ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 575 - (show annotations)
Sun Nov 21 12:55:42 2010 UTC (10 years, 11 months ago) by ph10
File size: 113527 byte(s)
Error occurred while calculating annotation data.
Documentation clarification.
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 The remainder of this document discusses the patterns that are supported by
56 PCRE when its main matching function, \fBpcre_exec()\fP, is used.
57 From release 6.0, PCRE offers a second matching function,
58 \fBpcre_dfa_exec()\fP, which matches using a different algorithm that is not
59 Perl-compatible. Some of the features discussed below are not available when
60 \fBpcre_dfa_exec()\fP is used. The advantages and disadvantages of the
61 alternative function, and how it differs from the normal function, are
62 discussed in the
63 .\" HREF
64 \fBpcrematching\fP
65 .\"
66 page.
67 .
68 .
69 .\" HTML <a name="newlines"></a>
71 .rs
72 .sp
73 PCRE supports five different conventions for indicating line breaks in
74 strings: a single CR (carriage return) character, a single LF (linefeed)
75 character, the two-character sequence CRLF, any of the three preceding, or any
76 Unicode newline sequence. The
77 .\" HREF
78 \fBpcreapi\fP
79 .\"
80 page has
81 .\" HTML <a href="pcreapi.html#newlines">
82 .\" </a>
83 further discussion
84 .\"
85 about newlines, and shows how to set the newline convention in the
86 \fIoptions\fP arguments for the compiling and matching functions.
87 .P
88 It is also possible to specify a newline convention by starting a pattern
89 string with one of the following five sequences:
90 .sp
91 (*CR) carriage return
92 (*LF) linefeed
93 (*CRLF) carriage return, followed by linefeed
94 (*ANYCRLF) any of the three above
95 (*ANY) all Unicode newline sequences
96 .sp
97 These override the default and the options given to \fBpcre_compile()\fP or
98 \fBpcre_compile2()\fP. For example, on a Unix system where LF is the default
99 newline sequence, the pattern
100 .sp
101 (*CR)a.b
102 .sp
103 changes the convention to CR. That pattern matches "a\enb" because LF is no
104 longer a newline. Note that these special settings, which are not
105 Perl-compatible, are recognized only at the very start of a pattern, and that
106 they must be in upper case. If more than one of them is present, the last one
107 is used.
108 .P
109 The newline convention affects the interpretation of the dot metacharacter when
110 PCRE_DOTALL is not set, and also the behaviour of \eN. However, it does not
111 affect what the \eR escape sequence matches. By default, this is any Unicode
112 newline sequence, for Perl compatibility. However, this can be changed; see the
113 description of \eR in the section entitled
114 .\" HTML <a href="#newlineseq">
115 .\" </a>
116 "Newline sequences"
117 .\"
118 below. A change of \eR setting can be combined with a change of newline
119 convention.
120 .
121 .
123 .rs
124 .sp
125 A regular expression is a pattern that is matched against a subject string from
126 left to right. Most characters stand for themselves in a pattern, and match the
127 corresponding characters in the subject. As a trivial example, the pattern
128 .sp
129 The quick brown fox
130 .sp
131 matches a portion of a subject string that is identical to itself. When
132 caseless matching is specified (the PCRE_CASELESS option), letters are matched
133 independently of case. In UTF-8 mode, PCRE always understands the concept of
134 case for characters whose values are less than 128, so caseless matching is
135 always possible. For characters with higher values, the concept of case is
136 supported if PCRE is compiled with Unicode property support, but not otherwise.
137 If you want to use caseless matching for characters 128 and above, you must
138 ensure that PCRE is compiled with Unicode property support as well as with
139 UTF-8 support.
140 .P
141 The power of regular expressions comes from the ability to include alternatives
142 and repetitions in the pattern. These are encoded in the pattern by the use of
143 \fImetacharacters\fP, which do not stand for themselves but instead are
144 interpreted in some special way.
145 .P
146 There are two different sets of metacharacters: those that are recognized
147 anywhere in the pattern except within square brackets, and those that are
148 recognized within square brackets. Outside square brackets, the metacharacters
149 are as follows:
150 .sp
151 \e general escape character with several uses
152 ^ assert start of string (or line, in multiline mode)
153 $ assert end of string (or line, in multiline mode)
154 . match any character except newline (by default)
155 [ start character class definition
156 | start of alternative branch
157 ( start subpattern
158 ) end subpattern
159 ? extends the meaning of (
160 also 0 or 1 quantifier
161 also quantifier minimizer
162 * 0 or more quantifier
163 + 1 or more quantifier
164 also "possessive quantifier"
165 { start min/max quantifier
166 .sp
167 Part of a pattern that is in square brackets is called a "character class". In
168 a character class the only metacharacters are:
169 .sp
170 \e general escape character
171 ^ negate the class, but only if the first character
172 - indicates character range
173 .\" JOIN
174 [ POSIX character class (only if followed by POSIX
175 syntax)
176 ] terminates the character class
177 .sp
178 The following sections describe the use of each of the metacharacters.
179 .
180 .
182 .rs
183 .sp
184 The backslash character has several uses. Firstly, if it is followed by a
185 character that is not a number or a letter, it takes away any special meaning
186 that character may have. This use of backslash as an escape character applies
187 both inside and outside character classes.
188 .P
189 For example, if you want to match a * character, you write \e* in the pattern.
190 This escaping action applies whether or not the following character would
191 otherwise be interpreted as a metacharacter, so it is always safe to precede a
192 non-alphanumeric with backslash to specify that it stands for itself. In
193 particular, if you want to match a backslash, you write \e\e.
194 .P
195 In UTF-8 mode, only ASCII numbers and letters have any special meaning after a
196 backslash. All other characters (in particular, those whose codepoints are
197 greater than 127) are treated as literals.
198 .P
199 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
200 pattern (other than in a character class) and characters between a # outside
201 a character class and the next newline are ignored. An escaping backslash can
202 be used to include a whitespace or # character as part of the pattern.
203 .P
204 If you want to remove the special meaning from a sequence of characters, you
205 can do so by putting them between \eQ and \eE. This is different from Perl in
206 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
207 Perl, $ and @ cause variable interpolation. Note the following examples:
208 .sp
209 Pattern PCRE matches Perl matches
210 .sp
211 .\" JOIN
212 \eQabc$xyz\eE abc$xyz abc followed by the
213 contents of $xyz
214 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
215 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
216 .sp
217 The \eQ...\eE sequence is recognized both inside and outside character classes.
218 An isolated \eE that is not preceded by \eQ is ignored.
219 .
220 .
221 .\" HTML <a name="digitsafterbackslash"></a>
222 .SS "Non-printing characters"
223 .rs
224 .sp
225 A second use of backslash provides a way of encoding non-printing characters
226 in patterns in a visible manner. There is no restriction on the appearance of
227 non-printing characters, apart from the binary zero that terminates a pattern,
228 but when a pattern is being prepared by text editing, it is often easier to use
229 one of the following escape sequences than the binary character it represents:
230 .sp
231 \ea alarm, that is, the BEL character (hex 07)
232 \ecx "control-x", where x is any ASCII character
233 \ee escape (hex 1B)
234 \ef formfeed (hex 0C)
235 \en linefeed (hex 0A)
236 \er carriage return (hex 0D)
237 \et tab (hex 09)
238 \eddd character with octal code ddd, or back reference
239 \exhh character with hex code hh
240 \ex{hhh..} character with hex code hhh..
241 .sp
242 The precise effect of \ecx is as follows: if x is a lower case letter, it
243 is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
244 Thus \ecz becomes hex 1A (z is 7A), but \ec{ becomes hex 3B ({ is 7B), while
245 \ec; becomes hex 7B (; is 3B). If the byte following \ec has a value greater
246 than 127, a compile-time error occurs. This locks out non-ASCII characters in
247 both byte mode and UTF-8 mode. (When PCRE is compiled in EBCDIC mode, all byte
248 values are valid. A lower case letter is converted to upper case, and then the
249 0xc0 bits are flipped.)
250 .P
251 After \ex, from zero to two hexadecimal digits are read (letters can be in
252 upper or lower case). Any number of hexadecimal digits may appear between \ex{
253 and }, but the value of the character code must be less than 256 in non-UTF-8
254 mode, and less than 2**31 in UTF-8 mode. That is, the maximum value in
255 hexadecimal is 7FFFFFFF. Note that this is bigger than the largest Unicode code
256 point, which is 10FFFF.
257 .P
258 If characters other than hexadecimal digits appear between \ex{ and }, or if
259 there is no terminating }, this form of escape is not recognized. Instead, the
260 initial \ex will be interpreted as a basic hexadecimal escape, with no
261 following digits, giving a character whose value is zero.
262 .P
263 Characters whose value is less than 256 can be defined by either of the two
264 syntaxes for \ex. There is no difference in the way they are handled. For
265 example, \exdc is exactly the same as \ex{dc}.
266 .P
267 After \e0 up to two further octal digits are read. If there are fewer than two
268 digits, just those that are present are used. Thus the sequence \e0\ex\e07
269 specifies two binary zeros followed by a BEL character (code value 7). Make
270 sure you supply two digits after the initial zero if the pattern character that
271 follows is itself an octal digit.
272 .P
273 The handling of a backslash followed by a digit other than 0 is complicated.
274 Outside a character class, PCRE reads it and any following digits as a decimal
275 number. If the number is less than 10, or if there have been at least that many
276 previous capturing left parentheses in the expression, the entire sequence is
277 taken as a \fIback reference\fP. A description of how this works is given
278 .\" HTML <a href="#backreferences">
279 .\" </a>
280 later,
281 .\"
282 following the discussion of
283 .\" HTML <a href="#subpattern">
284 .\" </a>
285 parenthesized subpatterns.
286 .\"
287 .P
288 Inside a character class, or if the decimal number is greater than 9 and there
289 have not been that many capturing subpatterns, PCRE re-reads up to three octal
290 digits following the backslash, and uses them to generate a data character. Any
291 subsequent digits stand for themselves. In non-UTF-8 mode, the value of a
292 character specified in octal must be less than \e400. In UTF-8 mode, values up
293 to \e777 are permitted. For example:
294 .sp
295 \e040 is another way of writing a space
296 .\" JOIN
297 \e40 is the same, provided there are fewer than 40
298 previous capturing subpatterns
299 \e7 is always a back reference
300 .\" JOIN
301 \e11 might be a back reference, or another way of
302 writing a tab
303 \e011 is always a tab
304 \e0113 is a tab followed by the character "3"
305 .\" JOIN
306 \e113 might be a back reference, otherwise the
307 character with octal code 113
308 .\" JOIN
309 \e377 might be a back reference, otherwise
310 the byte consisting entirely of 1 bits
311 .\" JOIN
312 \e81 is either a back reference, or a binary zero
313 followed by the two characters "8" and "1"
314 .sp
315 Note that octal values of 100 or greater must not be introduced by a leading
316 zero, because no more than three octal digits are ever read.
317 .P
318 All the sequences that define a single character value can be used both inside
319 and outside character classes. In addition, inside a character class, the
320 sequence \eb is interpreted as the backspace character (hex 08). The sequences
321 \eB, \eN, \eR, and \eX are not special inside a character class. Like any other
322 unrecognized escape sequences, they are treated as the literal characters "B",
323 "N", "R", and "X" by default, but cause an error if the PCRE_EXTRA option is
324 set. Outside a character class, these sequences have different meanings.
325 .
326 .
327 .SS "Absolute and relative back references"
328 .rs
329 .sp
330 The sequence \eg followed by an unsigned or a negative number, optionally
331 enclosed in braces, is an absolute or relative back reference. A named back
332 reference can be coded as \eg{name}. Back references are discussed
333 .\" HTML <a href="#backreferences">
334 .\" </a>
335 later,
336 .\"
337 following the discussion of
338 .\" HTML <a href="#subpattern">
339 .\" </a>
340 parenthesized subpatterns.
341 .\"
342 .
343 .
344 .SS "Absolute and relative subroutine calls"
345 .rs
346 .sp
347 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
348 a number enclosed either in angle brackets or single quotes, is an alternative
349 syntax for referencing a subpattern as a "subroutine". Details are discussed
350 .\" HTML <a href="#onigurumasubroutines">
351 .\" </a>
352 later.
353 .\"
354 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
355 synonymous. The former is a back reference; the latter is a
356 .\" HTML <a href="#subpatternsassubroutines">
357 .\" </a>
358 subroutine
359 .\"
360 call.
361 .
362 .
363 .\" HTML <a name="genericchartypes"></a>
364 .SS "Generic character types"
365 .rs
366 .sp
367 Another use of backslash is for specifying generic character types:
368 .sp
369 \ed any decimal digit
370 \eD any character that is not a decimal digit
371 \eh any horizontal whitespace character
372 \eH any character that is not a horizontal whitespace character
373 \es any whitespace character
374 \eS any character that is not a whitespace character
375 \ev any vertical whitespace character
376 \eV any character that is not a vertical whitespace character
377 \ew any "word" character
378 \eW any "non-word" character
379 .sp
380 There is also the single sequence \eN, which matches a non-newline character.
381 This is the same as
382 .\" HTML <a href="#fullstopdot">
383 .\" </a>
384 the "." metacharacter
385 .\"
386 when PCRE_DOTALL is not set.
