .sp
837 \eb matches at a word boundary
838 \eB matches when not at a word boundary
839 \eA matches at the start of the subject
840 \eZ matches at the end of the subject
841 also matches before a newline at the end of the subject
842 \ez matches only at the end of the subject
843 \eG matches at the first matching position in the subject
844 .sp
845 Inside a character class, \eb has a different meaning; it matches the backspace
846 character. If any other of these assertions appears in a character class, by
847 default it matches the corresponding literal character (for example, \eB
848 matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
849 escape sequence" error is generated instead.
850 .P
851 A word boundary is a position in the subject string where the current character
852 and the previous character do not both match \ew or \eW (i.e. one matches
853 \ew and the other matches \eW), or the start or end of the string if the
854 first or last character matches \ew, respectively. In UTF-8 mode, the meanings
855 of \ew and \eW can be changed by setting the PCRE_UCP option. When this is
856 done, it also affects \eb and \eB. Neither PCRE nor Perl has a separate "start
857 of word" or "end of word" metasequence. However, whatever follows \eb normally
858 determines which it is. For example, the fragment \eba matches "a" at the start
859 of a word.
860 .P
861 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
862 dollar (described in the next section) in that they only ever match at the very
863 start and end of the subject string, whatever options are set. Thus, they are
864 independent of multiline mode. These three assertions are not affected by the
865 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
866 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
867 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
868 at a point other than the beginning of the subject, \eA can never match. The
869 difference between \eZ and \ez is that \eZ matches before a newline at the end
870 of the string as well as at the very end, whereas \ez matches only at the end.
871 .P
872 The \eG assertion is true only when the current matching position is at the
873 start point of the match, as specified by the \fIstartoffset\fP argument of
874 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
875 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
876 arguments, you can mimic Perl's /g option, and it is in this kind of
877 implementation where \eG can be useful.
878 .P
879 Note, however, that PCRE's interpretation of \eG, as the start of the current
880 match, is subtly different from Perl's, which defines it as the end of the
881 previous match. In Perl, these can be different when the previously matched
882 string was empty. Because PCRE does just one match at a time, it cannot
883 reproduce this behaviour.
884 .P
885 If all the alternatives of a pattern begin with \eG, the expression is anchored
886 to the starting match position, and the "anchored" flag is set in the compiled
887 regular expression.
888 .
889 .
891 .rs
892 .sp
893 Outside a character class, in the default matching mode, the circumflex
894 character is an assertion that is true only if the current matching point is
895 at the start of the subject string. If the \fIstartoffset\fP argument of
896 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
897 option is unset. Inside a character class, circumflex has an entirely different
898 meaning
899 .\" HTML <a href="#characterclass">
900 .\" </a>
901 (see below).
902 .\"
903 .P
904 Circumflex need not be the first character of the pattern if a number of
905 alternatives are involved, but it should be the first thing in each alternative
906 in which it appears if the pattern is ever to match that branch. If all
907 possible alternatives start with a circumflex, that is, if the pattern is
908 constrained to match only at the start of the subject, it is said to be an
909 "anchored" pattern. (There are also other constructs that can cause a pattern
910 to be anchored.)
911 .P
912 A dollar character is an assertion that is true only if the current matching
913 point is at the end of the subject string, or immediately before a newline
914 at the end of the string (by default). Dollar need not be the last character of
915 the pattern if a number of alternatives are involved, but it should be the last
916 item in any branch in which it appears. Dollar has no special meaning in a
917 character class.
918 .P
919 The meaning of dollar can be changed so that it matches only at the very end of
920 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
921 does not affect the \eZ assertion.
922 .P
923 The meanings of the circumflex and dollar characters are changed if the
924 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
925 immediately after internal newlines as well as at the start of the subject
926 string. It does not match after a newline that ends the string. A dollar
927 matches before any newlines in the string, as well as at the very end, when
928 PCRE_MULTILINE is set. When newline is specified as the two-character
929 sequence CRLF, isolated CR and LF characters do not indicate newlines.
930 .P
931 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
932 \en represents a newline) in multiline mode, but not otherwise. Consequently,
933 patterns that are anchored in single line mode because all branches start with
934 ^ are not anchored in multiline mode, and a match for circumflex is possible
935 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
936 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
937 .P
938 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
939 end of the subject in both modes, and if all branches of a pattern start with
940 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
941 .
942 .
943 .\" HTML <a name="fullstopdot"></a>
945 .rs
946 .sp
947 Outside a character class, a dot in the pattern matches any one character in
948 the subject string except (by default) a character that signifies the end of a
949 line. In UTF-8 mode, the matched character may be more than one byte long.
950 .P
951 When a line ending is defined as a single character, dot never matches that
952 character; when the two-character sequence CRLF is used, dot does not match CR
953 if it is immediately followed by LF, but otherwise it matches all characters
954 (including isolated CRs and LFs). When any Unicode line endings are being
955 recognized, dot does not match CR or LF or any of the other line ending
956 characters.
957 .P
958 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
959 option is set, a dot matches any one character, without exception. If the
960 two-character sequence CRLF is present in the subject string, it takes two dots
961 to match it.
962 .P
963 The handling of dot is entirely independent of the handling of circumflex and
964 dollar, the only relationship being that they both involve newlines. Dot has no
965 special meaning in a character class.
966 .P
967 The escape sequence \eN behaves like a dot, except that it is not affected by
968 the PCRE_DOTALL option. In other words, it matches any character except one
969 that signifies the end of a line.
970 .
971 .
973 .rs
974 .sp
975 Outside a character class, the escape sequence \eC matches any one byte, both
976 in and out of UTF-8 mode. Unlike a dot, it always matches any line-ending
977 characters. The feature is provided in Perl in order to match individual bytes
978 in UTF-8 mode. Because it breaks up UTF-8 characters into individual bytes, the
979 rest of the string may start with a malformed UTF-8 character. For this reason,
980 the \eC escape sequence is best avoided.
981 .P
982 PCRE does not allow \eC to appear in lookbehind assertions
983 .\" HTML <a href="#lookbehind">
984 .\" </a>
985 (described below),
986 .\"
987 because in UTF-8 mode this would make it impossible to calculate the length of
988 the lookbehind.
989 .
990 .
991 .\" HTML <a name="characterclass"></a>
993 .rs
994 .sp
995 An opening square bracket introduces a character class, terminated by a closing
996 square bracket. A closing square bracket on its own is not special by default.
997 However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
998 bracket causes a compile-time error. If a closing square bracket is required as
999 a member of the class, it should be the first data character in the class
1000 (after an initial circumflex, if present) or escaped with a backslash.
1001 .P
1002 A character class matches a single character in the subject. In UTF-8 mode, the
1003 character may be more than one byte long. A matched character must be in the
1004 set of characters defined by the class, unless the first character in the class
1005 definition is a circumflex, in which case the subject character must not be in
1006 the set defined by the class. If a circumflex is actually required as a member
1007 of the class, ensure it is not the first character, or escape it with a
1008 backslash.
1009 .P
1010 For example, the character class [aeiou] matches any lower case vowel, while
1011 [^aeiou] matches any character that is not a lower case vowel. Note that a
1012 circumflex is just a convenient notation for specifying the characters that
1013 are in the class by enumerating those that are not. A class that starts with a
1014 circumflex is not an assertion; it still consumes a character from the subject
1015 string, and therefore it fails if the current pointer is at the end of the
1016 string.
1017 .P
1018 In UTF-8 mode, characters with values greater than 255 can be included in a
1019 class as a literal string of bytes, or by using the \ex{ escaping mechanism.
1020 .P
1021 When caseless matching is set, any letters in a class represent both their
1022 upper case and lower case versions, so for example, a caseless [aeiou] matches
1023 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1024 caseful version would. In UTF-8 mode, PCRE always understands the concept of
1025 case for characters whose values are less than 128, so caseless matching is
1026 always possible. For characters with higher values, the concept of case is
1027 supported if PCRE is compiled with Unicode property support, but not otherwise.
1028 If you want to use caseless matching in UTF8-mode for characters 128 and above,
1029 you must ensure that PCRE is compiled with Unicode property support as well as
1030 with UTF-8 support.
1031 .P
1032 Characters that might indicate line breaks are never treated in any special way
1033 when matching character classes, whatever line-ending sequence is in use, and
1034 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1035 such as [^a] always matches one of these characters.
1036 .P
1037 The minus (hyphen) character can be used to specify a range of characters in a
1038 character class. For example, [d-m] matches any letter between d and m,
1039 inclusive. If a minus character is required in a class, it must be escaped with
1040 a backslash or appear in a position where it cannot be interpreted as
1041 indicating a range, typically as the first or last character in the class.
1042 .P
1043 It is not possible to have the literal character "]" as the end character of a
1044 range. A pattern such as [W-]46] is interpreted as a class of two characters
1045 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1046 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1047 the end of range, so [W-\e]46] is interpreted as a class containing a range
1048 followed by two other characters. The octal or hexadecimal representation of
1049 "]" can also be used to end a range.
1050 .P
1051 Ranges operate in the collating sequence of character values. They can also be
1052 used for characters specified numerically, for example [\e000-\e037]. In UTF-8
1053 mode, ranges can include characters whose values are greater than 255, for
1054 example [\ex{100}-\ex{2ff}].
1055 .P
1056 If a range that includes letters is used when caseless matching is set, it
1057 matches the letters in either case. For example, [W-c] is equivalent to
1058 [][\e\e^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
1059 tables for a French locale are in use, [\exc8-\excb] matches accented E
1060 characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
1061 characters with values greater than 128 only when it is compiled with Unicode
1062 property support.
