ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 454 - (show annotations)
Tue Sep 22 09:42:11 2009 UTC (11 years, 10 months ago) by ph10
File size: 95798 byte(s)
Allow fixed-length subroutine calls in lookbehinds.
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, you must
25 build PCRE to include UTF-8 support, and then call \fBpcre_compile()\fP with
26 the PCRE_UTF8 option. There is also a special sequence that can be given at the
27 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 The remainder of this document discusses the patterns that are supported by
46 PCRE when its main matching function, \fBpcre_exec()\fP, is used.
47 From release 6.0, PCRE offers a second matching function,
48 \fBpcre_dfa_exec()\fP, which matches using a different algorithm that is not
49 Perl-compatible. Some of the features discussed below are not available when
50 \fBpcre_dfa_exec()\fP is used. The advantages and disadvantages of the
51 alternative function, and how it differs from the normal function, are
52 discussed in the
53 .\" HREF
54 \fBpcrematching\fP
55 .\"
56 page.
57 .
58 .
60 .rs
61 .sp
62 PCRE supports five different conventions for indicating line breaks in
63 strings: a single CR (carriage return) character, a single LF (linefeed)
64 character, the two-character sequence CRLF, any of the three preceding, or any
65 Unicode newline sequence. The
66 .\" HREF
67 \fBpcreapi\fP
68 .\"
69 page has
70 .\" HTML <a href="pcreapi.html#newlines">
71 .\" </a>
72 further discussion
73 .\"
74 about newlines, and shows how to set the newline convention in the
75 \fIoptions\fP arguments for the compiling and matching functions.
76 .P
77 It is also possible to specify a newline convention by starting a pattern
78 string with one of the following five sequences:
79 .sp
80 (*CR) carriage return
81 (*LF) linefeed
82 (*CRLF) carriage return, followed by linefeed
83 (*ANYCRLF) any of the three above
84 (*ANY) all Unicode newline sequences
85 .sp
86 These override the default and the options given to \fBpcre_compile()\fP. For
87 example, on a Unix system where LF is the default newline sequence, the pattern
88 .sp
89 (*CR)a.b
90 .sp
91 changes the convention to CR. That pattern matches "a\enb" because LF is no
92 longer a newline. Note that these special settings, which are not
93 Perl-compatible, are recognized only at the very start of a pattern, and that
94 they must be in upper case. If more than one of them is present, the last one
95 is used.
96 .P
97 The newline convention does not affect what the \eR escape sequence matches. By
98 default, this is any Unicode newline sequence, for Perl compatibility. However,
99 this can be changed; see the description of \eR in the section entitled
100 .\" HTML <a href="#newlineseq">
101 .\" </a>
102 "Newline sequences"
103 .\"
104 below. A change of \eR setting can be combined with a change of newline
105 convention.
106 .
107 .
109 .rs
110 .sp
111 A regular expression is a pattern that is matched against a subject string from
112 left to right. Most characters stand for themselves in a pattern, and match the
113 corresponding characters in the subject. As a trivial example, the pattern
114 .sp
115 The quick brown fox
116 .sp
117 matches a portion of a subject string that is identical to itself. When
118 caseless matching is specified (the PCRE_CASELESS option), letters are matched
119 independently of case. In UTF-8 mode, PCRE always understands the concept of
120 case for characters whose values are less than 128, so caseless matching is
121 always possible. For characters with higher values, the concept of case is
122 supported if PCRE is compiled with Unicode property support, but not otherwise.
123 If you want to use caseless matching for characters 128 and above, you must
124 ensure that PCRE is compiled with Unicode property support as well as with
125 UTF-8 support.
126 .P
127 The power of regular expressions comes from the ability to include alternatives
128 and repetitions in the pattern. These are encoded in the pattern by the use of
129 \fImetacharacters\fP, which do not stand for themselves but instead are
130 interpreted in some special way.
131 .P
132 There are two different sets of metacharacters: those that are recognized
133 anywhere in the pattern except within square brackets, and those that are
134 recognized within square brackets. Outside square brackets, the metacharacters
135 are as follows:
136 .sp
137 \e general escape character with several uses
138 ^ assert start of string (or line, in multiline mode)
139 $ assert end of string (or line, in multiline mode)
140 . match any character except newline (by default)
141 [ start character class definition
142 | start of alternative branch
143 ( start subpattern
144 ) end subpattern
145 ? extends the meaning of (
146 also 0 or 1 quantifier
147 also quantifier minimizer
148 * 0 or more quantifier
149 + 1 or more quantifier
150 also "possessive quantifier"
151 { start min/max quantifier
152 .sp
153 Part of a pattern that is in square brackets is called a "character class". In
154 a character class the only metacharacters are:
155 .sp
156 \e general escape character
157 ^ negate the class, but only if the first character
158 - indicates character range
159 .\" JOIN
160 [ POSIX character class (only if followed by POSIX
161 syntax)
162 ] terminates the character class
163 .sp
164 The following sections describe the use of each of the metacharacters.
165 .
166 .
168 .rs
169 .sp
170 The backslash character has several uses. Firstly, if it is followed by a
171 non-alphanumeric character, it takes away any special meaning that character
172 may have. This use of backslash as an escape character applies both inside and
173 outside character classes.
174 .P
175 For example, if you want to match a * character, you write \e* in the pattern.
176 This escaping action applies whether or not the following character would
177 otherwise be interpreted as a metacharacter, so it is always safe to precede a
178 non-alphanumeric with backslash to specify that it stands for itself. In
179 particular, if you want to match a backslash, you write \e\e.
180 .P
181 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
182 pattern (other than in a character class) and characters between a # outside
183 a character class and the next newline are ignored. An escaping backslash can
184 be used to include a whitespace or # character as part of the pattern.
185 .P
186 If you want to remove the special meaning from a sequence of characters, you
187 can do so by putting them between \eQ and \eE. This is different from Perl in
188 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
189 Perl, $ and @ cause variable interpolation. Note the following examples:
190 .sp
191 Pattern PCRE matches Perl matches
192 .sp
193 .\" JOIN
194 \eQabc$xyz\eE abc$xyz abc followed by the
195 contents of $xyz
196 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
197 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
198 .sp
199 The \eQ...\eE sequence is recognized both inside and outside character classes.
200 .
201 .
202 .\" HTML <a name="digitsafterbackslash"></a>
203 .SS "Non-printing characters"
204 .rs
205 .sp
206 A second use of backslash provides a way of encoding non-printing characters
207 in patterns in a visible manner. There is no restriction on the appearance of
208 non-printing characters, apart from the binary zero that terminates a pattern,
209 but when a pattern is being prepared by text editing, it is usually easier to
210 use one of the following escape sequences than the binary character it
211 represents:
212 .sp
213 \ea alarm, that is, the BEL character (hex 07)
214 \ecx "control-x", where x is any character
215 \ee escape (hex 1B)
216 \ef formfeed (hex 0C)
217 \en linefeed (hex 0A)
218 \er carriage return (hex 0D)
219 \et tab (hex 09)
220 \eddd character with octal code ddd, or backreference
221 \exhh character with hex code hh
222 \ex{hhh..} character with hex code hhh..
223 .sp
224 The precise effect of \ecx is as follows: if x is a lower case letter, it
225 is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
226 Thus \ecz becomes hex 1A, but \ec{ becomes hex 3B, while \ec; becomes hex
227 7B.
228 .P
229 After \ex, from zero to two hexadecimal digits are read (letters can be in
230 upper or lower case). Any number of hexadecimal digits may appear between \ex{
231 and }, but the value of the character code must be less than 256 in non-UTF-8
232 mode, and less than 2**31 in UTF-8 mode. That is, the maximum value in
233 hexadecimal is 7FFFFFFF. Note that this is bigger than the largest Unicode code
234 point, which is 10FFFF.
235 .P
236 If characters other than hexadecimal digits appear between \ex{ and }, or if
237 there is no terminating }, this form of escape is not recognized. Instead, the
238 initial \ex will be interpreted as a basic hexadecimal escape, with no
239 following digits, giving a character whose value is zero.
240 .P
241 Characters whose value is less than 256 can be defined by either of the two
242 syntaxes for \ex. There is no difference in the way they are handled. For
243 example, \exdc is exactly the same as \ex{dc}.
244 .P
245 After \e0 up to two further octal digits are read. If there are fewer than two
246 digits, just those that are present are used. Thus the sequence \e0\ex\e07
247 specifies two binary zeros followed by a BEL character (code value 7). Make
248 sure you supply two digits after the initial zero if the pattern character that
249 follows is itself an octal digit.
250 .P
251 The handling of a backslash followed by a digit other than 0 is complicated.
252 Outside a character class, PCRE reads it and any following digits as a decimal
253 number. If the number is less than 10, or if there have been at least that many
254 previous capturing left parentheses in the expression, the entire sequence is
255 taken as a \fIback reference\fP. A description of how this works is given
256 .\" HTML <a href="#backreferences">
257 .\" </a>
258 later,
259 .\"
260 following the discussion of
261 .\" HTML <a href="#subpattern">
262 .\" </a>
263 parenthesized subpatterns.
264 .\"
265 .P
266 Inside a character class, or if the decimal number is greater than 9 and there
267 have not been that many capturing subpatterns, PCRE re-reads up to three octal
268 digits following the backslash, and uses them to generate a data character. Any
269 subsequent digits stand for themselves. In non-UTF-8 mode, the value of a
270 character specified in octal must be less than \e400. In UTF-8 mode, values up
271 to \e777 are permitted. For example:
272 .sp
273 \e040 is another way of writing a space
274 .\" JOIN
275 \e40 is the same, provided there are fewer than 40
276 previous capturing subpatterns
277 \e7 is always a back reference
278 .\" JOIN
279 \e11 might be a back reference, or another way of
280 writing a tab
281 \e011 is always a tab
282 \e0113 is a tab followed by the character "3"
283 .\" JOIN
284 \e113 might be a back reference, otherwise the
285 character with octal code 113
286 .\" JOIN
287 \e377 might be a back reference, otherwise
288 the byte consisting entirely of 1 bits
289 .\" JOIN
290 \e81 is either a back reference, or a binary zero
291 followed by the two characters "8" and "1"
292 .sp
293 Note that octal values of 100 or greater must not be introduced by a leading
294 zero, because no more than three octal digits are ever read.
295 .P
296 All the sequences that define a single character value can be used both inside
297 and outside character classes. In addition, inside a character class, the
298 sequence \eb is interpreted as the backspace character (hex 08), and the
299 sequences \eR and \eX are interpreted as the characters "R" and "X",
300 respectively. Outside a character class, these sequences have different
301 meanings
302 .\" HTML <a href="#uniextseq">
303 .\" </a>
304 (see below).
305 .\"
306 .
307 .
308 .SS "Absolute and relative back references"
309 .rs
310 .sp
311 The sequence \eg followed by an unsigned or a negative number, optionally
312 enclosed in braces, is an absolute or relative back reference. A named back
313 reference can be coded as \eg{name}. Back references are discussed
314 .\" HTML <a href="#backreferences">
315 .\" </a>
316 later,
317 .\"
318 following the discussion of
319 .\" HTML <a href="#subpattern">
320 .\" </a>
321 parenthesized subpatterns.
322 .\"
323 .