387 .P
388 Each pair of lower and upper case escape sequences partitions the complete set
389 of characters into two disjoint sets. Any given character matches one, and only
390 one, of each pair. The sequences can appear both inside and outside character
391 classes. They each match one character of the appropriate type. If the current
392 matching point is at the end of the subject string, all of them fail, because
393 there is no character to match.
394 .P
395 For compatibility with Perl, \es does not match the VT character (code 11).
396 This makes it different from the the POSIX "space" class. The \es characters
397 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
398 included in a Perl script, \es may match the VT character. In PCRE, it never
399 does.
400 .P
401 A "word" character is an underscore or any character that is a letter or digit.
402 By default, the definition of letters and digits is controlled by PCRE's
403 low-valued character tables, and may vary if locale-specific matching is taking
404 place (see
405 .\" HTML <a href="pcreapi.html#localesupport">
406 .\" </a>
407 "Locale support"
408 .\"
409 in the
410 .\" HREF
411 \fBpcreapi\fP
412 .\"
413 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
414 or "french" in Windows, some character codes greater than 128 are used for
415 accented letters, and these are then matched by \ew. The use of locales with
416 Unicode is discouraged.
417 .P
418 By default, in UTF-8 mode, characters with values greater than 128 never match
419 \ed, \es, or \ew, and always match \eD, \eS, and \eW. These sequences retain
420 their original meanings from before UTF-8 support was available, mainly for
421 efficiency reasons. However, if PCRE is compiled with Unicode property support,
422 and the PCRE_UCP option is set, the behaviour is changed so that Unicode
423 properties are used to determine character types, as follows:
424 .sp
425 \ed any character that \ep{Nd} matches (decimal digit)
426 \es any character that \ep{Z} matches, plus HT, LF, FF, CR
427 \ew any character that \ep{L} or \ep{N} matches, plus underscore
428 .sp
429 The upper case escapes match the inverse sets of characters. Note that \ed
430 matches only decimal digits, whereas \ew matches any Unicode digit, as well as
431 any Unicode letter, and underscore. Note also that PCRE_UCP affects \eb, and
432 \eB because they are defined in terms of \ew and \eW. Matching these sequences
433 is noticeably slower when PCRE_UCP is set.
434 .P
435 The sequences \eh, \eH, \ev, and \eV are features that were added to Perl at
436 release 5.10. In contrast to the other sequences, which match only ASCII
437 characters by default, these always match certain high-valued codepoints in
438 UTF-8 mode, whether or not PCRE_UCP is set. The horizontal space characters
439 are:
440 .sp
441 U+0009 Horizontal tab
442 U+0020 Space
443 U+00A0 Non-break space
444 U+1680 Ogham space mark
445 U+180E Mongolian vowel separator
446 U+2000 En quad
447 U+2001 Em quad
448 U+2002 En space
449 U+2003 Em space
450 U+2004 Three-per-em space
451 U+2005 Four-per-em space
452 U+2006 Six-per-em space
453 U+2007 Figure space
454 U+2008 Punctuation space
455 U+2009 Thin space
456 U+200A Hair space
457 U+202F Narrow no-break space
458 U+205F Medium mathematical space
459 U+3000 Ideographic space
460 .sp
461 The vertical space characters are:
462 .sp
463 U+000A Linefeed
464 U+000B Vertical tab
465 U+000C Formfeed
466 U+000D Carriage return
467 U+0085 Next line
468 U+2028 Line separator
469 U+2029 Paragraph separator
470 .
471 .
472 .\" HTML <a name="newlineseq"></a>
473 .SS "Newline sequences"
474 .rs
475 .sp
476 Outside a character class, by default, the escape sequence \eR matches any
477 Unicode newline sequence. In non-UTF-8 mode \eR is equivalent to the following:
478 .sp
479 (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
480 .sp
481 This is an example of an "atomic group", details of which are given
482 .\" HTML <a href="#atomicgroup">
483 .\" </a>
484 below.
485 .\"
486 This particular group matches either the two-character sequence CR followed by
487 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
488 U+000B), FF (formfeed, U+000C), CR (carriage return, U+000D), or NEL (next
489 line, U+0085). The two-character sequence is treated as a single unit that
490 cannot be split.
491 .P
492 In UTF-8 mode, two additional characters whose codepoints are greater than 255
493 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
494 Unicode character property support is not needed for these characters to be
495 recognized.
496 .P
497 It is possible to restrict \eR to match only CR, LF, or CRLF (instead of the
498 complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
499 either at compile time or when the pattern is matched. (BSR is an abbrevation
500 for "backslash R".) This can be made the default when PCRE is built; if this is
501 the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
502 It is also possible to specify these settings by starting a pattern string with
503 one of the following sequences:
504 .sp
505 (*BSR_ANYCRLF) CR, LF, or CRLF only
506 (*BSR_UNICODE) any Unicode newline sequence
507 .sp
508 These override the default and the options given to \fBpcre_compile()\fP or
509 \fBpcre_compile2()\fP, but they can be overridden by options given to
510 \fBpcre_exec()\fP or \fBpcre_dfa_exec()\fP. Note that these special settings,
511 which are not Perl-compatible, are recognized only at the very start of a
512 pattern, and that they must be in upper case. If more than one of them is
513 present, the last one is used. They can be combined with a change of newline
514 convention; for example, a pattern can start with:
515 .sp
517 .sp
518 They can also be combined with the (*UTF8) or (*UCP) special sequences. Inside
519 a character class, \eR is treated as an unrecognized escape sequence, and so
520 matches the letter "R" by default, but causes an error if PCRE_EXTRA is set.
521 .
522 .
523 .\" HTML <a name="uniextseq"></a>
524 .SS Unicode character properties
525 .rs
526 .sp
527 When PCRE is built with Unicode character property support, three additional
528 escape sequences that match characters with specific properties are available.
529 When not in UTF-8 mode, these sequences are of course limited to testing
530 characters whose codepoints are less than 256, but they do work in this mode.
531 The extra escape sequences are:
532 .sp
533 \ep{\fIxx\fP} a character with the \fIxx\fP property
534 \eP{\fIxx\fP} a character without the \fIxx\fP property
535 \eX an extended Unicode sequence
536 .sp
537 The property names represented by \fIxx\fP above are limited to the Unicode
538 script names, the general category properties, "Any", which matches any
539 character (including newline), and some special PCRE properties (described
540 in the
541 .\" HTML <a href="#extraprops">
542 .\" </a>
543 next section).
544 .\"
545 Other Perl properties such as "InMusicalSymbols" are not currently supported by
546 PCRE. Note that \eP{Any} does not match any characters, so always causes a
547 match failure.
548 .P
549 Sets of Unicode characters are defined as belonging to certain scripts. A
550 character from one of these sets can be matched using a script name. For
551 example:
552 .sp
553 \ep{Greek}
554 \eP{Han}
555 .sp
556 Those that are not part of an identified script are lumped together as
557 "Common". The current list of scripts is:
558 .P
559 Arabic,
560 Armenian,
561 Avestan,
562 Balinese,
563 Bamum,
564 Bengali,
565 Bopomofo,
566 Braille,
567 Buginese,
568 Buhid,
569 Canadian_Aboriginal,
570 Carian,
571 Cham,
572 Cherokee,
573 Common,
574 Coptic,
575 Cuneiform,
576 Cypriot,
577 Cyrillic,
578 Deseret,
579 Devanagari,
580 Egyptian_Hieroglyphs,
581 Ethiopic,
582 Georgian,
583 Glagolitic,
584 Gothic,
585 Greek,
586 Gujarati,
587 Gurmukhi,
588 Han,
589 Hangul,
590 Hanunoo,
591 Hebrew,
592 Hiragana,
593 Imperial_Aramaic,
594 Inherited,
595 Inscriptional_Pahlavi,
596 Inscriptional_Parthian,
597 Javanese,
598 Kaithi,
599 Kannada,
600 Katakana,
601 Kayah_Li,
602 Kharoshthi,
603 Khmer,
604 Lao,
605 Latin,
606 Lepcha,
607 Limbu,
608 Linear_B,
609 Lisu,
610 Lycian,
611 Lydian,
612 Malayalam,
613 Meetei_Mayek,
614 Mongolian,
615 Myanmar,
616 New_Tai_Lue,
617 Nko,
618 Ogham,
619 Old_Italic,
620 Old_Persian,
621 Old_South_Arabian,
622 Old_Turkic,
623 Ol_Chiki,
624 Oriya,
625 Osmanya,
626 Phags_Pa,
627 Phoenician,
628 Rejang,
629 Runic,
630 Samaritan,
631 Saurashtra,
632 Shavian,
633 Sinhala,
634 Sundanese,
635 Syloti_Nagri,
636 Syriac,
637 Tagalog,
638 Tagbanwa,
639 Tai_Le,
640 Tai_Tham,
641 Tai_Viet,
642 Tamil,
643 Telugu,
644 Thaana,
645 Thai,
646 Tibetan,
647 Tifinagh,
648 Ugaritic,
649 Vai,
650 Yi.
651 .P
652 Each character has exactly one Unicode general category property, specified by
653 a two-letter abbreviation. For compatibility with Perl, negation can be
654 specified by including a circumflex between the opening brace and the property
655 name. For example, \ep{^Lu} is the same as \eP{Lu}.
656 .P
657 If only one letter is specified with \ep or \eP, it includes all the general
658 category properties that start with that letter. In this case, in the absence
659 of negation, the curly brackets in the escape sequence are optional; these two
660 examples have the same effect:
661 .sp
662 \ep{L}
663 \epL
664 .sp
665 The following general category property codes are supported:
666 .sp
667 C Other
668 Cc Control
669 Cf Format
670 Cn Unassigned
671 Co Private use
672 Cs Surrogate
673 .sp
674 L Letter
675 Ll Lower case letter
676 Lm Modifier letter
677 Lo Other letter
678 Lt Title case letter
679 Lu Upper case letter
680 .sp
681 M Mark
682 Mc Spacing mark
683 Me Enclosing mark
684 Mn Non-spacing mark
685 .sp
686 N Number
687 Nd Decimal number
688 Nl Letter number
689 No Other number
690 .sp
691 P Punctuation
692 Pc Connector punctuation
693 Pd Dash punctuation
694 Pe Close punctuation
695 Pf Final punctuation
696 Pi Initial punctuation
697 Po Other punctuation
698 Ps Open punctuation
699 .sp
700 S Symbol
701 Sc Currency symbol
702 Sk Modifier symbol
703 Sm Mathematical symbol
704 So Other symbol
705 .sp
706 Z Separator
707 Zl Line separator
708 Zp Paragraph separator
709 Zs Space separator
710 .sp
711 The special property L& is also supported: it matches a character that has
712 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
713 a modifier or "other".
714 .P
715 The Cs (Surrogate) property applies only to characters in the range U+D800 to
716 U+DFFF. Such characters are not valid in UTF-8 strings (see RFC 3629) and so
717 cannot be tested by PCRE, unless UTF-8 validity checking has been turned off
718 (see the discussion of PCRE_NO_UTF8_CHECK in the
719 .\" HREF
720 \fBpcreapi\fP
721 .\"
722 page). Perl does not support the Cs property.
723 .P
724 The long synonyms for property names that Perl supports (such as \ep{Letter})
725 are not supported by PCRE, nor is it permitted to prefix any of these
726 properties with "Is".
727 .P
728 No character that is in the Unicode table has the Cn (unassigned) property.
729 Instead, this property is assumed for any code point that is not in the
730 Unicode table.
731 .P
732 Specifying caseless matching does not affect these escape sequences. For
733 example, \ep{Lu} always matches only upper case letters.
734 .P
735 The \eX escape matches any number of Unicode characters that form an extended
736 Unicode sequence. \eX is equivalent to
737 .sp
738 (?>\ePM\epM*)
739 .sp
740 That is, it matches a character without the "mark" property, followed by zero
741 or more characters with the "mark" property, and treats the sequence as an
742 atomic group
743 .\" HTML <a href="#atomicgroup">
744 .\" </a>
745 (see below).
746 .\"
747 Characters with the "mark" property are typically accents that affect the
748 preceding character. None of them have codepoints less than 256, so in
749 non-UTF-8 mode \eX matches any one character.
750 .P
751 Matching characters by Unicode property is not fast, because PCRE has to search
752 a structure that contains data for over fifteen thousand characters. That is
753 why the traditional escape sequences such as \ed and \ew do not use Unicode
754 properties in PCRE by default, though you can make them do so by setting the
755 PCRE_UCP option for \fBpcre_compile()\fP or by starting the pattern with
756 (*UCP).
757 .
758 .
759 .\" HTML <a name="extraprops"></a>
760 .SS PCRE's additional properties
761 .rs
762 .sp
763 As well as the standard Unicode properties described in the previous
764 section, PCRE supports four more that make it possible to convert traditional
765 escape sequences such as \ew and \es and POSIX character classes to use Unicode
766 properties. PCRE uses these non-standard, non-Perl properties internally when
767 PCRE_UCP is set. They are:
768 .sp
769 Xan Any alphanumeric character
770 Xps Any POSIX space character
771 Xsp Any Perl space character
772 Xwd Any Perl "word" character
773 .sp
774 Xan matches characters that have either the L (letter) or the N (number)
775 property. Xps matches the characters tab, linefeed, vertical tab, formfeed, or
776 carriage return, and any other character that has the Z (separator) property.
777 Xsp is the same as Xps, except that vertical tab is excluded. Xwd matches the
778 same characters as Xan, plus underscore.
779 .
780 .
