ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 394 - (show annotations)
Wed Mar 18 16:38:23 2009 UTC (12 years, 7 months ago) by ph10
File size: 91652 byte(s)
Error occurred while calculating annotation data.
Add words about \b in UTF-8 mode.
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. How this affects pattern matching is mentioned in several
27 places below. There is also a summary of UTF-8 features in the
28 .\" HTML <a href="pcre.html#utf8support">
29 .\" </a>
30 section on UTF-8 support
31 .\"
32 in the main
33 .\" HREF
34 \fBpcre\fP
35 .\"
36 page.
37 .P
38 The remainder of this document discusses the patterns that are supported by
39 PCRE when its main matching function, \fBpcre_exec()\fP, is used.
40 From release 6.0, PCRE offers a second matching function,
41 \fBpcre_dfa_exec()\fP, which matches using a different algorithm that is not
42 Perl-compatible. Some of the features discussed below are not available when
43 \fBpcre_dfa_exec()\fP is used. The advantages and disadvantages of the
44 alternative function, and how it differs from the normal function, are
45 discussed in the
46 .\" HREF
47 \fBpcrematching\fP
48 .\"
49 page.
50 .
51 .
53 .rs
54 .sp
55 PCRE supports five different conventions for indicating line breaks in
56 strings: a single CR (carriage return) character, a single LF (linefeed)
57 character, the two-character sequence CRLF, any of the three preceding, or any
58 Unicode newline sequence. The
59 .\" HREF
60 \fBpcreapi\fP
61 .\"
62 page has
63 .\" HTML <a href="pcreapi.html#newlines">
64 .\" </a>
65 further discussion
66 .\"
67 about newlines, and shows how to set the newline convention in the
68 \fIoptions\fP arguments for the compiling and matching functions.
69 .P
70 It is also possible to specify a newline convention by starting a pattern
71 string with one of the following five sequences:
72 .sp
73 (*CR) carriage return
74 (*LF) linefeed
75 (*CRLF) carriage return, followed by linefeed
76 (*ANYCRLF) any of the three above
77 (*ANY) all Unicode newline sequences
78 .sp
79 These override the default and the options given to \fBpcre_compile()\fP. For
80 example, on a Unix system where LF is the default newline sequence, the pattern
81 .sp
82 (*CR)a.b
83 .sp
84 changes the convention to CR. That pattern matches "a\enb" because LF is no
85 longer a newline. Note that these special settings, which are not
86 Perl-compatible, are recognized only at the very start of a pattern, and that
87 they must be in upper case. If more than one of them is present, the last one
88 is used.
89 .P
90 The newline convention does not affect what the \eR escape sequence matches. By
91 default, this is any Unicode newline sequence, for Perl compatibility. However,
92 this can be changed; see the description of \eR in the section entitled
93 .\" HTML <a href="#newlineseq">
94 .\" </a>
95 "Newline sequences"
96 .\"
97 below. A change of \eR setting can be combined with a change of newline
98 convention.
99 .
100 .
102 .rs
103 .sp
104 A regular expression is a pattern that is matched against a subject string from
105 left to right. Most characters stand for themselves in a pattern, and match the
106 corresponding characters in the subject. As a trivial example, the pattern
107 .sp
108 The quick brown fox
109 .sp
110 matches a portion of a subject string that is identical to itself. When
111 caseless matching is specified (the PCRE_CASELESS option), letters are matched
112 independently of case. In UTF-8 mode, PCRE always understands the concept of
113 case for characters whose values are less than 128, so caseless matching is
114 always possible. For characters with higher values, the concept of case is
115 supported if PCRE is compiled with Unicode property support, but not otherwise.
116 If you want to use caseless matching for characters 128 and above, you must
117 ensure that PCRE is compiled with Unicode property support as well as with
118 UTF-8 support.
119 .P
120 The power of regular expressions comes from the ability to include alternatives
121 and repetitions in the pattern. These are encoded in the pattern by the use of
122 \fImetacharacters\fP, which do not stand for themselves but instead are
123 interpreted in some special way.
124 .P
125 There are two different sets of metacharacters: those that are recognized
126 anywhere in the pattern except within square brackets, and those that are
127 recognized within square brackets. Outside square brackets, the metacharacters
128 are as follows:
129 .sp
130 \e general escape character with several uses
131 ^ assert start of string (or line, in multiline mode)
132 $ assert end of string (or line, in multiline mode)
133 . match any character except newline (by default)
134 [ start character class definition
135 | start of alternative branch
136 ( start subpattern
137 ) end subpattern
138 ? extends the meaning of (
139 also 0 or 1 quantifier
140 also quantifier minimizer
141 * 0 or more quantifier
142 + 1 or more quantifier
143 also "possessive quantifier"
144 { start min/max quantifier
145 .sp
146 Part of a pattern that is in square brackets is called a "character class". In
147 a character class the only metacharacters are:
148 .sp
149 \e general escape character
150 ^ negate the class, but only if the first character
151 - indicates character range
152 .\" JOIN
153 [ POSIX character class (only if followed by POSIX
154 syntax)
155 ] terminates the character class
156 .sp
157 The following sections describe the use of each of the metacharacters.
158 .
159 .
161 .rs
162 .sp
163 The backslash character has several uses. Firstly, if it is followed by a
164 non-alphanumeric character, it takes away any special meaning that character
165 may have. This use of backslash as an escape character applies both inside and
166 outside character classes.
167 .P
168 For example, if you want to match a * character, you write \e* in the pattern.
169 This escaping action applies whether or not the following character would
170 otherwise be interpreted as a metacharacter, so it is always safe to precede a
171 non-alphanumeric with backslash to specify that it stands for itself. In
172 particular, if you want to match a backslash, you write \e\e.
173 .P
174 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
175 pattern (other than in a character class) and characters between a # outside
176 a character class and the next newline are ignored. An escaping backslash can
177 be used to include a whitespace or # character as part of the pattern.
178 .P
179 If you want to remove the special meaning from a sequence of characters, you
180 can do so by putting them between \eQ and \eE. This is different from Perl in
181 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
182 Perl, $ and @ cause variable interpolation. Note the following examples:
183 .sp
184 Pattern PCRE matches Perl matches
185 .sp
186 .\" JOIN
187 \eQabc$xyz\eE abc$xyz abc followed by the
188 contents of $xyz
189 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
190 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
191 .sp
192 The \eQ...\eE sequence is recognized both inside and outside character classes.
193 .
194 .
195 .\" HTML <a name="digitsafterbackslash"></a>
196 .SS "Non-printing characters"
197 .rs
198 .sp
199 A second use of backslash provides a way of encoding non-printing characters
200 in patterns in a visible manner. There is no restriction on the appearance of
201 non-printing characters, apart from the binary zero that terminates a pattern,
202 but when a pattern is being prepared by text editing, it is usually easier to
203 use one of the following escape sequences than the binary character it
204 represents:
205 .sp
206 \ea alarm, that is, the BEL character (hex 07)
207 \ecx "control-x", where x is any character
208 \ee escape (hex 1B)
209 \ef formfeed (hex 0C)
210 \en linefeed (hex 0A)
211 \er carriage return (hex 0D)
212 \et tab (hex 09)
213 \eddd character with octal code ddd, or backreference
214 \exhh character with hex code hh
215 \ex{hhh..} character with hex code hhh..
216 .sp
217 The precise effect of \ecx is as follows: if x is a lower case letter, it
218 is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
219 Thus \ecz becomes hex 1A, but \ec{ becomes hex 3B, while \ec; becomes hex
220 7B.
221 .P
222 After \ex, from zero to two hexadecimal digits are read (letters can be in
223 upper or lower case). Any number of hexadecimal digits may appear between \ex{
224 and }, but the value of the character code must be less than 256 in non-UTF-8
225 mode, and less than 2**31 in UTF-8 mode. That is, the maximum value in
226 hexadecimal is 7FFFFFFF. Note that this is bigger than the largest Unicode code
227 point, which is 10FFFF.
228 .P
229 If characters other than hexadecimal digits appear between \ex{ and }, or if
230 there is no terminating }, this form of escape is not recognized. Instead, the
231 initial \ex will be interpreted as a basic hexadecimal escape, with no
232 following digits, giving a character whose value is zero.
233 .P
234 Characters whose value is less than 256 can be defined by either of the two
235 syntaxes for \ex. There is no difference in the way they are handled. For
236 example, \exdc is exactly the same as \ex{dc}.
237 .P
238 After \e0 up to two further octal digits are read. If there are fewer than two
239 digits, just those that are present are used. Thus the sequence \e0\ex\e07
240 specifies two binary zeros followed by a BEL character (code value 7). Make
241 sure you supply two digits after the initial zero if the pattern character that
242 follows is itself an octal digit.
243 .P
244 The handling of a backslash followed by a digit other than 0 is complicated.
245 Outside a character class, PCRE reads it and any following digits as a decimal
246 number. If the number is less than 10, or if there have been at least that many
247 previous capturing left parentheses in the expression, the entire sequence is
248 taken as a \fIback reference\fP. A description of how this works is given
249 .\" HTML <a href="#backreferences">
250 .\" </a>
251 later,
252 .\"
253 following the discussion of
254 .\" HTML <a href="#subpattern">
255 .\" </a>
256 parenthesized subpatterns.
257 .\"
258 .P
259 Inside a character class, or if the decimal number is greater than 9 and there
260 have not been that many capturing subpatterns, PCRE re-reads up to three octal
261 digits following the backslash, and uses them to generate a data character. Any
262 subsequent digits stand for themselves. In non-UTF-8 mode, the value of a
263 character specified in octal must be less than \e400. In UTF-8 mode, values up
264 to \e777 are permitted. For example:
265 .sp
266 \e040 is another way of writing a space
267 .\" JOIN
268 \e40 is the same, provided there are fewer than 40
269 previous capturing subpatterns
270 \e7 is always a back reference
271 .\" JOIN
272 \e11 might be a back reference, or another way of
273 writing a tab
274 \e011 is always a tab
275 \e0113 is a tab followed by the character "3"
276 .\" JOIN
277 \e113 might be a back reference, otherwise the
278 character with octal code 113
279 .\" JOIN
280 \e377 might be a back reference, otherwise
281 the byte consisting entirely of 1 bits
282 .\" JOIN
283 \e81 is either a back reference, or a binary zero
284 followed by the two characters "8" and "1"
285 .sp
286 Note that octal values of 100 or greater must not be introduced by a leading
287 zero, because no more than three octal digits are ever read.
288 .P
289 All the sequences that define a single character value can be used both inside
290 and outside character classes. In addition, inside a character class, the
291 sequence \eb is interpreted as the backspace character (hex 08), and the
292 sequences \eR and \eX are interpreted as the characters "R" and "X",
293 respectively. Outside a character class, these sequences have different
294 meanings
295 .\" HTML <a href="#uniextseq">
296 .\" </a>
297 (see below).