324 .
325 .SS "Absolute and relative subroutine calls"
326 .rs
327 .sp
328 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
329 a number enclosed either in angle brackets or single quotes, is an alternative
330 syntax for referencing a subpattern as a "subroutine". Details are discussed
331 .\" HTML <a href="#onigurumasubroutines">
332 .\" </a>
333 later.
334 .\"
335 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
336 synonymous. The former is a back reference; the latter is a
337 .\" HTML <a href="#subpatternsassubroutines">
338 .\" </a>
339 subroutine
340 .\"
341 call.
342 .
343 .
344 .SS "Generic character types"
345 .rs
346 .sp
347 Another use of backslash is for specifying generic character types. The
348 following are always recognized:
349 .sp
350 \ed any decimal digit
351 \eD any character that is not a decimal digit
352 \eh any horizontal whitespace character
353 \eH any character that is not a horizontal whitespace character
354 \es any whitespace character
355 \eS any character that is not a whitespace character
356 \ev any vertical whitespace character
357 \eV any character that is not a vertical whitespace character
358 \ew any "word" character
359 \eW any "non-word" character
360 .sp
361 Each pair of escape sequences partitions the complete set of characters into
362 two disjoint sets. Any given character matches one, and only one, of each pair.
363 .P
364 These character type sequences can appear both inside and outside character
365 classes. They each match one character of the appropriate type. If the current
366 matching point is at the end of the subject string, all of them fail, since
367 there is no character to match.
368 .P
369 For compatibility with Perl, \es does not match the VT character (code 11).
370 This makes it different from the the POSIX "space" class. The \es characters
371 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
372 included in a Perl script, \es may match the VT character. In PCRE, it never
373 does.
374 .P
375 In UTF-8 mode, characters with values greater than 128 never match \ed, \es, or
376 \ew, and always match \eD, \eS, and \eW. This is true even when Unicode
377 character property support is available. These sequences retain their original
378 meanings from before UTF-8 support was available, mainly for efficiency
379 reasons. Note that this also affects \eb, because it is defined in terms of \ew
380 and \eW.
381 .P
382 The sequences \eh, \eH, \ev, and \eV are Perl 5.10 features. In contrast to the
383 other sequences, these do match certain high-valued codepoints in UTF-8 mode.
384 The horizontal space characters are:
385 .sp
386 U+0009 Horizontal tab
387 U+0020 Space
388 U+00A0 Non-break space
389 U+1680 Ogham space mark
390 U+180E Mongolian vowel separator
391 U+2000 En quad
392 U+2001 Em quad
393 U+2002 En space
394 U+2003 Em space
395 U+2004 Three-per-em space
396 U+2005 Four-per-em space
397 U+2006 Six-per-em space
398 U+2007 Figure space
399 U+2008 Punctuation space
400 U+2009 Thin space
401 U+200A Hair space
402 U+202F Narrow no-break space
403 U+205F Medium mathematical space
404 U+3000 Ideographic space
405 .sp
406 The vertical space characters are:
407 .sp
408 U+000A Linefeed
409 U+000B Vertical tab
410 U+000C Formfeed
411 U+000D Carriage return
412 U+0085 Next line
413 U+2028 Line separator
414 U+2029 Paragraph separator
415 .P
416 A "word" character is an underscore or any character less than 256 that is a
417 letter or digit. The definition of letters and digits is controlled by PCRE's
418 low-valued character tables, and may vary if locale-specific matching is taking
419 place (see
420 .\" HTML <a href="pcreapi.html#localesupport">
421 .\" </a>
422 "Locale support"
423 .\"
424 in the
425 .\" HREF
426 \fBpcreapi\fP
427 .\"
428 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
429 or "french" in Windows, some character codes greater than 128 are used for
430 accented letters, and these are matched by \ew. The use of locales with Unicode
431 is discouraged.
432 .
433 .
434 .\" HTML <a name="newlineseq"></a>
435 .SS "Newline sequences"
436 .rs
437 .sp
438 Outside a character class, by default, the escape sequence \eR matches any
439 Unicode newline sequence. This is a Perl 5.10 feature. In non-UTF-8 mode \eR is
440 equivalent to the following:
441 .sp
442 (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
443 .sp
444 This is an example of an "atomic group", details of which are given
445 .\" HTML <a href="#atomicgroup">
446 .\" </a>
447 below.
448 .\"
449 This particular group matches either the two-character sequence CR followed by
450 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
451 U+000B), FF (formfeed, U+000C), CR (carriage return, U+000D), or NEL (next
452 line, U+0085). The two-character sequence is treated as a single unit that
453 cannot be split.
454 .P
455 In UTF-8 mode, two additional characters whose codepoints are greater than 255
456 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
457 Unicode character property support is not needed for these characters to be
458 recognized.
459 .P
460 It is possible to restrict \eR to match only CR, LF, or CRLF (instead of the
461 complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
462 either at compile time or when the pattern is matched. (BSR is an abbrevation
463 for "backslash R".) This can be made the default when PCRE is built; if this is
464 the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
465 It is also possible to specify these settings by starting a pattern string with
466 one of the following sequences:
467 .sp
468 (*BSR_ANYCRLF) CR, LF, or CRLF only
469 (*BSR_UNICODE) any Unicode newline sequence
470 .sp
471 These override the default and the options given to \fBpcre_compile()\fP, but
472 they can be overridden by options given to \fBpcre_exec()\fP. Note that these
473 special settings, which are not Perl-compatible, are recognized only at the
474 very start of a pattern, and that they must be in upper case. If more than one
475 of them is present, the last one is used. They can be combined with a change of
476 newline convention, for example, a pattern can start with:
477 .sp
479 .sp
480 Inside a character class, \eR matches the letter "R".
481 .
482 .
483 .\" HTML <a name="uniextseq"></a>
484 .SS Unicode character properties
485 .rs
486 .sp
487 When PCRE is built with Unicode character property support, three additional
488 escape sequences that match characters with specific properties are available.
489 When not in UTF-8 mode, these sequences are of course limited to testing
490 characters whose codepoints are less than 256, but they do work in this mode.
491 The extra escape sequences are:
492 .sp
493 \ep{\fIxx\fP} a character with the \fIxx\fP property
494 \eP{\fIxx\fP} a character without the \fIxx\fP property
495 \eX an extended Unicode sequence
496 .sp
497 The property names represented by \fIxx\fP above are limited to the Unicode
498 script names, the general category properties, and "Any", which matches any
499 character (including newline). Other properties such as "InMusicalSymbols" are
500 not currently supported by PCRE. Note that \eP{Any} does not match any
501 characters, so always causes a match failure.
502 .P
503 Sets of Unicode characters are defined as belonging to certain scripts. A
504 character from one of these sets can be matched using a script name. For
505 example:
506 .sp
507 \ep{Greek}
508 \eP{Han}
509 .sp
510 Those that are not part of an identified script are lumped together as
511 "Common". The current list of scripts is:
512 .P
513 Arabic,
514 Armenian,
515 Balinese,
516 Bengali,
517 Bopomofo,
518 Braille,
519 Buginese,
520 Buhid,
521 Canadian_Aboriginal,
522 Cherokee,
523 Common,
524 Coptic,
525 Cuneiform,
526 Cypriot,
527 Cyrillic,
528 Deseret,
529 Devanagari,
530 Ethiopic,
531 Georgian,
532 Glagolitic,
533 Gothic,
534 Greek,
535 Gujarati,
536 Gurmukhi,
537 Han,
538 Hangul,
539 Hanunoo,
540 Hebrew,
541 Hiragana,
542 Inherited,
543 Kannada,
544 Katakana,
545 Kharoshthi,
546 Khmer,
547 Lao,
548 Latin,
549 Limbu,
550 Linear_B,
551 Malayalam,
552 Mongolian,
553 Myanmar,
554 New_Tai_Lue,
555 Nko,
556 Ogham,
557 Old_Italic,
558 Old_Persian,
559 Oriya,
560 Osmanya,
561 Phags_Pa,
562 Phoenician,
563 Runic,
564 Shavian,
565 Sinhala,
566 Syloti_Nagri,
567 Syriac,
568 Tagalog,
569 Tagbanwa,
570 Tai_Le,
571 Tamil,
572 Telugu,
573 Thaana,
574 Thai,
575 Tibetan,
576 Tifinagh,
577 Ugaritic,
578 Yi.
579 .P
580 Each character has exactly one general category property, specified by a
581 two-letter abbreviation. For compatibility with Perl, negation can be specified
582 by including a circumflex between the opening brace and the property name. For
583 example, \ep{^Lu} is the same as \eP{Lu}.
584 .P
585 If only one letter is specified with \ep or \eP, it includes all the general
586 category properties that start with that letter. In this case, in the absence
587 of negation, the curly brackets in the escape sequence are optional; these two
588 examples have the same effect:
589 .sp
590 \ep{L}
591 \epL
592 .sp
593 The following general category property codes are supported:
594 .sp
595 C Other
596 Cc Control
597 Cf Format
598 Cn Unassigned
599 Co Private use
600 Cs Surrogate
601 .sp
602 L Letter
603 Ll Lower case letter
604 Lm Modifier letter
605 Lo Other letter
606 Lt Title case letter
607 Lu Upper case letter
608 .sp
609 M Mark
610 Mc Spacing mark
611 Me Enclosing mark
612 Mn Non-spacing mark
613 .sp
614 N Number
615 Nd Decimal number
616 Nl Letter number
617 No Other number
618 .sp
619 P Punctuation
620 Pc Connector punctuation
621 Pd Dash punctuation
622 Pe Close punctuation
623 Pf Final punctuation
624 Pi Initial punctuation
625 Po Other punctuation
626 Ps Open punctuation
627 .sp
628 S Symbol
629 Sc Currency symbol
630 Sk Modifier symbol
631 Sm Mathematical symbol
632 So Other symbol
633 .sp
634 Z Separator
635 Zl Line separator
636 Zp Paragraph separator
637 Zs Space separator
638 .sp
639 The special property L& is also supported: it matches a character that has
640 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
641 a modifier or "other".
642 .P
643 The Cs (Surrogate) property applies only to characters in the range U+D800 to
644 U+DFFF. Such characters are not valid in UTF-8 strings (see RFC 3629) and so
645 cannot be tested by PCRE, unless UTF-8 validity checking has been turned off
646 (see the discussion of PCRE_NO_UTF8_CHECK in the
647 .\" HREF
648 \fBpcreapi\fP
649 .\"
650 page). Perl does not support the Cs property.
651 .P
652 The long synonyms for property names that Perl supports (such as \ep{Letter})
653 are not supported by PCRE, nor is it permitted to prefix any of these
654 properties with "Is".
655 .P
656 No character that is in the Unicode table has the Cn (unassigned) property.
657 Instead, this property is assumed for any code point that is not in the
658 Unicode table.
659 .P
660 Specifying caseless matching does not affect these escape sequences. For
661 example, \ep{Lu} always matches only upper case letters.
662 .P
663 The \eX escape matches any number of Unicode characters that form an extended
664 Unicode sequence. \eX is equivalent to
665 .sp
666 (?>\ePM\epM*)
667 .sp
668 That is, it matches a character without the "mark" property, followed by zero
669 or more characters with the "mark" property, and treats the sequence as an
670 atomic group
671 .\" HTML <a href="#atomicgroup">
672 .\" </a>
673 (see below).