1063 .P
1064 The character escape sequences \ed, \eD, \eh, \eH, \ep, \eP, \es, \eS, \ev,
1065 \eV, \ew, and \eW may appear in a character class, and add the characters that
1066 they match to the class. For example, [\edABCDEF] matches any hexadecimal
1067 digit. In UTF-8 mode, the PCRE_UCP option affects the meanings of \ed, \es, \ew
1068 and their upper case partners, just as it does when they appear outside a
1069 character class, as described in the section entitled
1070 .\" HTML <a href="#genericchartypes">
1071 .\" </a>
1072 "Generic character types"
1073 .\"
1074 above. The escape sequence \eb has a different meaning inside a character
1075 class; it matches the backspace character. The sequences \eB, \eN, \eR, and \eX
1076 are not special inside a character class. Like any other unrecognized escape
1077 sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1078 default, but cause an error if the PCRE_EXTRA option is set.
1079 .P
1080 A circumflex can conveniently be used with the upper case character types to
1081 specify a more restricted set of characters than the matching lower case type.
1082 For example, the class [^\eW_] matches any letter or digit, but not underscore,
1083 whereas [\ew] includes underscore. A positive character class should be read as
1084 "something OR something OR ..." and a negative class as "NOT something AND NOT
1085 something AND NOT ...".
1086 .P
1087 The only metacharacters that are recognized in character classes are backslash,
1088 hyphen (only where it can be interpreted as specifying a range), circumflex
1089 (only at the start), opening square bracket (only when it can be interpreted as
1090 introducing a POSIX class name - see the next section), and the terminating
1091 closing square bracket. However, escaping other non-alphanumeric characters
1092 does no harm.
1093 .
1094 .
1096 .rs
1097 .sp
1098 Perl supports the POSIX notation for character classes. This uses names
1099 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1100 this notation. For example,
1101 .sp
1102 [01[:alpha:]%]
1103 .sp
1104 matches "0", "1", any alphabetic character, or "%". The supported class names
1105 are:
1106 .sp
1107 alnum letters and digits
1108 alpha letters
1109 ascii character codes 0 - 127
1110 blank space or tab only
1111 cntrl control characters
1112 digit decimal digits (same as \ed)
1113 graph printing characters, excluding space
1114 lower lower case letters
1115 print printing characters, including space
1116 punct printing characters, excluding letters and digits and space
1117 space white space (not quite the same as \es)
1118 upper upper case letters
1119 word "word" characters (same as \ew)
1120 xdigit hexadecimal digits
1121 .sp
1122 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
1123 space (32). Notice that this list includes the VT character (code 11). This
1124 makes "space" different to \es, which does not include VT (for Perl
1125 compatibility).
1126 .P
1127 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1128 5.8. Another Perl extension is negation, which is indicated by a ^ character
1129 after the colon. For example,
1130 .sp
1131 [12[:^digit:]]
1132 .sp
1133 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1134 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1135 supported, and an error is given if they are encountered.
1136 .P
1137 By default, in UTF-8 mode, characters with values greater than 128 do not match
1138 any of the POSIX character classes. However, if the PCRE_UCP option is passed
1139 to \fBpcre_compile()\fP, some of the classes are changed so that Unicode
1140 character properties are used. This is achieved by replacing the POSIX classes
1141 by other sequences, as follows:
1142 .sp
1143 [:alnum:] becomes \ep{Xan}
1144 [:alpha:] becomes \ep{L}
1145 [:blank:] becomes \eh
1146 [:digit:] becomes \ep{Nd}
1147 [:lower:] becomes \ep{Ll}
1148 [:space:] becomes \ep{Xps}
1149 [:upper:] becomes \ep{Lu}
1150 [:word:] becomes \ep{Xwd}
1151 .sp
1152 Negated versions, such as [:^alpha:] use \eP instead of \ep. The other POSIX
1153 classes are unchanged, and match only characters with code points less than
1154 128.
1155 .
1156 .
1158 .rs
1159 .sp
1160 Vertical bar characters are used to separate alternative patterns. For example,
1161 the pattern
1162 .sp
1163 gilbert|sullivan
1164 .sp
1165 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1166 and an empty alternative is permitted (matching the empty string). The matching
1167 process tries each alternative in turn, from left to right, and the first one
1168 that succeeds is used. If the alternatives are within a subpattern
1169 .\" HTML <a href="#subpattern">
1170 .\" </a>
1171 (defined below),
1172 .\"
1173 "succeeds" means matching the rest of the main pattern as well as the
1174 alternative in the subpattern.
1175 .
1176 .
1178 .rs
1179 .sp
1180 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1181 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1182 the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1183 The option letters are
1184 .sp
1185 i for PCRE_CASELESS
1187 s for PCRE_DOTALL
1188 x for PCRE_EXTENDED
1189 .sp
1190 For example, (?im) sets caseless, multiline matching. It is also possible to
1191 unset these options by preceding the letter with a hyphen, and a combined
1192 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1193 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1194 permitted. If a letter appears both before and after the hyphen, the option is
1195 unset.
1196 .P
1197 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1198 changed in the same way as the Perl-compatible options by using the characters
1199 J, U and X respectively.
1200 .P
1201 When one of these option changes occurs at top level (that is, not inside
1202 subpattern parentheses), the change applies to the remainder of the pattern
1203 that follows. If the change is placed right at the start of a pattern, PCRE
1204 extracts it into the global options (and it will therefore show up in data
1205 extracted by the \fBpcre_fullinfo()\fP function).
1206 .P
1207 An option change within a subpattern (see below for a description of
1208 subpatterns) affects only that part of the subpattern that follows it, so
1209 .sp
1210 (a(?i)b)c
1211 .sp
1212 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1213 By this means, options can be made to have different settings in different
1214 parts of the pattern. Any changes made in one alternative do carry on
1215 into subsequent branches within the same subpattern. For example,
1216 .sp
1217 (a(?i)b|c)
1218 .sp
1219 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1220 branch is abandoned before the option setting. This is because the effects of
1221 option settings happen at compile time. There would be some very weird
1222 behaviour otherwise.
1223 .P
1224 \fBNote:\fP There are other PCRE-specific options that can be set by the
1225 application when the compile or match functions are called. In some cases the
1226 pattern can contain special leading sequences such as (*CRLF) to override what
1227 the application has set or what has been defaulted. Details are given in the
1228 section entitled
1229 .\" HTML <a href="#newlineseq">
1230 .\" </a>
1231 "Newline sequences"
1232 .\"
1233 above. There are also the (*UTF8) and (*UCP) leading sequences that can be used
1234 to set UTF-8 and Unicode property modes; they are equivalent to setting the
1235 PCRE_UTF8 and the PCRE_UCP options, respectively.
1236 .
1237 .
1238 .\" HTML <a name="subpattern"></a>
1240 .rs
1241 .sp
1242 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1243 Turning part of a pattern into a subpattern does two things:
1244 .sp
1245 1. It localizes a set of alternatives. For example, the pattern
1246 .sp
1247 cat(aract|erpillar|)
1248 .sp
1249 matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1250 match "cataract", "erpillar" or an empty string.
1251 .sp
1252 2. It sets up the subpattern as a capturing subpattern. This means that, when
1253 the whole pattern matches, that portion of the subject string that matched the
1254 subpattern is passed back to the caller via the \fIovector\fP argument of
1255 \fBpcre_exec()\fP. Opening parentheses are counted from left to right (starting
1256 from 1) to obtain numbers for the capturing subpatterns. For example, if the
1257 string "the red king" is matched against the pattern
1258 .sp
1259 the ((red|white) (king|queen))
1260 .sp
1261 the captured substrings are "red king", "red", and "king", and are numbered 1,
1262 2, and 3, respectively.
1263 .P
1264 The fact that plain parentheses fulfil two functions is not always helpful.
1265 There are often times when a grouping subpattern is required without a
1266 capturing requirement. If an opening parenthesis is followed by a question mark
1267 and a colon, the subpattern does not do any capturing, and is not counted when
1268 computing the number of any subsequent capturing subpatterns. For example, if
1269 the string "the white queen" is matched against the pattern
1270 .sp
1271 the ((?:red|white) (king|queen))
1272 .sp
1273 the captured substrings are "white queen" and "queen", and are numbered 1 and
1274 2. The maximum number of capturing subpatterns is 65535.
1275 .P
1276 As a convenient shorthand, if any option settings are required at the start of
1277 a non-capturing subpattern, the option letters may appear between the "?" and
1278 the ":". Thus the two patterns
1279 .sp
1280 (?i:saturday|sunday)
1281 (?:(?i)saturday|sunday)
1282 .sp
1283 match exactly the same set of strings. Because alternative branches are tried
1284 from left to right, and options are not reset until the end of the subpattern
1285 is reached, an option setting in one branch does affect subsequent branches, so
1286 the above patterns match "SUNDAY" as well as "Saturday".
1287 .
1288 .
1289 .\" HTML <a name="dupsubpatternnumber"></a>
1291 .rs
1292 .sp
1293 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1294 the same numbers for its capturing parentheses. Such a subpattern starts with
1295 (?| and is itself a non-capturing subpattern. For example, consider this
1296 pattern:
1297 .sp
1298 (?|(Sat)ur|(Sun))day
1299 .sp
1300 Because the two alternatives are inside a (?| group, both sets of capturing
1301 parentheses are numbered one. Thus, when the pattern matches, you can look
1302 at captured substring number one, whichever alternative matched. This construct
1303 is useful when you want to capture part, but not all, of one of a number of
1304 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1305 number is reset at the start of each branch. The numbers of any capturing
1306 parentheses that follow the subpattern start after the highest number used in
1307 any branch. The following example is taken from the Perl documentation. The
1308 numbers underneath show in which buffer the captured content will be stored.
1309 .sp
1310 # before ---------------branch-reset----------- after
1311 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1312 # 1 2 2 3 2 3 4
1313 .sp
1314 A back reference to a numbered subpattern uses the most recent value that is
1315 set for that number by any subpattern. The following pattern matches "abcabc"
1316 or "defdef":
1317 .sp
1318 /(?|(abc)|(def))\e1/
1319 .sp
1320 In contrast, a recursive or "subroutine" call to a numbered subpattern always
1321 refers to the first one in the pattern with the given number. The following
1322 pattern matches "abcabc" or "defabc":
1323 .sp
1324 /(?|(abc)|(def))(?1)/
1325 .sp
1326 If a
1327 .\" HTML <a href="#conditions">
1328 .\" </a>
1329 condition test
1330 .\"
1331 for a subpattern's having matched refers to a non-unique number, the test is
1332 true if any of the subpatterns of that number have matched.
