ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 227 - (show annotations)
Tue Aug 21 15:00:15 2007 UTC (13 years, 11 months ago) by ph10
File size: 87464 byte(s)
Error occurred while calculating annotation data.
Add (*CR) etc.
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. Perl's regular expressions are described in its own documentation, and
13 regular expressions in general are covered in a number of books, some of which
14 have copious examples. Jeffrey Friedl's "Mastering Regular Expressions",
15 published by O'Reilly, covers regular expressions in great detail. This
16 description of PCRE's regular expressions is intended as reference material.
17 .P
18 The original operation of PCRE was on strings of one-byte characters. However,
19 there is now also support for UTF-8 character strings. To use this, you must
20 build PCRE to include UTF-8 support, and then call \fBpcre_compile()\fP with
21 the PCRE_UTF8 option. How this affects pattern matching is mentioned in several
22 places below. There is also a summary of UTF-8 features in the
23 .\" HTML <a href="pcre.html#utf8support">
24 .\" </a>
25 section on UTF-8 support
26 .\"
27 in the main
28 .\" HREF
29 \fBpcre\fP
30 .\"
31 page.
32 .P
33 The remainder of this document discusses the patterns that are supported by
34 PCRE when its main matching function, \fBpcre_exec()\fP, is used.
35 From release 6.0, PCRE offers a second matching function,
36 \fBpcre_dfa_exec()\fP, which matches using a different algorithm that is not
37 Perl-compatible. Some of the features discussed below are not available when
38 \fBpcre_dfa_exec()\fP is used. The advantages and disadvantages of the
39 alternative function, and how it differs from the normal function, are
40 discussed in the
41 .\" HREF
42 \fBpcrematching\fP
43 .\"
44 page.
45 .
46 .
48 .rs
49 .sp
50 PCRE supports five different conventions for indicating line breaks in
51 strings: a single CR (carriage return) character, a single LF (linefeed)
52 character, the two-character sequence CRLF, any of the three preceding, or any
53 Unicode newline sequence. The
54 .\" HREF
55 \fBpcreapi\fP
56 .\"
57 page has
58 .\" HTML <a href="pcreapi.html#newlines">
59 .\" </a>
60 further discussion
61 .\"
62 about newlines, and shows how to set the newline convention in the
63 \fIoptions\fP arguments for the compiling and matching functions.
64 .P
65 It is also possible to specify a newline convention by starting a pattern
66 string with one of the following five sequences:
67 .sp
68 (*CR) carriage return
69 (*LF) linefeed
70 (*CRLF) carriage return, followed by linefeed
71 (*ANYCRLF) any of the three above
72 (*ANY) all Unicode newline sequences
73 .sp
74 These override the default and the options given to \fBpcre_compile()\fP. For
75 example, on a Unix system where LF is the default newline sequence, the pattern
76 .sp
77 (*CR)a.b
78 .sp
79 changes the convention to CR. That pattern matches "a\enb" because LF is no
80 longer a newline. Note that these special settings, which are not
81 Perl-compatible, are recognized only at the very start of a pattern, and that
82 they must be in upper case.
83 .
84 .
86 .rs
87 .sp
88 A regular expression is a pattern that is matched against a subject string from
89 left to right. Most characters stand for themselves in a pattern, and match the
90 corresponding characters in the subject. As a trivial example, the pattern
91 .sp
92 The quick brown fox
93 .sp
94 matches a portion of a subject string that is identical to itself. When
95 caseless matching is specified (the PCRE_CASELESS option), letters are matched
96 independently of case. In UTF-8 mode, PCRE always understands the concept of
97 case for characters whose values are less than 128, so caseless matching is
98 always possible. For characters with higher values, the concept of case is
99 supported if PCRE is compiled with Unicode property support, but not otherwise.
100 If you want to use caseless matching for characters 128 and above, you must
101 ensure that PCRE is compiled with Unicode property support as well as with
102 UTF-8 support.
103 .P
104 The power of regular expressions comes from the ability to include alternatives
105 and repetitions in the pattern. These are encoded in the pattern by the use of
106 \fImetacharacters\fP, which do not stand for themselves but instead are
107 interpreted in some special way.