781 .\" HTML <a name="resetmatchstart"></a>
782 .SS "Resetting the match start"
783 .rs
784 .sp
785 The escape sequence \eK causes any previously matched characters not to be
786 included in the final matched sequence. For example, the pattern:
787 .sp
788 foo\eKbar
789 .sp
790 matches "foobar", but reports that it has matched "bar". This feature is
791 similar to a lookbehind assertion
792 .\" HTML <a href="#lookbehind">
793 .\" </a>
794 (described below).
795 .\"
796 However, in this case, the part of the subject before the real match does not
797 have to be of fixed length, as lookbehind assertions do. The use of \eK does
798 not interfere with the setting of
799 .\" HTML <a href="#subpattern">
800 .\" </a>
801 captured substrings.
802 .\"
803 For example, when the pattern
804 .sp
805 (foo)\eKbar
806 .sp
807 matches "foobar", the first substring is still set to "foo".
808 .P
809 Perl documents that the use of \eK within assertions is "not well defined". In
810 PCRE, \eK is acted upon when it occurs inside positive assertions, but is
811 ignored in negative assertions.
812 .
813 .
814 .\" HTML <a name="smallassertions"></a>
815 .SS "Simple assertions"
816 .rs
817 .sp
818 The final use of backslash is for certain simple assertions. An assertion
819 specifies a condition that has to be met at a particular point in a match,
820 without consuming any characters from the subject string. The use of
821 subpatterns for more complicated assertions is described
822 .\" HTML <a href="#bigassertions">
823 .\" </a>
824 below.
825 .\"
826 The backslashed assertions are:
827 .sp
828 \eb matches at a word boundary
829 \eB matches when not at a word boundary
830 \eA matches at the start of the subject
831 \eZ matches at the end of the subject
832 also matches before a newline at the end of the subject
833 \ez matches only at the end of the subject
834 \eG matches at the first matching position in the subject
835 .sp
836 Inside a character class, \eb has a different meaning; it matches the backspace
837 character. If any other of these assertions appears in a character class, by
838 default it matches the corresponding literal character (for example, \eB
839 matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
840 escape sequence" error is generated instead.
841 .P
842 A word boundary is a position in the subject string where the current character
843 and the previous character do not both match \ew or \eW (i.e. one matches
844 \ew and the other matches \eW), or the start or end of the string if the
845 first or last character matches \ew, respectively. In UTF-8 mode, the meanings
846 of \ew and \eW can be changed by setting the PCRE_UCP option. When this is
847 done, it also affects \eb and \eB. Neither PCRE nor Perl has a separate "start
848 of word" or "end of word" metasequence. However, whatever follows \eb normally
849 determines which it is. For example, the fragment \eba matches "a" at the start
850 of a word.
851 .P
852 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
853 dollar (described in the next section) in that they only ever match at the very
854 start and end of the subject string, whatever options are set. Thus, they are
855 independent of multiline mode. These three assertions are not affected by the
856 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
857 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
858 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
859 at a point other than the beginning of the subject, \eA can never match. The
860 difference between \eZ and \ez is that \eZ matches before a newline at the end
861 of the string as well as at the very end, whereas \ez matches only at the end.
862 .P
863 The \eG assertion is true only when the current matching position is at the
864 start point of the match, as specified by the \fIstartoffset\fP argument of
865 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
866 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
867 arguments, you can mimic Perl's /g option, and it is in this kind of
868 implementation where \eG can be useful.
869 .P
870 Note, however, that PCRE's interpretation of \eG, as the start of the current
871 match, is subtly different from Perl's, which defines it as the end of the
872 previous match. In Perl, these can be different when the previously matched
873 string was empty. Because PCRE does just one match at a time, it cannot
874 reproduce this behaviour.
875 .P
876 If all the alternatives of a pattern begin with \eG, the expression is anchored
877 to the starting match position, and the "anchored" flag is set in the compiled
878 regular expression.
879 .
880 .
882 .rs
883 .sp
884 Outside a character class, in the default matching mode, the circumflex
885 character is an assertion that is true only if the current matching point is
886 at the start of the subject string. If the \fIstartoffset\fP argument of
887 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
888 option is unset. Inside a character class, circumflex has an entirely different
889 meaning
890 .\" HTML <a href="#characterclass">
891 .\" </a>
892 (see below).
893 .\"
894 .P
895 Circumflex need not be the first character of the pattern if a number of
896 alternatives are involved, but it should be the first thing in each alternative
897 in which it appears if the pattern is ever to match that branch. If all
898 possible alternatives start with a circumflex, that is, if the pattern is
899 constrained to match only at the start of the subject, it is said to be an
900 "anchored" pattern. (There are also other constructs that can cause a pattern
901 to be anchored.)
902 .P
903 A dollar character is an assertion that is true only if the current matching
904 point is at the end of the subject string, or immediately before a newline
905 at the end of the string (by default). Dollar need not be the last character of
906 the pattern if a number of alternatives are involved, but it should be the last
907 item in any branch in which it appears. Dollar has no special meaning in a
908 character class.
909 .P
910 The meaning of dollar can be changed so that it matches only at the very end of
911 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
912 does not affect the \eZ assertion.
913 .P
914 The meanings of the circumflex and dollar characters are changed if the
915 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
916 immediately after internal newlines as well as at the start of the subject
917 string. It does not match after a newline that ends the string. A dollar
918 matches before any newlines in the string, as well as at the very end, when
919 PCRE_MULTILINE is set. When newline is specified as the two-character
920 sequence CRLF, isolated CR and LF characters do not indicate newlines.
921 .P
922 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
923 \en represents a newline) in multiline mode, but not otherwise. Consequently,
924 patterns that are anchored in single line mode because all branches start with
925 ^ are not anchored in multiline mode, and a match for circumflex is possible
926 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
927 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
928 .P
929 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
930 end of the subject in both modes, and if all branches of a pattern start with
931 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
932 .
933 .
934 .\" HTML <a name="fullstopdot"></a>
936 .rs
937 .sp
938 Outside a character class, a dot in the pattern matches any one character in
939 the subject string except (by default) a character that signifies the end of a
940 line. In UTF-8 mode, the matched character may be more than one byte long.
941 .P
942 When a line ending is defined as a single character, dot never matches that
943 character; when the two-character sequence CRLF is used, dot does not match CR
944 if it is immediately followed by LF, but otherwise it matches all characters
945 (including isolated CRs and LFs). When any Unicode line endings are being
946 recognized, dot does not match CR or LF or any of the other line ending
947 characters.
948 .P
949 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
950 option is set, a dot matches any one character, without exception. If the
951 two-character sequence CRLF is present in the subject string, it takes two dots
952 to match it.
953 .P
954 The handling of dot is entirely independent of the handling of circumflex and
955 dollar, the only relationship being that they both involve newlines. Dot has no
956 special meaning in a character class.
957 .P
958 The escape sequence \eN behaves like a dot, except that it is not affected by
959 the PCRE_DOTALL option. In other words, it matches any character except one
960 that signifies the end of a line.
961 .
962 .
964 .rs
965 .sp
966 Outside a character class, the escape sequence \eC matches any one byte, both
967 in and out of UTF-8 mode. Unlike a dot, it always matches any line-ending
968 characters. The feature is provided in Perl in order to match individual bytes
969 in UTF-8 mode. Because it breaks up UTF-8 characters into individual bytes, the
970 rest of the string may start with a malformed UTF-8 character. For this reason,
971 the \eC escape sequence is best avoided.
972 .P
973 PCRE does not allow \eC to appear in lookbehind assertions
974 .\" HTML <a href="#lookbehind">
975 .\" </a>
976 (described below),
977 .\"
978 because in UTF-8 mode this would make it impossible to calculate the length of
979 the lookbehind.
980 .
981 .
982 .\" HTML <a name="characterclass"></a>
984 .rs
985 .sp
986 An opening square bracket introduces a character class, terminated by a closing
987 square bracket. A closing square bracket on its own is not special by default.
988 However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
989 bracket causes a compile-time error. If a closing square bracket is required as
990 a member of the class, it should be the first data character in the class
991 (after an initial circumflex, if present) or escaped with a backslash.
992 .P
993 A character class matches a single character in the subject. In UTF-8 mode, the
994 character may be more than one byte long. A matched character must be in the
995 set of characters defined by the class, unless the first character in the class
996 definition is a circumflex, in which case the subject character must not be in
997 the set defined by the class. If a circumflex is actually required as a member
998 of the class, ensure it is not the first character, or escape it with a
999 backslash.
1000 .P
1001 For example, the character class [aeiou] matches any lower case vowel, while
1002 [^aeiou] matches any character that is not a lower case vowel. Note that a
1003 circumflex is just a convenient notation for specifying the characters that
1004 are in the class by enumerating those that are not. A class that starts with a
1005 circumflex is not an assertion; it still consumes a character from the subject
1006 string, and therefore it fails if the current pointer is at the end of the
1007 string.
1008 .P
1009 In UTF-8 mode, characters with values greater than 255 can be included in a
1010 class as a literal string of bytes, or by using the \ex{ escaping mechanism.
1011 .P
1012 When caseless matching is set, any letters in a class represent both their
1013 upper case and lower case versions, so for example, a caseless [aeiou] matches
1014 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1015 caseful version would. In UTF-8 mode, PCRE always understands the concept of
1016 case for characters whose values are less than 128, so caseless matching is
1017 always possible. For characters with higher values, the concept of case is
1018 supported if PCRE is compiled with Unicode property support, but not otherwise.
1019 If you want to use caseless matching in UTF8-mode for characters 128 and above,
1020 you must ensure that PCRE is compiled with Unicode property support as well as
1021 with UTF-8 support.
1022 .P
1023 Characters that might indicate line breaks are never treated in any special way
1024 when matching character classes, whatever line-ending sequence is in use, and
1025 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1026 such as [^a] always matches one of these characters.
1027 .P
1028 The minus (hyphen) character can be used to specify a range of characters in a
1029 character class. For example, [d-m] matches any letter between d and m,
1030 inclusive. If a minus character is required in a class, it must be escaped with
1031 a backslash or appear in a position where it cannot be interpreted as
1032 indicating a range, typically as the first or last character in the class.
1033 .P
1034 It is not possible to have the literal character "]" as the end character of a
1035 range. A pattern such as [W-]46] is interpreted as a class of two characters
1036 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1037 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1038 the end of range, so [W-\e]46] is interpreted as a class containing a range
1039 followed by two other characters. The octal or hexadecimal representation of
1040 "]" can also be used to end a range.
1041 .P
1042 Ranges operate in the collating sequence of character values. They can also be
1043 used for characters specified numerically, for example [\e000-\e037]. In UTF-8
1044 mode, ranges can include characters whose values are greater than 255, for
1045 example [\ex{100}-\ex{2ff}].
1046 .P
1047 If a range that includes letters is used when caseless matching is set, it
1048 matches the letters in either case. For example, [W-c] is equivalent to
1049 [][\e\e^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
1050 tables for a French locale are in use, [\exc8-\excb] matches accented E
1051 characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
1052 characters with values greater than 128 only when it is compiled with Unicode
1053 property support.
1054 .P
1055 The character escape sequences \ed, \eD, \eh, \eH, \ep, \eP, \es, \eS, \ev,
1056 \eV, \ew, and \eW may appear in a character class, and add the characters that
1057 they match to the class. For example, [\edABCDEF] matches any hexadecimal
1058 digit. In UTF-8 mode, the PCRE_UCP option affects the meanings of \ed, \es, \ew
1059 and their upper case partners, just as it does when they appear outside a
1060 character class, as described in the section entitled
1061 .\" HTML <a href="#genericchartypes">
1062 .\" </a>
1063 "Generic character types"
1064 .\"
1065 above. The escape sequence \eb has a different meaning inside a character
1066 class; it matches the backspace character. The sequences \eB, \eN, \eR, and \eX
1067 are not special inside a character class. Like any other unrecognized escape
1068 sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1069 default, but cause an error if the PCRE_EXTRA option is set.
1070 .P
1071 A circumflex can conveniently be used with the upper case character types to
1072 specify a more restricted set of characters than the matching lower case type.
1073 For example, the class [^\eW_] matches any letter or digit, but not underscore,
1074 whereas [\ew] includes underscore. A positive character class should be read as
1075 "something OR something OR ..." and a negative class as "NOT something AND NOT
1076 something AND NOT ...".
1077 .P
1078 The only metacharacters that are recognized in character classes are backslash,
1079 hyphen (only where it can be interpreted as specifying a range), circumflex
1080 (only at the start), opening square bracket (only when it can be interpreted as
1081 introducing a POSIX class name - see the next section), and the terminating
1082 closing square bracket. However, escaping other non-alphanumeric characters
1083 does no harm.
1084 .
1085 .
1087 .rs
1088 .sp
1089 Perl supports the POSIX notation for character classes. This uses names
1090 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1091 this notation. For example,
1092 .sp
1093 [01[:alpha:]%]
1094 .sp
1095 matches "0", "1", any alphabetic character, or "%". The supported class names
1096 are:
1097 .sp
1098 alnum letters and digits
1099 alpha letters
1100 ascii character codes 0 - 127
1101 blank space or tab only
1102 cntrl control characters
1103 digit decimal digits (same as \ed)
1104 graph printing characters, excluding space
1105 lower lower case letters
1106 print printing characters, including space
1107 punct printing characters, excluding letters and digits and space
1108 space white space (not quite the same as \es)
1109 upper upper case letters
1110 word "word" characters (same as \ew)
1111 xdigit hexadecimal digits
1112 .sp
1113 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
1114 space (32). Notice that this list includes the VT character (code 11). This
1115 makes "space" different to \es, which does not include VT (for Perl
1116 compatibility).