298 .\"
299 .
300 .
301 .SS "Absolute and relative back references"
302 .rs
303 .sp
304 The sequence \eg followed by an unsigned or a negative number, optionally
305 enclosed in braces, is an absolute or relative back reference. A named back
306 reference can be coded as \eg{name}. Back references are discussed
307 .\" HTML <a href="#backreferences">
308 .\" </a>
309 later,
310 .\"
311 following the discussion of
312 .\" HTML <a href="#subpattern">
313 .\" </a>
314 parenthesized subpatterns.
315 .\"
316 .
317 .
318 .SS "Absolute and relative subroutine calls"
319 .rs
320 .sp
321 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
322 a number enclosed either in angle brackets or single quotes, is an alternative
323 syntax for referencing a subpattern as a "subroutine". Details are discussed
324 .\" HTML <a href="#onigurumasubroutines">
325 .\" </a>
326 later.
327 .\"
328 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
329 synonymous. The former is a back reference; the latter is a subroutine call.
330 .
331 .
332 .SS "Generic character types"
333 .rs
334 .sp
335 Another use of backslash is for specifying generic character types. The
336 following are always recognized:
337 .sp
338 \ed any decimal digit
339 \eD any character that is not a decimal digit
340 \eh any horizontal whitespace character
341 \eH any character that is not a horizontal whitespace character
342 \es any whitespace character
343 \eS any character that is not a whitespace character
344 \ev any vertical whitespace character
345 \eV any character that is not a vertical whitespace character
346 \ew any "word" character
347 \eW any "non-word" character
348 .sp
349 Each pair of escape sequences partitions the complete set of characters into
350 two disjoint sets. Any given character matches one, and only one, of each pair.
351 .P
352 These character type sequences can appear both inside and outside character
353 classes. They each match one character of the appropriate type. If the current
354 matching point is at the end of the subject string, all of them fail, since
355 there is no character to match.
356 .P
357 For compatibility with Perl, \es does not match the VT character (code 11).
358 This makes it different from the the POSIX "space" class. The \es characters
359 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
360 included in a Perl script, \es may match the VT character. In PCRE, it never
361 does.
362 .P
363 In UTF-8 mode, characters with values greater than 128 never match \ed, \es, or
364 \ew, and always match \eD, \eS, and \eW. This is true even when Unicode
365 character property support is available. These sequences retain their original
366 meanings from before UTF-8 support was available, mainly for efficiency
367 reasons. Note that this also affects \eb, because it is defined in terms of \ew
368 and \eW.
369 .P
370 The sequences \eh, \eH, \ev, and \eV are Perl 5.10 features. In contrast to the
371 other sequences, these do match certain high-valued codepoints in UTF-8 mode.
372 The horizontal space characters are:
373 .sp
374 U+0009 Horizontal tab
375 U+0020 Space
376 U+00A0 Non-break space
377 U+1680 Ogham space mark
378 U+180E Mongolian vowel separator
379 U+2000 En quad
380 U+2001 Em quad
381 U+2002 En space
382 U+2003 Em space
383 U+2004 Three-per-em space
384 U+2005 Four-per-em space
385 U+2006 Six-per-em space
386 U+2007 Figure space
387 U+2008 Punctuation space
388 U+2009 Thin space
389 U+200A Hair space
390 U+202F Narrow no-break space
391 U+205F Medium mathematical space
392 U+3000 Ideographic space
393 .sp
394 The vertical space characters are:
395 .sp
396 U+000A Linefeed
397 U+000B Vertical tab
398 U+000C Formfeed
399 U+000D Carriage return
400 U+0085 Next line
401 U+2028 Line separator
402 U+2029 Paragraph separator
403 .P
404 A "word" character is an underscore or any character less than 256 that is a
405 letter or digit. The definition of letters and digits is controlled by PCRE's
406 low-valued character tables, and may vary if locale-specific matching is taking
407 place (see
408 .\" HTML <a href="pcreapi.html#localesupport">
409 .\" </a>
410 "Locale support"
411 .\"
412 in the
413 .\" HREF
414 \fBpcreapi\fP
415 .\"
416 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
417 or "french" in Windows, some character codes greater than 128 are used for
418 accented letters, and these are matched by \ew. The use of locales with Unicode
419 is discouraged.
420 .
421 .
422 .\" HTML <a name="newlineseq"></a>
423 .SS "Newline sequences"
424 .rs
425 .sp
426 Outside a character class, by default, the escape sequence \eR matches any
427 Unicode newline sequence. This is a Perl 5.10 feature. In non-UTF-8 mode \eR is
428 equivalent to the following:
429 .sp
430 (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
431 .sp
432 This is an example of an "atomic group", details of which are given
433 .\" HTML <a href="#atomicgroup">
434 .\" </a>
435 below.
436 .\"
437 This particular group matches either the two-character sequence CR followed by
438 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
439 U+000B), FF (formfeed, U+000C), CR (carriage return, U+000D), or NEL (next
440 line, U+0085). The two-character sequence is treated as a single unit that
441 cannot be split.
442 .P
443 In UTF-8 mode, two additional characters whose codepoints are greater than 255
444 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
445 Unicode character property support is not needed for these characters to be
446 recognized.
447 .P
448 It is possible to restrict \eR to match only CR, LF, or CRLF (instead of the
449 complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
450 either at compile time or when the pattern is matched. (BSR is an abbrevation
451 for "backslash R".) This can be made the default when PCRE is built; if this is
452 the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
453 It is also possible to specify these settings by starting a pattern string with
454 one of the following sequences:
455 .sp
456 (*BSR_ANYCRLF) CR, LF, or CRLF only
457 (*BSR_UNICODE) any Unicode newline sequence
458 .sp
459 These override the default and the options given to \fBpcre_compile()\fP, but
460 they can be overridden by options given to \fBpcre_exec()\fP. Note that these
461 special settings, which are not Perl-compatible, are recognized only at the
462 very start of a pattern, and that they must be in upper case. If more than one
463 of them is present, the last one is used. They can be combined with a change of
464 newline convention, for example, a pattern can start with:
465 .sp
467 .sp
468 Inside a character class, \eR matches the letter "R".
469 .
470 .
471 .\" HTML <a name="uniextseq"></a>
472 .SS Unicode character properties
473 .rs
474 .sp
475 When PCRE is built with Unicode character property support, three additional
476 escape sequences that match characters with specific properties are available.
477 When not in UTF-8 mode, these sequences are of course limited to testing
478 characters whose codepoints are less than 256, but they do work in this mode.
479 The extra escape sequences are:
480 .sp
481 \ep{\fIxx\fP} a character with the \fIxx\fP property
482 \eP{\fIxx\fP} a character without the \fIxx\fP property
483 \eX an extended Unicode sequence
484 .sp
485 The property names represented by \fIxx\fP above are limited to the Unicode
486 script names, the general category properties, and "Any", which matches any
487 character (including newline). Other properties such as "InMusicalSymbols" are
488 not currently supported by PCRE. Note that \eP{Any} does not match any
489 characters, so always causes a match failure.
490 .P
491 Sets of Unicode characters are defined as belonging to certain scripts. A
492 character from one of these sets can be matched using a script name. For
493 example:
494 .sp
495 \ep{Greek}
496 \eP{Han}
497 .sp
498 Those that are not part of an identified script are lumped together as
499 "Common". The current list of scripts is:
500 .P
501 Arabic,
502 Armenian,
503 Balinese,
504 Bengali,
505 Bopomofo,
506 Braille,
507 Buginese,
508 Buhid,
509 Canadian_Aboriginal,
510 Cherokee,
511 Common,
512 Coptic,
513 Cuneiform,
514 Cypriot,
515 Cyrillic,
516 Deseret,
517 Devanagari,
518 Ethiopic,
519 Georgian,
520 Glagolitic,
521 Gothic,
522 Greek,
523 Gujarati,
524 Gurmukhi,
525 Han,
526 Hangul,
527 Hanunoo,
528 Hebrew,
529 Hiragana,
530 Inherited,
531 Kannada,
532 Katakana,
533 Kharoshthi,
534 Khmer,
535 Lao,
536 Latin,
537 Limbu,
538 Linear_B,
539 Malayalam,
540 Mongolian,
541 Myanmar,
542 New_Tai_Lue,
543 Nko,
544 Ogham,
545 Old_Italic,
546 Old_Persian,
547 Oriya,
548 Osmanya,
549 Phags_Pa,
550 Phoenician,
551 Runic,
552 Shavian,
553 Sinhala,
554 Syloti_Nagri,
555 Syriac,
556 Tagalog,
557 Tagbanwa,
558 Tai_Le,
559 Tamil,
560 Telugu,
561 Thaana,
562 Thai,
563 Tibetan,
564 Tifinagh,
565 Ugaritic,
566 Yi.
567 .P
568 Each character has exactly one general category property, specified by a
569 two-letter abbreviation. For compatibility with Perl, negation can be specified
570 by including a circumflex between the opening brace and the property name. For
571 example, \ep{^Lu} is the same as \eP{Lu}.
572 .P
573 If only one letter is specified with \ep or \eP, it includes all the general
574 category properties that start with that letter. In this case, in the absence
575 of negation, the curly brackets in the escape sequence are optional; these two
576 examples have the same effect:
577 .sp
578 \ep{L}
579 \epL
580 .sp
581 The following general category property codes are supported:
582 .sp
583 C Other
584 Cc Control
585 Cf Format
586 Cn Unassigned
587 Co Private use
588 Cs Surrogate
589 .sp
590 L Letter
591 Ll Lower case letter
592 Lm Modifier letter
593 Lo Other letter
594 Lt Title case letter
595 Lu Upper case letter
596 .sp
597 M Mark
598 Mc Spacing mark
599 Me Enclosing mark
600 Mn Non-spacing mark
601 .sp
602 N Number
603 Nd Decimal number
604 Nl Letter number
605 No Other number
606 .sp
607 P Punctuation
608 Pc Connector punctuation
609 Pd Dash punctuation
610 Pe Close punctuation
611 Pf Final punctuation
612 Pi Initial punctuation
613 Po Other punctuation
614 Ps Open punctuation
615 .sp
616 S Symbol
617 Sc Currency symbol
618 Sk Modifier symbol
619 Sm Mathematical symbol
620 So Other symbol
621 .sp
622 Z Separator
623 Zl Line separator
624 Zp Paragraph separator
625 Zs Space separator
626 .sp
627 The special property L& is also supported: it matches a character that has
628 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
629 a modifier or "other".