1333 .P
1334 An alternative approach to using this "branch reset" feature is to use
1335 duplicate named subpatterns, as described in the next section.
1336 .
1337 .
1339 .rs
1340 .sp
1341 Identifying capturing parentheses by number is simple, but it can be very hard
1342 to keep track of the numbers in complicated regular expressions. Furthermore,
1343 if an expression is modified, the numbers may change. To help with this
1344 difficulty, PCRE supports the naming of subpatterns. This feature was not
1345 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1346 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1347 the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1348 have different names, but PCRE does not.
1349 .P
1350 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1351 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1352 parentheses from other parts of the pattern, such as
1353 .\" HTML <a href="#backreferences">
1354 .\" </a>
1355 back references,
1356 .\"
1357 .\" HTML <a href="#recursion">
1358 .\" </a>
1359 recursion,
1360 .\"
1361 and
1362 .\" HTML <a href="#conditions">
1363 .\" </a>
1364 conditions,
1365 .\"
1366 can be made by name as well as by number.
1367 .P
1368 Names consist of up to 32 alphanumeric characters and underscores. Named
1369 capturing parentheses are still allocated numbers as well as names, exactly as
1370 if the names were not present. The PCRE API provides function calls for
1371 extracting the name-to-number translation table from a compiled pattern. There
1372 is also a convenience function for extracting a captured substring by name.
1373 .P
1374 By default, a name must be unique within a pattern, but it is possible to relax
1375 this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1376 names are also always permitted for subpatterns with the same number, set up as
1377 described in the previous section.) Duplicate names can be useful for patterns
1378 where only one instance of the named parentheses can match. Suppose you want to
1379 match the name of a weekday, either as a 3-letter abbreviation or as the full
1380 name, and in both cases you want to extract the abbreviation. This pattern
1381 (ignoring the line breaks) does the job:
1382 .sp
1383 (?<DN>Mon|Fri|Sun)(?:day)?|
1384 (?<DN>Tue)(?:sday)?|
1385 (?<DN>Wed)(?:nesday)?|
1386 (?<DN>Thu)(?:rsday)?|
1387 (?<DN>Sat)(?:urday)?
1388 .sp
1389 There are five capturing substrings, but only one is ever set after a match.
1390 (An alternative way of solving this problem is to use a "branch reset"
1391 subpattern, as described in the previous section.)
1392 .P
1393 The convenience function for extracting the data by name returns the substring
1394 for the first (and in this example, the only) subpattern of that name that
1395 matched. This saves searching to find which numbered subpattern it was.
1396 .P
1397 If you make a back reference to a non-unique named subpattern from elsewhere in
1398 the pattern, the one that corresponds to the first occurrence of the name is
1399 used. In the absence of duplicate numbers (see the previous section) this is
1400 the one with the lowest number. If you use a named reference in a condition
1401 test (see the
1402 .\"
1403 .\" HTML <a href="#conditions">
1404 .\" </a>
1405 section about conditions
1406 .\"
1407 below), either to check whether a subpattern has matched, or to check for
1408 recursion, all subpatterns with the same name are tested. If the condition is
1409 true for any one of them, the overall condition is true. This is the same
1410 behaviour as testing by number. For further details of the interfaces for
1411 handling named subpatterns, see the
1412 .\" HREF
1413 \fBpcreapi\fP
1414 .\"
1415 documentation.
1416 .P
1417 \fBWarning:\fP You cannot use different names to distinguish between two
1418 subpatterns with the same number because PCRE uses only the numbers when
1419 matching. For this reason, an error is given at compile time if different names
1420 are given to subpatterns with the same number. However, you can give the same
1421 name to subpatterns with the same number, even when PCRE_DUPNAMES is not set.
1422 .
1423 .
1425 .rs
1426 .sp
1427 Repetition is specified by quantifiers, which can follow any of the following
1428 items:
1429 .sp
1430 a literal data character
1431 the dot metacharacter
1432 the \eC escape sequence
1433 the \eX escape sequence (in UTF-8 mode with Unicode properties)
1434 the \eR escape sequence
1435 an escape such as \ed or \epL that matches a single character
1436 a character class
1437 a back reference (see next section)
1438 a parenthesized subpattern (unless it is an assertion)
1439 a recursive or "subroutine" call to a subpattern
1440 .sp
1441 The general repetition quantifier specifies a minimum and maximum number of
1442 permitted matches, by giving the two numbers in curly brackets (braces),
1443 separated by a comma. The numbers must be less than 65536, and the first must
1444 be less than or equal to the second. For example:
1445 .sp
1446 z{2,4}
1447 .sp
1448 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1449 character. If the second number is omitted, but the comma is present, there is
1450 no upper limit; if the second number and the comma are both omitted, the
1451 quantifier specifies an exact number of required matches. Thus
1452 .sp
1453 [aeiou]{3,}
1454 .sp
1455 matches at least 3 successive vowels, but may match many more, while
1456 .sp
1457 \ed{8}
1458 .sp
1459 matches exactly 8 digits. An opening curly bracket that appears in a position
1460 where a quantifier is not allowed, or one that does not match the syntax of a
1461 quantifier, is taken as a literal character. For example, {,6} is not a
1462 quantifier, but a literal string of four characters.
1463 .P
1464 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
1465 bytes. Thus, for example, \ex{100}{2} matches two UTF-8 characters, each of
1466 which is represented by a two-byte sequence. Similarly, when Unicode property
1467 support is available, \eX{3} matches three Unicode extended sequences, each of
1468 which may be several bytes long (and they may be of different lengths).
1469 .P
1470 The quantifier {0} is permitted, causing the expression to behave as if the
1471 previous item and the quantifier were not present. This may be useful for
1472 subpatterns that are referenced as
1473 .\" HTML <a href="#subpatternsassubroutines">
1474 .\" </a>
1475 subroutines
1476 .\"
1477 from elsewhere in the pattern (but see also the section entitled
1478 .\" HTML <a href="#subdefine">
1479 .\" </a>
1480 "Defining subpatterns for use by reference only"
1481 .\"
1482 below). Items other than subpatterns that have a {0} quantifier are omitted
1483 from the compiled pattern.
1484 .P
1485 For convenience, the three most common quantifiers have single-character
1486 abbreviations:
1487 .sp
1488 * is equivalent to {0,}
1489 + is equivalent to {1,}
1490 ? is equivalent to {0,1}
1491 .sp
1492 It is possible to construct infinite loops by following a subpattern that can
1493 match no characters with a quantifier that has no upper limit, for example:
1494 .sp
1495 (a?)*
1496 .sp
1497 Earlier versions of Perl and PCRE used to give an error at compile time for
1498 such patterns. However, because there are cases where this can be useful, such
1499 patterns are now accepted, but if any repetition of the subpattern does in fact
1500 match no characters, the loop is forcibly broken.
1501 .P
1502 By default, the quantifiers are "greedy", that is, they match as much as
1503 possible (up to the maximum number of permitted times), without causing the
1504 rest of the pattern to fail. The classic example of where this gives problems
1505 is in trying to match comments in C programs. These appear between /* and */
1506 and within the comment, individual * and / characters may appear. An attempt to
1507 match C comments by applying the pattern
1508 .sp
1509 /\e*.*\e*/
1510 .sp
1511 to the string
1512 .sp
1513 /* first comment */ not comment /* second comment */
1514 .sp
1515 fails, because it matches the entire string owing to the greediness of the .*
1516 item.
1517 .P
1518 However, if a quantifier is followed by a question mark, it ceases to be
1519 greedy, and instead matches the minimum number of times possible, so the
1520 pattern
1521 .sp
1522 /\e*.*?\e*/
1523 .sp
1524 does the right thing with the C comments. The meaning of the various
1525 quantifiers is not otherwise changed, just the preferred number of matches.
1526 Do not confuse this use of question mark with its use as a quantifier in its
1527 own right. Because it has two uses, it can sometimes appear doubled, as in
1528 .sp
1529 \ed??\ed
1530 .sp
1531 which matches one digit by preference, but can match two if that is the only
1532 way the rest of the pattern matches.
1533 .P
1534 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1535 the quantifiers are not greedy by default, but individual ones can be made
1536 greedy by following them with a question mark. In other words, it inverts the
1537 default behaviour.
1538 .P
1539 When a parenthesized subpattern is quantified with a minimum repeat count that
1540 is greater than 1 or with a limited maximum, more memory is required for the
1541 compiled pattern, in proportion to the size of the minimum or maximum.
1542 .P
1543 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1544 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1545 implicitly anchored, because whatever follows will be tried against every
1546 character position in the subject string, so there is no point in retrying the
1547 overall match at any position after the first. PCRE normally treats such a
1548 pattern as though it were preceded by \eA.
1549 .P
1550 In cases where it is known that the subject string contains no newlines, it is
1551 worth setting PCRE_DOTALL in order to obtain this optimization, or
1552 alternatively using ^ to indicate anchoring explicitly.
1553 .P
1554 However, there is one situation where the optimization cannot be used. When .*
1555 is inside capturing parentheses that are the subject of a back reference
1556 elsewhere in the pattern, a match at the start may fail where a later one
1557 succeeds. Consider, for example:
1558 .sp
1559 (.*)abc\e1
1560 .sp
1561 If the subject is "xyz123abc123" the match point is the fourth character. For
1562 this reason, such a pattern is not implicitly anchored.