674 .\"
675 Characters with the "mark" property are typically accents that affect the
676 preceding character. None of them have codepoints less than 256, so in
677 non-UTF-8 mode \eX matches any one character.
678 .P
679 Matching characters by Unicode property is not fast, because PCRE has to search
680 a structure that contains data for over fifteen thousand characters. That is
681 why the traditional escape sequences such as \ed and \ew do not use Unicode
682 properties in PCRE.
683 .
684 .
685 .\" HTML <a name="resetmatchstart"></a>
686 .SS "Resetting the match start"
687 .rs
688 .sp
689 The escape sequence \eK, which is a Perl 5.10 feature, causes any previously
690 matched characters not to be included in the final matched sequence. For
691 example, the pattern:
692 .sp
693 foo\eKbar
694 .sp
695 matches "foobar", but reports that it has matched "bar". This feature is
696 similar to a lookbehind assertion
697 .\" HTML <a href="#lookbehind">
698 .\" </a>
699 (described below).
700 .\"
701 However, in this case, the part of the subject before the real match does not
702 have to be of fixed length, as lookbehind assertions do. The use of \eK does
703 not interfere with the setting of
704 .\" HTML <a href="#subpattern">
705 .\" </a>
706 captured substrings.
707 .\"
708 For example, when the pattern
709 .sp
710 (foo)\eKbar
711 .sp
712 matches "foobar", the first substring is still set to "foo".
713 .
714 .
715 .\" HTML <a name="smallassertions"></a>
716 .SS "Simple assertions"
717 .rs
718 .sp
719 The final use of backslash is for certain simple assertions. An assertion
720 specifies a condition that has to be met at a particular point in a match,
721 without consuming any characters from the subject string. The use of
722 subpatterns for more complicated assertions is described
723 .\" HTML <a href="#bigassertions">
724 .\" </a>
725 below.
726 .\"
727 The backslashed assertions are:
728 .sp
729 \eb matches at a word boundary
730 \eB matches when not at a word boundary
731 \eA matches at the start of the subject
732 \eZ matches at the end of the subject
733 also matches before a newline at the end of the subject
734 \ez matches only at the end of the subject
735 \eG matches at the first matching position in the subject
736 .sp
737 These assertions may not appear in character classes (but note that \eb has a
738 different meaning, namely the backspace character, inside a character class).
739 .P
740 A word boundary is a position in the subject string where the current character
741 and the previous character do not both match \ew or \eW (i.e. one matches
742 \ew and the other matches \eW), or the start or end of the string if the
743 first or last character matches \ew, respectively.
744 .P
745 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
746 dollar (described in the next section) in that they only ever match at the very
747 start and end of the subject string, whatever options are set. Thus, they are
748 independent of multiline mode. These three assertions are not affected by the
749 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
750 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
751 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
752 at a point other than the beginning of the subject, \eA can never match. The
753 difference between \eZ and \ez is that \eZ matches before a newline at the end
754 of the string as well as at the very end, whereas \ez matches only at the end.
755 .P
756 The \eG assertion is true only when the current matching position is at the
757 start point of the match, as specified by the \fIstartoffset\fP argument of
758 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
759 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
760 arguments, you can mimic Perl's /g option, and it is in this kind of
761 implementation where \eG can be useful.
762 .P
763 Note, however, that PCRE's interpretation of \eG, as the start of the current
764 match, is subtly different from Perl's, which defines it as the end of the
765 previous match. In Perl, these can be different when the previously matched
766 string was empty. Because PCRE does just one match at a time, it cannot
767 reproduce this behaviour.
768 .P
769 If all the alternatives of a pattern begin with \eG, the expression is anchored
770 to the starting match position, and the "anchored" flag is set in the compiled
771 regular expression.
772 .
773 .
775 .rs
776 .sp
777 Outside a character class, in the default matching mode, the circumflex
778 character is an assertion that is true only if the current matching point is
779 at the start of the subject string. If the \fIstartoffset\fP argument of
780 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
781 option is unset. Inside a character class, circumflex has an entirely different
782 meaning
783 .\" HTML <a href="#characterclass">
784 .\" </a>
785 (see below).
786 .\"
787 .P
788 Circumflex need not be the first character of the pattern if a number of
789 alternatives are involved, but it should be the first thing in each alternative
790 in which it appears if the pattern is ever to match that branch. If all
791 possible alternatives start with a circumflex, that is, if the pattern is
792 constrained to match only at the start of the subject, it is said to be an
793 "anchored" pattern. (There are also other constructs that can cause a pattern
794 to be anchored.)
795 .P
796 A dollar character is an assertion that is true only if the current matching
797 point is at the end of the subject string, or immediately before a newline
798 at the end of the string (by default). Dollar need not be the last character of
799 the pattern if a number of alternatives are involved, but it should be the last
800 item in any branch in which it appears. Dollar has no special meaning in a
801 character class.
802 .P
803 The meaning of dollar can be changed so that it matches only at the very end of
804 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
805 does not affect the \eZ assertion.
806 .P
807 The meanings of the circumflex and dollar characters are changed if the
808 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
809 immediately after internal newlines as well as at the start of the subject
810 string. It does not match after a newline that ends the string. A dollar
811 matches before any newlines in the string, as well as at the very end, when
812 PCRE_MULTILINE is set. When newline is specified as the two-character
813 sequence CRLF, isolated CR and LF characters do not indicate newlines.
814 .P
815 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
816 \en represents a newline) in multiline mode, but not otherwise. Consequently,
817 patterns that are anchored in single line mode because all branches start with
818 ^ are not anchored in multiline mode, and a match for circumflex is possible
819 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
820 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
821 .P
822 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
823 end of the subject in both modes, and if all branches of a pattern start with
824 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
825 .
826 .
828 .rs
829 .sp
830 Outside a character class, a dot in the pattern matches any one character in
831 the subject string except (by default) a character that signifies the end of a
832 line. In UTF-8 mode, the matched character may be more than one byte long.
833 .P
834 When a line ending is defined as a single character, dot never matches that
835 character; when the two-character sequence CRLF is used, dot does not match CR
836 if it is immediately followed by LF, but otherwise it matches all characters
837 (including isolated CRs and LFs). When any Unicode line endings are being
838 recognized, dot does not match CR or LF or any of the other line ending
839 characters.
840 .P
841 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
842 option is set, a dot matches any one character, without exception. If the
843 two-character sequence CRLF is present in the subject string, it takes two dots
844 to match it.
845 .P
846 The handling of dot is entirely independent of the handling of circumflex and
847 dollar, the only relationship being that they both involve newlines. Dot has no
848 special meaning in a character class.
849 .
850 .
852 .rs
853 .sp
854 Outside a character class, the escape sequence \eC matches any one byte, both
855 in and out of UTF-8 mode. Unlike a dot, it always matches any line-ending
856 characters. The feature is provided in Perl in order to match individual bytes
857 in UTF-8 mode. Because it breaks up UTF-8 characters into individual bytes,
858 what remains in the string may be a malformed UTF-8 string. For this reason,
859 the \eC escape sequence is best avoided.
860 .P
861 PCRE does not allow \eC to appear in lookbehind assertions
862 .\" HTML <a href="#lookbehind">
863 .\" </a>
864 (described below),
865 .\"
866 because in UTF-8 mode this would make it impossible to calculate the length of
867 the lookbehind.
868 .
869 .
870 .\" HTML <a name="characterclass"></a>
872 .rs
873 .sp
874 An opening square bracket introduces a character class, terminated by a closing
875 square bracket. A closing square bracket on its own is not special. If a
876 closing square bracket is required as a member of the class, it should be the
877 first data character in the class (after an initial circumflex, if present) or
878 escaped with a backslash.
879 .P
880 A character class matches a single character in the subject. In UTF-8 mode, the
881 character may occupy more than one byte. A matched character must be in the set
882 of characters defined by the class, unless the first character in the class
883 definition is a circumflex, in which case the subject character must not be in
884 the set defined by the class. If a circumflex is actually required as a member
885 of the class, ensure it is not the first character, or escape it with a
886 backslash.
887 .P
888 For example, the character class [aeiou] matches any lower case vowel, while
889 [^aeiou] matches any character that is not a lower case vowel. Note that a
890 circumflex is just a convenient notation for specifying the characters that
891 are in the class by enumerating those that are not. A class that starts with a
892 circumflex is not an assertion: it still consumes a character from the subject
893 string, and therefore it fails if the current pointer is at the end of the
894 string.
895 .P
896 In UTF-8 mode, characters with values greater than 255 can be included in a
897 class as a literal string of bytes, or by using the \ex{ escaping mechanism.
898 .P
899 When caseless matching is set, any letters in a class represent both their
900 upper case and lower case versions, so for example, a caseless [aeiou] matches
901 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
902 caseful version would. In UTF-8 mode, PCRE always understands the concept of
903 case for characters whose values are less than 128, so caseless matching is
904 always possible. For characters with higher values, the concept of case is
905 supported if PCRE is compiled with Unicode property support, but not otherwise.
906 If you want to use caseless matching for characters 128 and above, you must
907 ensure that PCRE is compiled with Unicode property support as well as with
908 UTF-8 support.
909 .P
910 Characters that might indicate line breaks are never treated in any special way
911 when matching character classes, whatever line-ending sequence is in use, and
912 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
913 such as [^a] always matches one of these characters.
914 .P
915 The minus (hyphen) character can be used to specify a range of characters in a
916 character class. For example, [d-m] matches any letter between d and m,
917 inclusive. If a minus character is required in a class, it must be escaped with
918 a backslash or appear in a position where it cannot be interpreted as
919 indicating a range, typically as the first or last character in the class.
920 .P
921 It is not possible to have the literal character "]" as the end character of a
922 range. A pattern such as [W-]46] is interpreted as a class of two characters
923 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
924 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
925 the end of range, so [W-\e]46] is interpreted as a class containing a range
926 followed by two other characters. The octal or hexadecimal representation of
927 "]" can also be used to end a range.
928 .P
929 Ranges operate in the collating sequence of character values. They can also be
930 used for characters specified numerically, for example [\e000-\e037]. In UTF-8
931 mode, ranges can include characters whose values are greater than 255, for
932 example [\ex{100}-\ex{2ff}].
933 .P
934 If a range that includes letters is used when caseless matching is set, it
935 matches the letters in either case. For example, [W-c] is equivalent to
936 [][\e\e^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
937 tables for a French locale are in use, [\exc8-\excb] matches accented E
938 characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
939 characters with values greater than 128 only when it is compiled with Unicode
940 property support.
941 .P
942 The character types \ed, \eD, \ep, \eP, \es, \eS, \ew, and \eW may also appear
943 in a character class, and add the characters that they match to the class. For
944 example, [\edABCDEF] matches any hexadecimal digit. A circumflex can
945 conveniently be used with the upper case character types to specify a more
946 restricted set of characters than the matching lower case type. For example,
947 the class [^\eW_] matches any letter or digit, but not underscore.
948 .P
949 The only metacharacters that are recognized in character classes are backslash,
950 hyphen (only where it can be interpreted as specifying a range), circumflex
951 (only at the start), opening square bracket (only when it can be interpreted as
952 introducing a POSIX class name - see the next section), and the terminating
953 closing square bracket. However, escaping other non-alphanumeric characters
954 does no harm.