630 .P
631 The Cs (Surrogate) property applies only to characters in the range U+D800 to
632 U+DFFF. Such characters are not valid in UTF-8 strings (see RFC 3629) and so
633 cannot be tested by PCRE, unless UTF-8 validity checking has been turned off
634 (see the discussion of PCRE_NO_UTF8_CHECK in the
635 .\" HREF
636 \fBpcreapi\fP
637 .\"
638 page).
639 .P
640 The long synonyms for these properties that Perl supports (such as \ep{Letter})
641 are not supported by PCRE, nor is it permitted to prefix any of these
642 properties with "Is".
643 .P
644 No character that is in the Unicode table has the Cn (unassigned) property.
645 Instead, this property is assumed for any code point that is not in the
646 Unicode table.
647 .P
648 Specifying caseless matching does not affect these escape sequences. For
649 example, \ep{Lu} always matches only upper case letters.
650 .P
651 The \eX escape matches any number of Unicode characters that form an extended
652 Unicode sequence. \eX is equivalent to
653 .sp
654 (?>\ePM\epM*)
655 .sp
656 That is, it matches a character without the "mark" property, followed by zero
657 or more characters with the "mark" property, and treats the sequence as an
658 atomic group
659 .\" HTML <a href="#atomicgroup">
660 .\" </a>
661 (see below).
662 .\"
663 Characters with the "mark" property are typically accents that affect the
664 preceding character. None of them have codepoints less than 256, so in
665 non-UTF-8 mode \eX matches any one character.
666 .P
667 Matching characters by Unicode property is not fast, because PCRE has to search
668 a structure that contains data for over fifteen thousand characters. That is
669 why the traditional escape sequences such as \ed and \ew do not use Unicode
670 properties in PCRE.
671 .
672 .
673 .\" HTML <a name="resetmatchstart"></a>
674 .SS "Resetting the match start"
675 .rs
676 .sp
677 The escape sequence \eK, which is a Perl 5.10 feature, causes any previously
678 matched characters not to be included in the final matched sequence. For
679 example, the pattern:
680 .sp
681 foo\eKbar
682 .sp
683 matches "foobar", but reports that it has matched "bar". This feature is
684 similar to a lookbehind assertion
685 .\" HTML <a href="#lookbehind">
686 .\" </a>
687 (described below).
688 .\"
689 However, in this case, the part of the subject before the real match does not
690 have to be of fixed length, as lookbehind assertions do. The use of \eK does
691 not interfere with the setting of
692 .\" HTML <a href="#subpattern">
693 .\" </a>
694 captured substrings.
695 .\"
696 For example, when the pattern
697 .sp
698 (foo)\eKbar
699 .sp
700 matches "foobar", the first substring is still set to "foo".
701 .
702 .
703 .\" HTML <a name="smallassertions"></a>
704 .SS "Simple assertions"
705 .rs
706 .sp
707 The final use of backslash is for certain simple assertions. An assertion
708 specifies a condition that has to be met at a particular point in a match,
709 without consuming any characters from the subject string. The use of
710 subpatterns for more complicated assertions is described
711 .\" HTML <a href="#bigassertions">
712 .\" </a>
713 below.
714 .\"
715 The backslashed assertions are:
716 .sp
717 \eb matches at a word boundary
718 \eB matches when not at a word boundary
719 \eA matches at the start of the subject
720 \eZ matches at the end of the subject
721 also matches before a newline at the end of the subject
722 \ez matches only at the end of the subject
723 \eG matches at the first matching position in the subject
724 .sp
725 These assertions may not appear in character classes (but note that \eb has a
726 different meaning, namely the backspace character, inside a character class).
727 .P
728 A word boundary is a position in the subject string where the current character
729 and the previous character do not both match \ew or \eW (i.e. one matches
730 \ew and the other matches \eW), or the start or end of the string if the
731 first or last character matches \ew, respectively.
732 .P
733 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
734 dollar (described in the next section) in that they only ever match at the very
735 start and end of the subject string, whatever options are set. Thus, they are
736 independent of multiline mode. These three assertions are not affected by the
737 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
738 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
739 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
740 at a point other than the beginning of the subject, \eA can never match. The
741 difference between \eZ and \ez is that \eZ matches before a newline at the end
742 of the string as well as at the very end, whereas \ez matches only at the end.
743 .P
744 The \eG assertion is true only when the current matching position is at the
745 start point of the match, as specified by the \fIstartoffset\fP argument of
746 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
747 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
748 arguments, you can mimic Perl's /g option, and it is in this kind of
749 implementation where \eG can be useful.
750 .P
751 Note, however, that PCRE's interpretation of \eG, as the start of the current
752 match, is subtly different from Perl's, which defines it as the end of the
753 previous match. In Perl, these can be different when the previously matched
754 string was empty. Because PCRE does just one match at a time, it cannot
755 reproduce this behaviour.
756 .P
757 If all the alternatives of a pattern begin with \eG, the expression is anchored
758 to the starting match position, and the "anchored" flag is set in the compiled
759 regular expression.
760 .
761 .
763 .rs
764 .sp
765 Outside a character class, in the default matching mode, the circumflex
766 character is an assertion that is true only if the current matching point is
767 at the start of the subject string. If the \fIstartoffset\fP argument of
768 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
769 option is unset. Inside a character class, circumflex has an entirely different
770 meaning
771 .\" HTML <a href="#characterclass">
772 .\" </a>
773 (see below).
774 .\"
775 .P
776 Circumflex need not be the first character of the pattern if a number of
777 alternatives are involved, but it should be the first thing in each alternative
778 in which it appears if the pattern is ever to match that branch. If all
779 possible alternatives start with a circumflex, that is, if the pattern is
780 constrained to match only at the start of the subject, it is said to be an
781 "anchored" pattern. (There are also other constructs that can cause a pattern
782 to be anchored.)
783 .P
784 A dollar character is an assertion that is true only if the current matching
785 point is at the end of the subject string, or immediately before a newline
786 at the end of the string (by default). Dollar need not be the last character of
787 the pattern if a number of alternatives are involved, but it should be the last
788 item in any branch in which it appears. Dollar has no special meaning in a
789 character class.
790 .P
791 The meaning of dollar can be changed so that it matches only at the very end of
792 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
793 does not affect the \eZ assertion.
794 .P
795 The meanings of the circumflex and dollar characters are changed if the
796 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
797 immediately after internal newlines as well as at the start of the subject
798 string. It does not match after a newline that ends the string. A dollar
799 matches before any newlines in the string, as well as at the very end, when
800 PCRE_MULTILINE is set. When newline is specified as the two-character
801 sequence CRLF, isolated CR and LF characters do not indicate newlines.
802 .P
803 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
804 \en represents a newline) in multiline mode, but not otherwise. Consequently,
805 patterns that are anchored in single line mode because all branches start with
806 ^ are not anchored in multiline mode, and a match for circumflex is possible
807 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
808 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
809 .P
810 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
811 end of the subject in both modes, and if all branches of a pattern start with
812 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
813 .
814 .
816 .rs
817 .sp
818 Outside a character class, a dot in the pattern matches any one character in
819 the subject string except (by default) a character that signifies the end of a
820 line. In UTF-8 mode, the matched character may be more than one byte long.
821 .P
822 When a line ending is defined as a single character, dot never matches that
823 character; when the two-character sequence CRLF is used, dot does not match CR
824 if it is immediately followed by LF, but otherwise it matches all characters
825 (including isolated CRs and LFs). When any Unicode line endings are being
826 recognized, dot does not match CR or LF or any of the other line ending
827 characters.
828 .P
829 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
830 option is set, a dot matches any one character, without exception. If the
831 two-character sequence CRLF is present in the subject string, it takes two dots
832 to match it.
833 .P
834 The handling of dot is entirely independent of the handling of circumflex and
835 dollar, the only relationship being that they both involve newlines. Dot has no
836 special meaning in a character class.
837 .
838 .
840 .rs
841 .sp
842 Outside a character class, the escape sequence \eC matches any one byte, both
843 in and out of UTF-8 mode. Unlike a dot, it always matches any line-ending
844 characters. The feature is provided in Perl in order to match individual bytes
845 in UTF-8 mode. Because it breaks up UTF-8 characters into individual bytes,
846 what remains in the string may be a malformed UTF-8 string. For this reason,
847 the \eC escape sequence is best avoided.
848 .P
849 PCRE does not allow \eC to appear in lookbehind assertions
850 .\" HTML <a href="#lookbehind">
851 .\" </a>
852 (described below),
853 .\"
854 because in UTF-8 mode this would make it impossible to calculate the length of
855 the lookbehind.
856 .
857 .
858 .\" HTML <a name="characterclass"></a>
860 .rs
861 .sp
862 An opening square bracket introduces a character class, terminated by a closing
863 square bracket. A closing square bracket on its own is not special. If a
864 closing square bracket is required as a member of the class, it should be the
865 first data character in the class (after an initial circumflex, if present) or
866 escaped with a backslash.
867 .P
868 A character class matches a single character in the subject. In UTF-8 mode, the
869 character may occupy more than one byte. A matched character must be in the set
870 of characters defined by the class, unless the first character in the class
871 definition is a circumflex, in which case the subject character must not be in
872 the set defined by the class. If a circumflex is actually required as a member
873 of the class, ensure it is not the first character, or escape it with a
874 backslash.
875 .P
876 For example, the character class [aeiou] matches any lower case vowel, while
877 [^aeiou] matches any character that is not a lower case vowel. Note that a
878 circumflex is just a convenient notation for specifying the characters that
879 are in the class by enumerating those that are not. A class that starts with a
880 circumflex is not an assertion: it still consumes a character from the subject
881 string, and therefore it fails if the current pointer is at the end of the
882 string.
883 .P
884 In UTF-8 mode, characters with values greater than 255 can be included in a
885 class as a literal string of bytes, or by using the \ex{ escaping mechanism.
886 .P
887 When caseless matching is set, any letters in a class represent both their
888 upper case and lower case versions, so for example, a caseless [aeiou] matches
889 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
890 caseful version would. In UTF-8 mode, PCRE always understands the concept of
891 case for characters whose values are less than 128, so caseless matching is
892 always possible. For characters with higher values, the concept of case is
893 supported if PCRE is compiled with Unicode property support, but not otherwise.
894 If you want to use caseless matching for characters 128 and above, you must
895 ensure that PCRE is compiled with Unicode property support as well as with
896 UTF-8 support.
897 .P
898 Characters that might indicate line breaks are never treated in any special way
899 when matching character classes, whatever line-ending sequence is in use, and
900 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
901 such as [^a] always matches one of these characters.
902 .P
903 The minus (hyphen) character can be used to specify a range of characters in a
904 character class. For example, [d-m] matches any letter between d and m,
905 inclusive. If a minus character is required in a class, it must be escaped with
906 a backslash or appear in a position where it cannot be interpreted as
907 indicating a range, typically as the first or last character in the class.