108 .P
109 There are two different sets of metacharacters: those that are recognized
110 anywhere in the pattern except within square brackets, and those that are
111 recognized within square brackets. Outside square brackets, the metacharacters
112 are as follows:
113 .sp
114 \e general escape character with several uses
115 ^ assert start of string (or line, in multiline mode)
116 $ assert end of string (or line, in multiline mode)
117 . match any character except newline (by default)
118 [ start character class definition
119 | start of alternative branch
120 ( start subpattern
121 ) end subpattern
122 ? extends the meaning of (
123 also 0 or 1 quantifier
124 also quantifier minimizer
125 * 0 or more quantifier
126 + 1 or more quantifier
127 also "possessive quantifier"
128 { start min/max quantifier
129 .sp
130 Part of a pattern that is in square brackets is called a "character class". In
131 a character class the only metacharacters are:
132 .sp
133 \e general escape character
134 ^ negate the class, but only if the first character
135 - indicates character range
136 .\" JOIN
137 [ POSIX character class (only if followed by POSIX
138 syntax)
139 ] terminates the character class
140 .sp
141 The following sections describe the use of each of the metacharacters.
142 .
143 .
145 .rs
146 .sp
147 The backslash character has several uses. Firstly, if it is followed by a
148 non-alphanumeric character, it takes away any special meaning that character
149 may have. This use of backslash as an escape character applies both inside and
150 outside character classes.
151 .P
152 For example, if you want to match a * character, you write \e* in the pattern.
153 This escaping action applies whether or not the following character would
154 otherwise be interpreted as a metacharacter, so it is always safe to precede a
155 non-alphanumeric with backslash to specify that it stands for itself. In
156 particular, if you want to match a backslash, you write \e\e.
157 .P
158 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
159 pattern (other than in a character class) and characters between a # outside
160 a character class and the next newline are ignored. An escaping backslash can
161 be used to include a whitespace or # character as part of the pattern.
162 .P
163 If you want to remove the special meaning from a sequence of characters, you
164 can do so by putting them between \eQ and \eE. This is different from Perl in
165 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
166 Perl, $ and @ cause variable interpolation. Note the following examples:
167 .sp
168 Pattern PCRE matches Perl matches
169 .sp
170 .\" JOIN
171 \eQabc$xyz\eE abc$xyz abc followed by the
172 contents of $xyz
173 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
174 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
175 .sp
176 The \eQ...\eE sequence is recognized both inside and outside character classes.
177 .
178 .
179 .\" HTML <a name="digitsafterbackslash"></a>
180 .SS "Non-printing characters"
181 .rs
182 .sp
183 A second use of backslash provides a way of encoding non-printing characters
184 in patterns in a visible manner. There is no restriction on the appearance of
185 non-printing characters, apart from the binary zero that terminates a pattern,
186 but when a pattern is being prepared by text editing, it is usually easier to
187 use one of the following escape sequences than the binary character it
188 represents:
189 .sp
190 \ea alarm, that is, the BEL character (hex 07)
191 \ecx "control-x", where x is any character
192 \ee escape (hex 1B)
193 \ef formfeed (hex 0C)
194 \en linefeed (hex 0A)
195 \er carriage return (hex 0D)
196 \et tab (hex 09)
197 \eddd character with octal code ddd, or backreference
198 \exhh character with hex code hh
199 \ex{hhh..} character with hex code hhh..
200 .sp
201 The precise effect of \ecx is as follows: if x is a lower case letter, it
202 is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
203 Thus \ecz becomes hex 1A, but \ec{ becomes hex 3B, while \ec; becomes hex
204 7B.
205 .P
206 After \ex, from zero to two hexadecimal digits are read (letters can be in
207 upper or lower case). Any number of hexadecimal digits may appear between \ex{
208 and }, but the value of the character code must be less than 256 in non-UTF-8
209 mode, and less than 2**31 in UTF-8 mode. That is, the maximum value in
210 hexadecimal is 7FFFFFFF. Note that this is bigger than the largest Unicode code
211 point, which is 10FFFF.