955 .
956 .
958 .rs
959 .sp
960 Perl supports the POSIX notation for character classes. This uses names
961 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
962 this notation. For example,
963 .sp
964 [01[:alpha:]%]
965 .sp
966 matches "0", "1", any alphabetic character, or "%". The supported class names
967 are
968 .sp
969 alnum letters and digits
970 alpha letters
971 ascii character codes 0 - 127
972 blank space or tab only
973 cntrl control characters
974 digit decimal digits (same as \ed)
975 graph printing characters, excluding space
976 lower lower case letters
977 print printing characters, including space
978 punct printing characters, excluding letters and digits
979 space white space (not quite the same as \es)
980 upper upper case letters
981 word "word" characters (same as \ew)
982 xdigit hexadecimal digits
983 .sp
984 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
985 space (32). Notice that this list includes the VT character (code 11). This
986 makes "space" different to \es, which does not include VT (for Perl
987 compatibility).
988 .P
989 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
990 5.8. Another Perl extension is negation, which is indicated by a ^ character
991 after the colon. For example,
992 .sp
993 [12[:^digit:]]
994 .sp
995 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
996 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
997 supported, and an error is given if they are encountered.
998 .P
999 In UTF-8 mode, characters with values greater than 128 do not match any of
1000 the POSIX character classes.
1001 .
1002 .
1004 .rs
1005 .sp
1006 Vertical bar characters are used to separate alternative patterns. For example,
1007 the pattern
1008 .sp
1009 gilbert|sullivan
1010 .sp
1011 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1012 and an empty alternative is permitted (matching the empty string). The matching
1013 process tries each alternative in turn, from left to right, and the first one
1014 that succeeds is used. If the alternatives are within a subpattern
1015 .\" HTML <a href="#subpattern">
1016 .\" </a>
1017 (defined below),
1018 .\"
1019 "succeeds" means matching the rest of the main pattern as well as the
1020 alternative in the subpattern.
1021 .
1022 .
1024 .rs
1025 .sp
1026 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1027 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1028 the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1029 The option letters are
1030 .sp
1031 i for PCRE_CASELESS
1033 s for PCRE_DOTALL
1034 x for PCRE_EXTENDED
1035 .sp
1036 For example, (?im) sets caseless, multiline matching. It is also possible to
1037 unset these options by preceding the letter with a hyphen, and a combined
1038 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1039 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1040 permitted. If a letter appears both before and after the hyphen, the option is
1041 unset.
1042 .P
1043 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1044 changed in the same way as the Perl-compatible options by using the characters
1045 J, U and X respectively.
1046 .P
1047 When one of these option changes occurs at top level (that is, not inside
1048 subpattern parentheses), the change applies to the remainder of the pattern
1049 that follows. If the change is placed right at the start of a pattern, PCRE
1050 extracts it into the global options (and it will therefore show up in data
1051 extracted by the \fBpcre_fullinfo()\fP function).
1052 .P
1053 An option change within a subpattern (see below for a description of
1054 subpatterns) affects only that part of the current pattern that follows it, so
1055 .sp
1056 (a(?i)b)c
1057 .sp
1058 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1059 By this means, options can be made to have different settings in different
1060 parts of the pattern. Any changes made in one alternative do carry on
1061 into subsequent branches within the same subpattern. For example,
1062 .sp
1063 (a(?i)b|c)
1064 .sp
1065 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1066 branch is abandoned before the option setting. This is because the effects of
1067 option settings happen at compile time. There would be some very weird
1068 behaviour otherwise.
1069 .P
1070 \fBNote:\fP There are other PCRE-specific options that can be set by the
1071 application when the compile or match functions are called. In some cases the
1072 pattern can contain special leading sequences such as (*CRLF) to override what
1073 the application has set or what has been defaulted. Details are given in the
1074 section entitled
1075 .\" HTML <a href="#newlineseq">
1076 .\" </a>
1077 "Newline sequences"
1078 .\"
1079 above. There is also the (*UTF8) leading sequence that can be used to set UTF-8
1080 mode; this is equivalent to setting the PCRE_UTF8 option.
1081 .
1082 .
1083 .\" HTML <a name="subpattern"></a>
1085 .rs
1086 .sp
1087 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1088 Turning part of a pattern into a subpattern does two things:
1089 .sp
1090 1. It localizes a set of alternatives. For example, the pattern
1091 .sp
1092 cat(aract|erpillar|)
1093 .sp
1094 matches one of the words "cat", "cataract", or "caterpillar". Without the
1095 parentheses, it would match "cataract", "erpillar" or an empty string.
1096 .sp
1097 2. It sets up the subpattern as a capturing subpattern. This means that, when
1098 the whole pattern matches, that portion of the subject string that matched the
1099 subpattern is passed back to the caller via the \fIovector\fP argument of
1100 \fBpcre_exec()\fP. Opening parentheses are counted from left to right (starting
1101 from 1) to obtain numbers for the capturing subpatterns.
1102 .P
1103 For example, if the string "the red king" is matched against the pattern
1104 .sp
1105 the ((red|white) (king|queen))
1106 .sp
1107 the captured substrings are "red king", "red", and "king", and are numbered 1,
1108 2, and 3, respectively.
1109 .P
1110 The fact that plain parentheses fulfil two functions is not always helpful.
1111 There are often times when a grouping subpattern is required without a
1112 capturing requirement. If an opening parenthesis is followed by a question mark
1113 and a colon, the subpattern does not do any capturing, and is not counted when
1114 computing the number of any subsequent capturing subpatterns. For example, if
1115 the string "the white queen" is matched against the pattern
1116 .sp
1117 the ((?:red|white) (king|queen))
1118 .sp
1119 the captured substrings are "white queen" and "queen", and are numbered 1 and
1120 2. The maximum number of capturing subpatterns is 65535.
1121 .P
1122 As a convenient shorthand, if any option settings are required at the start of
1123 a non-capturing subpattern, the option letters may appear between the "?" and
1124 the ":". Thus the two patterns
1125 .sp
1126 (?i:saturday|sunday)
1127 (?:(?i)saturday|sunday)
1128 .sp
1129 match exactly the same set of strings. Because alternative branches are tried
1130 from left to right, and options are not reset until the end of the subpattern
1131 is reached, an option setting in one branch does affect subsequent branches, so
1132 the above patterns match "SUNDAY" as well as "Saturday".
1133 .
1134 .
1136 .rs
1137 .sp
1138 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1139 the same numbers for its capturing parentheses. Such a subpattern starts with
1140 (?| and is itself a non-capturing subpattern. For example, consider this
1141 pattern:
1142 .sp
1143 (?|(Sat)ur|(Sun))day
1144 .sp
1145 Because the two alternatives are inside a (?| group, both sets of capturing
1146 parentheses are numbered one. Thus, when the pattern matches, you can look
1147 at captured substring number one, whichever alternative matched. This construct
1148 is useful when you want to capture part, but not all, of one of a number of
1149 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1150 number is reset at the start of each branch. The numbers of any capturing
1151 buffers that follow the subpattern start after the highest number used in any
1152 branch. The following example is taken from the Perl documentation.
1153 The numbers underneath show in which buffer the captured content will be
1154 stored.
1155 .sp
1156 # before ---------------branch-reset----------- after
1157 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1158 # 1 2 2 3 2 3 4
1159 .sp
1160 A backreference or a recursive call to a numbered subpattern always refers to
1161 the first one in the pattern with the given number.
1162 .P
1163 An alternative approach to using this "branch reset" feature is to use
1164 duplicate named subpatterns, as described in the next section.
1165 .
1166 .
1168 .rs
1169 .sp
1170 Identifying capturing parentheses by number is simple, but it can be very hard
1171 to keep track of the numbers in complicated regular expressions. Furthermore,
1172 if an expression is modified, the numbers may change. To help with this
1173 difficulty, PCRE supports the naming of subpatterns. This feature was not
1174 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1175 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1176 the Perl and the Python syntax.
1177 .P
1178 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1179 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1180 parentheses from other parts of the pattern, such as
1181 .\" HTML <a href="#backreferences">
1182 .\" </a>
1183 backreferences,
1184 .\"
1185 .\" HTML <a href="#recursion">
1186 .\" </a>
1187 recursion,
1188 .\"
1189 and
1190 .\" HTML <a href="#conditions">
1191 .\" </a>
1192 conditions,
1193 .\"
1194 can be made by name as well as by number.
1195 .P
1196 Names consist of up to 32 alphanumeric characters and underscores. Named
1197 capturing parentheses are still allocated numbers as well as names, exactly as
1198 if the names were not present. The PCRE API provides function calls for
1199 extracting the name-to-number translation table from a compiled pattern. There
1200 is also a convenience function for extracting a captured substring by name.
1201 .P
1202 By default, a name must be unique within a pattern, but it is possible to relax
1203 this constraint by setting the PCRE_DUPNAMES option at compile time. This can
1204 be useful for patterns where only one instance of the named parentheses can
1205 match. Suppose you want to match the name of a weekday, either as a 3-letter
1206 abbreviation or as the full name, and in both cases you want to extract the
1207 abbreviation. This pattern (ignoring the line breaks) does the job:
1208 .sp
1209 (?<DN>Mon|Fri|Sun)(?:day)?|
1210 (?<DN>Tue)(?:sday)?|
1211 (?<DN>Wed)(?:nesday)?|
1212 (?<DN>Thu)(?:rsday)?|
1213 (?<DN>Sat)(?:urday)?
1214 .sp
1215 There are five capturing substrings, but only one is ever set after a match.
1216 (An alternative way of solving this problem is to use a "branch reset"
1217 subpattern, as described in the previous section.)
1218 .P
1219 The convenience function for extracting the data by name returns the substring
1220 for the first (and in this example, the only) subpattern of that name that
1221 matched. This saves searching to find which numbered subpattern it was. If you
1222 make a reference to a non-unique named subpattern from elsewhere in the
1223 pattern, the one that corresponds to the lowest number is used. For further
1224 details of the interfaces for handling named subpatterns, see the
1225 .\" HREF
1226 \fBpcreapi\fP
1227 .\"
1228 documentation.
1229 .P
1230 \fBWarning:\fP You cannot use different names to distinguish between two
1231 subpatterns with the same number (see the previous section) because PCRE uses
1232 only the numbers when matching.
1233 .
1234 .
1236 .rs
1237 .sp
1238 Repetition is specified by quantifiers, which can follow any of the following
1239 items:
1240 .sp
1241 a literal data character
1242 the dot metacharacter
1243 the \eC escape sequence
1244 the \eX escape sequence (in UTF-8 mode with Unicode properties)
1245 the \eR escape sequence
1246 an escape such as \ed that matches a single character
1247 a character class
1248 a back reference (see next section)
1249 a parenthesized subpattern (unless it is an assertion)
1250 .sp
1251 The general repetition quantifier specifies a minimum and maximum number of
1252 permitted matches, by giving the two numbers in curly brackets (braces),
1253 separated by a comma. The numbers must be less than 65536, and the first must
1254 be less than or equal to the second. For example:
1255 .sp
1256 z{2,4}
1257 .sp
1258 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1259 character. If the second number is omitted, but the comma is present, there is
1260 no upper limit; if the second number and the comma are both omitted, the
1261 quantifier specifies an exact number of required matches. Thus
1262 .sp
1263 [aeiou]{3,}
1264 .sp
1265 matches at least 3 successive vowels, but may match many more, while
1266 .sp
1267 \ed{8}
1268 .sp
1269 matches exactly 8 digits. An opening curly bracket that appears in a position
1270 where a quantifier is not allowed, or one that does not match the syntax of a
1271 quantifier, is taken as a literal character. For example, {,6} is not a
1272 quantifier, but a literal string of four characters.