1563 .P
1564 When a capturing subpattern is repeated, the value captured is the substring
1565 that matched the final iteration. For example, after
1566 .sp
1567 (tweedle[dume]{3}\es*)+
1568 .sp
1569 has matched "tweedledum tweedledee" the value of the captured substring is
1570 "tweedledee". However, if there are nested capturing subpatterns, the
1571 corresponding captured values may have been set in previous iterations. For
1572 example, after
1573 .sp
1574 /(a|(b))+/
1575 .sp
1576 matches "aba" the value of the second captured substring is "b".
1577 .
1578 .
1579 .\" HTML <a name="atomicgroup"></a>
1581 .rs
1582 .sp
1583 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1584 repetition, failure of what follows normally causes the repeated item to be
1585 re-evaluated to see if a different number of repeats allows the rest of the
1586 pattern to match. Sometimes it is useful to prevent this, either to change the
1587 nature of the match, or to cause it fail earlier than it otherwise might, when
1588 the author of the pattern knows there is no point in carrying on.
1589 .P
1590 Consider, for example, the pattern \ed+foo when applied to the subject line
1591 .sp
1592 123456bar
1593 .sp
1594 After matching all 6 digits and then failing to match "foo", the normal
1595 action of the matcher is to try again with only 5 digits matching the \ed+
1596 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1597 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1598 that once a subpattern has matched, it is not to be re-evaluated in this way.
1599 .P
1600 If we use atomic grouping for the previous example, the matcher gives up
1601 immediately on failing to match "foo" the first time. The notation is a kind of
1602 special parenthesis, starting with (?> as in this example:
1603 .sp
1604 (?>\ed+)foo
1605 .sp
1606 This kind of parenthesis "locks up" the part of the pattern it contains once
1607 it has matched, and a failure further into the pattern is prevented from
1608 backtracking into it. Backtracking past it to previous items, however, works as
1609 normal.
1610 .P
1611 An alternative description is that a subpattern of this type matches the string
1612 of characters that an identical standalone pattern would match, if anchored at
1613 the current point in the subject string.
1614 .P
1615 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1616 the above example can be thought of as a maximizing repeat that must swallow
1617 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1618 number of digits they match in order to make the rest of the pattern match,
1619 (?>\ed+) can only match an entire sequence of digits.
1620 .P
1621 Atomic groups in general can of course contain arbitrarily complicated
1622 subpatterns, and can be nested. However, when the subpattern for an atomic
1623 group is just a single repeated item, as in the example above, a simpler
1624 notation, called a "possessive quantifier" can be used. This consists of an
1625 additional + character following a quantifier. Using this notation, the
1626 previous example can be rewritten as
1627 .sp
1628 \ed++foo
1629 .sp
1630 Note that a possessive quantifier can be used with an entire group, for
1631 example:
1632 .sp
1633 (abc|xyz){2,3}+
1634 .sp
1635 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1636 option is ignored. They are a convenient notation for the simpler forms of
1637 atomic group. However, there is no difference in the meaning of a possessive
1638 quantifier and the equivalent atomic group, though there may be a performance
1639 difference; possessive quantifiers should be slightly faster.
1640 .P
1641 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1642 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1643 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1644 package, and PCRE copied it from there. It ultimately found its way into Perl
1645 at release 5.10.
1646 .P
1647 PCRE has an optimization that automatically "possessifies" certain simple
1648 pattern constructs. For example, the sequence A+B is treated as A++B because
1649 there is no point in backtracking into a sequence of A's when B must follow.
1650 .P
1651 When a pattern contains an unlimited repeat inside a subpattern that can itself
1652 be repeated an unlimited number of times, the use of an atomic group is the
1653 only way to avoid some failing matches taking a very long time indeed. The
1654 pattern
1655 .sp
1656 (\eD+|<\ed+>)*[!?]
1657 .sp
1658 matches an unlimited number of substrings that either consist of non-digits, or
1659 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1660 quickly. However, if it is applied to
1661 .sp
1662 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1663 .sp
1664 it takes a long time before reporting failure. This is because the string can
1665 be divided between the internal \eD+ repeat and the external * repeat in a
1666 large number of ways, and all have to be tried. (The example uses [!?] rather
1667 than a single character at the end, because both PCRE and Perl have an
1668 optimization that allows for fast failure when a single character is used. They
1669 remember the last single character that is required for a match, and fail early
1670 if it is not present in the string.) If the pattern is changed so that it uses
1671 an atomic group, like this:
1672 .sp
1673 ((?>\eD+)|<\ed+>)*[!?]
1674 .sp
1675 sequences of non-digits cannot be broken, and failure happens quickly.
1676 .
1677 .
1678 .\" HTML <a name="backreferences"></a>
1680 .rs
1681 .sp
1682 Outside a character class, a backslash followed by a digit greater than 0 (and
1683 possibly further digits) is a back reference to a capturing subpattern earlier
1684 (that is, to its left) in the pattern, provided there have been that many
1685 previous capturing left parentheses.
1686 .P
1687 However, if the decimal number following the backslash is less than 10, it is
1688 always taken as a back reference, and causes an error only if there are not
1689 that many capturing left parentheses in the entire pattern. In other words, the
1690 parentheses that are referenced need not be to the left of the reference for
1691 numbers less than 10. A "forward back reference" of this type can make sense
1692 when a repetition is involved and the subpattern to the right has participated
1693 in an earlier iteration.
1694 .P
1695 It is not possible to have a numerical "forward back reference" to a subpattern
1696 whose number is 10 or more using this syntax because a sequence such as \e50 is
1697 interpreted as a character defined in octal. See the subsection entitled
1698 "Non-printing characters"
1699 .\" HTML <a href="#digitsafterbackslash">
1700 .\" </a>
1701 above
1702 .\"
1703 for further details of the handling of digits following a backslash. There is
1704 no such problem when named parentheses are used. A back reference to any
1705 subpattern is possible using named parentheses (see below).
1706 .P
1707 Another way of avoiding the ambiguity inherent in the use of digits following a
1708 backslash is to use the \eg escape sequence. This escape must be followed by an
1709 unsigned number or a negative number, optionally enclosed in braces. These
1710 examples are all identical:
1711 .sp
1712 (ring), \e1
1713 (ring), \eg1
1714 (ring), \eg{1}
1715 .sp
1716 An unsigned number specifies an absolute reference without the ambiguity that
1717 is present in the older syntax. It is also useful when literal digits follow
1718 the reference. A negative number is a relative reference. Consider this
1719 example:
1720 .sp
1721 (abc(def)ghi)\eg{-1}
1722 .sp
1723 The sequence \eg{-1} is a reference to the most recently started capturing
1724 subpattern before \eg, that is, is it equivalent to \e2 in this example.
1725 Similarly, \eg{-2} would be equivalent to \e1. The use of relative references
1726 can be helpful in long patterns, and also in patterns that are created by
1727 joining together fragments that contain references within themselves.
1728 .P
1729 A back reference matches whatever actually matched the capturing subpattern in
1730 the current subject string, rather than anything matching the subpattern
1731 itself (see
1732 .\" HTML <a href="#subpatternsassubroutines">
1733 .\" </a>
1734 "Subpatterns as subroutines"
1735 .\"
1736 below for a way of doing that). So the pattern
1737 .sp
1738 (sens|respons)e and \e1ibility
1739 .sp
1740 matches "sense and sensibility" and "response and responsibility", but not
1741 "sense and responsibility". If caseful matching is in force at the time of the
1742 back reference, the case of letters is relevant. For example,
1743 .sp
1744 ((?i)rah)\es+\e1
1745 .sp
1746 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1747 capturing subpattern is matched caselessly.
1748 .P
1749 There are several different ways of writing back references to named
1750 subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
1751 \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1752 back reference syntax, in which \eg can be used for both numeric and named
1753 references, is also supported. We could rewrite the above example in any of
1754 the following ways:
1755 .sp
1756 (?<p1>(?i)rah)\es+\ek<p1>
1757 (?'p1'(?i)rah)\es+\ek{p1}
1758 (?P<p1>(?i)rah)\es+(?P=p1)
1759 (?<p1>(?i)rah)\es+\eg{p1}
1760 .sp
1761 A subpattern that is referenced by name may appear in the pattern before or
1762 after the reference.
1763 .P
1764 There may be more than one back reference to the same subpattern. If a
1765 subpattern has not actually been used in a particular match, any back
1766 references to it always fail by default. For example, the pattern
1767 .sp
1768 (a|(bc))\e2
1769 .sp
1770 always fails if it starts to match "a" rather than "bc". However, if the
1771 PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
1772 unset value matches an empty string.
1773 .P
1774 Because there may be many capturing parentheses in a pattern, all digits
1775 following a backslash are taken as part of a potential back reference number.
1776 If the pattern continues with a digit character, some delimiter must be used to
1777 terminate the back reference. If the PCRE_EXTENDED option is set, this can be
1778 whitespace. Otherwise, the \eg{ syntax or an empty comment (see
1779 .\" HTML <a href="#comments">
1780 .\" </a>
1781 "Comments"
1782 .\"
1783 below) can be used.
1784 .
1785 .SS "Recursive back references"
1786 .rs
1787 .sp
1788 A back reference that occurs inside the parentheses to which it refers fails
1789 when the subpattern is first used, so, for example, (a\e1) never matches.
1790 However, such references can be useful inside repeated subpatterns. For
1791 example, the pattern
1792 .sp
1793 (a|b\e1)+
1794 .sp
1795 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1796 the subpattern, the back reference matches the character string corresponding
1797 to the previous iteration. In order for this to work, the pattern must be such
1798 that the first iteration does not need to match the back reference. This can be
1799 done using alternation, as in the example above, or by a quantifier with a
1800 minimum of zero.
1801 .P
1802 Back references of this type cause the group that they reference to be treated
1803 as an
1804 .\" HTML <a href="#atomicgroup">
1805 .\" </a>
1806 atomic group.
1807 .\"
1808 Once the whole group has been matched, a subsequent matching failure cannot
1809 cause backtracking into the middle of the group.
1810 .
1811 .