212 .P
213 If characters other than hexadecimal digits appear between \ex{ and }, or if
214 there is no terminating }, this form of escape is not recognized. Instead, the
215 initial \ex will be interpreted as a basic hexadecimal escape, with no
216 following digits, giving a character whose value is zero.
217 .P
218 Characters whose value is less than 256 can be defined by either of the two
219 syntaxes for \ex. There is no difference in the way they are handled. For
220 example, \exdc is exactly the same as \ex{dc}.
221 .P
222 After \e0 up to two further octal digits are read. If there are fewer than two
223 digits, just those that are present are used. Thus the sequence \e0\ex\e07
224 specifies two binary zeros followed by a BEL character (code value 7). Make
225 sure you supply two digits after the initial zero if the pattern character that
226 follows is itself an octal digit.
227 .P
228 The handling of a backslash followed by a digit other than 0 is complicated.
229 Outside a character class, PCRE reads it and any following digits as a decimal
230 number. If the number is less than 10, or if there have been at least that many
231 previous capturing left parentheses in the expression, the entire sequence is
232 taken as a \fIback reference\fP. A description of how this works is given
233 .\" HTML <a href="#backreferences">
234 .\" </a>
235 later,
236 .\"
237 following the discussion of
238 .\" HTML <a href="#subpattern">
239 .\" </a>
240 parenthesized subpatterns.
241 .\"
242 .P
243 Inside a character class, or if the decimal number is greater than 9 and there
244 have not been that many capturing subpatterns, PCRE re-reads up to three octal
245 digits following the backslash, and uses them to generate a data character. Any
246 subsequent digits stand for themselves. In non-UTF-8 mode, the value of a
247 character specified in octal must be less than \e400. In UTF-8 mode, values up
248 to \e777 are permitted. For example:
249 .sp
250 \e040 is another way of writing a space
251 .\" JOIN
252 \e40 is the same, provided there are fewer than 40
253 previous capturing subpatterns
254 \e7 is always a back reference
255 .\" JOIN
256 \e11 might be a back reference, or another way of
257 writing a tab
258 \e011 is always a tab
259 \e0113 is a tab followed by the character "3"
260 .\" JOIN
261 \e113 might be a back reference, otherwise the
262 character with octal code 113
263 .\" JOIN
264 \e377 might be a back reference, otherwise
265 the byte consisting entirely of 1 bits
266 .\" JOIN
267 \e81 is either a back reference, or a binary zero
268 followed by the two characters "8" and "1"
269 .sp
270 Note that octal values of 100 or greater must not be introduced by a leading
271 zero, because no more than three octal digits are ever read.
272 .P
273 All the sequences that define a single character value can be used both inside
274 and outside character classes. In addition, inside a character class, the
275 sequence \eb is interpreted as the backspace character (hex 08), and the
276 sequences \eR and \eX are interpreted as the characters "R" and "X",
277 respectively. Outside a character class, these sequences have different
278 meanings
279 .\" HTML <a href="#uniextseq">
280 .\" </a>
281 (see below).
282 .\"
283 .
284 .
285 .SS "Absolute and relative back references"
286 .rs
287 .sp
288 The sequence \eg followed by an unsigned or a negative number, optionally
289 enclosed in braces, is an absolute or relative back reference. A named back
290 reference can be coded as \eg{name}. Back references are discussed
291 .\" HTML <a href="#backreferences">
292 .\" </a>
293 later,
294 .\"
295 following the discussion of
296 .\" HTML <a href="#subpattern">
297 .\" </a>
298 parenthesized subpatterns.
299 .\"
300 .
301 .
302 .SS "Generic character types"
303 .rs
304 .sp
305 Another use of backslash is for specifying generic character types. The
306 following are always recognized:
307 .sp
308 \ed any decimal digit
309 \eD any character that is not a decimal digit
310 \eh any horizontal whitespace character
311 \eH any character that is not a horizontal whitespace character
312 \es any whitespace character
313 \eS any character that is not a whitespace character
314 \ev any vertical whitespace character
315 \eV any character that is not a vertical whitespace character
316 \ew any "word" character
317 \eW any "non-word" character
318 .sp
319 Each pair of escape sequences partitions the complete set of characters into
320 two disjoint sets. Any given character matches one, and only one, of each pair.