1273 .P
1274 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
1275 bytes. Thus, for example, \ex{100}{2} matches two UTF-8 characters, each of
1276 which is represented by a two-byte sequence. Similarly, when Unicode property
1277 support is available, \eX{3} matches three Unicode extended sequences, each of
1278 which may be several bytes long (and they may be of different lengths).
1279 .P
1280 The quantifier {0} is permitted, causing the expression to behave as if the
1281 previous item and the quantifier were not present. This may be useful for
1282 subpatterns that are referenced as
1283 .\" HTML <a href="#subpatternsassubroutines">
1284 .\" </a>
1285 subroutines
1286 .\"
1287 from elsewhere in the pattern. Items other than subpatterns that have a {0}
1288 quantifier are omitted from the compiled pattern.
1289 .P
1290 For convenience, the three most common quantifiers have single-character
1291 abbreviations:
1292 .sp
1293 * is equivalent to {0,}
1294 + is equivalent to {1,}
1295 ? is equivalent to {0,1}
1296 .sp
1297 It is possible to construct infinite loops by following a subpattern that can
1298 match no characters with a quantifier that has no upper limit, for example:
1299 .sp
1300 (a?)*
1301 .sp
1302 Earlier versions of Perl and PCRE used to give an error at compile time for
1303 such patterns. However, because there are cases where this can be useful, such
1304 patterns are now accepted, but if any repetition of the subpattern does in fact
1305 match no characters, the loop is forcibly broken.
1306 .P
1307 By default, the quantifiers are "greedy", that is, they match as much as
1308 possible (up to the maximum number of permitted times), without causing the
1309 rest of the pattern to fail. The classic example of where this gives problems
1310 is in trying to match comments in C programs. These appear between /* and */
1311 and within the comment, individual * and / characters may appear. An attempt to
1312 match C comments by applying the pattern
1313 .sp
1314 /\e*.*\e*/
1315 .sp
1316 to the string
1317 .sp
1318 /* first comment */ not comment /* second comment */
1319 .sp
1320 fails, because it matches the entire string owing to the greediness of the .*
1321 item.
1322 .P
1323 However, if a quantifier is followed by a question mark, it ceases to be
1324 greedy, and instead matches the minimum number of times possible, so the
1325 pattern
1326 .sp
1327 /\e*.*?\e*/
1328 .sp
1329 does the right thing with the C comments. The meaning of the various
1330 quantifiers is not otherwise changed, just the preferred number of matches.
1331 Do not confuse this use of question mark with its use as a quantifier in its
1332 own right. Because it has two uses, it can sometimes appear doubled, as in
1333 .sp
1334 \ed??\ed
1335 .sp
1336 which matches one digit by preference, but can match two if that is the only
1337 way the rest of the pattern matches.
1338 .P
1339 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1340 the quantifiers are not greedy by default, but individual ones can be made
1341 greedy by following them with a question mark. In other words, it inverts the
1342 default behaviour.
1343 .P
1344 When a parenthesized subpattern is quantified with a minimum repeat count that
1345 is greater than 1 or with a limited maximum, more memory is required for the
1346 compiled pattern, in proportion to the size of the minimum or maximum.
1347 .P
1348 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1349 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1350 implicitly anchored, because whatever follows will be tried against every
1351 character position in the subject string, so there is no point in retrying the
1352 overall match at any position after the first. PCRE normally treats such a
1353 pattern as though it were preceded by \eA.
1354 .P
1355 In cases where it is known that the subject string contains no newlines, it is
1356 worth setting PCRE_DOTALL in order to obtain this optimization, or
1357 alternatively using ^ to indicate anchoring explicitly.
1358 .P
1359 However, there is one situation where the optimization cannot be used. When .*
1360 is inside capturing parentheses that are the subject of a backreference
1361 elsewhere in the pattern, a match at the start may fail where a later one
1362 succeeds. Consider, for example:
1363 .sp
1364 (.*)abc\e1
1365 .sp
1366 If the subject is "xyz123abc123" the match point is the fourth character. For
1367 this reason, such a pattern is not implicitly anchored.
1368 .P
1369 When a capturing subpattern is repeated, the value captured is the substring
1370 that matched the final iteration. For example, after
1371 .sp
1372 (tweedle[dume]{3}\es*)+
1373 .sp
1374 has matched "tweedledum tweedledee" the value of the captured substring is
1375 "tweedledee". However, if there are nested capturing subpatterns, the
1376 corresponding captured values may have been set in previous iterations. For
1377 example, after
1378 .sp
1379 /(a|(b))+/
1380 .sp
1381 matches "aba" the value of the second captured substring is "b".
1382 .
1383 .
1384 .\" HTML <a name="atomicgroup"></a>
1386 .rs
1387 .sp
1388 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1389 repetition, failure of what follows normally causes the repeated item to be
1390 re-evaluated to see if a different number of repeats allows the rest of the
1391 pattern to match. Sometimes it is useful to prevent this, either to change the
1392 nature of the match, or to cause it fail earlier than it otherwise might, when
1393 the author of the pattern knows there is no point in carrying on.
1394 .P
1395 Consider, for example, the pattern \ed+foo when applied to the subject line
1396 .sp
1397 123456bar
1398 .sp
1399 After matching all 6 digits and then failing to match "foo", the normal
1400 action of the matcher is to try again with only 5 digits matching the \ed+
1401 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1402 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1403 that once a subpattern has matched, it is not to be re-evaluated in this way.
1404 .P
1405 If we use atomic grouping for the previous example, the matcher gives up
1406 immediately on failing to match "foo" the first time. The notation is a kind of
1407 special parenthesis, starting with (?> as in this example:
1408 .sp
1409 (?>\ed+)foo
1410 .sp
1411 This kind of parenthesis "locks up" the part of the pattern it contains once
1412 it has matched, and a failure further into the pattern is prevented from
1413 backtracking into it. Backtracking past it to previous items, however, works as
1414 normal.
1415 .P
1416 An alternative description is that a subpattern of this type matches the string
1417 of characters that an identical standalone pattern would match, if anchored at
1418 the current point in the subject string.
1419 .P
1420 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1421 the above example can be thought of as a maximizing repeat that must swallow
1422 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1423 number of digits they match in order to make the rest of the pattern match,
1424 (?>\ed+) can only match an entire sequence of digits.
1425 .P
1426 Atomic groups in general can of course contain arbitrarily complicated
1427 subpatterns, and can be nested. However, when the subpattern for an atomic
1428 group is just a single repeated item, as in the example above, a simpler
1429 notation, called a "possessive quantifier" can be used. This consists of an
1430 additional + character following a quantifier. Using this notation, the
1431 previous example can be rewritten as
1432 .sp
1433 \ed++foo
1434 .sp
1435 Note that a possessive quantifier can be used with an entire group, for
1436 example:
1437 .sp
1438 (abc|xyz){2,3}+
1439 .sp
1440 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1441 option is ignored. They are a convenient notation for the simpler forms of
1442 atomic group. However, there is no difference in the meaning of a possessive
1443 quantifier and the equivalent atomic group, though there may be a performance
1444 difference; possessive quantifiers should be slightly faster.
1445 .P
1446 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1447 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1448 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1449 package, and PCRE copied it from there. It ultimately found its way into Perl
1450 at release 5.10.
1451 .P
1452 PCRE has an optimization that automatically "possessifies" certain simple
1453 pattern constructs. For example, the sequence A+B is treated as A++B because
1454 there is no point in backtracking into a sequence of A's when B must follow.
1455 .P
1456 When a pattern contains an unlimited repeat inside a subpattern that can itself
1457 be repeated an unlimited number of times, the use of an atomic group is the
1458 only way to avoid some failing matches taking a very long time indeed. The
1459 pattern
1460 .sp
1461 (\eD+|<\ed+>)*[!?]
1462 .sp
1463 matches an unlimited number of substrings that either consist of non-digits, or
1464 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1465 quickly. However, if it is applied to
1466 .sp
1467 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1468 .sp
1469 it takes a long time before reporting failure. This is because the string can
1470 be divided between the internal \eD+ repeat and the external * repeat in a
1471 large number of ways, and all have to be tried. (The example uses [!?] rather
1472 than a single character at the end, because both PCRE and Perl have an
1473 optimization that allows for fast failure when a single character is used. They
1474 remember the last single character that is required for a match, and fail early
1475 if it is not present in the string.) If the pattern is changed so that it uses
1476 an atomic group, like this:
1477 .sp
1478 ((?>\eD+)|<\ed+>)*[!?]
1479 .sp
1480 sequences of non-digits cannot be broken, and failure happens quickly.
1481 .
1482 .
1483 .\" HTML <a name="backreferences"></a>
1485 .rs
1486 .sp
1487 Outside a character class, a backslash followed by a digit greater than 0 (and
1488 possibly further digits) is a back reference to a capturing subpattern earlier
1489 (that is, to its left) in the pattern, provided there have been that many
1490 previous capturing left parentheses.
1491 .P
1492 However, if the decimal number following the backslash is less than 10, it is
1493 always taken as a back reference, and causes an error only if there are not
1494 that many capturing left parentheses in the entire pattern. In other words, the
1495 parentheses that are referenced need not be to the left of the reference for
1496 numbers less than 10. A "forward back reference" of this type can make sense
1497 when a repetition is involved and the subpattern to the right has participated
1498 in an earlier iteration.
1499 .P
1500 It is not possible to have a numerical "forward back reference" to a subpattern
1501 whose number is 10 or more using this syntax because a sequence such as \e50 is
1502 interpreted as a character defined in octal. See the subsection entitled
1503 "Non-printing characters"
1504 .\" HTML <a href="#digitsafterbackslash">
1505 .\" </a>
1506 above
1507 .\"
1508 for further details of the handling of digits following a backslash. There is
1509 no such problem when named parentheses are used. A back reference to any
1510 subpattern is possible using named parentheses (see below).
1511 .P
1512 Another way of avoiding the ambiguity inherent in the use of digits following a
1513 backslash is to use the \eg escape sequence, which is a feature introduced in
1514 Perl 5.10. This escape must be followed by an unsigned number or a negative
1515 number, optionally enclosed in braces. These examples are all identical:
1516 .sp
1517 (ring), \e1
1518 (ring), \eg1
1519 (ring), \eg{1}
1520 .sp
1521 An unsigned number specifies an absolute reference without the ambiguity that
1522 is present in the older syntax. It is also useful when literal digits follow
1523 the reference. A negative number is a relative reference. Consider this
1524 example:
1525 .sp
1526 (abc(def)ghi)\eg{-1}
1527 .sp
1528 The sequence \eg{-1} is a reference to the most recently started capturing
1529 subpattern before \eg, that is, is it equivalent to \e2. Similarly, \eg{-2}
1530 would be equivalent to \e1. The use of relative references can be helpful in
1531 long patterns, and also in patterns that are created by joining together
1532 fragments that contain references within themselves.