1117 .P
1118 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1119 5.8. Another Perl extension is negation, which is indicated by a ^ character
1120 after the colon. For example,
1121 .sp
1122 [12[:^digit:]]
1123 .sp
1124 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1125 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1126 supported, and an error is given if they are encountered.
1127 .P
1128 By default, in UTF-8 mode, characters with values greater than 128 do not match
1129 any of the POSIX character classes. However, if the PCRE_UCP option is passed
1130 to \fBpcre_compile()\fP, some of the classes are changed so that Unicode
1131 character properties are used. This is achieved by replacing the POSIX classes
1132 by other sequences, as follows:
1133 .sp
1134 [:alnum:] becomes \ep{Xan}
1135 [:alpha:] becomes \ep{L}
1136 [:blank:] becomes \eh
1137 [:digit:] becomes \ep{Nd}
1138 [:lower:] becomes \ep{Ll}
1139 [:space:] becomes \ep{Xps}
1140 [:upper:] becomes \ep{Lu}
1141 [:word:] becomes \ep{Xwd}
1142 .sp
1143 Negated versions, such as [:^alpha:] use \eP instead of \ep. The other POSIX
1144 classes are unchanged, and match only characters with code points less than
1145 128.
1146 .
1147 .
1149 .rs
1150 .sp
1151 Vertical bar characters are used to separate alternative patterns. For example,
1152 the pattern
1153 .sp
1154 gilbert|sullivan
1155 .sp
1156 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1157 and an empty alternative is permitted (matching the empty string). The matching
1158 process tries each alternative in turn, from left to right, and the first one
1159 that succeeds is used. If the alternatives are within a subpattern
1160 .\" HTML <a href="#subpattern">
1161 .\" </a>
1162 (defined below),
1163 .\"
1164 "succeeds" means matching the rest of the main pattern as well as the
1165 alternative in the subpattern.
1166 .
1167 .
1169 .rs
1170 .sp
1171 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1172 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1173 the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1174 The option letters are
1175 .sp
1176 i for PCRE_CASELESS
1178 s for PCRE_DOTALL
1179 x for PCRE_EXTENDED
1180 .sp
1181 For example, (?im) sets caseless, multiline matching. It is also possible to
1182 unset these options by preceding the letter with a hyphen, and a combined
1183 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1184 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1185 permitted. If a letter appears both before and after the hyphen, the option is
1186 unset.
1187 .P
1188 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1189 changed in the same way as the Perl-compatible options by using the characters
1190 J, U and X respectively.
1191 .P
1192 When one of these option changes occurs at top level (that is, not inside
1193 subpattern parentheses), the change applies to the remainder of the pattern
1194 that follows. If the change is placed right at the start of a pattern, PCRE
1195 extracts it into the global options (and it will therefore show up in data
1196 extracted by the \fBpcre_fullinfo()\fP function).
1197 .P
1198 An option change within a subpattern (see below for a description of
1199 subpatterns) affects only that part of the subpattern that follows it, so
1200 .sp
1201 (a(?i)b)c
1202 .sp
1203 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1204 By this means, options can be made to have different settings in different
1205 parts of the pattern. Any changes made in one alternative do carry on
1206 into subsequent branches within the same subpattern. For example,
1207 .sp
1208 (a(?i)b|c)
1209 .sp
1210 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1211 branch is abandoned before the option setting. This is because the effects of
1212 option settings happen at compile time. There would be some very weird
1213 behaviour otherwise.
1214 .P
1215 \fBNote:\fP There are other PCRE-specific options that can be set by the
1216 application when the compile or match functions are called. In some cases the
1217 pattern can contain special leading sequences such as (*CRLF) to override what
1218 the application has set or what has been defaulted. Details are given in the
1219 section entitled
1220 .\" HTML <a href="#newlineseq">
1221 .\" </a>
1222 "Newline sequences"
1223 .\"
1224 above. There are also the (*UTF8) and (*UCP) leading sequences that can be used
1225 to set UTF-8 and Unicode property modes; they are equivalent to setting the
1226 PCRE_UTF8 and the PCRE_UCP options, respectively.
1227 .
1228 .
1229 .\" HTML <a name="subpattern"></a>
1231 .rs
1232 .sp
1233 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1234 Turning part of a pattern into a subpattern does two things:
1235 .sp
1236 1. It localizes a set of alternatives. For example, the pattern
1237 .sp
1238 cat(aract|erpillar|)
1239 .sp
1240 matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1241 match "cataract", "erpillar" or an empty string.
1242 .sp
1243 2. It sets up the subpattern as a capturing subpattern. This means that, when
1244 the whole pattern matches, that portion of the subject string that matched the
1245 subpattern is passed back to the caller via the \fIovector\fP argument of
1246 \fBpcre_exec()\fP. Opening parentheses are counted from left to right (starting
1247 from 1) to obtain numbers for the capturing subpatterns. For example, if the
1248 string "the red king" is matched against the pattern
1249 .sp
1250 the ((red|white) (king|queen))
1251 .sp
1252 the captured substrings are "red king", "red", and "king", and are numbered 1,
1253 2, and 3, respectively.
1254 .P
1255 The fact that plain parentheses fulfil two functions is not always helpful.
1256 There are often times when a grouping subpattern is required without a
1257 capturing requirement. If an opening parenthesis is followed by a question mark
1258 and a colon, the subpattern does not do any capturing, and is not counted when
1259 computing the number of any subsequent capturing subpatterns. For example, if
1260 the string "the white queen" is matched against the pattern
1261 .sp
1262 the ((?:red|white) (king|queen))
1263 .sp
1264 the captured substrings are "white queen" and "queen", and are numbered 1 and
1265 2. The maximum number of capturing subpatterns is 65535.
1266 .P
1267 As a convenient shorthand, if any option settings are required at the start of
1268 a non-capturing subpattern, the option letters may appear between the "?" and
1269 the ":". Thus the two patterns
1270 .sp
1271 (?i:saturday|sunday)
1272 (?:(?i)saturday|sunday)
1273 .sp
1274 match exactly the same set of strings. Because alternative branches are tried
1275 from left to right, and options are not reset until the end of the subpattern
1276 is reached, an option setting in one branch does affect subsequent branches, so
1277 the above patterns match "SUNDAY" as well as "Saturday".
1278 .
1279 .
1280 .\" HTML <a name="dupsubpatternnumber"></a>
1282 .rs
1283 .sp
1284 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1285 the same numbers for its capturing parentheses. Such a subpattern starts with
1286 (?| and is itself a non-capturing subpattern. For example, consider this
1287 pattern:
1288 .sp
1289 (?|(Sat)ur|(Sun))day
1290 .sp
1291 Because the two alternatives are inside a (?| group, both sets of capturing
1292 parentheses are numbered one. Thus, when the pattern matches, you can look
1293 at captured substring number one, whichever alternative matched. This construct
1294 is useful when you want to capture part, but not all, of one of a number of
1295 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1296 number is reset at the start of each branch. The numbers of any capturing
1297 parentheses that follow the subpattern start after the highest number used in
1298 any branch. The following example is taken from the Perl documentation. The
1299 numbers underneath show in which buffer the captured content will be stored.
1300 .sp
1301 # before ---------------branch-reset----------- after
1302 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1303 # 1 2 2 3 2 3 4
1304 .sp
1305 A back reference to a numbered subpattern uses the most recent value that is
1306 set for that number by any subpattern. The following pattern matches "abcabc"
1307 or "defdef":
1308 .sp
1309 /(?|(abc)|(def))\e1/
1310 .sp
1311 In contrast, a recursive or "subroutine" call to a numbered subpattern always
1312 refers to the first one in the pattern with the given number. The following
1313 pattern matches "abcabc" or "defabc":
1314 .sp
1315 /(?|(abc)|(def))(?1)/
1316 .sp
1317 If a
1318 .\" HTML <a href="#conditions">
1319 .\" </a>
1320 condition test
1321 .\"
1322 for a subpattern's having matched refers to a non-unique number, the test is
1323 true if any of the subpatterns of that number have matched.
1324 .P
1325 An alternative approach to using this "branch reset" feature is to use
1326 duplicate named subpatterns, as described in the next section.
1327 .
1328 .
1330 .rs
1331 .sp
1332 Identifying capturing parentheses by number is simple, but it can be very hard
1333 to keep track of the numbers in complicated regular expressions. Furthermore,
1334 if an expression is modified, the numbers may change. To help with this
1335 difficulty, PCRE supports the naming of subpatterns. This feature was not
1336 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1337 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1338 the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1339 have different names, but PCRE does not.
1340 .P
1341 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1342 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1343 parentheses from other parts of the pattern, such as
1344 .\" HTML <a href="#backreferences">
1345 .\" </a>
1346 back references,
1347 .\"
1348 .\" HTML <a href="#recursion">
1349 .\" </a>
1350 recursion,
1351 .\"
1352 and
1353 .\" HTML <a href="#conditions">
1354 .\" </a>
1355 conditions,
1356 .\"
1357 can be made by name as well as by number.
1358 .P
1359 Names consist of up to 32 alphanumeric characters and underscores. Named
1360 capturing parentheses are still allocated numbers as well as names, exactly as
1361 if the names were not present. The PCRE API provides function calls for
1362 extracting the name-to-number translation table from a compiled pattern. There
1363 is also a convenience function for extracting a captured substring by name.
1364 .P
1365 By default, a name must be unique within a pattern, but it is possible to relax
1366 this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1367 names are also always permitted for subpatterns with the same number, set up as
1368 described in the previous section.) Duplicate names can be useful for patterns
1369 where only one instance of the named parentheses can match. Suppose you want to
1370 match the name of a weekday, either as a 3-letter abbreviation or as the full
1371 name, and in both cases you want to extract the abbreviation. This pattern
1372 (ignoring the line breaks) does the job:
1373 .sp
1374 (?<DN>Mon|Fri|Sun)(?:day)?|
1375 (?<DN>Tue)(?:sday)?|
1376 (?<DN>Wed)(?:nesday)?|
1377 (?<DN>Thu)(?:rsday)?|
1378 (?<DN>Sat)(?:urday)?
1379 .sp
1380 There are five capturing substrings, but only one is ever set after a match.
1381 (An alternative way of solving this problem is to use a "branch reset"
1382 subpattern, as described in the previous section.)
1383 .P
1384 The convenience function for extracting the data by name returns the substring
1385 for the first (and in this example, the only) subpattern of that name that
1386 matched. This saves searching to find which numbered subpattern it was.
1387 .P
1388 If you make a back reference to a non-unique named subpattern from elsewhere in
1389 the pattern, the one that corresponds to the first occurrence of the name is
1390 used. In the absence of duplicate numbers (see the previous section) this is
1391 the one with the lowest number. If you use a named reference in a condition
1392 test (see the
1393 .\"
1394 .\" HTML <a href="#conditions">
1395 .\" </a>
1396 section about conditions
1397 .\"
1398 below), either to check whether a subpattern has matched, or to check for
1399 recursion, all subpatterns with the same name are tested. If the condition is
1400 true for any one of them, the overall condition is true. This is the same
1401 behaviour as testing by number. For further details of the interfaces for
1402 handling named subpatterns, see the
1403 .\" HREF
1404 \fBpcreapi\fP
1405 .\"
1406 documentation.
1407 .P
1408 \fBWarning:\fP You cannot use different names to distinguish between two
1409 subpatterns with the same number because PCRE uses only the numbers when
1410 matching. For this reason, an error is given at compile time if different names
1411 are given to subpatterns with the same number. However, you can give the same
1412 name to subpatterns with the same number, even when PCRE_DUPNAMES is not set.
1413 .
1414 .
1416 .rs
1417 .sp
1418 Repetition is specified by quantifiers, which can follow any of the following
1419 items:
1420 .sp
1421 a literal data character
1422 the dot metacharacter
1423 the \eC escape sequence
1424 the \eX escape sequence (in UTF-8 mode with Unicode properties)
1425 the \eR escape sequence
1426 an escape such as \ed or \epL that matches a single character
1427 a character class
1428 a back reference (see next section)
1429 a parenthesized subpattern (unless it is an assertion)
1430 a recursive or "subroutine" call to a subpattern
1431 .sp
1432 The general repetition quantifier specifies a minimum and maximum number of
1433 permitted matches, by giving the two numbers in curly brackets (braces),
1434 separated by a comma. The numbers must be less than 65536, and the first must
1435 be less than or equal to the second. For example:
1436 .sp
1437 z{2,4}
1438 .sp
1439 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1440 character. If the second number is omitted, but the comma is present, there is
1441 no upper limit; if the second number and the comma are both omitted, the
1442 quantifier specifies an exact number of required matches. Thus
1443 .sp
1444 [aeiou]{3,}
1445 .sp
1446 matches at least 3 successive vowels, but may match many more, while
1447 .sp
1448 \ed{8}
1449 .sp
1450 matches exactly 8 digits. An opening curly bracket that appears in a position
1451 where a quantifier is not allowed, or one that does not match the syntax of a
1452 quantifier, is taken as a literal character. For example, {,6} is not a
1453 quantifier, but a literal string of four characters.