321 .P
322 These character type sequences can appear both inside and outside character
323 classes. They each match one character of the appropriate type. If the current
324 matching point is at the end of the subject string, all of them fail, since
325 there is no character to match.
326 .P
327 For compatibility with Perl, \es does not match the VT character (code 11).
328 This makes it different from the the POSIX "space" class. The \es characters
329 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
330 included in a Perl script, \es may match the VT character. In PCRE, it never
331 does.
332 .P
333 In UTF-8 mode, characters with values greater than 128 never match \ed, \es, or
334 \ew, and always match \eD, \eS, and \eW. This is true even when Unicode
335 character property support is available. These sequences retain their original
336 meanings from before UTF-8 support was available, mainly for efficiency
337 reasons.
338 .P
339 The sequences \eh, \eH, \ev, and \eV are Perl 5.10 features. In contrast to the
340 other sequences, these do match certain high-valued codepoints in UTF-8 mode.
341 The horizontal space characters are:
342 .sp
343 U+0009 Horizontal tab
344 U+0020 Space
345 U+00A0 Non-break space
346 U+1680 Ogham space mark
347 U+180E Mongolian vowel separator
348 U+2000 En quad
349 U+2001 Em quad
350 U+2002 En space
351 U+2003 Em space
352 U+2004 Three-per-em space
353 U+2005 Four-per-em space
354 U+2006 Six-per-em space
355 U+2007 Figure space
356 U+2008 Punctuation space
357 U+2009 Thin space
358 U+200A Hair space
359 U+202F Narrow no-break space
360 U+205F Medium mathematical space
361 U+3000 Ideographic space
362 .sp
363 The vertical space characters are:
364 .sp
365 U+000A Linefeed
366 U+000B Vertical tab
367 U+000C Formfeed
368 U+000D Carriage return
369 U+0085 Next line
370 U+2028 Line separator
371 U+2029 Paragraph separator
372 .P
373 A "word" character is an underscore or any character less than 256 that is a
374 letter or digit. The definition of letters and digits is controlled by PCRE's
375 low-valued character tables, and may vary if locale-specific matching is taking
376 place (see
377 .\" HTML <a href="pcreapi.html#localesupport">
378 .\" </a>
379 "Locale support"
380 .\"
381 in the
382 .\" HREF
383 \fBpcreapi\fP
384 .\"
385 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
386 or "french" in Windows, some character codes greater than 128 are used for
387 accented letters, and these are matched by \ew. The use of locales with Unicode
388 is discouraged.
389 .
390 .
391 .SS "Newline sequences"
392 .rs
393 .sp
394 Outside a character class, the escape sequence \eR matches any Unicode newline
395 sequence. This is a Perl 5.10 feature. In non-UTF-8 mode \eR is equivalent to
396 the following:
397 .sp
398 (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
399 .sp
400 This is an example of an "atomic group", details of which are given
401 .\" HTML <a href="#atomicgroup">
402 .\" </a>
403 below.
404 .\"
405 This particular group matches either the two-character sequence CR followed by
406 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
407 U+000B), FF (formfeed, U+000C), CR (carriage return, U+000D), or NEL (next
408 line, U+0085). The two-character sequence is treated as a single unit that
409 cannot be split.
410 .P
411 In UTF-8 mode, two additional characters whose codepoints are greater than 255
412 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
413 Unicode character property support is not needed for these characters to be
414 recognized.
415 .P
416 Inside a character class, \eR matches the letter "R".
417 .
418 .
419 .\" HTML <a name="uniextseq"></a>
420 .SS Unicode character properties
421 .rs
422 .sp
423 When PCRE is built with Unicode character property support, three additional
424 escape sequences that match characters with specific properties are available.