1533 .P
1534 A back reference matches whatever actually matched the capturing subpattern in
1535 the current subject string, rather than anything matching the subpattern
1536 itself (see
1537 .\" HTML <a href="#subpatternsassubroutines">
1538 .\" </a>
1539 "Subpatterns as subroutines"
1540 .\"
1541 below for a way of doing that). So the pattern
1542 .sp
1543 (sens|respons)e and \e1ibility
1544 .sp
1545 matches "sense and sensibility" and "response and responsibility", but not
1546 "sense and responsibility". If caseful matching is in force at the time of the
1547 back reference, the case of letters is relevant. For example,
1548 .sp
1549 ((?i)rah)\es+\e1
1550 .sp
1551 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1552 capturing subpattern is matched caselessly.
1553 .P
1554 There are several different ways of writing back references to named
1555 subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
1556 \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1557 back reference syntax, in which \eg can be used for both numeric and named
1558 references, is also supported. We could rewrite the above example in any of
1559 the following ways:
1560 .sp
1561 (?<p1>(?i)rah)\es+\ek<p1>
1562 (?'p1'(?i)rah)\es+\ek{p1}
1563 (?P<p1>(?i)rah)\es+(?P=p1)
1564 (?<p1>(?i)rah)\es+\eg{p1}
1565 .sp
1566 A subpattern that is referenced by name may appear in the pattern before or
1567 after the reference.
1568 .P
1569 There may be more than one back reference to the same subpattern. If a
1570 subpattern has not actually been used in a particular match, any back
1571 references to it always fail. For example, the pattern
1572 .sp
1573 (a|(bc))\e2
1574 .sp
1575 always fails if it starts to match "a" rather than "bc". Because there may be
1576 many capturing parentheses in a pattern, all digits following the backslash are
1577 taken as part of a potential back reference number. If the pattern continues
1578 with a digit character, some delimiter must be used to terminate the back
1579 reference. If the PCRE_EXTENDED option is set, this can be whitespace.
1580 Otherwise an empty comment (see
1581 .\" HTML <a href="#comments">
1582 .\" </a>
1583 "Comments"
1584 .\"
1585 below) can be used.
1586 .P
1587 A back reference that occurs inside the parentheses to which it refers fails
1588 when the subpattern is first used, so, for example, (a\e1) never matches.
1589 However, such references can be useful inside repeated subpatterns. For
1590 example, the pattern
1591 .sp
1592 (a|b\e1)+
1593 .sp
1594 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1595 the subpattern, the back reference matches the character string corresponding
1596 to the previous iteration. In order for this to work, the pattern must be such
1597 that the first iteration does not need to match the back reference. This can be
1598 done using alternation, as in the example above, or by a quantifier with a
1599 minimum of zero.
1600 .
1601 .
1602 .\" HTML <a name="bigassertions"></a>
1604 .rs
1605 .sp
1606 An assertion is a test on the characters following or preceding the current
1607 matching point that does not actually consume any characters. The simple
1608 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
1609 .\" HTML <a href="#smallassertions">
1610 .\" </a>
1611 above.
1612 .\"
1613 .P
1614 More complicated assertions are coded as subpatterns. There are two kinds:
1615 those that look ahead of the current position in the subject string, and those
1616 that look behind it. An assertion subpattern is matched in the normal way,
1617 except that it does not cause the current matching position to be changed.
1618 .P
1619 Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1620 because it makes no sense to assert the same thing several times. If any kind
1621 of assertion contains capturing subpatterns within it, these are counted for
1622 the purposes of numbering the capturing subpatterns in the whole pattern.
1623 However, substring capturing is carried out only for positive assertions,
1624 because it does not make sense for negative assertions.
1625 .
1626 .
1627 .SS "Lookahead assertions"
1628 .rs
1629 .sp
1630 Lookahead assertions start with (?= for positive assertions and (?! for
1631 negative assertions. For example,
1632 .sp
1633 \ew+(?=;)
1634 .sp
1635 matches a word followed by a semicolon, but does not include the semicolon in
1636 the match, and
1637 .sp
1638 foo(?!bar)
1639 .sp
1640 matches any occurrence of "foo" that is not followed by "bar". Note that the
1641 apparently similar pattern
1642 .sp
1643 (?!foo)bar
1644 .sp
1645 does not find an occurrence of "bar" that is preceded by something other than
1646 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1647 (?!foo) is always true when the next three characters are "bar". A
1648 lookbehind assertion is needed to achieve the other effect.
1649 .P
1650 If you want to force a matching failure at some point in a pattern, the most
1651 convenient way to do it is with (?!) because an empty string always matches, so
1652 an assertion that requires there not to be an empty string must always fail.
1653 .
1654 .
1655 .\" HTML <a name="lookbehind"></a>
1656 .SS "Lookbehind assertions"
1657 .rs
1658 .sp
1659 Lookbehind assertions start with (?<= for positive assertions and (?<! for
1660 negative assertions. For example,
1661 .sp
1662 (?<!foo)bar
1663 .sp
1664 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1665 a lookbehind assertion are restricted such that all the strings it matches must
1666 have a fixed length. However, if there are several top-level alternatives, they
1667 do not all have to have the same fixed length. Thus
1668 .sp
1669 (?<=bullock|donkey)
1670 .sp
1671 is permitted, but
1672 .sp
1673 (?<!dogs?|cats?)
1674 .sp
1675 causes an error at compile time. Branches that match different length strings
1676 are permitted only at the top level of a lookbehind assertion. This is an
1677 extension compared with Perl (5.8 and 5.10), which requires all branches to
1678 match the same length of string. An assertion such as
1679 .sp
1680 (?<=ab(c|de))
1681 .sp
1682 is not permitted, because its single top-level branch can match two different
1683 lengths, but it is acceptable to PCRE if rewritten to use two top-level
1684 branches:
1685 .sp
1686 (?<=abc|abde)
1687 .sp
1688 In some cases, the Perl 5.10 escape sequence \eK
1689 .\" HTML <a href="#resetmatchstart">
1690 .\" </a>
1691 (see above)
1692 .\"
1693 can be used instead of a lookbehind assertion to get round the fixed-length
1694 restriction.
1695 .P
1696 The implementation of lookbehind assertions is, for each alternative, to
1697 temporarily move the current position back by the fixed length and then try to
1698 match. If there are insufficient characters before the current position, the
1699 assertion fails.
1700 .P
1701 PCRE does not allow the \eC escape (which matches a single byte in UTF-8 mode)
1702 to appear in lookbehind assertions, because it makes it impossible to calculate
1703 the length of the lookbehind. The \eX and \eR escapes, which can match
1704 different numbers of bytes, are also not permitted.
1705 .P
1706 .\" HTML <a href="#subpatternsassubroutines">
1707 .\" </a>
1708 "Subroutine"
1709 .\"
1710 calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
1711 as the subpattern matches a fixed-length string.
1712 .\" HTML <a href="#recursion">
1713 .\" </a>
1714 Recursion,
1715 .\"
1716 however, is not supported.
1717 .P
1718 Possessive quantifiers can be used in conjunction with lookbehind assertions to
1719 specify efficient matching at the end of the subject string. Consider a simple
1720 pattern such as
1721 .sp
1722 abcd$
1723 .sp
1724 when applied to a long string that does not match. Because matching proceeds
1725 from left to right, PCRE will look for each "a" in the subject and then see if
1726 what follows matches the rest of the pattern. If the pattern is specified as
1727 .sp
1728 ^.*abcd$
1729 .sp
1730 the initial .* matches the entire string at first, but when this fails (because
1731 there is no following "a"), it backtracks to match all but the last character,
1732 then all but the last two characters, and so on. Once again the search for "a"
1733 covers the entire string, from right to left, so we are no better off. However,
1734 if the pattern is written as
1735 .sp
1736 ^.*+(?<=abcd)
1737 .sp
1738 there can be no backtracking for the .*+ item; it can match only the entire
1739 string. The subsequent lookbehind assertion does a single test on the last four
1740 characters. If it fails, the match fails immediately. For long strings, this
1741 approach makes a significant difference to the processing time.
1742 .
1743 .
1744 .SS "Using multiple assertions"
1745 .rs
1746 .sp
1747 Several assertions (of any sort) may occur in succession. For example,
1748 .sp
1749 (?<=\ed{3})(?<!999)foo
1750 .sp
1751 matches "foo" preceded by three digits that are not "999". Notice that each of
1752 the assertions is applied independently at the same point in the subject
1753 string. First there is a check that the previous three characters are all
1754 digits, and then there is a check that the same three characters are not "999".
1755 This pattern does \fInot\fP match "foo" preceded by six characters, the first
1756 of which are digits and the last three of which are not "999". For example, it
1757 doesn't match "123abcfoo". A pattern to do that is
1758 .sp
1759 (?<=\ed{3}...)(?<!999)foo
1760 .sp
1761 This time the first assertion looks at the preceding six characters, checking
1762 that the first three are digits, and then the second assertion checks that the
1763 preceding three characters are not "999".
1764 .P
1765 Assertions can be nested in any combination. For example,
1766 .sp
1767 (?<=(?<!foo)bar)baz
1768 .sp
1769 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1770 preceded by "foo", while
1771 .sp
1772 (?<=\ed{3}(?!999)...)foo
1773 .sp
1774 is another pattern that matches "foo" preceded by three digits and any three
1775 characters that are not "999".
1776 .
1777 .
1778 .\" HTML <a name="conditions"></a>
1780 .rs
1781 .sp
1782 It is possible to cause the matching process to obey a subpattern
1783 conditionally or to choose between two alternative subpatterns, depending on
1784 the result of an assertion, or whether a previous capturing subpattern matched
1785 or not. The two possible forms of conditional subpattern are
1786 .sp
1787 (?(condition)yes-pattern)
1788 (?(condition)yes-pattern|no-pattern)
1789 .sp
1790 If the condition is satisfied, the yes-pattern is used; otherwise the
1791 no-pattern (if present) is used. If there are more than two alternatives in the
1792 subpattern, a compile-time error occurs.
1793 .P
1794 There are four kinds of condition: references to subpatterns, references to
1795 recursion, a pseudo-condition called DEFINE, and assertions.
1796 .
1797 .SS "Checking for a used subpattern by number"
1798 .rs
1799 .sp
1800 If the text between the parentheses consists of a sequence of digits, the
1801 condition is true if the capturing subpattern of that number has previously
1802 matched. An alternative notation is to precede the digits with a plus or minus
1803 sign. In this case, the subpattern number is relative rather than absolute.
1804 The most recently opened parentheses can be referenced by (?(-1), the next most
1805 recent by (?(-2), and so on. In looping constructs it can also make sense to
1806 refer to subsequent groups with constructs such as (?(+2).