1812 .\" HTML <a name="bigassertions"></a>
1814 .rs
1815 .sp
1816 An assertion is a test on the characters following or preceding the current
1817 matching point that does not actually consume any characters. The simple
1818 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
1819 .\" HTML <a href="#smallassertions">
1820 .\" </a>
1821 above.
1822 .\"
1823 .P
1824 More complicated assertions are coded as subpatterns. There are two kinds:
1825 those that look ahead of the current position in the subject string, and those
1826 that look behind it. An assertion subpattern is matched in the normal way,
1827 except that it does not cause the current matching position to be changed.
1828 .P
1829 Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1830 because it makes no sense to assert the same thing several times. If any kind
1831 of assertion contains capturing subpatterns within it, these are counted for
1832 the purposes of numbering the capturing subpatterns in the whole pattern.
1833 However, substring capturing is carried out only for positive assertions,
1834 because it does not make sense for negative assertions.
1835 .
1836 .
1837 .SS "Lookahead assertions"
1838 .rs
1839 .sp
1840 Lookahead assertions start with (?= for positive assertions and (?! for
1841 negative assertions. For example,
1842 .sp
1843 \ew+(?=;)
1844 .sp
1845 matches a word followed by a semicolon, but does not include the semicolon in
1846 the match, and
1847 .sp
1848 foo(?!bar)
1849 .sp
1850 matches any occurrence of "foo" that is not followed by "bar". Note that the
1851 apparently similar pattern
1852 .sp
1853 (?!foo)bar
1854 .sp
1855 does not find an occurrence of "bar" that is preceded by something other than
1856 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1857 (?!foo) is always true when the next three characters are "bar". A
1858 lookbehind assertion is needed to achieve the other effect.
1859 .P
1860 If you want to force a matching failure at some point in a pattern, the most
1861 convenient way to do it is with (?!) because an empty string always matches, so
1862 an assertion that requires there not to be an empty string must always fail.
1863 The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
1864 .
1865 .
1866 .\" HTML <a name="lookbehind"></a>
1867 .SS "Lookbehind assertions"
1868 .rs
1869 .sp
1870 Lookbehind assertions start with (?<= for positive assertions and (?<! for
1871 negative assertions. For example,
1872 .sp
1873 (?<!foo)bar
1874 .sp
1875 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1876 a lookbehind assertion are restricted such that all the strings it matches must
1877 have a fixed length. However, if there are several top-level alternatives, they
1878 do not all have to have the same fixed length. Thus
1879 .sp
1880 (?<=bullock|donkey)
1881 .sp
1882 is permitted, but
1883 .sp
1884 (?<!dogs?|cats?)
1885 .sp
1886 causes an error at compile time. Branches that match different length strings
1887 are permitted only at the top level of a lookbehind assertion. This is an
1888 extension compared with Perl, which requires all branches to match the same
1889 length of string. An assertion such as
1890 .sp
1891 (?<=ab(c|de))
1892 .sp
1893 is not permitted, because its single top-level branch can match two different
1894 lengths, but it is acceptable to PCRE if rewritten to use two top-level
1895 branches:
1896 .sp
1897 (?<=abc|abde)
1898 .sp
1899 In some cases, the escape sequence \eK
1900 .\" HTML <a href="#resetmatchstart">
1901 .\" </a>
1902 (see above)
1903 .\"
1904 can be used instead of a lookbehind assertion to get round the fixed-length
1905 restriction.
1906 .P
1907 The implementation of lookbehind assertions is, for each alternative, to
1908 temporarily move the current position back by the fixed length and then try to
1909 match. If there are insufficient characters before the current position, the
1910 assertion fails.
1911 .P
1912 PCRE does not allow the \eC escape (which matches a single byte in UTF-8 mode)
1913 to appear in lookbehind assertions, because it makes it impossible to calculate
1914 the length of the lookbehind. The \eX and \eR escapes, which can match
1915 different numbers of bytes, are also not permitted.
1916 .P
1917 .\" HTML <a href="#subpatternsassubroutines">
1918 .\" </a>
1919 "Subroutine"
1920 .\"
1921 calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
1922 as the subpattern matches a fixed-length string.
1923 .\" HTML <a href="#recursion">
1924 .\" </a>
1925 Recursion,
1926 .\"
1927 however, is not supported.
1928 .P
1929 Possessive quantifiers can be used in conjunction with lookbehind assertions to
1930 specify efficient matching of fixed-length strings at the end of subject
1931 strings. Consider a simple pattern such as
1932 .sp
1933 abcd$
1934 .sp
1935 when applied to a long string that does not match. Because matching proceeds
1936 from left to right, PCRE will look for each "a" in the subject and then see if
1937 what follows matches the rest of the pattern. If the pattern is specified as
1938 .sp
1939 ^.*abcd$
1940 .sp
1941 the initial .* matches the entire string at first, but when this fails (because
1942 there is no following "a"), it backtracks to match all but the last character,
1943 then all but the last two characters, and so on. Once again the search for "a"
1944 covers the entire string, from right to left, so we are no better off. However,
1945 if the pattern is written as
1946 .sp
1947 ^.*+(?<=abcd)
1948 .sp
1949 there can be no backtracking for the .*+ item; it can match only the entire
1950 string. The subsequent lookbehind assertion does a single test on the last four
1951 characters. If it fails, the match fails immediately. For long strings, this
1952 approach makes a significant difference to the processing time.
1953 .
1954 .
1955 .SS "Using multiple assertions"
1956 .rs
1957 .sp
1958 Several assertions (of any sort) may occur in succession. For example,
1959 .sp
1960 (?<=\ed{3})(?<!999)foo
1961 .sp
1962 matches "foo" preceded by three digits that are not "999". Notice that each of
1963 the assertions is applied independently at the same point in the subject
1964 string. First there is a check that the previous three characters are all
1965 digits, and then there is a check that the same three characters are not "999".
1966 This pattern does \fInot\fP match "foo" preceded by six characters, the first
1967 of which are digits and the last three of which are not "999". For example, it
1968 doesn't match "123abcfoo". A pattern to do that is
1969 .sp
1970 (?<=\ed{3}...)(?<!999)foo
1971 .sp
1972 This time the first assertion looks at the preceding six characters, checking
1973 that the first three are digits, and then the second assertion checks that the
1974 preceding three characters are not "999".
1975 .P
1976 Assertions can be nested in any combination. For example,
1977 .sp
1978 (?<=(?<!foo)bar)baz
1979 .sp
1980 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1981 preceded by "foo", while
1982 .sp
1983 (?<=\ed{3}(?!999)...)foo
1984 .sp
1985 is another pattern that matches "foo" preceded by three digits and any three
1986 characters that are not "999".
1987 .
1988 .
1989 .\" HTML <a name="conditions"></a>
1991 .rs
1992 .sp
1993 It is possible to cause the matching process to obey a subpattern
1994 conditionally or to choose between two alternative subpatterns, depending on
1995 the result of an assertion, or whether a specific capturing subpattern has
1996 already been matched. The two possible forms of conditional subpattern are:
1997 .sp
1998 (?(condition)yes-pattern)
1999 (?(condition)yes-pattern|no-pattern)
2000 .sp
2001 If the condition is satisfied, the yes-pattern is used; otherwise the
2002 no-pattern (if present) is used. If there are more than two alternatives in the
2003 subpattern, a compile-time error occurs. Each of the two alternatives may
2004 itself contain nested subpatterns of any form, including conditional
2005 subpatterns; the restriction to two alternatives applies only at the level of
2006 the condition. This pattern fragment is an example where the alternatives are
2007 complex:
2008 .sp
2009 (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
2010 .sp
2011 .P
2012 There are four kinds of condition: references to subpatterns, references to
2013 recursion, a pseudo-condition called DEFINE, and assertions.
2014 .
2015 .SS "Checking for a used subpattern by number"
2016 .rs
2017 .sp
2018 If the text between the parentheses consists of a sequence of digits, the
2019 condition is true if a capturing subpattern of that number has previously
2020 matched. If there is more than one capturing subpattern with the same number
2021 (see the earlier
2022 .\"
2023 .\" HTML <a href="#recursion">
2024 .\" </a>
2025 section about duplicate subpattern numbers),
2026 .\"
2027 the condition is true if any of them have matched. An alternative notation is
2028 to precede the digits with a plus or minus sign. In this case, the subpattern
2029 number is relative rather than absolute. The most recently opened parentheses
2030 can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
2031 loops it can also make sense to refer to subsequent groups. The next
2032 parentheses to be opened can be referenced as (?(+1), and so on. (The value
2033 zero in any of these forms is not used; it provokes a compile-time error.)
2034 .P
2035 Consider the following pattern, which contains non-significant white space to
2036 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2037 three parts for ease of discussion:
2038 .sp
2039 ( \e( )? [^()]+ (?(1) \e) )
2040 .sp
2041 The first part matches an optional opening parenthesis, and if that
2042 character is present, sets it as the first captured substring. The second part
2043 matches one or more characters that are not parentheses. The third part is a
2044 conditional subpattern that tests whether or not the first set of parentheses
2045 matched. If they did, that is, if subject started with an opening parenthesis,
2046 the condition is true, and so the yes-pattern is executed and a closing
2047 parenthesis is required. Otherwise, since no-pattern is not present, the
2048 subpattern matches nothing. In other words, this pattern matches a sequence of
2049 non-parentheses, optionally enclosed in parentheses.
2050 .P
2051 If you were embedding this pattern in a larger one, you could use a relative
2052 reference:
2053 .sp
2054 ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
2055 .sp
2056 This makes the fragment independent of the parentheses in the larger pattern.
2057 .
2058 .SS "Checking for a used subpattern by name"
2059 .rs
2060 .sp
2061 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
2062 subpattern by name. For compatibility with earlier versions of PCRE, which had
2063 this facility before Perl, the syntax (?(name)...) is also recognized. However,
2064 there is a possible ambiguity with this syntax, because subpattern names may
2065 consist entirely of digits. PCRE looks first for a named subpattern; if it
2066 cannot find one and the name consists entirely of digits, PCRE looks for a
2067 subpattern of that number, which must be greater than zero. Using subpattern
2068 names that consist entirely of digits is not recommended.