425 When not in UTF-8 mode, these sequences are of course limited to testing
426 characters whose codepoints are less than 256, but they do work in this mode.
427 The extra escape sequences are:
428 .sp
429 \ep{\fIxx\fP} a character with the \fIxx\fP property
430 \eP{\fIxx\fP} a character without the \fIxx\fP property
431 \eX an extended Unicode sequence
432 .sp
433 The property names represented by \fIxx\fP above are limited to the Unicode
434 script names, the general category properties, and "Any", which matches any
435 character (including newline). Other properties such as "InMusicalSymbols" are
436 not currently supported by PCRE. Note that \eP{Any} does not match any
437 characters, so always causes a match failure.
438 .P
439 Sets of Unicode characters are defined as belonging to certain scripts. A
440 character from one of these sets can be matched using a script name. For
441 example:
442 .sp
443 \ep{Greek}
444 \eP{Han}
445 .sp
446 Those that are not part of an identified script are lumped together as
447 "Common". The current list of scripts is:
448 .P
449 Arabic,
450 Armenian,
451 Balinese,
452 Bengali,
453 Bopomofo,
454 Braille,
455 Buginese,
456 Buhid,
457 Canadian_Aboriginal,
458 Cherokee,
459 Common,
460 Coptic,
461 Cuneiform,
462 Cypriot,
463 Cyrillic,
464 Deseret,
465 Devanagari,
466 Ethiopic,
467 Georgian,
468 Glagolitic,
469 Gothic,
470 Greek,
471 Gujarati,
472 Gurmukhi,
473 Han,
474 Hangul,
475 Hanunoo,
476 Hebrew,
477 Hiragana,
478 Inherited,
479 Kannada,
480 Katakana,
481 Kharoshthi,
482 Khmer,
483 Lao,
484 Latin,
485 Limbu,
486 Linear_B,
487 Malayalam,
488 Mongolian,
489 Myanmar,
490 New_Tai_Lue,
491 Nko,
492 Ogham,
493 Old_Italic,
494 Old_Persian,
495 Oriya,
496 Osmanya,
497 Phags_Pa,
498 Phoenician,
499 Runic,
500 Shavian,
501 Sinhala,
502 Syloti_Nagri,
503 Syriac,
504 Tagalog,
505 Tagbanwa,
506 Tai_Le,
507 Tamil,
508 Telugu,
509 Thaana,
510 Thai,
511 Tibetan,
512 Tifinagh,
513 Ugaritic,
514 Yi.
515 .P
516 Each character has exactly one general category property, specified by a
517 two-letter abbreviation. For compatibility with Perl, negation can be specified
518 by including a circumflex between the opening brace and the property name. For
519 example, \ep{^Lu} is the same as \eP{Lu}.
520 .P
521 If only one letter is specified with \ep or \eP, it includes all the general
522 category properties that start with that letter. In this case, in the absence
523 of negation, the curly brackets in the escape sequence are optional; these two
524 examples have the same effect:
525 .sp
526 \ep{L}
527 \epL
528 .sp
529 The following general category property codes are supported:
530 .sp
531 C Other
532 Cc Control
533 Cf Format
534 Cn Unassigned
535 Co Private use
536 Cs Surrogate
537 .sp
538 L Letter
539 Ll Lower case letter
540 Lm Modifier letter
541 Lo Other letter
542 Lt Title case letter
543 Lu Upper case letter
544 .sp
545 M Mark
546 Mc Spacing mark
547 Me Enclosing mark
548 Mn Non-spacing mark
549 .sp
550 N Number
551 Nd Decimal number
552 Nl Letter number
553 No Other number
554 .sp
555 P Punctuation
556 Pc Connector punctuation
557 Pd Dash punctuation
558 Pe Close punctuation
559 Pf Final punctuation
560 Pi Initial punctuation
561 Po Other punctuation
562 Ps Open punctuation
563 .sp
564 S Symbol
565 Sc Currency symbol
566 Sk Modifier symbol
567 Sm Mathematical symbol
568 So Other symbol
569 .sp
570 Z Separator
571 Zl Line separator
572 Zp Paragraph separator
573 Zs Space separator
574 .sp
575 The special property L& is also supported: it matches a character that has
576 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
577 a modifier or "other".