1807 .P
1808 Consider the following pattern, which contains non-significant white space to
1809 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
1810 three parts for ease of discussion:
1811 .sp
1812 ( \e( )? [^()]+ (?(1) \e) )
1813 .sp
1814 The first part matches an optional opening parenthesis, and if that
1815 character is present, sets it as the first captured substring. The second part
1816 matches one or more characters that are not parentheses. The third part is a
1817 conditional subpattern that tests whether the first set of parentheses matched
1818 or not. If they did, that is, if subject started with an opening parenthesis,
1819 the condition is true, and so the yes-pattern is executed and a closing
1820 parenthesis is required. Otherwise, since no-pattern is not present, the
1821 subpattern matches nothing. In other words, this pattern matches a sequence of
1822 non-parentheses, optionally enclosed in parentheses.
1823 .P
1824 If you were embedding this pattern in a larger one, you could use a relative
1825 reference:
1826 .sp
1827 ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
1828 .sp
1829 This makes the fragment independent of the parentheses in the larger pattern.
1830 .
1831 .SS "Checking for a used subpattern by name"
1832 .rs
1833 .sp
1834 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
1835 subpattern by name. For compatibility with earlier versions of PCRE, which had
1836 this facility before Perl, the syntax (?(name)...) is also recognized. However,
1837 there is a possible ambiguity with this syntax, because subpattern names may
1838 consist entirely of digits. PCRE looks first for a named subpattern; if it
1839 cannot find one and the name consists entirely of digits, PCRE looks for a
1840 subpattern of that number, which must be greater than zero. Using subpattern
1841 names that consist entirely of digits is not recommended.
1842 .P
1843 Rewriting the above example to use a named subpattern gives this:
1844 .sp
1845 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
1846 .sp
1847 .
1848 .SS "Checking for pattern recursion"
1849 .rs
1850 .sp
1851 If the condition is the string (R), and there is no subpattern with the name R,
1852 the condition is true if a recursive call to the whole pattern or any
1853 subpattern has been made. If digits or a name preceded by ampersand follow the
1854 letter R, for example:
1855 .sp
1856 (?(R3)...) or (?(R&name)...)
1857 .sp
1858 the condition is true if the most recent recursion is into the subpattern whose
1859 number or name is given. This condition does not check the entire recursion
1860 stack.
1861 .P
1862 At "top level", all these recursion test conditions are false.
1863 .\" HTML <a href="#recursion">
1864 .\" </a>
1865 Recursive patterns
1866 .\"
1867 are described below.
1868 .
1869 .SS "Defining subpatterns for use by reference only"
1870 .rs
1871 .sp
1872 If the condition is the string (DEFINE), and there is no subpattern with the
1873 name DEFINE, the condition is always false. In this case, there may be only one
1874 alternative in the subpattern. It is always skipped if control reaches this
1875 point in the pattern; the idea of DEFINE is that it can be used to define
1876 "subroutines" that can be referenced from elsewhere. (The use of
1877 .\" HTML <a href="#subpatternsassubroutines">
1878 .\" </a>
1879 "subroutines"
1880 .\"
1881 is described below.) For example, a pattern to match an IPv4 address could be
1882 written like this (ignore whitespace and line breaks):
1883 .sp
1884 (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
1885 \eb (?&byte) (\e.(?&byte)){3} \eb
1886 .sp
1887 The first part of the pattern is a DEFINE group inside which a another group
1888 named "byte" is defined. This matches an individual component of an IPv4
1889 address (a number less than 256). When matching takes place, this part of the
1890 pattern is skipped because DEFINE acts like a false condition.
1891 .P
1892 The rest of the pattern uses references to the named group to match the four
1893 dot-separated components of an IPv4 address, insisting on a word boundary at
1894 each end.
1895 .
1896 .SS "Assertion conditions"
1897 .rs
1898 .sp
1899 If the condition is not in any of the above formats, it must be an assertion.
1900 This may be a positive or negative lookahead or lookbehind assertion. Consider
1901 this pattern, again containing non-significant white space, and with the two
1902 alternatives on the second line:
1903 .sp
1904 (?(?=[^a-z]*[a-z])
1905 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
1906 .sp
1907 The condition is a positive lookahead assertion that matches an optional
1908 sequence of non-letters followed by a letter. In other words, it tests for the
1909 presence of at least one letter in the subject. If a letter is found, the
1910 subject is matched against the first alternative; otherwise it is matched
1911 against the second. This pattern matches strings in one of the two forms
1912 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
1913 .
1914 .
1915 .\" HTML <a name="comments"></a>
1917 .rs
1918 .sp
1919 The sequence (?# marks the start of a comment that continues up to the next
1920 closing parenthesis. Nested parentheses are not permitted. The characters
1921 that make up a comment play no part in the pattern matching at all.
1922 .P
1923 If the PCRE_EXTENDED option is set, an unescaped # character outside a
1924 character class introduces a comment that continues to immediately after the
1925 next newline in the pattern.
1926 .
1927 .
1928 .\" HTML <a name="recursion"></a>
1930 .rs
1931 .sp
1932 Consider the problem of matching a string in parentheses, allowing for
1933 unlimited nested parentheses. Without the use of recursion, the best that can
1934 be done is to use a pattern that matches up to some fixed depth of nesting. It
1935 is not possible to handle an arbitrary nesting depth.
1936 .P
1937 For some time, Perl has provided a facility that allows regular expressions to
1938 recurse (amongst other things). It does this by interpolating Perl code in the
1939 expression at run time, and the code can refer to the expression itself. A Perl
1940 pattern using code interpolation to solve the parentheses problem can be
1941 created like this:
1942 .sp
1943 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
1944 .sp
1945 The (?p{...}) item interpolates Perl code at run time, and in this case refers
1946 recursively to the pattern in which it appears.
1947 .P
1948 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
1949 supports special syntax for recursion of the entire pattern, and also for
1950 individual subpattern recursion. After its introduction in PCRE and Python,
1951 this kind of recursion was subsequently introduced into Perl at release 5.10.
1952 .P
1953 A special item that consists of (? followed by a number greater than zero and a
1954 closing parenthesis is a recursive call of the subpattern of the given number,
1955 provided that it occurs inside that subpattern. (If not, it is a
1956 .\" HTML <a href="#subpatternsassubroutines">
1957 .\" </a>
1958 "subroutine"
1959 .\"
1960 call, which is described in the next section.) The special item (?R) or (?0) is
1961 a recursive call of the entire regular expression.
1962 .P
1963 This PCRE pattern solves the nested parentheses problem (assume the
1964 PCRE_EXTENDED option is set so that white space is ignored):
1965 .sp
1966 \e( ( (?>[^()]+) | (?R) )* \e)
1967 .sp
1968 First it matches an opening parenthesis. Then it matches any number of
1969 substrings which can either be a sequence of non-parentheses, or a recursive
1970 match of the pattern itself (that is, a correctly parenthesized substring).
1971 Finally there is a closing parenthesis.
1972 .P
1973 If this were part of a larger pattern, you would not want to recurse the entire
1974 pattern, so instead you could use this:
1975 .sp
1976 ( \e( ( (?>[^()]+) | (?1) )* \e) )
1977 .sp
1978 We have put the pattern into parentheses, and caused the recursion to refer to
1979 them instead of the whole pattern.
1980 .P
1981 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
1982 is made easier by the use of relative references. (A Perl 5.10 feature.)
1983 Instead of (?1) in the pattern above you can write (?-2) to refer to the second
1984 most recently opened parentheses preceding the recursion. In other words, a
1985 negative number counts capturing parentheses leftwards from the point at which
1986 it is encountered.
1987 .P
1988 It is also possible to refer to subsequently opened parentheses, by writing
1989 references such as (?+2). However, these cannot be recursive because the
1990 reference is not inside the parentheses that are referenced. They are always
1991 .\" HTML <a href="#subpatternsassubroutines">
1992 .\" </a>
1993 "subroutine"
1994 .\"
1995 calls, as described in the next section.
1996 .P
1997 An alternative approach is to use named parentheses instead. The Perl syntax
1998 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
1999 could rewrite the above example as follows:
2000 .sp
2001 (?<pn> \e( ( (?>[^()]+) | (?&pn) )* \e) )
2002 .sp
2003 If there is more than one subpattern with the same name, the earliest one is
2004 used.
2005 .P
2006 This particular example pattern that we have been looking at contains nested
2007 unlimited repeats, and so the use of atomic grouping for matching strings of
2008 non-parentheses is important when applying the pattern to strings that do not
2009 match. For example, when this pattern is applied to
2010 .sp
2011 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2012 .sp
2013 it yields "no match" quickly. However, if atomic grouping is not used,
2014 the match runs for a very long time indeed because there are so many different
2015 ways the + and * repeats can carve up the subject, and all have to be tested
2016 before failure can be reported.
2017 .P
2018 At the end of a match, the values set for any capturing subpatterns are those
2019 from the outermost level of the recursion at which the subpattern value is set.
2020 If you want to obtain intermediate values, a callout function can be used (see
2021 below and the
2022 .\" HREF
2023 \fBpcrecallout\fP
2024 .\"
2025 documentation). If the pattern above is matched against
2026 .sp
2027 (ab(cd)ef)
2028 .sp
2029 the value for the capturing parentheses is "ef", which is the last value taken
2030 on at the top level. If additional parentheses are added, giving
2031 .sp
2032 \e( ( ( (?>[^()]+) | (?R) )* ) \e)
2033 ^ ^
2034 ^ ^
2035 .sp
2036 the string they capture is "ab(cd)ef", the contents of the top level
2037 parentheses. If there are more than 15 capturing parentheses in a pattern, PCRE
2038 has to obtain extra memory to store data during a recursion, which it does by
2039 using \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no
2040 memory can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2041 .P
2042 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2043 Consider this pattern, which matches text in angle brackets, allowing for
2044 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2045 recursing), whereas any characters are permitted at the outer level.
2046 .sp
2047 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
2048 .sp
2049 In this pattern, (?(R) is the start of a conditional subpattern, with two
2050 different alternatives for the recursive and non-recursive cases. The (?R) item
2051 is the actual recursive call.
2052 .
2053 .
2054 .\" HTML <a name="recursiondifference"></a>
2055 .SS "Recursion difference from Perl"
2056 .rs
2057 .sp
2058 In PCRE (like Python, but unlike Perl), a recursive subpattern call is always
2059 treated as an atomic group. That is, once it has matched some of the subject
2060 string, it is never re-entered, even if it contains untried alternatives and
2061 there is a subsequent matching failure. This can be illustrated by the
2062 following pattern, which purports to match a palindromic string that contains
2063 an odd number of characters (for example, "a", "aba", "abcba", "abcdcba"):
2064 .sp
2065 ^(.|(.)(?1)\e2)$
2066 .sp
2067 The idea is that it either matches a single character, or two identical
2068 characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2069 it does not if the pattern is longer than three characters. Consider the
2070 subject string "abcba":
2071 .P
2072 At the top level, the first character is matched, but as it is not at the end
2073 of the string, the first alternative fails; the second alternative is taken
2074 and the recursion kicks in. The recursive call to subpattern 1 successfully
2075 matches the next character ("b"). (Note that the beginning and end of line
2076 tests are not part of the recursion).