908 .P
909 It is not possible to have the literal character "]" as the end character of a
910 range. A pattern such as [W-]46] is interpreted as a class of two characters
911 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
912 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
913 the end of range, so [W-\e]46] is interpreted as a class containing a range
914 followed by two other characters. The octal or hexadecimal representation of
915 "]" can also be used to end a range.
916 .P
917 Ranges operate in the collating sequence of character values. They can also be
918 used for characters specified numerically, for example [\e000-\e037]. In UTF-8
919 mode, ranges can include characters whose values are greater than 255, for
920 example [\ex{100}-\ex{2ff}].
921 .P
922 If a range that includes letters is used when caseless matching is set, it
923 matches the letters in either case. For example, [W-c] is equivalent to
924 [][\e\e^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
925 tables for a French locale are in use, [\exc8-\excb] matches accented E
926 characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
927 characters with values greater than 128 only when it is compiled with Unicode
928 property support.
929 .P
930 The character types \ed, \eD, \ep, \eP, \es, \eS, \ew, and \eW may also appear
931 in a character class, and add the characters that they match to the class. For
932 example, [\edABCDEF] matches any hexadecimal digit. A circumflex can
933 conveniently be used with the upper case character types to specify a more
934 restricted set of characters than the matching lower case type. For example,
935 the class [^\eW_] matches any letter or digit, but not underscore.
936 .P
937 The only metacharacters that are recognized in character classes are backslash,
938 hyphen (only where it can be interpreted as specifying a range), circumflex
939 (only at the start), opening square bracket (only when it can be interpreted as
940 introducing a POSIX class name - see the next section), and the terminating
941 closing square bracket. However, escaping other non-alphanumeric characters
942 does no harm.
943 .
944 .
946 .rs
947 .sp
948 Perl supports the POSIX notation for character classes. This uses names
949 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
950 this notation. For example,
951 .sp
952 [01[:alpha:]%]
953 .sp
954 matches "0", "1", any alphabetic character, or "%". The supported class names
955 are
956 .sp
957 alnum letters and digits
958 alpha letters
959 ascii character codes 0 - 127
960 blank space or tab only
961 cntrl control characters
962 digit decimal digits (same as \ed)
963 graph printing characters, excluding space
964 lower lower case letters
965 print printing characters, including space
966 punct printing characters, excluding letters and digits
967 space white space (not quite the same as \es)
968 upper upper case letters
969 word "word" characters (same as \ew)
970 xdigit hexadecimal digits
971 .sp
972 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
973 space (32). Notice that this list includes the VT character (code 11). This
974 makes "space" different to \es, which does not include VT (for Perl
975 compatibility).
976 .P
977 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
978 5.8. Another Perl extension is negation, which is indicated by a ^ character
979 after the colon. For example,
980 .sp
981 [12[:^digit:]]
982 .sp
983 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
984 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
985 supported, and an error is given if they are encountered.
986 .P
987 In UTF-8 mode, characters with values greater than 128 do not match any of
988 the POSIX character classes.
989 .
990 .
992 .rs
993 .sp
994 Vertical bar characters are used to separate alternative patterns. For example,
995 the pattern
996 .sp
997 gilbert|sullivan
998 .sp
999 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1000 and an empty alternative is permitted (matching the empty string). The matching
1001 process tries each alternative in turn, from left to right, and the first one
1002 that succeeds is used. If the alternatives are within a subpattern
1003 .\" HTML <a href="#subpattern">
1004 .\" </a>
1005 (defined below),
1006 .\"
1007 "succeeds" means matching the rest of the main pattern as well as the
1008 alternative in the subpattern.
1009 .
1010 .
1012 .rs
1013 .sp
1014 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1015 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1016 the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1017 The option letters are
1018 .sp
1019 i for PCRE_CASELESS
1021 s for PCRE_DOTALL
1022 x for PCRE_EXTENDED
1023 .sp
1024 For example, (?im) sets caseless, multiline matching. It is also possible to
1025 unset these options by preceding the letter with a hyphen, and a combined
1026 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1027 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1028 permitted. If a letter appears both before and after the hyphen, the option is
1029 unset.
1030 .P
1031 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1032 changed in the same way as the Perl-compatible options by using the characters
1033 J, U and X respectively.
1034 .P
1035 When an option change occurs at top level (that is, not inside subpattern
1036 parentheses), the change applies to the remainder of the pattern that follows.
1037 If the change is placed right at the start of a pattern, PCRE extracts it into
1038 the global options (and it will therefore show up in data extracted by the
1039 \fBpcre_fullinfo()\fP function).
1040 .P
1041 An option change within a subpattern (see below for a description of
1042 subpatterns) affects only that part of the current pattern that follows it, so
1043 .sp
1044 (a(?i)b)c
1045 .sp
1046 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1047 By this means, options can be made to have different settings in different
1048 parts of the pattern. Any changes made in one alternative do carry on
1049 into subsequent branches within the same subpattern. For example,
1050 .sp
1051 (a(?i)b|c)
1052 .sp
1053 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1054 branch is abandoned before the option setting. This is because the effects of
1055 option settings happen at compile time. There would be some very weird
1056 behaviour otherwise.
1057 .P
1058 \fBNote:\fP There are other PCRE-specific options that can be set by the
1059 application when the compile or match functions are called. In some cases the
1060 pattern can contain special leading sequences to override what the application
1061 has set or what has been defaulted. Details are given in the section entitled
1062 .\" HTML <a href="#newlineseq">
1063 .\" </a>
1064 "Newline sequences"
1065 .\"
1066 above.
1067 .
1068 .
1069 .\" HTML <a name="subpattern"></a>
1071 .rs
1072 .sp
1073 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1074 Turning part of a pattern into a subpattern does two things:
1075 .sp
1076 1. It localizes a set of alternatives. For example, the pattern
1077 .sp
1078 cat(aract|erpillar|)
1079 .sp
1080 matches one of the words "cat", "cataract", or "caterpillar". Without the
1081 parentheses, it would match "cataract", "erpillar" or an empty string.
1082 .sp
1083 2. It sets up the subpattern as a capturing subpattern. This means that, when
1084 the whole pattern matches, that portion of the subject string that matched the
1085 subpattern is passed back to the caller via the \fIovector\fP argument of
1086 \fBpcre_exec()\fP. Opening parentheses are counted from left to right (starting
1087 from 1) to obtain numbers for the capturing subpatterns.
1088 .P
1089 For example, if the string "the red king" is matched against the pattern
1090 .sp
1091 the ((red|white) (king|queen))
1092 .sp
1093 the captured substrings are "red king", "red", and "king", and are numbered 1,
1094 2, and 3, respectively.
1095 .P
1096 The fact that plain parentheses fulfil two functions is not always helpful.
1097 There are often times when a grouping subpattern is required without a
1098 capturing requirement. If an opening parenthesis is followed by a question mark
1099 and a colon, the subpattern does not do any capturing, and is not counted when
1100 computing the number of any subsequent capturing subpatterns. For example, if
1101 the string "the white queen" is matched against the pattern
1102 .sp
1103 the ((?:red|white) (king|queen))
1104 .sp
1105 the captured substrings are "white queen" and "queen", and are numbered 1 and
1106 2. The maximum number of capturing subpatterns is 65535.
1107 .P
1108 As a convenient shorthand, if any option settings are required at the start of
1109 a non-capturing subpattern, the option letters may appear between the "?" and
1110 the ":". Thus the two patterns
1111 .sp
1112 (?i:saturday|sunday)
1113 (?:(?i)saturday|sunday)
1114 .sp
1115 match exactly the same set of strings. Because alternative branches are tried
1116 from left to right, and options are not reset until the end of the subpattern
1117 is reached, an option setting in one branch does affect subsequent branches, so
1118 the above patterns match "SUNDAY" as well as "Saturday".
1119 .
1120 .
1122 .rs
1123 .sp
1124 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1125 the same numbers for its capturing parentheses. Such a subpattern starts with
1126 (?| and is itself a non-capturing subpattern. For example, consider this
1127 pattern:
1128 .sp
1129 (?|(Sat)ur|(Sun))day
1130 .sp
1131 Because the two alternatives are inside a (?| group, both sets of capturing
1132 parentheses are numbered one. Thus, when the pattern matches, you can look
1133 at captured substring number one, whichever alternative matched. This construct
1134 is useful when you want to capture part, but not all, of one of a number of
1135 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1136 number is reset at the start of each branch. The numbers of any capturing
1137 buffers that follow the subpattern start after the highest number used in any
1138 branch. The following example is taken from the Perl documentation.
1139 The numbers underneath show in which buffer the captured content will be
1140 stored.
1141 .sp
1142 # before ---------------branch-reset----------- after
1143 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1144 # 1 2 2 3 2 3 4
1145 .sp
1146 A backreference or a recursive call to a numbered subpattern always refers to
1147 the first one in the pattern with the given number.
1148 .P
1149 An alternative approach to using this "branch reset" feature is to use
1150 duplicate named subpatterns, as described in the next section.
1151 .
1152 .
1154 .rs
1155 .sp
1156 Identifying capturing parentheses by number is simple, but it can be very hard
1157 to keep track of the numbers in complicated regular expressions. Furthermore,
1158 if an expression is modified, the numbers may change. To help with this
1159 difficulty, PCRE supports the naming of subpatterns. This feature was not
1160 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1161 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1162 the Perl and the Python syntax.
1163 .P
1164 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1165 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1166 parentheses from other parts of the pattern, such as
1167 .\" HTML <a href="#backreferences">
1168 .\" </a>
1169 backreferences,
1170 .\"
1171 .\" HTML <a href="#recursion">
1172 .\" </a>
1173 recursion,
1174 .\"
1175 and
1176 .\" HTML <a href="#conditions">
1177 .\" </a>
1178 conditions,
1179 .\"
1180 can be made by name as well as by number.
1181 .P
1182 Names consist of up to 32 alphanumeric characters and underscores. Named
1183 capturing parentheses are still allocated numbers as well as names, exactly as
1184 if the names were not present. The PCRE API provides function calls for
1185 extracting the name-to-number translation table from a compiled pattern. There
1186 is also a convenience function for extracting a captured substring by name.
1187 .P
1188 By default, a name must be unique within a pattern, but it is possible to relax
1189 this constraint by setting the PCRE_DUPNAMES option at compile time. This can
1190 be useful for patterns where only one instance of the named parentheses can
1191 match. Suppose you want to match the name of a weekday, either as a 3-letter
1192 abbreviation or as the full name, and in both cases you want to extract the
1193 abbreviation. This pattern (ignoring the line breaks) does the job:
1194 .sp
1195 (?<DN>Mon|Fri|Sun)(?:day)?|
1196 (?<DN>Tue)(?:sday)?|
1197 (?<DN>Wed)(?:nesday)?|
1198 (?<DN>Thu)(?:rsday)?|
1199 (?<DN>Sat)(?:urday)?