1454 .P
1455 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
1456 bytes. Thus, for example, \ex{100}{2} matches two UTF-8 characters, each of
1457 which is represented by a two-byte sequence. Similarly, when Unicode property
1458 support is available, \eX{3} matches three Unicode extended sequences, each of
1459 which may be several bytes long (and they may be of different lengths).
1460 .P
1461 The quantifier {0} is permitted, causing the expression to behave as if the
1462 previous item and the quantifier were not present. This may be useful for
1463 subpatterns that are referenced as
1464 .\" HTML <a href="#subpatternsassubroutines">
1465 .\" </a>
1466 subroutines
1467 .\"
1468 from elsewhere in the pattern (but see also the section entitled
1469 .\" HTML <a href="#subdefine">
1470 .\" </a>
1471 "Defining subpatterns for use by reference only"
1472 .\"
1473 below). Items other than subpatterns that have a {0} quantifier are omitted
1474 from the compiled pattern.
1475 .P
1476 For convenience, the three most common quantifiers have single-character
1477 abbreviations:
1478 .sp
1479 * is equivalent to {0,}
1480 + is equivalent to {1,}
1481 ? is equivalent to {0,1}
1482 .sp
1483 It is possible to construct infinite loops by following a subpattern that can
1484 match no characters with a quantifier that has no upper limit, for example:
1485 .sp
1486 (a?)*
1487 .sp
1488 Earlier versions of Perl and PCRE used to give an error at compile time for
1489 such patterns. However, because there are cases where this can be useful, such
1490 patterns are now accepted, but if any repetition of the subpattern does in fact
1491 match no characters, the loop is forcibly broken.
1492 .P
1493 By default, the quantifiers are "greedy", that is, they match as much as
1494 possible (up to the maximum number of permitted times), without causing the
1495 rest of the pattern to fail. The classic example of where this gives problems
1496 is in trying to match comments in C programs. These appear between /* and */
1497 and within the comment, individual * and / characters may appear. An attempt to
1498 match C comments by applying the pattern
1499 .sp
1500 /\e*.*\e*/
1501 .sp
1502 to the string
1503 .sp
1504 /* first comment */ not comment /* second comment */
1505 .sp
1506 fails, because it matches the entire string owing to the greediness of the .*
1507 item.
1508 .P
1509 However, if a quantifier is followed by a question mark, it ceases to be
1510 greedy, and instead matches the minimum number of times possible, so the
1511 pattern
1512 .sp
1513 /\e*.*?\e*/
1514 .sp
1515 does the right thing with the C comments. The meaning of the various
1516 quantifiers is not otherwise changed, just the preferred number of matches.
1517 Do not confuse this use of question mark with its use as a quantifier in its
1518 own right. Because it has two uses, it can sometimes appear doubled, as in
1519 .sp
1520 \ed??\ed
1521 .sp
1522 which matches one digit by preference, but can match two if that is the only
1523 way the rest of the pattern matches.
1524 .P
1525 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1526 the quantifiers are not greedy by default, but individual ones can be made
1527 greedy by following them with a question mark. In other words, it inverts the
1528 default behaviour.
1529 .P
1530 When a parenthesized subpattern is quantified with a minimum repeat count that
1531 is greater than 1 or with a limited maximum, more memory is required for the
1532 compiled pattern, in proportion to the size of the minimum or maximum.
1533 .P
1534 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1535 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1536 implicitly anchored, because whatever follows will be tried against every
1537 character position in the subject string, so there is no point in retrying the
1538 overall match at any position after the first. PCRE normally treats such a
1539 pattern as though it were preceded by \eA.
1540 .P
1541 In cases where it is known that the subject string contains no newlines, it is
1542 worth setting PCRE_DOTALL in order to obtain this optimization, or
1543 alternatively using ^ to indicate anchoring explicitly.
1544 .P
1545 However, there is one situation where the optimization cannot be used. When .*
1546 is inside capturing parentheses that are the subject of a back reference
1547 elsewhere in the pattern, a match at the start may fail where a later one
1548 succeeds. Consider, for example:
1549 .sp
1550 (.*)abc\e1
1551 .sp
1552 If the subject is "xyz123abc123" the match point is the fourth character. For
1553 this reason, such a pattern is not implicitly anchored.
1554 .P
1555 When a capturing subpattern is repeated, the value captured is the substring
1556 that matched the final iteration. For example, after
1557 .sp
1558 (tweedle[dume]{3}\es*)+
1559 .sp
1560 has matched "tweedledum tweedledee" the value of the captured substring is
1561 "tweedledee". However, if there are nested capturing subpatterns, the
1562 corresponding captured values may have been set in previous iterations. For
1563 example, after
1564 .sp
1565 /(a|(b))+/
1566 .sp
1567 matches "aba" the value of the second captured substring is "b".
1568 .
1569 .
1570 .\" HTML <a name="atomicgroup"></a>
1572 .rs
1573 .sp
1574 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1575 repetition, failure of what follows normally causes the repeated item to be
1576 re-evaluated to see if a different number of repeats allows the rest of the
1577 pattern to match. Sometimes it is useful to prevent this, either to change the
1578 nature of the match, or to cause it fail earlier than it otherwise might, when
1579 the author of the pattern knows there is no point in carrying on.
1580 .P
1581 Consider, for example, the pattern \ed+foo when applied to the subject line
1582 .sp
1583 123456bar
1584 .sp
1585 After matching all 6 digits and then failing to match "foo", the normal
1586 action of the matcher is to try again with only 5 digits matching the \ed+
1587 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1588 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1589 that once a subpattern has matched, it is not to be re-evaluated in this way.
1590 .P
1591 If we use atomic grouping for the previous example, the matcher gives up
1592 immediately on failing to match "foo" the first time. The notation is a kind of
1593 special parenthesis, starting with (?> as in this example:
1594 .sp
1595 (?>\ed+)foo
1596 .sp
1597 This kind of parenthesis "locks up" the part of the pattern it contains once
1598 it has matched, and a failure further into the pattern is prevented from
1599 backtracking into it. Backtracking past it to previous items, however, works as
1600 normal.
1601 .P
1602 An alternative description is that a subpattern of this type matches the string
1603 of characters that an identical standalone pattern would match, if anchored at
1604 the current point in the subject string.
1605 .P
1606 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1607 the above example can be thought of as a maximizing repeat that must swallow
1608 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1609 number of digits they match in order to make the rest of the pattern match,
1610 (?>\ed+) can only match an entire sequence of digits.
1611 .P
1612 Atomic groups in general can of course contain arbitrarily complicated
1613 subpatterns, and can be nested. However, when the subpattern for an atomic
1614 group is just a single repeated item, as in the example above, a simpler
1615 notation, called a "possessive quantifier" can be used. This consists of an
1616 additional + character following a quantifier. Using this notation, the
1617 previous example can be rewritten as
1618 .sp
1619 \ed++foo
1620 .sp
1621 Note that a possessive quantifier can be used with an entire group, for
1622 example:
1623 .sp
1624 (abc|xyz){2,3}+
1625 .sp
1626 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1627 option is ignored. They are a convenient notation for the simpler forms of
1628 atomic group. However, there is no difference in the meaning of a possessive
1629 quantifier and the equivalent atomic group, though there may be a performance
1630 difference; possessive quantifiers should be slightly faster.
1631 .P
1632 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1633 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1634 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1635 package, and PCRE copied it from there. It ultimately found its way into Perl
1636 at release 5.10.
1637 .P
1638 PCRE has an optimization that automatically "possessifies" certain simple
1639 pattern constructs. For example, the sequence A+B is treated as A++B because
1640 there is no point in backtracking into a sequence of A's when B must follow.
1641 .P
1642 When a pattern contains an unlimited repeat inside a subpattern that can itself
1643 be repeated an unlimited number of times, the use of an atomic group is the
1644 only way to avoid some failing matches taking a very long time indeed. The
1645 pattern
1646 .sp
1647 (\eD+|<\ed+>)*[!?]
1648 .sp
1649 matches an unlimited number of substrings that either consist of non-digits, or
1650 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1651 quickly. However, if it is applied to
1652 .sp
1653 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1654 .sp
1655 it takes a long time before reporting failure. This is because the string can
1656 be divided between the internal \eD+ repeat and the external * repeat in a
1657 large number of ways, and all have to be tried. (The example uses [!?] rather
1658 than a single character at the end, because both PCRE and Perl have an
1659 optimization that allows for fast failure when a single character is used. They
1660 remember the last single character that is required for a match, and fail early
1661 if it is not present in the string.) If the pattern is changed so that it uses
1662 an atomic group, like this:
1663 .sp
1664 ((?>\eD+)|<\ed+>)*[!?]
1665 .sp
1666 sequences of non-digits cannot be broken, and failure happens quickly.
1667 .
1668 .
1669 .\" HTML <a name="backreferences"></a>
1671 .rs
1672 .sp
1673 Outside a character class, a backslash followed by a digit greater than 0 (and
1674 possibly further digits) is a back reference to a capturing subpattern earlier
1675 (that is, to its left) in the pattern, provided there have been that many
1676 previous capturing left parentheses.
1677 .P
1678 However, if the decimal number following the backslash is less than 10, it is
1679 always taken as a back reference, and causes an error only if there are not
1680 that many capturing left parentheses in the entire pattern. In other words, the
1681 parentheses that are referenced need not be to the left of the reference for
1682 numbers less than 10. A "forward back reference" of this type can make sense
1683 when a repetition is involved and the subpattern to the right has participated
1684 in an earlier iteration.
1685 .P
1686 It is not possible to have a numerical "forward back reference" to a subpattern
1687 whose number is 10 or more using this syntax because a sequence such as \e50 is
1688 interpreted as a character defined in octal. See the subsection entitled
1689 "Non-printing characters"
1690 .\" HTML <a href="#digitsafterbackslash">
1691 .\" </a>
1692 above
1693 .\"
1694 for further details of the handling of digits following a backslash. There is
1695 no such problem when named parentheses are used. A back reference to any
1696 subpattern is possible using named parentheses (see below).
1697 .P
1698 Another way of avoiding the ambiguity inherent in the use of digits following a
1699 backslash is to use the \eg escape sequence. This escape must be followed by an
1700 unsigned number or a negative number, optionally enclosed in braces. These
1701 examples are all identical:
1702 .sp
1703 (ring), \e1
1704 (ring), \eg1
1705 (ring), \eg{1}
1706 .sp
1707 An unsigned number specifies an absolute reference without the ambiguity that
1708 is present in the older syntax. It is also useful when literal digits follow
1709 the reference. A negative number is a relative reference. Consider this
1710 example:
1711 .sp
1712 (abc(def)ghi)\eg{-1}
1713 .sp
1714 The sequence \eg{-1} is a reference to the most recently started capturing
1715 subpattern before \eg, that is, is it equivalent to \e2 in this example.
1716 Similarly, \eg{-2} would be equivalent to \e1. The use of relative references
1717 can be helpful in long patterns, and also in patterns that are created by
1718 joining together fragments that contain references within themselves.
1719 .P
1720 A back reference matches whatever actually matched the capturing subpattern in
1721 the current subject string, rather than anything matching the subpattern
1722 itself (see
1723 .\" HTML <a href="#subpatternsassubroutines">
1724 .\" </a>
1725 "Subpatterns as subroutines"
1726 .\"
1727 below for a way of doing that). So the pattern
1728 .sp
1729 (sens|respons)e and \e1ibility
1730 .sp
1731 matches "sense and sensibility" and "response and responsibility", but not
1732 "sense and responsibility". If caseful matching is in force at the time of the
1733 back reference, the case of letters is relevant. For example,
1734 .sp
1735 ((?i)rah)\es+\e1
1736 .sp
1737 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1738 capturing subpattern is matched caselessly.
1739 .P
1740 There are several different ways of writing back references to named
1741 subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
1742 \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1743 back reference syntax, in which \eg can be used for both numeric and named
1744 references, is also supported. We could rewrite the above example in any of
1745 the following ways:
1746 .sp
1747 (?<p1>(?i)rah)\es+\ek<p1>
1748 (?'p1'(?i)rah)\es+\ek{p1}
1749 (?P<p1>(?i)rah)\es+(?P=p1)
1750 (?<p1>(?i)rah)\es+\eg{p1}
1751 .sp
1752 A subpattern that is referenced by name may appear in the pattern before or
1753 after the reference.
1754 .P
1755 There may be more than one back reference to the same subpattern. If a
1756 subpattern has not actually been used in a particular match, any back
1757 references to it always fail by default. For example, the pattern
1758 .sp
1759 (a|(bc))\e2
1760 .sp
1761 always fails if it starts to match "a" rather than "bc". However, if the
1762 PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
1763 unset value matches an empty string.
1764 .P
1765 Because there may be many capturing parentheses in a pattern, all digits
1766 following a backslash are taken as part of a potential back reference number.
1767 If the pattern continues with a digit character, some delimiter must be used to
1768 terminate the back reference. If the PCRE_EXTENDED option is set, this can be
1769 whitespace. Otherwise, the \eg{ syntax or an empty comment (see
1770 .\" HTML <a href="#comments">
1771 .\" </a>
1772 "Comments"
1773 .\"
1774 below) can be used.
1775 .
1776 .SS "Recursive back references"
1777 .rs
1778 .sp
1779 A back reference that occurs inside the parentheses to which it refers fails
1780 when the subpattern is first used, so, for example, (a\e1) never matches.