2069 .P
2070 Rewriting the above example to use a named subpattern gives this:
2071 .sp
2072 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
2073 .sp
2074 If the name used in a condition of this kind is a duplicate, the test is
2075 applied to all subpatterns of the same name, and is true if any one of them has
2076 matched.
2077 .
2078 .SS "Checking for pattern recursion"
2079 .rs
2080 .sp
2081 If the condition is the string (R), and there is no subpattern with the name R,
2082 the condition is true if a recursive call to the whole pattern or any
2083 subpattern has been made. If digits or a name preceded by ampersand follow the
2084 letter R, for example:
2085 .sp
2086 (?(R3)...) or (?(R&name)...)
2087 .sp
2088 the condition is true if the most recent recursion is into a subpattern whose
2089 number or name is given. This condition does not check the entire recursion
2090 stack. If the name used in a condition of this kind is a duplicate, the test is
2091 applied to all subpatterns of the same name, and is true if any one of them is
2092 the most recent recursion.
2093 .P
2094 At "top level", all these recursion test conditions are false.
2095 .\" HTML <a href="#recursion">
2096 .\" </a>
2097 The syntax for recursive patterns
2098 .\"
2099 is described below.
2100 .
2101 .\" HTML <a name="subdefine"></a>
2102 .SS "Defining subpatterns for use by reference only"
2103 .rs
2104 .sp
2105 If the condition is the string (DEFINE), and there is no subpattern with the
2106 name DEFINE, the condition is always false. In this case, there may be only one
2107 alternative in the subpattern. It is always skipped if control reaches this
2108 point in the pattern; the idea of DEFINE is that it can be used to define
2109 "subroutines" that can be referenced from elsewhere. (The use of
2110 .\" HTML <a href="#subpatternsassubroutines">
2111 .\" </a>
2112 "subroutines"
2113 .\"
2114 is described below.) For example, a pattern to match an IPv4 address such as
2115 "" could be written like this (ignore whitespace and line
2116 breaks):
2117 .sp
2118 (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
2119 \eb (?&byte) (\e.(?&byte)){3} \eb
2120 .sp
2121 The first part of the pattern is a DEFINE group inside which a another group
2122 named "byte" is defined. This matches an individual component of an IPv4
2123 address (a number less than 256). When matching takes place, this part of the
2124 pattern is skipped because DEFINE acts like a false condition. The rest of the
2125 pattern uses references to the named group to match the four dot-separated
2126 components of an IPv4 address, insisting on a word boundary at each end.
2127 .
2128 .SS "Assertion conditions"
2129 .rs
2130 .sp
2131 If the condition is not in any of the above formats, it must be an assertion.
2132 This may be a positive or negative lookahead or lookbehind assertion. Consider
2133 this pattern, again containing non-significant white space, and with the two
2134 alternatives on the second line:
2135 .sp
2136 (?(?=[^a-z]*[a-z])
2137 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
2138 .sp
2139 The condition is a positive lookahead assertion that matches an optional
2140 sequence of non-letters followed by a letter. In other words, it tests for the
2141 presence of at least one letter in the subject. If a letter is found, the
2142 subject is matched against the first alternative; otherwise it is matched
2143 against the second. This pattern matches strings in one of the two forms
2144 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2145 .
2146 .
2147 .\" HTML <a name="comments"></a>
2149 .rs
2150 .sp
2151 There are two ways of including comments in patterns that are processed by
2152 PCRE. In both cases, the start of the comment must not be in a character class,
2153 nor in the middle of any other sequence of related characters such as (?: or a
2154 subpattern name or number. The characters that make up a comment play no part
2155 in the pattern matching.
2156 .P
2157 The sequence (?# marks the start of a comment that continues up to the next
2158 closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2159 option is set, an unescaped # character also introduces a comment, which in
2160 this case continues to immediately after the next newline character or
2161 character sequence in the pattern. Which characters are interpreted as newlines
2162 is controlled by the options passed to \fBpcre_compile()\fP or by a special
2163 sequence at the start of the pattern, as described in the section entitled
2164 .\" HTML <a href="#newlines">
2165 .\" </a>
2166 "Newline conventions"
2167 .\"
2168 above. Note that the end of this type of comment is a literal newline sequence
2169 in the pattern; escape sequences that happen to represent a newline do not
2170 count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2171 default newline convention is in force:
2172 .sp
2173 abc #comment \en still comment
2174 .sp
2175 On encountering the # character, \fBpcre_compile()\fP skips along, looking for
2176 a newline in the pattern. The sequence \en is still literal at this stage, so
2177 it does not terminate the comment. Only an actual character with the code value
2178 0x0a (the default newline) does so.
2179 .
2180 .
2181 .\" HTML <a name="recursion"></a>
2183 .rs
2184 .sp
2185 Consider the problem of matching a string in parentheses, allowing for
2186 unlimited nested parentheses. Without the use of recursion, the best that can
2187 be done is to use a pattern that matches up to some fixed depth of nesting. It
2188 is not possible to handle an arbitrary nesting depth.
2189 .P
2190 For some time, Perl has provided a facility that allows regular expressions to
2191 recurse (amongst other things). It does this by interpolating Perl code in the
2192 expression at run time, and the code can refer to the expression itself. A Perl
2193 pattern using code interpolation to solve the parentheses problem can be
2194 created like this:
2195 .sp
2196 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
2197 .sp
2198 The (?p{...}) item interpolates Perl code at run time, and in this case refers
2199 recursively to the pattern in which it appears.
2200 .P
2201 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2202 supports special syntax for recursion of the entire pattern, and also for
2203 individual subpattern recursion. After its introduction in PCRE and Python,
2204 this kind of recursion was subsequently introduced into Perl at release 5.10.
2205 .P
2206 A special item that consists of (? followed by a number greater than zero and a
2207 closing parenthesis is a recursive call of the subpattern of the given number,
2208 provided that it occurs inside that subpattern. (If not, it is a
2209 .\" HTML <a href="#subpatternsassubroutines">
2210 .\" </a>
2211 "subroutine"
2212 .\"
2213 call, which is described in the next section.) The special item (?R) or (?0) is
2214 a recursive call of the entire regular expression.
2215 .P
2216 This PCRE pattern solves the nested parentheses problem (assume the
2217 PCRE_EXTENDED option is set so that white space is ignored):
2218 .sp
2219 \e( ( [^()]++ | (?R) )* \e)
2220 .sp
2221 First it matches an opening parenthesis. Then it matches any number of
2222 substrings which can either be a sequence of non-parentheses, or a recursive
2223 match of the pattern itself (that is, a correctly parenthesized substring).
2224 Finally there is a closing parenthesis. Note the use of a possessive quantifier
2225 to avoid backtracking into sequences of non-parentheses.
2226 .P
2227 If this were part of a larger pattern, you would not want to recurse the entire
2228 pattern, so instead you could use this:
2229 .sp
2230 ( \e( ( [^()]++ | (?1) )* \e) )
2231 .sp
2232 We have put the pattern into parentheses, and caused the recursion to refer to
2233 them instead of the whole pattern.
2234 .P
2235 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2236 is made easier by the use of relative references. Instead of (?1) in the
2237 pattern above you can write (?-2) to refer to the second most recently opened
2238 parentheses preceding the recursion. In other words, a negative number counts
2239 capturing parentheses leftwards from the point at which it is encountered.
2240 .P
2241 It is also possible to refer to subsequently opened parentheses, by writing
2242 references such as (?+2). However, these cannot be recursive because the
2243 reference is not inside the parentheses that are referenced. They are always
2244 .\" HTML <a href="#subpatternsassubroutines">
2245 .\" </a>
2246 "subroutine"
2247 .\"
2248 calls, as described in the next section.
2249 .P
2250 An alternative approach is to use named parentheses instead. The Perl syntax
2251 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
2252 could rewrite the above example as follows:
2253 .sp
2254 (?<pn> \e( ( [^()]++ | (?&pn) )* \e) )
2255 .sp
2256 If there is more than one subpattern with the same name, the earliest one is
2257 used.
2258 .P
2259 This particular example pattern that we have been looking at contains nested
2260 unlimited repeats, and so the use of a possessive quantifier for matching
2261 strings of non-parentheses is important when applying the pattern to strings
2262 that do not match. For example, when this pattern is applied to
2263 .sp
2264 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2265 .sp
2266 it yields "no match" quickly. However, if a possessive quantifier is not used,
2267 the match runs for a very long time indeed because there are so many different
2268 ways the + and * repeats can carve up the subject, and all have to be tested
2269 before failure can be reported.
2270 .P
2271 At the end of a match, the values of capturing parentheses are those from
2272 the outermost level. If you want to obtain intermediate values, a callout
2273 function can be used (see below and the
2274 .\" HREF
2275 \fBpcrecallout\fP
2276 .\"
2277 documentation). If the pattern above is matched against
2278 .sp
2279 (ab(cd)ef)
2280 .sp
2281 the value for the inner capturing parentheses (numbered 2) is "ef", which is
2282 the last value taken on at the top level. If a capturing subpattern is not
2283 matched at the top level, its final value is unset, even if it is (temporarily)
2284 set at a deeper level.
2285 .P
2286 If there are more than 15 capturing parentheses in a pattern, PCRE has to
2287 obtain extra memory to store data during a recursion, which it does by using
2288 \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no memory can
2289 be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2290 .P
2291 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2292 Consider this pattern, which matches text in angle brackets, allowing for
2293 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2294 recursing), whereas any characters are permitted at the outer level.
2295 .sp
2296 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
2297 .sp
2298 In this pattern, (?(R) is the start of a conditional subpattern, with two
2299 different alternatives for the recursive and non-recursive cases. The (?R) item
2300 is the actual recursive call.
2301 .
2302 .