1200 .sp
1201 There are five capturing substrings, but only one is ever set after a match.
1202 (An alternative way of solving this problem is to use a "branch reset"
1203 subpattern, as described in the previous section.)
1204 .P
1205 The convenience function for extracting the data by name returns the substring
1206 for the first (and in this example, the only) subpattern of that name that
1207 matched. This saves searching to find which numbered subpattern it was. If you
1208 make a reference to a non-unique named subpattern from elsewhere in the
1209 pattern, the one that corresponds to the lowest number is used. For further
1210 details of the interfaces for handling named subpatterns, see the
1211 .\" HREF
1212 \fBpcreapi\fP
1213 .\"
1214 documentation.
1215 .P
1216 \fBWarning:\fP You cannot use different names to distinguish between two
1217 subpatterns with the same number (see the previous section) because PCRE uses
1218 only the numbers when matching.
1219 .
1220 .
1222 .rs
1223 .sp
1224 Repetition is specified by quantifiers, which can follow any of the following
1225 items:
1226 .sp
1227 a literal data character
1228 the dot metacharacter
1229 the \eC escape sequence
1230 the \eX escape sequence (in UTF-8 mode with Unicode properties)
1231 the \eR escape sequence
1232 an escape such as \ed that matches a single character
1233 a character class
1234 a back reference (see next section)
1235 a parenthesized subpattern (unless it is an assertion)
1236 .sp
1237 The general repetition quantifier specifies a minimum and maximum number of
1238 permitted matches, by giving the two numbers in curly brackets (braces),
1239 separated by a comma. The numbers must be less than 65536, and the first must
1240 be less than or equal to the second. For example:
1241 .sp
1242 z{2,4}
1243 .sp
1244 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1245 character. If the second number is omitted, but the comma is present, there is
1246 no upper limit; if the second number and the comma are both omitted, the
1247 quantifier specifies an exact number of required matches. Thus
1248 .sp
1249 [aeiou]{3,}
1250 .sp
1251 matches at least 3 successive vowels, but may match many more, while
1252 .sp
1253 \ed{8}
1254 .sp
1255 matches exactly 8 digits. An opening curly bracket that appears in a position
1256 where a quantifier is not allowed, or one that does not match the syntax of a
1257 quantifier, is taken as a literal character. For example, {,6} is not a
1258 quantifier, but a literal string of four characters.
1259 .P
1260 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
1261 bytes. Thus, for example, \ex{100}{2} matches two UTF-8 characters, each of
1262 which is represented by a two-byte sequence. Similarly, when Unicode property
1263 support is available, \eX{3} matches three Unicode extended sequences, each of
1264 which may be several bytes long (and they may be of different lengths).
1265 .P
1266 The quantifier {0} is permitted, causing the expression to behave as if the
1267 previous item and the quantifier were not present. This may be useful for
1268 subpatterns that are referenced as
1269 .\" HTML <a href="#subpatternsassubroutines">
1270 .\" </a>
1271 subroutines
1272 .\"
1273 from elsewhere in the pattern. Items other than subpatterns that have a {0}
1274 quantifier are omitted from the compiled pattern.
1275 .P
1276 For convenience, the three most common quantifiers have single-character
1277 abbreviations:
1278 .sp
1279 * is equivalent to {0,}
1280 + is equivalent to {1,}
1281 ? is equivalent to {0,1}
1282 .sp
1283 It is possible to construct infinite loops by following a subpattern that can
1284 match no characters with a quantifier that has no upper limit, for example:
1285 .sp
1286 (a?)*
1287 .sp
1288 Earlier versions of Perl and PCRE used to give an error at compile time for
1289 such patterns. However, because there are cases where this can be useful, such
1290 patterns are now accepted, but if any repetition of the subpattern does in fact
1291 match no characters, the loop is forcibly broken.
1292 .P
1293 By default, the quantifiers are "greedy", that is, they match as much as
1294 possible (up to the maximum number of permitted times), without causing the
1295 rest of the pattern to fail. The classic example of where this gives problems
1296 is in trying to match comments in C programs. These appear between /* and */
1297 and within the comment, individual * and / characters may appear. An attempt to
1298 match C comments by applying the pattern
1299 .sp
1300 /\e*.*\e*/
1301 .sp
1302 to the string
1303 .sp
1304 /* first comment */ not comment /* second comment */
1305 .sp
1306 fails, because it matches the entire string owing to the greediness of the .*
1307 item.
1308 .P
1309 However, if a quantifier is followed by a question mark, it ceases to be
1310 greedy, and instead matches the minimum number of times possible, so the
1311 pattern
1312 .sp
1313 /\e*.*?\e*/
1314 .sp
1315 does the right thing with the C comments. The meaning of the various
1316 quantifiers is not otherwise changed, just the preferred number of matches.
1317 Do not confuse this use of question mark with its use as a quantifier in its
1318 own right. Because it has two uses, it can sometimes appear doubled, as in
1319 .sp
1320 \ed??\ed
1321 .sp
1322 which matches one digit by preference, but can match two if that is the only
1323 way the rest of the pattern matches.
1324 .P
1325 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1326 the quantifiers are not greedy by default, but individual ones can be made
1327 greedy by following them with a question mark. In other words, it inverts the
1328 default behaviour.
1329 .P
1330 When a parenthesized subpattern is quantified with a minimum repeat count that
1331 is greater than 1 or with a limited maximum, more memory is required for the
1332 compiled pattern, in proportion to the size of the minimum or maximum.
1333 .P
1334 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1335 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1336 implicitly anchored, because whatever follows will be tried against every
1337 character position in the subject string, so there is no point in retrying the
1338 overall match at any position after the first. PCRE normally treats such a
1339 pattern as though it were preceded by \eA.
1340 .P
1341 In cases where it is known that the subject string contains no newlines, it is
1342 worth setting PCRE_DOTALL in order to obtain this optimization, or
1343 alternatively using ^ to indicate anchoring explicitly.
1344 .P
1345 However, there is one situation where the optimization cannot be used. When .*
1346 is inside capturing parentheses that are the subject of a backreference
1347 elsewhere in the pattern, a match at the start may fail where a later one
1348 succeeds. Consider, for example:
1349 .sp
1350 (.*)abc\e1
1351 .sp
1352 If the subject is "xyz123abc123" the match point is the fourth character. For
1353 this reason, such a pattern is not implicitly anchored.
1354 .P
1355 When a capturing subpattern is repeated, the value captured is the substring
1356 that matched the final iteration. For example, after
1357 .sp
1358 (tweedle[dume]{3}\es*)+
1359 .sp
1360 has matched "tweedledum tweedledee" the value of the captured substring is
1361 "tweedledee". However, if there are nested capturing subpatterns, the
1362 corresponding captured values may have been set in previous iterations. For
1363 example, after
1364 .sp
1365 /(a|(b))+/
1366 .sp
1367 matches "aba" the value of the second captured substring is "b".
1368 .
1369 .
1370 .\" HTML <a name="atomicgroup"></a>
1372 .rs
1373 .sp
1374 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1375 repetition, failure of what follows normally causes the repeated item to be
1376 re-evaluated to see if a different number of repeats allows the rest of the
1377 pattern to match. Sometimes it is useful to prevent this, either to change the
1378 nature of the match, or to cause it fail earlier than it otherwise might, when
1379 the author of the pattern knows there is no point in carrying on.
1380 .P
1381 Consider, for example, the pattern \ed+foo when applied to the subject line
1382 .sp
1383 123456bar
1384 .sp
1385 After matching all 6 digits and then failing to match "foo", the normal
1386 action of the matcher is to try again with only 5 digits matching the \ed+
1387 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1388 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1389 that once a subpattern has matched, it is not to be re-evaluated in this way.
1390 .P
1391 If we use atomic grouping for the previous example, the matcher gives up
1392 immediately on failing to match "foo" the first time. The notation is a kind of
1393 special parenthesis, starting with (?> as in this example:
1394 .sp
1395 (?>\ed+)foo
1396 .sp
1397 This kind of parenthesis "locks up" the part of the pattern it contains once
1398 it has matched, and a failure further into the pattern is prevented from
1399 backtracking into it. Backtracking past it to previous items, however, works as
1400 normal.
1401 .P
1402 An alternative description is that a subpattern of this type matches the string
1403 of characters that an identical standalone pattern would match, if anchored at
1404 the current point in the subject string.
1405 .P
1406 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1407 the above example can be thought of as a maximizing repeat that must swallow
1408 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1409 number of digits they match in order to make the rest of the pattern match,
1410 (?>\ed+) can only match an entire sequence of digits.
1411 .P
1412 Atomic groups in general can of course contain arbitrarily complicated
1413 subpatterns, and can be nested. However, when the subpattern for an atomic
1414 group is just a single repeated item, as in the example above, a simpler
1415 notation, called a "possessive quantifier" can be used. This consists of an
1416 additional + character following a quantifier. Using this notation, the
1417 previous example can be rewritten as
1418 .sp
1419 \ed++foo
1420 .sp
1421 Note that a possessive quantifier can be used with an entire group, for
1422 example:
1423 .sp
1424 (abc|xyz){2,3}+
1425 .sp
1426 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1427 option is ignored. They are a convenient notation for the simpler forms of
1428 atomic group. However, there is no difference in the meaning of a possessive
1429 quantifier and the equivalent atomic group, though there may be a performance
1430 difference; possessive quantifiers should be slightly faster.
1431 .P
1432 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1433 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1434 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1435 package, and PCRE copied it from there. It ultimately found its way into Perl
1436 at release 5.10.
1437 .P
1438 PCRE has an optimization that automatically "possessifies" certain simple
1439 pattern constructs. For example, the sequence A+B is treated as A++B because
1440 there is no point in backtracking into a sequence of A's when B must follow.
1441 .P
1442 When a pattern contains an unlimited repeat inside a subpattern that can itself
1443 be repeated an unlimited number of times, the use of an atomic group is the
1444 only way to avoid some failing matches taking a very long time indeed. The
1445 pattern
1446 .sp
1447 (\eD+|<\ed+>)*[!?]
1448 .sp
1449 matches an unlimited number of substrings that either consist of non-digits, or
1450 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1451 quickly. However, if it is applied to
1452 .sp
1453 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1454 .sp
1455 it takes a long time before reporting failure. This is because the string can
1456 be divided between the internal \eD+ repeat and the external * repeat in a
1457 large number of ways, and all have to be tried. (The example uses [!?] rather
1458 than a single character at the end, because both PCRE and Perl have an
1459 optimization that allows for fast failure when a single character is used. They
1460 remember the last single character that is required for a match, and fail early
1461 if it is not present in the string.) If the pattern is changed so that it uses
1462 an atomic group, like this:
1463 .sp
1464 ((?>\eD+)|<\ed+>)*[!?]