578 .P
579 The Cs (Surrogate) property applies only to characters in the range U+D800 to
580 U+DFFF. Such characters are not valid in UTF-8 strings (see RFC 3629) and so
581 cannot be tested by PCRE, unless UTF-8 validity checking has been turned off
582 (see the discussion of PCRE_NO_UTF8_CHECK in the
583 .\" HREF
584 \fBpcreapi\fP
585 .\"
586 page).
587 .P
588 The long synonyms for these properties that Perl supports (such as \ep{Letter})
589 are not supported by PCRE, nor is it permitted to prefix any of these
590 properties with "Is".
591 .P
592 No character that is in the Unicode table has the Cn (unassigned) property.
593 Instead, this property is assumed for any code point that is not in the
594 Unicode table.
595 .P
596 Specifying caseless matching does not affect these escape sequences. For
597 example, \ep{Lu} always matches only upper case letters.
598 .P
599 The \eX escape matches any number of Unicode characters that form an extended
600 Unicode sequence. \eX is equivalent to
601 .sp
602 (?>\ePM\epM*)
603 .sp
604 That is, it matches a character without the "mark" property, followed by zero
605 or more characters with the "mark" property, and treats the sequence as an
606 atomic group
607 .\" HTML <a href="#atomicgroup">
608 .\" </a>
609 (see below).
610 .\"
611 Characters with the "mark" property are typically accents that affect the
612 preceding character. None of them have codepoints less than 256, so in
613 non-UTF-8 mode \eX matches any one character.
614 .P
615 Matching characters by Unicode property is not fast, because PCRE has to search
616 a structure that contains data for over fifteen thousand characters. That is
617 why the traditional escape sequences such as \ed and \ew do not use Unicode
618 properties in PCRE.
619 .
620 .
621 .\" HTML <a name="resetmatchstart"></a>
622 .SS "Resetting the match start"
623 .rs
624 .sp
625 The escape sequence \eK, which is a Perl 5.10 feature, causes any previously
626 matched characters not to be included in the final matched sequence. For
627 example, the pattern:
628 .sp
629 foo\eKbar
630 .sp
631 matches "foobar", but reports that it has matched "bar". This feature is
632 similar to a lookbehind assertion
633 .\" HTML <a href="#lookbehind">
634 .\" </a>
635 (described below).
636 .\"
637 However, in this case, the part of the subject before the real match does not
638 have to be of fixed length, as lookbehind assertions do. The use of \eK does
639 not interfere with the setting of
640 .\" HTML <a href="#subpattern">
641 .\" </a>
642 captured substrings.
643 .\"
644 For example, when the pattern
645 .sp
646 (foo)\eKbar
647 .sp
648 matches "foobar", the first substring is still set to "foo".
649 .
650 .
651 .\" HTML <a name="smallassertions"></a>
652 .SS "Simple assertions"
653 .rs
654 .sp
655 The final use of backslash is for certain simple assertions. An assertion
656 specifies a condition that has to be met at a particular point in a match,
657 without consuming any characters from the subject string. The use of
658 subpatterns for more complicated assertions is described
659 .\" HTML <a href="#bigassertions">
660 .\" </a>
661 below.
662 .\"
663 The backslashed assertions are:
664 .sp
665 \eb matches at a word boundary
666 \eB matches when not at a word boundary
667 \eA matches at the start of the subject
668 \eZ matches at the end of the subject
669 also matches before a newline at the end of the subject
670 \ez matches only at the end of the subject
671 \eG matches at the first matching position in the subject
672 .sp
673 These assertions may not appear in character classes (but note that \eb has a
674 different meaning, namely the backspace character, inside a character class).
675 .P
676 A word boundary is a position in the subject string where the current character
677 and the previous character do not both match \ew or \eW (i.e. one matches
678 \ew and the other matches \eW), or the start or end of the string if the
679 first or last character matches \ew, respectively.