2303 .\" HTML <a name="recursiondifference"></a>
2304 .SS "Recursion difference from Perl"
2305 .rs
2306 .sp
2307 In PCRE (like Python, but unlike Perl), a recursive subpattern call is always
2308 treated as an atomic group. That is, once it has matched some of the subject
2309 string, it is never re-entered, even if it contains untried alternatives and
2310 there is a subsequent matching failure. This can be illustrated by the
2311 following pattern, which purports to match a palindromic string that contains
2312 an odd number of characters (for example, "a", "aba", "abcba", "abcdcba"):
2313 .sp
2314 ^(.|(.)(?1)\e2)$
2315 .sp
2316 The idea is that it either matches a single character, or two identical
2317 characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2318 it does not if the pattern is longer than three characters. Consider the
2319 subject string "abcba":
2320 .P
2321 At the top level, the first character is matched, but as it is not at the end
2322 of the string, the first alternative fails; the second alternative is taken
2323 and the recursion kicks in. The recursive call to subpattern 1 successfully
2324 matches the next character ("b"). (Note that the beginning and end of line
2325 tests are not part of the recursion).
2326 .P
2327 Back at the top level, the next character ("c") is compared with what
2328 subpattern 2 matched, which was "a". This fails. Because the recursion is
2329 treated as an atomic group, there are now no backtracking points, and so the
2330 entire match fails. (Perl is able, at this point, to re-enter the recursion and
2331 try the second alternative.) However, if the pattern is written with the
2332 alternatives in the other order, things are different:
2333 .sp
2334 ^((.)(?1)\e2|.)$
2335 .sp
2336 This time, the recursing alternative is tried first, and continues to recurse
2337 until it runs out of characters, at which point the recursion fails. But this
2338 time we do have another alternative to try at the higher level. That is the big
2339 difference: in the previous case the remaining alternative is at a deeper
2340 recursion level, which PCRE cannot use.
2341 .P
2342 To change the pattern so that it matches all palindromic strings, not just
2343 those with an odd number of characters, it is tempting to change the pattern to
2344 this:
2345 .sp
2346 ^((.)(?1)\e2|.?)$
2347 .sp
2348 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2349 deeper recursion has matched a single character, it cannot be entered again in
2350 order to match an empty string. The solution is to separate the two cases, and
2351 write out the odd and even cases as alternatives at the higher level:
2352 .sp
2353 ^(?:((.)(?1)\e2|)|((.)(?3)\e4|.))
2354 .sp
2355 If you want to match typical palindromic phrases, the pattern has to ignore all
2356 non-word characters, which can be done like this:
2357 .sp
2358 ^\eW*+(?:((.)\eW*+(?1)\eW*+\e2|)|((.)\eW*+(?3)\eW*+\e4|\eW*+.\eW*+))\eW*+$
2359 .sp
2360 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2361 man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2362 the use of the possessive quantifier *+ to avoid backtracking into sequences of
2363 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2364 more) to match typical phrases, and Perl takes so long that you think it has
2365 gone into a loop.
2366 .P
2367 \fBWARNING\fP: The palindrome-matching patterns above work only if the subject
2368 string does not start with a palindrome that is shorter than the entire string.
2369 For example, although "abcba" is correctly matched, if the subject is "ababa",
2370 PCRE finds the palindrome "aba" at the start, then fails at top level because
2371 the end of the string does not follow. Once again, it cannot jump back into the
2372 recursion to try other alternatives, so the entire match fails.
2373 .
2374 .
2375 .\" HTML <a name="subpatternsassubroutines"></a>
2377 .rs
2378 .sp
2379 If the syntax for a recursive subpattern reference (either by number or by
2380 name) is used outside the parentheses to which it refers, it operates like a
2381 subroutine in a programming language. The "called" subpattern may be defined
2382 before or after the reference. A numbered reference can be absolute or
2383 relative, as in these examples:
2384 .sp
2385 (...(absolute)...)...(?2)...
2386 (...(relative)...)...(?-1)...
2387 (...(?+1)...(relative)...
2388 .sp
2389 An earlier example pointed out that the pattern
2390 .sp
2391 (sens|respons)e and \e1ibility
2392 .sp
2393 matches "sense and sensibility" and "response and responsibility", but not
2394 "sense and responsibility". If instead the pattern
2395 .sp
2396 (sens|respons)e and (?1)ibility
2397 .sp
2398 is used, it does match "sense and responsibility" as well as the other two
2399 strings. Another example is given in the discussion of DEFINE above.
2400 .P
2401 Like recursive subpatterns, a subroutine call is always treated as an atomic
2402 group. That is, once it has matched some of the subject string, it is never
2403 re-entered, even if it contains untried alternatives and there is a subsequent
2404 matching failure. Any capturing parentheses that are set during the subroutine
2405 call revert to their previous values afterwards.
2406 .P
2407 When a subpattern is used as a subroutine, processing options such as
2408 case-independence are fixed when the subpattern is defined. They cannot be
2409 changed for different calls. For example, consider this pattern:
2410 .sp
2411 (abc)(?i:(?-1))
2412 .sp
2413 It matches "abcabc". It does not match "abcABC" because the change of
2414 processing option does not affect the called subpattern.
2415 .
2416 .
2417 .\" HTML <a name="onigurumasubroutines"></a>
2419 .rs
2420 .sp
2421 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
2422 a number enclosed either in angle brackets or single quotes, is an alternative
2423 syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2424 are two of the examples used above, rewritten using this syntax:
2425 .sp
2426 (?<pn> \e( ( (?>[^()]+) | \eg<pn> )* \e) )
2427 (sens|respons)e and \eg'1'ibility
2428 .sp
2429 PCRE supports an extension to Oniguruma: if a number is preceded by a
2430 plus or a minus sign it is taken as a relative reference. For example:
2431 .sp
2432 (abc)(?i:\eg<-1>)
2433 .sp
2434 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
2435 synonymous. The former is a back reference; the latter is a subroutine call.
2436 .
2437 .
2439 .rs
2440 .sp
2441 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2442 code to be obeyed in the middle of matching a regular expression. This makes it
2443 possible, amongst other things, to extract different substrings that match the
2444 same pair of parentheses when there is a repetition.
2445 .P
2446 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2447 code. The feature is called "callout". The caller of PCRE provides an external
2448 function by putting its entry point in the global variable \fIpcre_callout\fP.
2449 By default, this variable contains NULL, which disables all calling out.
2450 .P
2451 Within a regular expression, (?C) indicates the points at which the external
2452 function is to be called. If you want to identify different callout points, you
2453 can put a number less than 256 after the letter C. The default value is zero.
2454 For example, this pattern has two callout points:
2455 .sp
2456 (?C1)abc(?C2)def
2457 .sp
2458 If the PCRE_AUTO_CALLOUT flag is passed to \fBpcre_compile()\fP, callouts are
2459 automatically installed before each item in the pattern. They are all numbered
2460 255.
2461 .P
2462 During matching, when PCRE reaches a callout point (and \fIpcre_callout\fP is
2463 set), the external function is called. It is provided with the number of the
2464 callout, the position in the pattern, and, optionally, one item of data
2465 originally supplied by the caller of \fBpcre_exec()\fP. The callout function
2466 may cause matching to proceed, to backtrack, or to fail altogether. A complete
2467 description of the interface to the callout function is given in the
2468 .\" HREF
2469 \fBpcrecallout\fP
2470 .\"
2471 documentation.
2472 .
2473 .
2474 .\" HTML <a name="backtrackcontrol"></a>
2476 .rs
2477 .sp
2478 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2479 are described in the Perl documentation as "experimental and subject to change
2480 or removal in a future version of Perl". It goes on to say: "Their usage in
2481 production code should be noted to avoid problems during upgrades." The same
2482 remarks apply to the PCRE features described in this section.
2483 .P
2484 Since these verbs are specifically related to backtracking, most of them can be
2485 used only when the pattern is to be matched using \fBpcre_exec()\fP, which uses
2486 a backtracking algorithm. With the exception of (*FAIL), which behaves like a
2487 failing negative assertion, they cause an error if encountered by
2488 \fBpcre_dfa_exec()\fP.
2489 .P
2490 If any of these verbs are used in an assertion or subroutine subpattern
2491 (including recursive subpatterns), their effect is confined to that subpattern;
2492 it does not extend to the surrounding pattern. Note that such subpatterns are
2493 processed as anchored at the point where they are tested.
2494 .P
2495 The new verbs make use of what was previously invalid syntax: an opening
2496 parenthesis followed by an asterisk. They are generally of the form
2497 (*VERB) or (*VERB:NAME). Some may take either form, with differing behaviour,
2498 depending on whether or not an argument is present. An name is a sequence of
2499 letters, digits, and underscores. If the name is empty, that is, if the closing
2500 parenthesis immediately follows the colon, the effect is as if the colon were
2501 not there. Any number of these verbs may occur in a pattern.
2502 .P
2503 PCRE contains some optimizations that are used to speed up matching by running
2504 some checks at the start of each match attempt. For example, it may know the
2505 minimum length of matching subject, or that a particular character must be
2506 present. When one of these optimizations suppresses the running of a match, any
2507 included backtracking verbs will not, of course, be processed. You can suppress
2508 the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2509 when calling \fBpcre_compile()\fP or \fBpcre_exec()\fP, or by starting the
2510 pattern with (*NO_START_OPT).
2511 .
2512 .
2513 .SS "Verbs that act immediately"
2514 .rs
2515 .sp
2516 The following verbs act as soon as they are encountered. They may not be
2517 followed by a name.
2518 .sp
2519 (*ACCEPT)
2520 .sp
2521 This verb causes the match to end successfully, skipping the remainder of the
2522 pattern. When inside a recursion, only the innermost pattern is ended
2523 immediately. If (*ACCEPT) is inside capturing parentheses, the data so far is
2524 captured. (This feature was added to PCRE at release 8.00.) For example:
2525 .sp
2526 A((?:A|B(*ACCEPT)|C)D)
2527 .sp
2528 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2529 the outer parentheses.