1781 However, such references can be useful inside repeated subpatterns. For
1782 example, the pattern
1783 .sp
1784 (a|b\e1)+
1785 .sp
1786 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1787 the subpattern, the back reference matches the character string corresponding
1788 to the previous iteration. In order for this to work, the pattern must be such
1789 that the first iteration does not need to match the back reference. This can be
1790 done using alternation, as in the example above, or by a quantifier with a
1791 minimum of zero.
1792 .P
1793 Back references of this type cause the group that they reference to be treated
1794 as an
1795 .\" HTML <a href="#atomicgroup">
1796 .\" </a>
1797 atomic group.
1798 .\"
1799 Once the whole group has been matched, a subsequent matching failure cannot
1800 cause backtracking into the middle of the group.
1801 .
1802 .
1803 .\" HTML <a name="bigassertions"></a>
1805 .rs
1806 .sp
1807 An assertion is a test on the characters following or preceding the current
1808 matching point that does not actually consume any characters. The simple
1809 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
1810 .\" HTML <a href="#smallassertions">
1811 .\" </a>
1812 above.
1813 .\"
1814 .P
1815 More complicated assertions are coded as subpatterns. There are two kinds:
1816 those that look ahead of the current position in the subject string, and those
1817 that look behind it. An assertion subpattern is matched in the normal way,
1818 except that it does not cause the current matching position to be changed.
1819 .P
1820 Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1821 because it makes no sense to assert the same thing several times. If any kind
1822 of assertion contains capturing subpatterns within it, these are counted for
1823 the purposes of numbering the capturing subpatterns in the whole pattern.
1824 However, substring capturing is carried out only for positive assertions,
1825 because it does not make sense for negative assertions.
1826 .
1827 .
1828 .SS "Lookahead assertions"
1829 .rs
1830 .sp
1831 Lookahead assertions start with (?= for positive assertions and (?! for
1832 negative assertions. For example,
1833 .sp
1834 \ew+(?=;)
1835 .sp
1836 matches a word followed by a semicolon, but does not include the semicolon in
1837 the match, and
1838 .sp
1839 foo(?!bar)
1840 .sp
1841 matches any occurrence of "foo" that is not followed by "bar". Note that the
1842 apparently similar pattern
1843 .sp
1844 (?!foo)bar
1845 .sp
1846 does not find an occurrence of "bar" that is preceded by something other than
1847 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1848 (?!foo) is always true when the next three characters are "bar". A
1849 lookbehind assertion is needed to achieve the other effect.
1850 .P
1851 If you want to force a matching failure at some point in a pattern, the most
1852 convenient way to do it is with (?!) because an empty string always matches, so
1853 an assertion that requires there not to be an empty string must always fail.
1854 The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
1855 .
1856 .
1857 .\" HTML <a name="lookbehind"></a>
1858 .SS "Lookbehind assertions"
1859 .rs
1860 .sp
1861 Lookbehind assertions start with (?<= for positive assertions and (?<! for
1862 negative assertions. For example,
1863 .sp
1864 (?<!foo)bar
1865 .sp
1866 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1867 a lookbehind assertion are restricted such that all the strings it matches must
1868 have a fixed length. However, if there are several top-level alternatives, they
1869 do not all have to have the same fixed length. Thus
1870 .sp
1871 (?<=bullock|donkey)
1872 .sp
1873 is permitted, but
1874 .sp
1875 (?<!dogs?|cats?)
1876 .sp
1877 causes an error at compile time. Branches that match different length strings
1878 are permitted only at the top level of a lookbehind assertion. This is an
1879 extension compared with Perl, which requires all branches to match the same
1880 length of string. An assertion such as
1881 .sp
1882 (?<=ab(c|de))
1883 .sp
1884 is not permitted, because its single top-level branch can match two different
1885 lengths, but it is acceptable to PCRE if rewritten to use two top-level
1886 branches:
1887 .sp
1888 (?<=abc|abde)
1889 .sp
1890 In some cases, the escape sequence \eK
1891 .\" HTML <a href="#resetmatchstart">
1892 .\" </a>
1893 (see above)
1894 .\"
1895 can be used instead of a lookbehind assertion to get round the fixed-length
1896 restriction.
1897 .P
1898 The implementation of lookbehind assertions is, for each alternative, to
1899 temporarily move the current position back by the fixed length and then try to
1900 match. If there are insufficient characters before the current position, the
1901 assertion fails.
1902 .P
1903 PCRE does not allow the \eC escape (which matches a single byte in UTF-8 mode)
1904 to appear in lookbehind assertions, because it makes it impossible to calculate
1905 the length of the lookbehind. The \eX and \eR escapes, which can match
1906 different numbers of bytes, are also not permitted.
1907 .P
1908 .\" HTML <a href="#subpatternsassubroutines">
1909 .\" </a>
1910 "Subroutine"
1911 .\"
1912 calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
1913 as the subpattern matches a fixed-length string.
1914 .\" HTML <a href="#recursion">
1915 .\" </a>
1916 Recursion,
1917 .\"
1918 however, is not supported.
1919 .P
1920 Possessive quantifiers can be used in conjunction with lookbehind assertions to
1921 specify efficient matching of fixed-length strings at the end of subject
1922 strings. Consider a simple pattern such as
1923 .sp
1924 abcd$
1925 .sp
1926 when applied to a long string that does not match. Because matching proceeds
1927 from left to right, PCRE will look for each "a" in the subject and then see if
1928 what follows matches the rest of the pattern. If the pattern is specified as
1929 .sp
1930 ^.*abcd$
1931 .sp
1932 the initial .* matches the entire string at first, but when this fails (because
1933 there is no following "a"), it backtracks to match all but the last character,
1934 then all but the last two characters, and so on. Once again the search for "a"
1935 covers the entire string, from right to left, so we are no better off. However,
1936 if the pattern is written as
1937 .sp
1938 ^.*+(?<=abcd)
1939 .sp
1940 there can be no backtracking for the .*+ item; it can match only the entire
1941 string. The subsequent lookbehind assertion does a single test on the last four
1942 characters. If it fails, the match fails immediately. For long strings, this
1943 approach makes a significant difference to the processing time.
1944 .
1945 .
1946 .SS "Using multiple assertions"
1947 .rs
1948 .sp
1949 Several assertions (of any sort) may occur in succession. For example,
1950 .sp
1951 (?<=\ed{3})(?<!999)foo
1952 .sp
1953 matches "foo" preceded by three digits that are not "999". Notice that each of
1954 the assertions is applied independently at the same point in the subject
1955 string. First there is a check that the previous three characters are all
1956 digits, and then there is a check that the same three characters are not "999".
1957 This pattern does \fInot\fP match "foo" preceded by six characters, the first
1958 of which are digits and the last three of which are not "999". For example, it
1959 doesn't match "123abcfoo". A pattern to do that is
1960 .sp
1961 (?<=\ed{3}...)(?<!999)foo
1962 .sp
1963 This time the first assertion looks at the preceding six characters, checking
1964 that the first three are digits, and then the second assertion checks that the
1965 preceding three characters are not "999".
1966 .P
1967 Assertions can be nested in any combination. For example,
1968 .sp
1969 (?<=(?<!foo)bar)baz
1970 .sp
1971 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1972 preceded by "foo", while
1973 .sp
1974 (?<=\ed{3}(?!999)...)foo
1975 .sp
1976 is another pattern that matches "foo" preceded by three digits and any three
1977 characters that are not "999".
1978 .
1979 .
1980 .\" HTML <a name="conditions"></a>
1982 .rs
1983 .sp
1984 It is possible to cause the matching process to obey a subpattern
1985 conditionally or to choose between two alternative subpatterns, depending on
1986 the result of an assertion, or whether a specific capturing subpattern has
1987 already been matched. The two possible forms of conditional subpattern are:
1988 .sp
1989 (?(condition)yes-pattern)
1990 (?(condition)yes-pattern|no-pattern)
1991 .sp
1992 If the condition is satisfied, the yes-pattern is used; otherwise the
1993 no-pattern (if present) is used. If there are more than two alternatives in the
1994 subpattern, a compile-time error occurs. Each of the two alternatives may
1995 itself contain nested subpatterns of any form, including conditional
1996 subpatterns; the restriction to two alternatives applies only at the level of
1997 the condition. This pattern fragment is an example where the alternatives are
1998 complex:
1999 .sp
2000 (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
2001 .sp
2002 .P
2003 There are four kinds of condition: references to subpatterns, references to
2004 recursion, a pseudo-condition called DEFINE, and assertions.
2005 .
2006 .SS "Checking for a used subpattern by number"
2007 .rs
2008 .sp
2009 If the text between the parentheses consists of a sequence of digits, the
2010 condition is true if a capturing subpattern of that number has previously
2011 matched. If there is more than one capturing subpattern with the same number
2012 (see the earlier
2013 .\"
2014 .\" HTML <a href="#recursion">
2015 .\" </a>
2016 section about duplicate subpattern numbers),
2017 .\"
2018 the condition is true if any of them have matched. An alternative notation is
2019 to precede the digits with a plus or minus sign. In this case, the subpattern
2020 number is relative rather than absolute. The most recently opened parentheses
2021 can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
2022 loops it can also make sense to refer to subsequent groups. The next
2023 parentheses to be opened can be referenced as (?(+1), and so on. (The value
2024 zero in any of these forms is not used; it provokes a compile-time error.)
2025 .P
2026 Consider the following pattern, which contains non-significant white space to
2027 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2028 three parts for ease of discussion:
2029 .sp
2030 ( \e( )? [^()]+ (?(1) \e) )
2031 .sp
2032 The first part matches an optional opening parenthesis, and if that
2033 character is present, sets it as the first captured substring. The second part
2034 matches one or more characters that are not parentheses. The third part is a
2035 conditional subpattern that tests whether or not the first set of parentheses
2036 matched. If they did, that is, if subject started with an opening parenthesis,
2037 the condition is true, and so the yes-pattern is executed and a closing
2038 parenthesis is required. Otherwise, since no-pattern is not present, the
2039 subpattern matches nothing. In other words, this pattern matches a sequence of
2040 non-parentheses, optionally enclosed in parentheses.
2041 .P
2042 If you were embedding this pattern in a larger one, you could use a relative
2043 reference:
2044 .sp
2045 ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
2046 .sp
2047 This makes the fragment independent of the parentheses in the larger pattern.
2048 .
2049 .SS "Checking for a used subpattern by name"
2050 .rs
2051 .sp
2052 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
2053 subpattern by name. For compatibility with earlier versions of PCRE, which had
2054 this facility before Perl, the syntax (?(name)...) is also recognized. However,
2055 there is a possible ambiguity with this syntax, because subpattern names may
2056 consist entirely of digits. PCRE looks first for a named subpattern; if it
2057 cannot find one and the name consists entirely of digits, PCRE looks for a
2058 subpattern of that number, which must be greater than zero. Using subpattern
2059 names that consist entirely of digits is not recommended.
2060 .P
2061 Rewriting the above example to use a named subpattern gives this:
2062 .sp
2063 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
2064 .sp
2065 If the name used in a condition of this kind is a duplicate, the test is
2066 applied to all subpatterns of the same name, and is true if any one of them has
2067 matched.
2068 .
2069 .SS "Checking for pattern recursion"
2070 .rs
2071 .sp
2072 If the condition is the string (R), and there is no subpattern with the name R,
2073 the condition is true if a recursive call to the whole pattern or any
2074 subpattern has been made. If digits or a name preceded by ampersand follow the
2075 letter R, for example:
2076 .sp
2077 (?(R3)...) or (?(R&name)...)
2078 .sp
2079 the condition is true if the most recent recursion is into a subpattern whose
2080 number or name is given. This condition does not check the entire recursion
2081 stack. If the name used in a condition of this kind is a duplicate, the test is
2082 applied to all subpatterns of the same name, and is true if any one of them is
2083 the most recent recursion.
2084 .P
2085 At "top level", all these recursion test conditions are false.
2086 .\" HTML <a href="#recursion">
2087 .\" </a>
2088 The syntax for recursive patterns
2089 .\"
2090 is described below.
2091 .
2092 .\" HTML <a name="subdefine"></a>
2093 .SS "Defining subpatterns for use by reference only"
2094 .rs
2095 .sp
2096 If the condition is the string (DEFINE), and there is no subpattern with the
2097 name DEFINE, the condition is always false. In this case, there may be only one
2098 alternative in the subpattern. It is always skipped if control reaches this
2099 point in the pattern; the idea of DEFINE is that it can be used to define
2100 "subroutines" that can be referenced from elsewhere. (The use of
2101 .\" HTML <a href="#subpatternsassubroutines">
2102 .\" </a>
2103 "subroutines"
2104 .\"
2105 is described below.) For example, a pattern to match an IPv4 address such as
2106 "" could be written like this (ignore whitespace and line
2107 breaks):
2108 .sp
2109 (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
2110 \eb (?&byte) (\e.(?&byte)){3} \eb
2111 .sp
2112 The first part of the pattern is a DEFINE group inside which a another group
2113 named "byte" is defined. This matches an individual component of an IPv4
2114 address (a number less than 256). When matching takes place, this part of the
2115 pattern is skipped because DEFINE acts like a false condition. The rest of the
2116 pattern uses references to the named group to match the four dot-separated
2117 components of an IPv4 address, insisting on a word boundary at each end.
2118 .