1465 .sp
1466 sequences of non-digits cannot be broken, and failure happens quickly.
1467 .
1468 .
1469 .\" HTML <a name="backreferences"></a>
1471 .rs
1472 .sp
1473 Outside a character class, a backslash followed by a digit greater than 0 (and
1474 possibly further digits) is a back reference to a capturing subpattern earlier
1475 (that is, to its left) in the pattern, provided there have been that many
1476 previous capturing left parentheses.
1477 .P
1478 However, if the decimal number following the backslash is less than 10, it is
1479 always taken as a back reference, and causes an error only if there are not
1480 that many capturing left parentheses in the entire pattern. In other words, the
1481 parentheses that are referenced need not be to the left of the reference for
1482 numbers less than 10. A "forward back reference" of this type can make sense
1483 when a repetition is involved and the subpattern to the right has participated
1484 in an earlier iteration.
1485 .P
1486 It is not possible to have a numerical "forward back reference" to a subpattern
1487 whose number is 10 or more using this syntax because a sequence such as \e50 is
1488 interpreted as a character defined in octal. See the subsection entitled
1489 "Non-printing characters"
1490 .\" HTML <a href="#digitsafterbackslash">
1491 .\" </a>
1492 above
1493 .\"
1494 for further details of the handling of digits following a backslash. There is
1495 no such problem when named parentheses are used. A back reference to any
1496 subpattern is possible using named parentheses (see below).
1497 .P
1498 Another way of avoiding the ambiguity inherent in the use of digits following a
1499 backslash is to use the \eg escape sequence, which is a feature introduced in
1500 Perl 5.10. This escape must be followed by an unsigned number or a negative
1501 number, optionally enclosed in braces. These examples are all identical:
1502 .sp
1503 (ring), \e1
1504 (ring), \eg1
1505 (ring), \eg{1}
1506 .sp
1507 An unsigned number specifies an absolute reference without the ambiguity that
1508 is present in the older syntax. It is also useful when literal digits follow
1509 the reference. A negative number is a relative reference. Consider this
1510 example:
1511 .sp
1512 (abc(def)ghi)\eg{-1}
1513 .sp
1514 The sequence \eg{-1} is a reference to the most recently started capturing
1515 subpattern before \eg, that is, is it equivalent to \e2. Similarly, \eg{-2}
1516 would be equivalent to \e1. The use of relative references can be helpful in
1517 long patterns, and also in patterns that are created by joining together
1518 fragments that contain references within themselves.
1519 .P
1520 A back reference matches whatever actually matched the capturing subpattern in
1521 the current subject string, rather than anything matching the subpattern
1522 itself (see
1523 .\" HTML <a href="#subpatternsassubroutines">
1524 .\" </a>
1525 "Subpatterns as subroutines"
1526 .\"
1527 below for a way of doing that). So the pattern
1528 .sp
1529 (sens|respons)e and \e1ibility
1530 .sp
1531 matches "sense and sensibility" and "response and responsibility", but not
1532 "sense and responsibility". If caseful matching is in force at the time of the
1533 back reference, the case of letters is relevant. For example,
1534 .sp
1535 ((?i)rah)\es+\e1
1536 .sp
1537 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1538 capturing subpattern is matched caselessly.
1539 .P
1540 There are several different ways of writing back references to named
1541 subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
1542 \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1543 back reference syntax, in which \eg can be used for both numeric and named
1544 references, is also supported. We could rewrite the above example in any of
1545 the following ways:
1546 .sp
1547 (?<p1>(?i)rah)\es+\ek<p1>
1548 (?'p1'(?i)rah)\es+\ek{p1}
1549 (?P<p1>(?i)rah)\es+(?P=p1)
1550 (?<p1>(?i)rah)\es+\eg{p1}
1551 .sp
1552 A subpattern that is referenced by name may appear in the pattern before or
1553 after the reference.
1554 .P
1555 There may be more than one back reference to the same subpattern. If a
1556 subpattern has not actually been used in a particular match, any back
1557 references to it always fail. For example, the pattern
1558 .sp
1559 (a|(bc))\e2
1560 .sp
1561 always fails if it starts to match "a" rather than "bc". Because there may be
1562 many capturing parentheses in a pattern, all digits following the backslash are
1563 taken as part of a potential back reference number. If the pattern continues
1564 with a digit character, some delimiter must be used to terminate the back
1565 reference. If the PCRE_EXTENDED option is set, this can be whitespace.
1566 Otherwise an empty comment (see
1567 .\" HTML <a href="#comments">
1568 .\" </a>
1569 "Comments"
1570 .\"
1571 below) can be used.
1572 .P
1573 A back reference that occurs inside the parentheses to which it refers fails
1574 when the subpattern is first used, so, for example, (a\e1) never matches.
1575 However, such references can be useful inside repeated subpatterns. For
1576 example, the pattern
1577 .sp
1578 (a|b\e1)+
1579 .sp
1580 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1581 the subpattern, the back reference matches the character string corresponding
1582 to the previous iteration. In order for this to work, the pattern must be such
1583 that the first iteration does not need to match the back reference. This can be
1584 done using alternation, as in the example above, or by a quantifier with a
1585 minimum of zero.
1586 .
1587 .
1588 .\" HTML <a name="bigassertions"></a>
1590 .rs
1591 .sp
1592 An assertion is a test on the characters following or preceding the current
1593 matching point that does not actually consume any characters. The simple
1594 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
1595 .\" HTML <a href="#smallassertions">
1596 .\" </a>
1597 above.
1598 .\"
1599 .P
1600 More complicated assertions are coded as subpatterns. There are two kinds:
1601 those that look ahead of the current position in the subject string, and those
1602 that look behind it. An assertion subpattern is matched in the normal way,
1603 except that it does not cause the current matching position to be changed.
1604 .P
1605 Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1606 because it makes no sense to assert the same thing several times. If any kind
1607 of assertion contains capturing subpatterns within it, these are counted for
1608 the purposes of numbering the capturing subpatterns in the whole pattern.
1609 However, substring capturing is carried out only for positive assertions,
1610 because it does not make sense for negative assertions.
1611 .
1612 .
1613 .SS "Lookahead assertions"
1614 .rs
1615 .sp
1616 Lookahead assertions start with (?= for positive assertions and (?! for
1617 negative assertions. For example,
1618 .sp
1619 \ew+(?=;)
1620 .sp
1621 matches a word followed by a semicolon, but does not include the semicolon in
1622 the match, and
1623 .sp
1624 foo(?!bar)
1625 .sp
1626 matches any occurrence of "foo" that is not followed by "bar". Note that the
1627 apparently similar pattern
1628 .sp
1629 (?!foo)bar
1630 .sp
1631 does not find an occurrence of "bar" that is preceded by something other than
1632 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1633 (?!foo) is always true when the next three characters are "bar". A
1634 lookbehind assertion is needed to achieve the other effect.
1635 .P
1636 If you want to force a matching failure at some point in a pattern, the most
1637 convenient way to do it is with (?!) because an empty string always matches, so
1638 an assertion that requires there not to be an empty string must always fail.
1639 .
1640 .
1641 .\" HTML <a name="lookbehind"></a>
1642 .SS "Lookbehind assertions"
1643 .rs
1644 .sp
1645 Lookbehind assertions start with (?<= for positive assertions and (?<! for
1646 negative assertions. For example,
1647 .sp
1648 (?<!foo)bar
1649 .sp
1650 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1651 a lookbehind assertion are restricted such that all the strings it matches must
1652 have a fixed length. However, if there are several top-level alternatives, they
1653 do not all have to have the same fixed length. Thus
1654 .sp
1655 (?<=bullock|donkey)
1656 .sp
1657 is permitted, but
1658 .sp
1659 (?<!dogs?|cats?)
1660 .sp
1661 causes an error at compile time. Branches that match different length strings
1662 are permitted only at the top level of a lookbehind assertion. This is an
1663 extension compared with Perl (at least for 5.8), which requires all branches to
1664 match the same length of string. An assertion such as
1665 .sp
1666 (?<=ab(c|de))
1667 .sp
1668 is not permitted, because its single top-level branch can match two different
1669 lengths, but it is acceptable if rewritten to use two top-level branches:
1670 .sp
1671 (?<=abc|abde)
1672 .sp
1673 In some cases, the Perl 5.10 escape sequence \eK
1674 .\" HTML <a href="#resetmatchstart">
1675 .\" </a>
1676 (see above)
1677 .\"
1678 can be used instead of a lookbehind assertion; this is not restricted to a
1679 fixed-length.
1680 .P
1681 The implementation of lookbehind assertions is, for each alternative, to
1682 temporarily move the current position back by the fixed length and then try to
1683 match. If there are insufficient characters before the current position, the
1684 assertion fails.
1685 .P
1686 PCRE does not allow the \eC escape (which matches a single byte in UTF-8 mode)
1687 to appear in lookbehind assertions, because it makes it impossible to calculate
1688 the length of the lookbehind. The \eX and \eR escapes, which can match
1689 different numbers of bytes, are also not permitted.
1690 .P
1691 Possessive quantifiers can be used in conjunction with lookbehind assertions to
1692 specify efficient matching at the end of the subject string. Consider a simple
1693 pattern such as
1694 .sp
1695 abcd$
1696 .sp
1697 when applied to a long string that does not match. Because matching proceeds
1698 from left to right, PCRE will look for each "a" in the subject and then see if
1699 what follows matches the rest of the pattern. If the pattern is specified as
1700 .sp
1701 ^.*abcd$
1702 .sp
1703 the initial .* matches the entire string at first, but when this fails (because
1704 there is no following "a"), it backtracks to match all but the last character,
1705 then all but the last two characters, and so on. Once again the search for "a"
1706 covers the entire string, from right to left, so we are no better off. However,
1707 if the pattern is written as
1708 .sp
1709 ^.*+(?<=abcd)
1710 .sp
1711 there can be no backtracking for the .*+ item; it can match only the entire
1712 string. The subsequent lookbehind assertion does a single test on the last four
1713 characters. If it fails, the match fails immediately. For long strings, this
1714 approach makes a significant difference to the processing time.
1715 .
1716 .
1717 .SS "Using multiple assertions"
1718 .rs
1719 .sp
1720 Several assertions (of any sort) may occur in succession. For example,
1721 .sp
1722 (?<=\ed{3})(?<!999)foo
1723 .sp
1724 matches "foo" preceded by three digits that are not "999". Notice that each of
1725 the assertions is applied independently at the same point in the subject
1726 string. First there is a check that the previous three characters are all
1727 digits, and then there is a check that the same three characters are not "999".