2530 .sp
2531 (*FAIL) or (*F)
2532 .sp
2533 This verb causes the match to fail, forcing backtracking to occur. It is
2534 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2535 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2536 Perl features that are not present in PCRE. The nearest equivalent is the
2537 callout feature, as for example in this pattern:
2538 .sp
2539 a+(?C)(*FAIL)
2540 .sp
2541 A match with the string "aaaa" always fails, but the callout is taken before
2542 each backtrack happens (in this example, 10 times).
2543 .
2544 .
2545 .SS "Recording which path was taken"
2546 .rs
2547 .sp
2548 There is one verb whose main purpose is to track how a match was arrived at,
2549 though it also has a secondary use in conjunction with advancing the match
2550 starting point (see (*SKIP) below).
2551 .sp
2552 (*MARK:NAME) or (*:NAME)
2553 .sp
2554 A name is always required with this verb. There may be as many instances of
2555 (*MARK) as you like in a pattern, and their names do not have to be unique.
2556 .P
2557 When a match succeeds, the name of the last-encountered (*MARK) is passed back
2558 to the caller via the \fIpcre_extra\fP data structure, as described in the
2559 .\" HTML <a href="pcreapi.html#extradata">
2560 .\" </a>
2561 section on \fIpcre_extra\fP
2562 .\"
2563 in the
2564 .\" HREF
2565 \fBpcreapi\fP
2566 .\"
2567 documentation. No data is returned for a partial match. Here is an example of
2568 \fBpcretest\fP output, where the /K modifier requests the retrieval and
2569 outputting of (*MARK) data:
2570 .sp
2571 /X(*MARK:A)Y|X(*MARK:B)Z/K
2572 XY
2573 0: XY
2574 MK: A
2575 XZ
2576 0: XZ
2577 MK: B
2578 .sp
2579 The (*MARK) name is tagged with "MK:" in this output, and in this example it
2580 indicates which of the two alternatives matched. This is a more efficient way
2581 of obtaining this information than putting each alternative in its own
2582 capturing parentheses.
2583 .P
2584 A name may also be returned after a failed match if the final path through the
2585 pattern involves (*MARK). However, unless (*MARK) used in conjunction with
2586 (*COMMIT), this is unlikely to happen for an unanchored pattern because, as the
2587 starting point for matching is advanced, the final check is often with an empty
2588 string, causing a failure before (*MARK) is reached. For example:
2589 .sp
2590 /X(*MARK:A)Y|X(*MARK:B)Z/K
2591 XP
2592 No match
2593 .sp
2594 There are three potential starting points for this match (starting with X,
2595 starting with P, and with an empty string). If the pattern is anchored, the
2596 result is different:
2597 .sp
2598 /^X(*MARK:A)Y|^X(*MARK:B)Z/K
2599 XP
2600 No match, mark = B
2601 .sp
2602 PCRE's start-of-match optimizations can also interfere with this. For example,
2603 if, as a result of a call to \fBpcre_study()\fP, it knows the minimum
2604 subject length for a match, a shorter subject will not be scanned at all.
2605 .P
2606 Note that similar anomalies (though different in detail) exist in Perl, no
2607 doubt for the same reasons. The use of (*MARK) data after a failed match of an
2608 unanchored pattern is not recommended, unless (*COMMIT) is involved.
2609 .
2610 .
2611 .SS "Verbs that act after backtracking"
2612 .rs
2613 .sp
2614 The following verbs do nothing when they are encountered. Matching continues
2615 with what follows, but if there is no subsequent match, causing a backtrack to
2616 the verb, a failure is forced. That is, backtracking cannot pass to the left of
2617 the verb. However, when one of these verbs appears inside an atomic group, its
2618 effect is confined to that group, because once the group has been matched,
2619 there is never any backtracking into it. In this situation, backtracking can
2620 "jump back" to the left of the entire atomic group. (Remember also, as stated
2621 above, that this localization also applies in subroutine calls and assertions.)
2622 .P
2623 These verbs differ in exactly what kind of failure occurs when backtracking
2624 reaches them.
2625 .sp
2626 (*COMMIT)
2627 .sp
2628 This verb, which may not be followed by a name, causes the whole match to fail
2629 outright if the rest of the pattern does not match. Even if the pattern is
2630 unanchored, no further attempts to find a match by advancing the starting point
2631 take place. Once (*COMMIT) has been passed, \fBpcre_exec()\fP is committed to
2632 finding a match at the current starting point, or not at all. For example:
2633 .sp
2634 a+(*COMMIT)b
2635 .sp
2636 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2637 dynamic anchor, or "I've started, so I must finish." The name of the most
2638 recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
2639 match failure.
2640 .P
2641 Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
2642 unless PCRE's start-of-match optimizations are turned off, as shown in this
2643 \fBpcretest\fP example:
2644 .sp
2645 /(*COMMIT)abc/
2646 xyzabc
2647 0: abc
2648 xyzabc\eY
2649 No match
2650 .sp
2651 PCRE knows that any match must start with "a", so the optimization skips along
2652 the subject to "a" before running the first match attempt, which succeeds. When
2653 the optimization is disabled by the \eY escape in the second subject, the match
2654 starts at "x" and so the (*COMMIT) causes it to fail without trying any other
2655 starting points.
2656 .sp
2657 (*PRUNE) or (*PRUNE:NAME)
2658 .sp
2659 This verb causes the match to fail at the current starting position in the
2660 subject if the rest of the pattern does not match. If the pattern is
2661 unanchored, the normal "bumpalong" advance to the next starting character then
2662 happens. Backtracking can occur as usual to the left of (*PRUNE), before it is
2663 reached, or when matching to the right of (*PRUNE), but if there is no match to
2664 the right, backtracking cannot cross (*PRUNE). In simple cases, the use of
2665 (*PRUNE) is just an alternative to an atomic group or possessive quantifier,
2666 but there are some uses of (*PRUNE) that cannot be expressed in any other way.
2667 The behaviour of (*PRUNE:NAME) is the same as (*MARK:NAME)(*PRUNE) when the
2668 match fails completely; the name is passed back if this is the final attempt.
2669 (*PRUNE:NAME) does not pass back a name if the match succeeds. In an anchored
2670 pattern (*PRUNE) has the same effect as (*COMMIT).
2671 .sp
2672 (*SKIP)
2673 .sp
2674 This verb, when given without a name, is like (*PRUNE), except that if the
2675 pattern is unanchored, the "bumpalong" advance is not to the next character,
2676 but to the position in the subject where (*SKIP) was encountered. (*SKIP)
2677 signifies that whatever text was matched leading up to it cannot be part of a
2678 successful match. Consider:
2679 .sp
2680 a+(*SKIP)b
2681 .sp
2682 If the subject is "aaaac...", after the first match attempt fails (starting at
2683 the first character in the string), the starting point skips on to start the
2684 next attempt at "c". Note that a possessive quantifer does not have the same
2685 effect as this example; although it would suppress backtracking during the
2686 first match attempt, the second attempt would start at the second character
2687 instead of skipping on to "c".
2688 .sp
2689 (*SKIP:NAME)
2690 .sp
2691 When (*SKIP) has an associated name, its behaviour is modified. If the
2692 following pattern fails to match, the previous path through the pattern is
2693 searched for the most recent (*MARK) that has the same name. If one is found,
2694 the "bumpalong" advance is to the subject position that corresponds to that
2695 (*MARK) instead of to where (*SKIP) was encountered. If no (*MARK) with a
2696 matching name is found, normal "bumpalong" of one character happens (the
2697 (*SKIP) is ignored).
2698 .sp
2699 (*THEN) or (*THEN:NAME)
2700 .sp
2701 This verb causes a skip to the next alternation in the innermost enclosing
2702 group if the rest of the pattern does not match. That is, it cancels pending
2703 backtracking, but only within the current alternation. Its name comes from the
2704 observation that it can be used for a pattern-based if-then-else block:
2705 .sp
2706 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2707 .sp
2708 If the COND1 pattern matches, FOO is tried (and possibly further items after
2709 the end of the group if FOO succeeds); on failure the matcher skips to the
2710 second alternative and tries COND2, without backtracking into COND1. The
2711 behaviour of (*THEN:NAME) is exactly the same as (*MARK:NAME)(*THEN) if the
2712 overall match fails. If (*THEN) is not directly inside an alternation, it acts
2713 like (*PRUNE).
2714 .
2715 .P
2716 The above verbs provide four different "strengths" of control when subsequent
2717 matching fails. (*THEN) is the weakest, carrying on the match at the next
2718 alternation. (*PRUNE) comes next, failing the match at the current starting
2719 position, but allowing an advance to the next character (for an unanchored
2720 pattern). (*SKIP) is similar, except that the advance may be more than one
2721 character. (*COMMIT) is the strongest, causing the entire match to fail.
2722 .P
2723 If more than one is present in a pattern, the "stongest" one wins. For example,
2724 consider this pattern, where A, B, etc. are complex pattern fragments:
2725 .sp
2726 (A(*COMMIT)B(*THEN)C|D)
2727 .sp
2728 Once A has matched, PCRE is committed to this match, at the current starting
2729 position. If subsequently B matches, but C does not, the normal (*THEN) action
2730 of trying the next alternation (that is, D) does not happen because (*COMMIT)
2731 overrides.
2732 .
2733 .
2734 .SH "SEE ALSO"
2735 .rs
2736 .sp
2737 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3),
2738 \fBpcresyntax\fP(3), \fBpcre\fP(3).
2739 .
2740 .
2742 .rs
2743 .sp
2744 .nf
2745 Philip Hazel
2746 University Computing Service
2747 Cambridge CB2 3QH, England.
2748 .fi
2749 .
2750 .
2752 .rs
2753 .sp
2754 .nf
2755 Last updated: 12 June 2011
2756 Copyright (c) 1997-2011 University of Cambridge.
2757 .fi


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

  ViewVC Help
Powered by ViewVC 1.1.5