2119 .SS "Assertion conditions"
2120 .rs
2121 .sp
2122 If the condition is not in any of the above formats, it must be an assertion.
2123 This may be a positive or negative lookahead or lookbehind assertion. Consider
2124 this pattern, again containing non-significant white space, and with the two
2125 alternatives on the second line:
2126 .sp
2127 (?(?=[^a-z]*[a-z])
2128 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
2129 .sp
2130 The condition is a positive lookahead assertion that matches an optional
2131 sequence of non-letters followed by a letter. In other words, it tests for the
2132 presence of at least one letter in the subject. If a letter is found, the
2133 subject is matched against the first alternative; otherwise it is matched
2134 against the second. This pattern matches strings in one of the two forms
2135 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2136 .
2137 .
2138 .\" HTML <a name="comments"></a>
2140 .rs
2141 .sp
2142 There are two ways of including comments in patterns that are processed by
2143 PCRE. In both cases, the start of the comment must not be in a character class,
2144 nor in the middle of any other sequence of related characters such as (?: or a
2145 subpattern name or number. The characters that make up a comment play no part
2146 in the pattern matching.
2147 .P
2148 The sequence (?# marks the start of a comment that continues up to the next
2149 closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2150 option is set, an unescaped # character also introduces a comment, which in
2151 this case continues to immediately after the next newline character or
2152 character sequence in the pattern. Which characters are interpreted as newlines
2153 is controlled by the options passed to \fBpcre_compile()\fP or by a special
2154 sequence at the start of the pattern, as described in the section entitled
2155 .\" HTML <a href="#newlines">
2156 .\" </a>
2157 "Newline conventions"
2158 .\"
2159 above. Note that the end of this type of comment is a literal newline sequence
2160 in the pattern; escape sequences that happen to represent a newline do not
2161 count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2162 default newline convention is in force:
2163 .sp
2164 abc #comment \en still comment
2165 .sp
2166 On encountering the # character, \fBpcre_compile()\fP skips along, looking for
2167 a newline in the pattern. The sequence \en is still literal at this stage, so
2168 it does not terminate the comment. Only an actual character with the code value
2169 0x0a (the default newline) does so.
2170 .
2171 .
2172 .\" HTML <a name="recursion"></a>
2174 .rs
2175 .sp
2176 Consider the problem of matching a string in parentheses, allowing for
2177 unlimited nested parentheses. Without the use of recursion, the best that can
2178 be done is to use a pattern that matches up to some fixed depth of nesting. It
2179 is not possible to handle an arbitrary nesting depth.
2180 .P
2181 For some time, Perl has provided a facility that allows regular expressions to
2182 recurse (amongst other things). It does this by interpolating Perl code in the
2183 expression at run time, and the code can refer to the expression itself. A Perl
2184 pattern using code interpolation to solve the parentheses problem can be
2185 created like this:
2186 .sp
2187 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
2188 .sp
2189 The (?p{...}) item interpolates Perl code at run time, and in this case refers
2190 recursively to the pattern in which it appears.
2191 .P
2192 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2193 supports special syntax for recursion of the entire pattern, and also for
2194 individual subpattern recursion. After its introduction in PCRE and Python,
2195 this kind of recursion was subsequently introduced into Perl at release 5.10.
2196 .P
2197 A special item that consists of (? followed by a number greater than zero and a
2198 closing parenthesis is a recursive call of the subpattern of the given number,
2199 provided that it occurs inside that subpattern. (If not, it is a
2200 .\" HTML <a href="#subpatternsassubroutines">
2201 .\" </a>
2202 "subroutine"
2203 .\"
2204 call, which is described in the next section.) The special item (?R) or (?0) is
2205 a recursive call of the entire regular expression.
2206 .P
2207 This PCRE pattern solves the nested parentheses problem (assume the
2208 PCRE_EXTENDED option is set so that white space is ignored):
2209 .sp
2210 \e( ( [^()]++ | (?R) )* \e)
2211 .sp
2212 First it matches an opening parenthesis. Then it matches any number of
2213 substrings which can either be a sequence of non-parentheses, or a recursive
2214 match of the pattern itself (that is, a correctly parenthesized substring).
2215 Finally there is a closing parenthesis. Note the use of a possessive quantifier
2216 to avoid backtracking into sequences of non-parentheses.
2217 .P
2218 If this were part of a larger pattern, you would not want to recurse the entire
2219 pattern, so instead you could use this:
2220 .sp
2221 ( \e( ( [^()]++ | (?1) )* \e) )
2222 .sp
2223 We have put the pattern into parentheses, and caused the recursion to refer to
2224 them instead of the whole pattern.
2225 .P
2226 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2227 is made easier by the use of relative references. Instead of (?1) in the
2228 pattern above you can write (?-2) to refer to the second most recently opened
2229 parentheses preceding the recursion. In other words, a negative number counts
2230 capturing parentheses leftwards from the point at which it is encountered.
2231 .P
2232 It is also possible to refer to subsequently opened parentheses, by writing
2233 references such as (?+2). However, these cannot be recursive because the
2234 reference is not inside the parentheses that are referenced. They are always
2235 .\" HTML <a href="#subpatternsassubroutines">
2236 .\" </a>
2237 "subroutine"
2238 .\"
2239 calls, as described in the next section.
2240 .P
2241 An alternative approach is to use named parentheses instead. The Perl syntax
2242 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
2243 could rewrite the above example as follows:
2244 .sp
2245 (?<pn> \e( ( [^()]++ | (?&pn) )* \e) )
2246 .sp
2247 If there is more than one subpattern with the same name, the earliest one is
2248 used.
2249 .P
2250 This particular example pattern that we have been looking at contains nested
2251 unlimited repeats, and so the use of a possessive quantifier for matching
2252 strings of non-parentheses is important when applying the pattern to strings
2253 that do not match. For example, when this pattern is applied to
2254 .sp
2255 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2256 .sp
2257 it yields "no match" quickly. However, if a possessive quantifier is not used,
2258 the match runs for a very long time indeed because there are so many different
2259 ways the + and * repeats can carve up the subject, and all have to be tested
2260 before failure can be reported.
2261 .P
2262 At the end of a match, the values of capturing parentheses are those from
2263 the outermost level. If you want to obtain intermediate values, a callout
2264 function can be used (see below and the
2265 .\" HREF
2266 \fBpcrecallout\fP
2267 .\"
2268 documentation). If the pattern above is matched against
2269 .sp
2270 (ab(cd)ef)
2271 .sp
2272 the value for the inner capturing parentheses (numbered 2) is "ef", which is
2273 the last value taken on at the top level. If a capturing subpattern is not
2274 matched at the top level, its final value is unset, even if it is (temporarily)
2275 set at a deeper level.
2276 .P
2277 If there are more than 15 capturing parentheses in a pattern, PCRE has to
2278 obtain extra memory to store data during a recursion, which it does by using
2279 \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no memory can
2280 be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2281 .P
2282 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2283 Consider this pattern, which matches text in angle brackets, allowing for
2284 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2285 recursing), whereas any characters are permitted at the outer level.
2286 .sp
2287 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
2288 .sp
2289 In this pattern, (?(R) is the start of a conditional subpattern, with two
2290 different alternatives for the recursive and non-recursive cases. The (?R) item
2291 is the actual recursive call.
2292 .
2293 .
2294 .\" HTML <a name="recursiondifference"></a>
2295 .SS "Recursion difference from Perl"
2296 .rs
2297 .sp
2298 In PCRE (like Python, but unlike Perl), a recursive subpattern call is always
2299 treated as an atomic group. That is, once it has matched some of the subject
2300 string, it is never re-entered, even if it contains untried alternatives and
2301 there is a subsequent matching failure. This can be illustrated by the
2302 following pattern, which purports to match a palindromic string that contains
2303 an odd number of characters (for example, "a", "aba", "abcba", "abcdcba"):
2304 .sp
2305 ^(.|(.)(?1)\e2)$
2306 .sp
2307 The idea is that it either matches a single character, or two identical
2308 characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2309 it does not if the pattern is longer than three characters. Consider the
2310 subject string "abcba":
2311 .P
2312 At the top level, the first character is matched, but as it is not at the end
2313 of the string, the first alternative fails; the second alternative is taken
2314 and the recursion kicks in. The recursive call to subpattern 1 successfully
2315 matches the next character ("b"). (Note that the beginning and end of line
2316 tests are not part of the recursion).
2317 .P
2318 Back at the top level, the next character ("c") is compared with what
2319 subpattern 2 matched, which was "a". This fails. Because the recursion is
2320 treated as an atomic group, there are now no backtracking points, and so the
2321 entire match fails. (Perl is able, at this point, to re-enter the recursion and
2322 try the second alternative.) However, if the pattern is written with the
2323 alternatives in the other order, things are different:
2324 .sp
2325 ^((.)(?1)\e2|.)$
2326 .sp
2327 This time, the recursing alternative is tried first, and continues to recurse
2328 until it runs out of characters, at which point the recursion fails. But this
2329 time we do have another alternative to try at the higher level. That is the big
2330 difference: in the previous case the remaining alternative is at a deeper
2331 recursion level, which PCRE cannot use.
2332 .P
2333 To change the pattern so that it matches all palindromic strings, not just
2334 those with an odd number of characters, it is tempting to change the pattern to
2335 this:
2336 .sp
2337 ^((.)(?1)\e2|.?)$
2338 .sp
2339 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2340 deeper recursion has matched a single character, it cannot be entered again in
2341 order to match an empty string. The solution is to separate the two cases, and
2342 write out the odd and even cases as alternatives at the higher level:
2343 .sp
2344 ^(?:((.)(?1)\e2|)|((.)(?3)\e4|.))
2345 .sp
2346 If you want to match typical palindromic phrases, the pattern has to ignore all
2347 non-word characters, which can be done like this:
2348 .sp
2349 ^\eW*+(?:((.)\eW*+(?1)\eW*+\e2|)|((.)\eW*+(?3)\eW*+\e4|\eW*+.\eW*+))\eW*+$
2350 .sp
2351 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2352 man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2353 the use of the possessive quantifier *+ to avoid backtracking into sequences of
2354 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2355 more) to match typical phrases, and Perl takes so long that you think it has
2356 gone into a loop.
2357 .P
2358 \fBWARNING\fP: The palindrome-matching patterns above work only if the subject
2359 string does not start with a palindrome that is shorter than the entire string.
2360 For example, although "abcba" is correctly matched, if the subject is "ababa",
2361 PCRE finds the palindrome "aba" at the start, then fails at top level because
2362 the end of the string does not follow. Once again, it cannot jump back into the
2363 recursion to try other alternatives, so the entire match fails.
2364 .
2365 .
2366 .\" HTML <a name="subpatternsassubroutines"></a>
2368 .rs
2369 .sp
2370 If the syntax for a recursive subpattern reference (either by number or by
2371 name) is used outside the parentheses to which it refers, it operates like a
2372 subroutine in a programming language. The "called" subpattern may be defined
2373 before or after the reference. A numbered reference can be absolute or
2374 relative, as in these examples:
2375 .sp
2376 (...(absolute)...)...(?2)...
2377 (...(relative)...)...(?-1)...
2378 (...(?+1)...(relative)...
2379 .sp
2380 An earlier example pointed out that the pattern
2381 .sp
2382 (sens|respons)e and \e1ibility
2383 .sp
2384 matches "sense and sensibility" and "response and responsibility", but not
2385 "sense and responsibility". If instead the pattern
2386 .sp
2387 (sens|respons)e and (?1)ibility
2388 .sp
2389 is used, it does match "sense and responsibility" as well as the other two
2390 strings. Another example is given in the discussion of DEFINE above.
2391 .P
2392 Like recursive subpatterns, a subroutine call is always treated as an atomic
2393 group. That is, once it has matched some of the subject string, it is never
2394 re-entered, even if it contains untried alternatives and there is a subsequent
2395 matching failure. Any capturing parentheses that are set during the subroutine
2396 call revert to their previous values afterwards.
2397 .P
2398 When a subpattern is used as a subroutine, processing options such as
2399 case-independence are fixed when the subpattern is defined. They cannot be
2400 changed for different calls. For example, consider this pattern:
2401 .sp
2402 (abc)(?i:(?-1))
2403 .sp
2404 It matches "abcabc". It does not match "abcABC" because the change of
2405 processing option does not affect the called subpattern.
2406 .
2407 .
2408 .\" HTML <a name="onigurumasubroutines"></a>
2410 .rs
2411 .sp
2412 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
2413 a number enclosed either in angle brackets or single quotes, is an alternative
2414 syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2415 are two of the examples used above, rewritten using this syntax:
2416 .sp
2417 (?<pn> \e( ( (?>[^()]+) | \eg<pn> )* \e) )
2418 (sens|respons)e and \eg'1'ibility
2419 .sp
2420 PCRE supports an extension to Oniguruma: if a number is preceded by a
2421 plus or a minus sign it is taken as a relative reference. For example:
2422 .sp
2423 (abc)(?i:\eg<-1>)
2424 .sp
2425 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
2426 synonymous. The former is a back reference; the latter is a subroutine call.
2427 .
2428 .
2430 .rs
2431 .sp
2432 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2433 code to be obeyed in the middle of matching a regular expression. This makes it
2434 possible, amongst other things, to extract different substrings that match the
2435 same pair of parentheses when there is a repetition.
2436 .P
2437 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2438 code. The feature is called "callout". The caller of PCRE provides an external
2439 function by putting its entry point in the global variable \fIpcre_callout\fP.