1728 This pattern does \fInot\fP match "foo" preceded by six characters, the first
1729 of which are digits and the last three of which are not "999". For example, it
1730 doesn't match "123abcfoo". A pattern to do that is
1731 .sp
1732 (?<=\ed{3}...)(?<!999)foo
1733 .sp
1734 This time the first assertion looks at the preceding six characters, checking
1735 that the first three are digits, and then the second assertion checks that the
1736 preceding three characters are not "999".
1737 .P
1738 Assertions can be nested in any combination. For example,
1739 .sp
1740 (?<=(?<!foo)bar)baz
1741 .sp
1742 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1743 preceded by "foo", while
1744 .sp
1745 (?<=\ed{3}(?!999)...)foo
1746 .sp
1747 is another pattern that matches "foo" preceded by three digits and any three
1748 characters that are not "999".
1749 .
1750 .
1751 .\" HTML <a name="conditions"></a>
1753 .rs
1754 .sp
1755 It is possible to cause the matching process to obey a subpattern
1756 conditionally or to choose between two alternative subpatterns, depending on
1757 the result of an assertion, or whether a previous capturing subpattern matched
1758 or not. The two possible forms of conditional subpattern are
1759 .sp
1760 (?(condition)yes-pattern)
1761 (?(condition)yes-pattern|no-pattern)
1762 .sp
1763 If the condition is satisfied, the yes-pattern is used; otherwise the
1764 no-pattern (if present) is used. If there are more than two alternatives in the
1765 subpattern, a compile-time error occurs.
1766 .P
1767 There are four kinds of condition: references to subpatterns, references to
1768 recursion, a pseudo-condition called DEFINE, and assertions.
1769 .
1770 .SS "Checking for a used subpattern by number"
1771 .rs
1772 .sp
1773 If the text between the parentheses consists of a sequence of digits, the
1774 condition is true if the capturing subpattern of that number has previously
1775 matched. An alternative notation is to precede the digits with a plus or minus
1776 sign. In this case, the subpattern number is relative rather than absolute.
1777 The most recently opened parentheses can be referenced by (?(-1), the next most
1778 recent by (?(-2), and so on. In looping constructs it can also make sense to
1779 refer to subsequent groups with constructs such as (?(+2).
1780 .P
1781 Consider the following pattern, which contains non-significant white space to
1782 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
1783 three parts for ease of discussion:
1784 .sp
1785 ( \e( )? [^()]+ (?(1) \e) )
1786 .sp
1787 The first part matches an optional opening parenthesis, and if that
1788 character is present, sets it as the first captured substring. The second part
1789 matches one or more characters that are not parentheses. The third part is a
1790 conditional subpattern that tests whether the first set of parentheses matched
1791 or not. If they did, that is, if subject started with an opening parenthesis,
1792 the condition is true, and so the yes-pattern is executed and a closing
1793 parenthesis is required. Otherwise, since no-pattern is not present, the
1794 subpattern matches nothing. In other words, this pattern matches a sequence of
1795 non-parentheses, optionally enclosed in parentheses.
1796 .P
1797 If you were embedding this pattern in a larger one, you could use a relative
1798 reference:
1799 .sp
1800 ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
1801 .sp
1802 This makes the fragment independent of the parentheses in the larger pattern.
1803 .
1804 .SS "Checking for a used subpattern by name"
1805 .rs
1806 .sp
1807 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
1808 subpattern by name. For compatibility with earlier versions of PCRE, which had
1809 this facility before Perl, the syntax (?(name)...) is also recognized. However,
1810 there is a possible ambiguity with this syntax, because subpattern names may
1811 consist entirely of digits. PCRE looks first for a named subpattern; if it
1812 cannot find one and the name consists entirely of digits, PCRE looks for a
1813 subpattern of that number, which must be greater than zero. Using subpattern
1814 names that consist entirely of digits is not recommended.
1815 .P
1816 Rewriting the above example to use a named subpattern gives this:
1817 .sp
1818 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
1819 .sp
1820 .
1821 .SS "Checking for pattern recursion"
1822 .rs
1823 .sp
1824 If the condition is the string (R), and there is no subpattern with the name R,
1825 the condition is true if a recursive call to the whole pattern or any
1826 subpattern has been made. If digits or a name preceded by ampersand follow the
1827 letter R, for example:
1828 .sp
1829 (?(R3)...) or (?(R&name)...)
1830 .sp
1831 the condition is true if the most recent recursion is into the subpattern whose
1832 number or name is given. This condition does not check the entire recursion
1833 stack.
1834 .P
1835 At "top level", all these recursion test conditions are false. Recursive
1836 patterns are described below.
1837 .
1838 .SS "Defining subpatterns for use by reference only"
1839 .rs
1840 .sp
1841 If the condition is the string (DEFINE), and there is no subpattern with the
1842 name DEFINE, the condition is always false. In this case, there may be only one
1843 alternative in the subpattern. It is always skipped if control reaches this
1844 point in the pattern; the idea of DEFINE is that it can be used to define
1845 "subroutines" that can be referenced from elsewhere. (The use of "subroutines"
1846 is described below.) For example, a pattern to match an IPv4 address could be
1847 written like this (ignore whitespace and line breaks):
1848 .sp
1849 (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
1850 \eb (?&byte) (\e.(?&byte)){3} \eb
1851 .sp
1852 The first part of the pattern is a DEFINE group inside which a another group
1853 named "byte" is defined. This matches an individual component of an IPv4
1854 address (a number less than 256). When matching takes place, this part of the
1855 pattern is skipped because DEFINE acts like a false condition.
1856 .P
1857 The rest of the pattern uses references to the named group to match the four
1858 dot-separated components of an IPv4 address, insisting on a word boundary at
1859 each end.
1860 .
1861 .SS "Assertion conditions"
1862 .rs
1863 .sp
1864 If the condition is not in any of the above formats, it must be an assertion.
1865 This may be a positive or negative lookahead or lookbehind assertion. Consider
1866 this pattern, again containing non-significant white space, and with the two
1867 alternatives on the second line:
1868 .sp
1869 (?(?=[^a-z]*[a-z])
1870 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
1871 .sp
1872 The condition is a positive lookahead assertion that matches an optional
1873 sequence of non-letters followed by a letter. In other words, it tests for the
1874 presence of at least one letter in the subject. If a letter is found, the
1875 subject is matched against the first alternative; otherwise it is matched
1876 against the second. This pattern matches strings in one of the two forms
1877 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
1878 .
1879 .
1880 .\" HTML <a name="comments"></a>
1882 .rs
1883 .sp
1884 The sequence (?# marks the start of a comment that continues up to the next
1885 closing parenthesis. Nested parentheses are not permitted. The characters
1886 that make up a comment play no part in the pattern matching at all.
1887 .P
1888 If the PCRE_EXTENDED option is set, an unescaped # character outside a
1889 character class introduces a comment that continues to immediately after the
1890 next newline in the pattern.
1891 .
1892 .
1893 .\" HTML <a name="recursion"></a>
1895 .rs
1896 .sp
1897 Consider the problem of matching a string in parentheses, allowing for
1898 unlimited nested parentheses. Without the use of recursion, the best that can
1899 be done is to use a pattern that matches up to some fixed depth of nesting. It
1900 is not possible to handle an arbitrary nesting depth.
1901 .P
1902 For some time, Perl has provided a facility that allows regular expressions to
1903 recurse (amongst other things). It does this by interpolating Perl code in the
1904 expression at run time, and the code can refer to the expression itself. A Perl
1905 pattern using code interpolation to solve the parentheses problem can be
1906 created like this:
1907 .sp
1908 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
1909 .sp
1910 The (?p{...}) item interpolates Perl code at run time, and in this case refers
1911 recursively to the pattern in which it appears.
1912 .P
1913 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
1914 supports special syntax for recursion of the entire pattern, and also for
1915 individual subpattern recursion. After its introduction in PCRE and Python,
1916 this kind of recursion was introduced into Perl at release 5.10.
1917 .P
1918 A special item that consists of (? followed by a number greater than zero and a
1919 closing parenthesis is a recursive call of the subpattern of the given number,
1920 provided that it occurs inside that subpattern. (If not, it is a "subroutine"
1921 call, which is described in the next section.) The special item (?R) or (?0) is
1922 a recursive call of the entire regular expression.
1923 .P
1924 In PCRE (like Python, but unlike Perl), a recursive subpattern call is always
1925 treated as an atomic group. That is, once it has matched some of the subject
1926 string, it is never re-entered, even if it contains untried alternatives and
1927 there is a subsequent matching failure.
1928 .P
1929 This PCRE pattern solves the nested parentheses problem (assume the
1930 PCRE_EXTENDED option is set so that white space is ignored):
1931 .sp
1932 \e( ( (?>[^()]+) | (?R) )* \e)
1933 .sp
1934 First it matches an opening parenthesis. Then it matches any number of
1935 substrings which can either be a sequence of non-parentheses, or a recursive
1936 match of the pattern itself (that is, a correctly parenthesized substring).
1937 Finally there is a closing parenthesis.
1938 .P
1939 If this were part of a larger pattern, you would not want to recurse the entire
1940 pattern, so instead you could use this:
1941 .sp
1942 ( \e( ( (?>[^()]+) | (?1) )* \e) )
1943 .sp
1944 We have put the pattern into parentheses, and caused the recursion to refer to
1945 them instead of the whole pattern.
1946 .P
1947 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
1948 is made easier by the use of relative references. (A Perl 5.10 feature.)
1949 Instead of (?1) in the pattern above you can write (?-2) to refer to the second
1950 most recently opened parentheses preceding the recursion. In other words, a
1951 negative number counts capturing parentheses leftwards from the point at which
1952 it is encountered.
1953 .P
1954 It is also possible to refer to subsequently opened parentheses, by writing
1955 references such as (?+2). However, these cannot be recursive because the
1956 reference is not inside the parentheses that are referenced. They are always
1957 "subroutine" calls, as described in the next section.
1958 .P
1959 An alternative approach is to use named parentheses instead. The Perl syntax
1960 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
1961 could rewrite the above example as follows:
1962 .sp
1963 (?<pn> \e( ( (?>[^()]+) | (?&pn) )* \e) )
1964 .sp
1965 If there is more than one subpattern with the same name, the earliest one is
1966 used.
1967 .P
1968 This particular example pattern that we have been looking at contains nested
1969 unlimited repeats, and so the use of atomic grouping for matching strings of
1970 non-parentheses is important when applying the pattern to strings that do not
1971 match. For example, when this pattern is applied to
1972 .sp
1973 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
1974 .sp
1975 it yields "no match" quickly. However, if atomic grouping is not used,
1976 the match runs for a very long time indeed because there are so many different
1977 ways the + and * repeats can carve up the subject, and all have to be tested
1978 before failure can be reported.
1979 .P
1980 At the end of a match, the values set for any capturing subpatterns are those
1981 from the outermost level of the recursion at which the subpattern value is set.