2077 .P
2078 Back at the top level, the next character ("c") is compared with what
2079 subpattern 2 matched, which was "a". This fails. Because the recursion is
2080 treated as an atomic group, there are now no backtracking points, and so the
2081 entire match fails. (Perl is able, at this point, to re-enter the recursion and
2082 try the second alternative.) However, if the pattern is written with the
2083 alternatives in the other order, things are different:
2084 .sp
2085 ^((.)(?1)\e2|.)$
2086 .sp
2087 This time, the recursing alternative is tried first, and continues to recurse
2088 until it runs out of characters, at which point the recursion fails. But this
2089 time we do have another alternative to try at the higher level. That is the big
2090 difference: in the previous case the remaining alternative is at a deeper
2091 recursion level, which PCRE cannot use.
2092 .P
2093 To change the pattern so that matches all palindromic strings, not just those
2094 with an odd number of characters, it is tempting to change the pattern to this:
2095 .sp
2096 ^((.)(?1)\e2|.?)$
2097 .sp
2098 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2099 deeper recursion has matched a single character, it cannot be entered again in
2100 order to match an empty string. The solution is to separate the two cases, and
2101 write out the odd and even cases as alternatives at the higher level:
2102 .sp
2103 ^(?:((.)(?1)\e2|)|((.)(?3)\e4|.))
2104 .sp
2105 If you want to match typical palindromic phrases, the pattern has to ignore all
2106 non-word characters, which can be done like this:
2107 .sp
2108 ^\eW*+(?:((.)\eW*+(?1)\eW*+\e2|)|((.)\eW*+(?3)\eW*+\4|\eW*+.\eW*+))\eW*+$
2109 .sp
2110 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2111 man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2112 the use of the possessive quantifier *+ to avoid backtracking into sequences of
2113 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2114 more) to match typical phrases, and Perl takes so long that you think it has
2115 gone into a loop.
2116 .
2117 .
2118 .\" HTML <a name="subpatternsassubroutines"></a>
2120 .rs
2121 .sp
2122 If the syntax for a recursive subpattern reference (either by number or by
2123 name) is used outside the parentheses to which it refers, it operates like a
2124 subroutine in a programming language. The "called" subpattern may be defined
2125 before or after the reference. A numbered reference can be absolute or
2126 relative, as in these examples:
2127 .sp
2128 (...(absolute)...)...(?2)...
2129 (...(relative)...)...(?-1)...
2130 (...(?+1)...(relative)...
2131 .sp
2132 An earlier example pointed out that the pattern
2133 .sp
2134 (sens|respons)e and \e1ibility
2135 .sp
2136 matches "sense and sensibility" and "response and responsibility", but not
2137 "sense and responsibility". If instead the pattern
2138 .sp
2139 (sens|respons)e and (?1)ibility
2140 .sp
2141 is used, it does match "sense and responsibility" as well as the other two
2142 strings. Another example is given in the discussion of DEFINE above.
2143 .P
2144 Like recursive subpatterns, a "subroutine" call is always treated as an atomic
2145 group. That is, once it has matched some of the subject string, it is never
2146 re-entered, even if it contains untried alternatives and there is a subsequent
2147 matching failure.
2148 .P
2149 When a subpattern is used as a subroutine, processing options such as
2150 case-independence are fixed when the subpattern is defined. They cannot be
2151 changed for different calls. For example, consider this pattern:
2152 .sp
2153 (abc)(?i:(?-1))
2154 .sp
2155 It matches "abcabc". It does not match "abcABC" because the change of
2156 processing option does not affect the called subpattern.
2157 .
2158 .
2159 .\" HTML <a name="onigurumasubroutines"></a>
2161 .rs
2162 .sp
2163 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
2164 a number enclosed either in angle brackets or single quotes, is an alternative
2165 syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2166 are two of the examples used above, rewritten using this syntax:
2167 .sp
2168 (?<pn> \e( ( (?>[^()]+) | \eg<pn> )* \e) )
2169 (sens|respons)e and \eg'1'ibility
2170 .sp
2171 PCRE supports an extension to Oniguruma: if a number is preceded by a
2172 plus or a minus sign it is taken as a relative reference. For example:
2173 .sp
2174 (abc)(?i:\eg<-1>)
2175 .sp
2176 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
2177 synonymous. The former is a back reference; the latter is a subroutine call.
2178 .
2179 .
2181 .rs
2182 .sp
2183 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2184 code to be obeyed in the middle of matching a regular expression. This makes it
2185 possible, amongst other things, to extract different substrings that match the
2186 same pair of parentheses when there is a repetition.
2187 .P
2188 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2189 code. The feature is called "callout". The caller of PCRE provides an external
2190 function by putting its entry point in the global variable \fIpcre_callout\fP.
2191 By default, this variable contains NULL, which disables all calling out.
2192 .P
2193 Within a regular expression, (?C) indicates the points at which the external
2194 function is to be called. If you want to identify different callout points, you
2195 can put a number less than 256 after the letter C. The default value is zero.
2196 For example, this pattern has two callout points:
2197 .sp
2198 (?C1)abc(?C2)def
2199 .sp
2200 If the PCRE_AUTO_CALLOUT flag is passed to \fBpcre_compile()\fP, callouts are
2201 automatically installed before each item in the pattern. They are all numbered
2202 255.
2203 .P
2204 During matching, when PCRE reaches a callout point (and \fIpcre_callout\fP is
2205 set), the external function is called. It is provided with the number of the
2206 callout, the position in the pattern, and, optionally, one item of data
2207 originally supplied by the caller of \fBpcre_exec()\fP. The callout function
2208 may cause matching to proceed, to backtrack, or to fail altogether. A complete
2209 description of the interface to the callout function is given in the
2210 .\" HREF
2211 \fBpcrecallout\fP
2212 .\"
2213 documentation.
2214 .
2215 .
2217 .rs
2218 .sp
2219 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2220 are described in the Perl documentation as "experimental and subject to change
2221 or removal in a future version of Perl". It goes on to say: "Their usage in
2222 production code should be noted to avoid problems during upgrades." The same
2223 remarks apply to the PCRE features described in this section.
2224 .P
2225 Since these verbs are specifically related to backtracking, most of them can be
2226 used only when the pattern is to be matched using \fBpcre_exec()\fP, which uses
2227 a backtracking algorithm. With the exception of (*FAIL), which behaves like a
2228 failing negative assertion, they cause an error if encountered by
2229 \fBpcre_dfa_exec()\fP.
2230 .P
2231 If any of these verbs are used in an assertion subpattern, their effect is
2232 confined to that subpattern; it does not extend to the surrounding pattern.
2233 Note that assertion subpatterns are processed as anchored at the point where
2234 they are tested.
2235 .P
2236 The new verbs make use of what was previously invalid syntax: an opening
2237 parenthesis followed by an asterisk. In Perl, they are generally of the form
2238 (*VERB:ARG) but PCRE does not support the use of arguments, so its general
2239 form is just (*VERB). Any number of these verbs may occur in a pattern. There
2240 are two kinds:
2241 .
2242 .SS "Verbs that act immediately"
2243 .rs
2244 .sp
2245 The following verbs act as soon as they are encountered:
2246 .sp
2247 (*ACCEPT)
2248 .sp
2249 This verb causes the match to end successfully, skipping the remainder of the
2250 pattern. When inside a recursion, only the innermost pattern is ended
2251 immediately. If the (*ACCEPT) is inside capturing parentheses, the data so far
2252 is captured. (This feature was added to PCRE at release 8.00.) For example:
2253 .sp
2254 A((?:A|B(*ACCEPT)|C)D)
2255 .sp
2256 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2257 the outer parentheses.
2258 .sp
2259 (*FAIL) or (*F)
2260 .sp
2261 This verb causes the match to fail, forcing backtracking to occur. It is
2262 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2263 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2264 Perl features that are not present in PCRE. The nearest equivalent is the
2265 callout feature, as for example in this pattern:
2266 .sp
2267 a+(?C)(*FAIL)
2268 .sp
2269 A match with the string "aaaa" always fails, but the callout is taken before
2270 each backtrack happens (in this example, 10 times).
2271 .
2272 .SS "Verbs that act after backtracking"
2273 .rs
2274 .sp
2275 The following verbs do nothing when they are encountered. Matching continues
2276 with what follows, but if there is no subsequent match, a failure is forced.
2277 The verbs differ in exactly what kind of failure occurs.
2278 .sp
2279 (*COMMIT)
2280 .sp
2281 This verb causes the whole match to fail outright if the rest of the pattern
2282 does not match. Even if the pattern is unanchored, no further attempts to find
2283 a match by advancing the start point take place. Once (*COMMIT) has been
2284 passed, \fBpcre_exec()\fP is committed to finding a match at the current
2285 starting point, or not at all. For example:
2286 .sp
2287 a+(*COMMIT)b
2288 .sp
2289 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2290 dynamic anchor, or "I've started, so I must finish."
2291 .sp
2292 (*PRUNE)
2293 .sp
2294 This verb causes the match to fail at the current position if the rest of the
2295 pattern does not match. If the pattern is unanchored, the normal "bumpalong"
2296 advance to the next starting character then happens. Backtracking can occur as
2297 usual to the left of (*PRUNE), or when matching to the right of (*PRUNE), but
2298 if there is no match to the right, backtracking cannot cross (*PRUNE).
2299 In simple cases, the use of (*PRUNE) is just an alternative to an atomic
2300 group or possessive quantifier, but there are some uses of (*PRUNE) that cannot
2301 be expressed in any other way.
2302 .sp
2303 (*SKIP)
2304 .sp
2305 This verb is like (*PRUNE), except that if the pattern is unanchored, the
2306 "bumpalong" advance is not to the next character, but to the position in the
2307 subject where (*SKIP) was encountered. (*SKIP) signifies that whatever text
2308 was matched leading up to it cannot be part of a successful match. Consider:
2309 .sp
2310 a+(*SKIP)b
2311 .sp
2312 If the subject is "aaaac...", after the first match attempt fails (starting at
2313 the first character in the string), the starting point skips on to start the
2314 next attempt at "c". Note that a possessive quantifer does not have the same
2315 effect in this example; although it would suppress backtracking during the
2316 first match attempt, the second attempt would start at the second character
2317 instead of skipping on to "c".
2318 .sp
2319 (*THEN)
2320 .sp
2321 This verb causes a skip to the next alternation if the rest of the pattern does
2322 not match. That is, it cancels pending backtracking, but only within the
2323 current alternation. Its name comes from the observation that it can be used
2324 for a pattern-based if-then-else block:
2325 .sp
2326 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2327 .sp
2328 If the COND1 pattern matches, FOO is tried (and possibly further items after
2329 the end of the group if FOO succeeds); on failure the matcher skips to the
2330 second alternative and tries COND2, without backtracking into COND1. If (*THEN)
2331 is used outside of any alternation, it acts exactly like (*PRUNE).
2332 .
2333 .
2334 .SH "SEE ALSO"
2335 .rs
2336 .sp
2337 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3), \fBpcre\fP(3).
2338 .
2339 .
2341 .rs
2342 .sp
2343 .nf
2344 Philip Hazel
2345 University Computing Service
2346 Cambridge CB2 3QH, England.
2347 .fi
2348 .
2349 .
2351 .rs
2352 .sp
2353 .nf
2354 Last updated: 22 September 2009
2355 Copyright (c) 1997-2009 University of Cambridge.
2356 .fi


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

  ViewVC Help
Powered by ViewVC 1.1.5