2440 By default, this variable contains NULL, which disables all calling out.
2441 .P
2442 Within a regular expression, (?C) indicates the points at which the external
2443 function is to be called. If you want to identify different callout points, you
2444 can put a number less than 256 after the letter C. The default value is zero.
2445 For example, this pattern has two callout points:
2446 .sp
2447 (?C1)abc(?C2)def
2448 .sp
2449 If the PCRE_AUTO_CALLOUT flag is passed to \fBpcre_compile()\fP, callouts are
2450 automatically installed before each item in the pattern. They are all numbered
2451 255.
2452 .P
2453 During matching, when PCRE reaches a callout point (and \fIpcre_callout\fP is
2454 set), the external function is called. It is provided with the number of the
2455 callout, the position in the pattern, and, optionally, one item of data
2456 originally supplied by the caller of \fBpcre_exec()\fP. The callout function
2457 may cause matching to proceed, to backtrack, or to fail altogether. A complete
2458 description of the interface to the callout function is given in the
2459 .\" HREF
2460 \fBpcrecallout\fP
2461 .\"
2462 documentation.
2463 .
2464 .
2465 .\" HTML <a name="backtrackcontrol"></a>
2467 .rs
2468 .sp
2469 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2470 are described in the Perl documentation as "experimental and subject to change
2471 or removal in a future version of Perl". It goes on to say: "Their usage in
2472 production code should be noted to avoid problems during upgrades." The same
2473 remarks apply to the PCRE features described in this section.
2474 .P
2475 Since these verbs are specifically related to backtracking, most of them can be
2476 used only when the pattern is to be matched using \fBpcre_exec()\fP, which uses
2477 a backtracking algorithm. With the exception of (*FAIL), which behaves like a
2478 failing negative assertion, they cause an error if encountered by
2479 \fBpcre_dfa_exec()\fP.
2480 .P
2481 If any of these verbs are used in an assertion or subroutine subpattern
2482 (including recursive subpatterns), their effect is confined to that subpattern;
2483 it does not extend to the surrounding pattern. Note that such subpatterns are
2484 processed as anchored at the point where they are tested.
2485 .P
2486 The new verbs make use of what was previously invalid syntax: an opening
2487 parenthesis followed by an asterisk. They are generally of the form
2488 (*VERB) or (*VERB:NAME). Some may take either form, with differing behaviour,
2489 depending on whether or not an argument is present. An name is a sequence of
2490 letters, digits, and underscores. If the name is empty, that is, if the closing
2491 parenthesis immediately follows the colon, the effect is as if the colon were
2492 not there. Any number of these verbs may occur in a pattern.
2493 .P
2494 PCRE contains some optimizations that are used to speed up matching by running
2495 some checks at the start of each match attempt. For example, it may know the
2496 minimum length of matching subject, or that a particular character must be
2497 present. When one of these optimizations suppresses the running of a match, any
2498 included backtracking verbs will not, of course, be processed. You can suppress
2499 the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2500 when calling \fBpcre_exec()\fP.
2501 .
2502 .
2503 .SS "Verbs that act immediately"
2504 .rs
2505 .sp
2506 The following verbs act as soon as they are encountered. They may not be
2507 followed by a name.
2508 .sp
2509 (*ACCEPT)
2510 .sp
2511 This verb causes the match to end successfully, skipping the remainder of the
2512 pattern. When inside a recursion, only the innermost pattern is ended
2513 immediately. If (*ACCEPT) is inside capturing parentheses, the data so far is
2514 captured. (This feature was added to PCRE at release 8.00.) For example:
2515 .sp
2516 A((?:A|B(*ACCEPT)|C)D)
2517 .sp
2518 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2519 the outer parentheses.
2520 .sp
2521 (*FAIL) or (*F)
2522 .sp
2523 This verb causes the match to fail, forcing backtracking to occur. It is
2524 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2525 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2526 Perl features that are not present in PCRE. The nearest equivalent is the
2527 callout feature, as for example in this pattern:
2528 .sp
2529 a+(?C)(*FAIL)
2530 .sp
2531 A match with the string "aaaa" always fails, but the callout is taken before
2532 each backtrack happens (in this example, 10 times).
2533 .
2534 .
2535 .SS "Recording which path was taken"
2536 .rs
2537 .sp
2538 There is one verb whose main purpose is to track how a match was arrived at,
2539 though it also has a secondary use in conjunction with advancing the match
2540 starting point (see (*SKIP) below).
2541 .sp
2542 (*MARK:NAME) or (*:NAME)
2543 .sp
2544 A name is always required with this verb. There may be as many instances of
2545 (*MARK) as you like in a pattern, and their names do not have to be unique.
2546 .P
2547 When a match succeeds, the name of the last-encountered (*MARK) is passed back
2548 to the caller via the \fIpcre_extra\fP data structure, as described in the
2549 .\" HTML <a href="pcreapi.html#extradata">
2550 .\" </a>
2551 section on \fIpcre_extra\fP
2552 .\"
2553 in the
2554 .\" HREF
2555 \fBpcreapi\fP
2556 .\"
2557 documentation. No data is returned for a partial match. Here is an example of
2558 \fBpcretest\fP output, where the /K modifier requests the retrieval and
2559 outputting of (*MARK) data:
2560 .sp
2561 /X(*MARK:A)Y|X(*MARK:B)Z/K
2562 XY
2563 0: XY
2564 MK: A
2565 XZ
2566 0: XZ
2567 MK: B
2568 .sp
2569 The (*MARK) name is tagged with "MK:" in this output, and in this example it
2570 indicates which of the two alternatives matched. This is a more efficient way
2571 of obtaining this information than putting each alternative in its own
2572 capturing parentheses.
2573 .P
2574 A name may also be returned after a failed match if the final path through the
2575 pattern involves (*MARK). However, unless (*MARK) used in conjunction with
2576 (*COMMIT), this is unlikely to happen for an unanchored pattern because, as the
2577 starting point for matching is advanced, the final check is often with an empty
2578 string, causing a failure before (*MARK) is reached. For example:
2579 .sp
2580 /X(*MARK:A)Y|X(*MARK:B)Z/K
2581 XP
2582 No match
2583 .sp
2584 There are three potential starting points for this match (starting with X,
2585 starting with P, and with an empty string). If the pattern is anchored, the
2586 result is different:
2587 .sp
2588 /^X(*MARK:A)Y|^X(*MARK:B)Z/K
2589 XP
2590 No match, mark = B
2591 .sp
2592 PCRE's start-of-match optimizations can also interfere with this. For example,
2593 if, as a result of a call to \fBpcre_study()\fP, it knows the minimum
2594 subject length for a match, a shorter subject will not be scanned at all.
2595 .P
2596 Note that similar anomalies (though different in detail) exist in Perl, no
2597 doubt for the same reasons. The use of (*MARK) data after a failed match of an
2598 unanchored pattern is not recommended, unless (*COMMIT) is involved.
2599 .
2600 .
2601 .SS "Verbs that act after backtracking"
2602 .rs
2603 .sp
2604 The following verbs do nothing when they are encountered. Matching continues
2605 with what follows, but if there is no subsequent match, causing a backtrack to
2606 the verb, a failure is forced. That is, backtracking cannot pass to the left of
2607 the verb. However, when one of these verbs appears inside an atomic group, its
2608 effect is confined to that group, because once the group has been matched,
2609 there is never any backtracking into it. In this situation, backtracking can
2610 "jump back" to the left of the entire atomic group. (Remember also, as stated
2611 above, that this localization also applies in subroutine calls and assertions.)
2612 .P
2613 These verbs differ in exactly what kind of failure occurs when backtracking
2614 reaches them.
2615 .sp
2616 (*COMMIT)
2617 .sp
2618 This verb, which may not be followed by a name, causes the whole match to fail
2619 outright if the rest of the pattern does not match. Even if the pattern is
2620 unanchored, no further attempts to find a match by advancing the starting point
2621 take place. Once (*COMMIT) has been passed, \fBpcre_exec()\fP is committed to
2622 finding a match at the current starting point, or not at all. For example:
2623 .sp
2624 a+(*COMMIT)b
2625 .sp
2626 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2627 dynamic anchor, or "I've started, so I must finish." The name of the most
2628 recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
2629 match failure.
2630 .P
2631 Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
2632 unless PCRE's start-of-match optimizations are turned off, as shown in this
2633 \fBpcretest\fP example:
2634 .sp
2635 /(*COMMIT)abc/
2636 xyzabc
2637 0: abc
2638 xyzabc\eY
2639 No match
2640 .sp
2641 PCRE knows that any match must start with "a", so the optimization skips along
2642 the subject to "a" before running the first match attempt, which succeeds. When
2643 the optimization is disabled by the \eY escape in the second subject, the match
2644 starts at "x" and so the (*COMMIT) causes it to fail without trying any other
2645 starting points.
2646 .sp
2647 (*PRUNE) or (*PRUNE:NAME)
2648 .sp
2649 This verb causes the match to fail at the current starting position in the
2650 subject if the rest of the pattern does not match. If the pattern is
2651 unanchored, the normal "bumpalong" advance to the next starting character then
2652 happens. Backtracking can occur as usual to the left of (*PRUNE), before it is
2653 reached, or when matching to the right of (*PRUNE), but if there is no match to
2654 the right, backtracking cannot cross (*PRUNE). In simple cases, the use of
2655 (*PRUNE) is just an alternative to an atomic group or possessive quantifier,
2656 but there are some uses of (*PRUNE) that cannot be expressed in any other way.
2657 The behaviour of (*PRUNE:NAME) is the same as (*MARK:NAME)(*PRUNE) when the
2658 match fails completely; the name is passed back if this is the final attempt.
2659 (*PRUNE:NAME) does not pass back a name if the match succeeds. In an anchored
2660 pattern (*PRUNE) has the same effect as (*COMMIT).
2661 .sp
2662 (*SKIP)
2663 .sp
2664 This verb, when given without a name, is like (*PRUNE), except that if the
2665 pattern is unanchored, the "bumpalong" advance is not to the next character,
2666 but to the position in the subject where (*SKIP) was encountered. (*SKIP)
2667 signifies that whatever text was matched leading up to it cannot be part of a
2668 successful match. Consider:
2669 .sp
2670 a+(*SKIP)b
2671 .sp
2672 If the subject is "aaaac...", after the first match attempt fails (starting at
2673 the first character in the string), the starting point skips on to start the
2674 next attempt at "c". Note that a possessive quantifer does not have the same
2675 effect as this example; although it would suppress backtracking during the
2676 first match attempt, the second attempt would start at the second character
2677 instead of skipping on to "c".
2678 .sp
2679 (*SKIP:NAME)
2680 .sp
2681 When (*SKIP) has an associated name, its behaviour is modified. If the
2682 following pattern fails to match, the previous path through the pattern is
2683 searched for the most recent (*MARK) that has the same name. If one is found,
2684 the "bumpalong" advance is to the subject position that corresponds to that
2685 (*MARK) instead of to where (*SKIP) was encountered. If no (*MARK) with a
2686 matching name is found, normal "bumpalong" of one character happens (the
2687 (*SKIP) is ignored).
2688 .sp
2689 (*THEN) or (*THEN:NAME)
2690 .sp
2691 This verb causes a skip to the next alternation in the innermost enclosing
2692 group if the rest of the pattern does not match. That is, it cancels pending
2693 backtracking, but only within the current alternation. Its name comes from the
2694 observation that it can be used for a pattern-based if-then-else block:
2695 .sp
2696 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2697 .sp
2698 If the COND1 pattern matches, FOO is tried (and possibly further items after
2699 the end of the group if FOO succeeds); on failure the matcher skips to the
2700 second alternative and tries COND2, without backtracking into COND1. The
2701 behaviour of (*THEN:NAME) is exactly the same as (*MARK:NAME)(*THEN) if the
2702 overall match fails. If (*THEN) is not directly inside an alternation, it acts
2703 like (*PRUNE).
2704 .
2705 .P
2706 The above verbs provide four different "strengths" of control when subsequent
2707 matching fails. (*THEN) is the weakest, carrying on the match at the next
2708 alternation. (*PRUNE) comes next, failing the match at the current starting
2709 position, but allowing an advance to the next character (for an unanchored
2710 pattern). (*SKIP) is similar, except that the advance may be more than one
2711 character. (*COMMIT) is the strongest, causing the entire match to fail.
2712 .P
2713 If more than one is present in a pattern, the "stongest" one wins. For example,
2714 consider this pattern, where A, B, etc. are complex pattern fragments:
2715 .sp
2716 (A(*COMMIT)B(*THEN)C|D)
2717 .sp
2718 Once A has matched, PCRE is committed to this match, at the current starting
2719 position. If subsequently B matches, but C does not, the normal (*THEN) action
2720 of trying the next alternation (that is, D) does not happen because (*COMMIT)
2721 overrides.
2722 .
2723 .
2724 .SH "SEE ALSO"
2725 .rs
2726 .sp
2727 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3),
2728 \fBpcresyntax\fP(3), \fBpcre\fP(3).
2729 .
2730 .
2732 .rs
2733 .sp
2734 .nf
2735 Philip Hazel
2736 University Computing Service
2737 Cambridge CB2 3QH, England.
2738 .fi
2739 .
2740 .
2742 .rs
2743 .sp
2744 .nf
2745 Last updated: 21 November 2010
2746 Copyright (c) 1997-2010 University of Cambridge.
2747 .fi


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

  ViewVC Help
Powered by ViewVC 1.1.5