1982 If you want to obtain intermediate values, a callout function can be used (see
1983 below and the
1984 .\" HREF
1985 \fBpcrecallout\fP
1986 .\"
1987 documentation). If the pattern above is matched against
1988 .sp
1989 (ab(cd)ef)
1990 .sp
1991 the value for the capturing parentheses is "ef", which is the last value taken
1992 on at the top level. If additional parentheses are added, giving
1993 .sp
1994 \e( ( ( (?>[^()]+) | (?R) )* ) \e)
1995 ^ ^
1996 ^ ^
1997 .sp
1998 the string they capture is "ab(cd)ef", the contents of the top level
1999 parentheses. If there are more than 15 capturing parentheses in a pattern, PCRE
2000 has to obtain extra memory to store data during a recursion, which it does by
2001 using \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no
2002 memory can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2003 .P
2004 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2005 Consider this pattern, which matches text in angle brackets, allowing for
2006 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2007 recursing), whereas any characters are permitted at the outer level.
2008 .sp
2009 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
2010 .sp
2011 In this pattern, (?(R) is the start of a conditional subpattern, with two
2012 different alternatives for the recursive and non-recursive cases. The (?R) item
2013 is the actual recursive call.
2014 .
2015 .
2016 .\" HTML <a name="subpatternsassubroutines"></a>
2018 .rs
2019 .sp
2020 If the syntax for a recursive subpattern reference (either by number or by
2021 name) is used outside the parentheses to which it refers, it operates like a
2022 subroutine in a programming language. The "called" subpattern may be defined
2023 before or after the reference. A numbered reference can be absolute or
2024 relative, as in these examples:
2025 .sp
2026 (...(absolute)...)...(?2)...
2027 (...(relative)...)...(?-1)...
2028 (...(?+1)...(relative)...
2029 .sp
2030 An earlier example pointed out that the pattern
2031 .sp
2032 (sens|respons)e and \e1ibility
2033 .sp
2034 matches "sense and sensibility" and "response and responsibility", but not
2035 "sense and responsibility". If instead the pattern
2036 .sp
2037 (sens|respons)e and (?1)ibility
2038 .sp
2039 is used, it does match "sense and responsibility" as well as the other two
2040 strings. Another example is given in the discussion of DEFINE above.
2041 .P
2042 Like recursive subpatterns, a "subroutine" call is always treated as an atomic
2043 group. That is, once it has matched some of the subject string, it is never
2044 re-entered, even if it contains untried alternatives and there is a subsequent
2045 matching failure.
2046 .P
2047 When a subpattern is used as a subroutine, processing options such as
2048 case-independence are fixed when the subpattern is defined. They cannot be
2049 changed for different calls. For example, consider this pattern:
2050 .sp
2051 (abc)(?i:(?-1))
2052 .sp
2053 It matches "abcabc". It does not match "abcABC" because the change of
2054 processing option does not affect the called subpattern.
2055 .
2056 .
2057 .\" HTML <a name="onigurumasubroutines"></a>
2059 .rs
2060 .sp
2061 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
2062 a number enclosed either in angle brackets or single quotes, is an alternative
2063 syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2064 are two of the examples used above, rewritten using this syntax:
2065 .sp
2066 (?<pn> \e( ( (?>[^()]+) | \eg<pn> )* \e) )
2067 (sens|respons)e and \eg'1'ibility
2068 .sp
2069 PCRE supports an extension to Oniguruma: if a number is preceded by a
2070 plus or a minus sign it is taken as a relative reference. For example:
2071 .sp
2072 (abc)(?i:\eg<-1>)
2073 .sp
2074 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
2075 synonymous. The former is a back reference; the latter is a subroutine call.
2076 .
2077 .
2079 .rs
2080 .sp
2081 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2082 code to be obeyed in the middle of matching a regular expression. This makes it
2083 possible, amongst other things, to extract different substrings that match the
2084 same pair of parentheses when there is a repetition.
2085 .P
2086 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2087 code. The feature is called "callout". The caller of PCRE provides an external
2088 function by putting its entry point in the global variable \fIpcre_callout\fP.
2089 By default, this variable contains NULL, which disables all calling out.
2090 .P
2091 Within a regular expression, (?C) indicates the points at which the external
2092 function is to be called. If you want to identify different callout points, you
2093 can put a number less than 256 after the letter C. The default value is zero.
2094 For example, this pattern has two callout points:
2095 .sp
2096 (?C1)abc(?C2)def
2097 .sp
2098 If the PCRE_AUTO_CALLOUT flag is passed to \fBpcre_compile()\fP, callouts are
2099 automatically installed before each item in the pattern. They are all numbered
2100 255.
2101 .P
2102 During matching, when PCRE reaches a callout point (and \fIpcre_callout\fP is
2103 set), the external function is called. It is provided with the number of the
2104 callout, the position in the pattern, and, optionally, one item of data
2105 originally supplied by the caller of \fBpcre_exec()\fP. The callout function
2106 may cause matching to proceed, to backtrack, or to fail altogether. A complete
2107 description of the interface to the callout function is given in the
2108 .\" HREF
2109 \fBpcrecallout\fP
2110 .\"
2111 documentation.
2112 .
2113 .
2115 .rs
2116 .sp
2117 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2118 are described in the Perl documentation as "experimental and subject to change
2119 or removal in a future version of Perl". It goes on to say: "Their usage in
2120 production code should be noted to avoid problems during upgrades." The same
2121 remarks apply to the PCRE features described in this section.
2122 .P
2123 Since these verbs are specifically related to backtracking, most of them can be
2124 used only when the pattern is to be matched using \fBpcre_exec()\fP, which uses
2125 a backtracking algorithm. With the exception of (*FAIL), which behaves like a
2126 failing negative assertion, they cause an error if encountered by
2127 \fBpcre_dfa_exec()\fP.
2128 .P
2129 The new verbs make use of what was previously invalid syntax: an opening
2130 parenthesis followed by an asterisk. In Perl, they are generally of the form
2131 (*VERB:ARG) but PCRE does not support the use of arguments, so its general
2132 form is just (*VERB). Any number of these verbs may occur in a pattern. There
2133 are two kinds:
2134 .
2135 .SS "Verbs that act immediately"
2136 .rs
2137 .sp
2138 The following verbs act as soon as they are encountered:
2139 .sp
2140 (*ACCEPT)
2141 .sp
2142 This verb causes the match to end successfully, skipping the remainder of the
2143 pattern. When inside a recursion, only the innermost pattern is ended
2144 immediately. PCRE differs from Perl in what happens if the (*ACCEPT) is inside
2145 capturing parentheses. In Perl, the data so far is captured: in PCRE no data is
2146 captured. For example:
2147 .sp
2148 A(A|B(*ACCEPT)|C)D
2149 .sp
2150 This matches "AB", "AAD", or "ACD", but when it matches "AB", no data is
2151 captured.
2152 .sp
2153 (*FAIL) or (*F)
2154 .sp
2155 This verb causes the match to fail, forcing backtracking to occur. It is
2156 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2157 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2158 Perl features that are not present in PCRE. The nearest equivalent is the
2159 callout feature, as for example in this pattern:
2160 .sp
2161 a+(?C)(*FAIL)
2162 .sp
2163 A match with the string "aaaa" always fails, but the callout is taken before
2164 each backtrack happens (in this example, 10 times).
2165 .
2166 .SS "Verbs that act after backtracking"
2167 .rs
2168 .sp
2169 The following verbs do nothing when they are encountered. Matching continues
2170 with what follows, but if there is no subsequent match, a failure is forced.
2171 The verbs differ in exactly what kind of failure occurs.
2172 .sp
2173 (*COMMIT)
2174 .sp
2175 This verb causes the whole match to fail outright if the rest of the pattern
2176 does not match. Even if the pattern is unanchored, no further attempts to find
2177 a match by advancing the start point take place. Once (*COMMIT) has been
2178 passed, \fBpcre_exec()\fP is committed to finding a match at the current
2179 starting point, or not at all. For example:
2180 .sp
2181 a+(*COMMIT)b
2182 .sp
2183 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2184 dynamic anchor, or "I've started, so I must finish."
2185 .sp
2186 (*PRUNE)
2187 .sp
2188 This verb causes the match to fail at the current position if the rest of the
2189 pattern does not match. If the pattern is unanchored, the normal "bumpalong"
2190 advance to the next starting character then happens. Backtracking can occur as
2191 usual to the left of (*PRUNE), or when matching to the right of (*PRUNE), but
2192 if there is no match to the right, backtracking cannot cross (*PRUNE).
2193 In simple cases, the use of (*PRUNE) is just an alternative to an atomic
2194 group or possessive quantifier, but there are some uses of (*PRUNE) that cannot
2195 be expressed in any other way.
2196 .sp
2197 (*SKIP)
2198 .sp
2199 This verb is like (*PRUNE), except that if the pattern is unanchored, the
2200 "bumpalong" advance is not to the next character, but to the position in the
2201 subject where (*SKIP) was encountered. (*SKIP) signifies that whatever text
2202 was matched leading up to it cannot be part of a successful match. Consider:
2203 .sp
2204 a+(*SKIP)b
2205 .sp
2206 If the subject is "aaaac...", after the first match attempt fails (starting at
2207 the first character in the string), the starting point skips on to start the
2208 next attempt at "c". Note that a possessive quantifer does not have the same
2209 effect in this example; although it would suppress backtracking during the
2210 first match attempt, the second attempt would start at the second character
2211 instead of skipping on to "c".
2212 .sp
2213 (*THEN)
2214 .sp
2215 This verb causes a skip to the next alternation if the rest of the pattern does
2216 not match. That is, it cancels pending backtracking, but only within the
2217 current alternation. Its name comes from the observation that it can be used
2218 for a pattern-based if-then-else block:
2219 .sp
2220 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2221 .sp
2222 If the COND1 pattern matches, FOO is tried (and possibly further items after
2223 the end of the group if FOO succeeds); on failure the matcher skips to the
2224 second alternative and tries COND2, without backtracking into COND1. If (*THEN)
2225 is used outside of any alternation, it acts exactly like (*PRUNE).
2226 .
2227 .
2228 .SH "SEE ALSO"
2229 .rs
2230 .sp
2231 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3), \fBpcre\fP(3).
2232 .
2233 .
2235 .rs
2236 .sp
2237 .nf
2238 Philip Hazel
2239 University Computing Service
2240 Cambridge CB2 3QH, England.
2241 .fi
2242 .
2243 .
2245 .rs
2246 .sp
2247 .nf
2248 Last updated: 18 March 2009
2249 Copyright (c) 1997-2009 University of Cambridge.
2250 .fi


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

  ViewVC Help
Powered by ViewVC 1.1.5