680 .P
681 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
682 dollar (described in the next section) in that they only ever match at the very
683 start and end of the subject string, whatever options are set. Thus, they are
684 independent of multiline mode. These three assertions are not affected by the
685 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
686 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
687 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
688 at a point other than the beginning of the subject, \eA can never match. The
689 difference between \eZ and \ez is that \eZ matches before a newline at the end
690 of the string as well as at the very end, whereas \ez matches only at the end.
691 .P
692 The \eG assertion is true only when the current matching position is at the
693 start point of the match, as specified by the \fIstartoffset\fP argument of
694 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
695 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
696 arguments, you can mimic Perl's /g option, and it is in this kind of
697 implementation where \eG can be useful.
698 .P
699 Note, however, that PCRE's interpretation of \eG, as the start of the current
700 match, is subtly different from Perl's, which defines it as the end of the
701 previous match. In Perl, these can be different when the previously matched
702 string was empty. Because PCRE does just one match at a time, it cannot
703 reproduce this behaviour.
704 .P
705 If all the alternatives of a pattern begin with \eG, the expression is anchored
706 to the starting match position, and the "anchored" flag is set in the compiled
707 regular expression.
708 .
709 .
711 .rs
712 .sp
713 Outside a character class, in the default matching mode, the circumflex
714 character is an assertion that is true only if the current matching point is
715 at the start of the subject string. If the \fIstartoffset\fP argument of
716 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
717 option is unset. Inside a character class, circumflex has an entirely different
718 meaning
719 .\" HTML <a href="#characterclass">
720 .\" </a>
721 (see below).
722 .\"
723 .P
724 Circumflex need not be the first character of the pattern if a number of
725 alternatives are involved, but it should be the first thing in each alternative
726 in which it appears if the pattern is ever to match that branch. If all
727 possible alternatives start with a circumflex, that is, if the pattern is
728 constrained to match only at the start of the subject, it is said to be an
729 "anchored" pattern. (There are also other constructs that can cause a pattern
730 to be anchored.)
731 .P
732 A dollar character is an assertion that is true only if the current matching
733 point is at the end of the subject string, or immediately before a newline
734 at the end of the string (by default). Dollar need not be the last character of
735 the pattern if a number of alternatives are involved, but it should be the last
736 item in any branch in which it appears. Dollar has no special meaning in a
737 character class.
738 .P
739 The meaning of dollar can be changed so that it matches only at the very end of
740 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
741 does not affect the \eZ assertion.
742 .P
743 The meanings of the circumflex and dollar characters are changed if the
744 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
745 immediately after internal newlines as well as at the start of the subject
746 string. It does not match after a newline that ends the string. A dollar
747 matches before any newlines in the string, as well as at the very end, when
748 PCRE_MULTILINE is set. When newline is specified as the two-character
749 sequence CRLF, isolated CR and LF characters do not indicate newlines.
750 .P
751 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
752 \en represents a newline) in multiline mode, but not otherwise. Consequently,
753 patterns that are anchored in single line mode because all branches start with
754 ^ are not anchored in multiline mode, and a match for circumflex is possible
755 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
756 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
757 .P
758 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
759 end of the subject in both modes, and if all branches of a pattern start with
760 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
761 .
762 .
764 .rs
765 .sp
766 Outside a character class, a dot in the pattern matches any one character in
767 the subject string except (by default) a character that signifies the end of a
768 line. In UTF-8 mode, the matched character may be more than one byte long.
769 .P
770 When a line ending is defined as a single character, dot never matches that
771 character; when the two-character sequence CRLF is used, dot does not match CR
772 if it is immediately followed by LF, but otherwise it matches all characters
773 (including isolated CRs and LFs). When any Unicode line endings are being
774 recognized, dot does not match CR or LF or any of the other line ending
775 characters.
776 .P
777 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
778 option is set, a dot matches any one character, without exception. If the
779 two-character sequence CRLF is present in the subject string, it takes two dots
780 to match it.
781 .P
782 The handling of dot is entirely independent of the handling of circumflex and
783 dollar, the only relationship being that they both involve newlines. Dot has no
784 special meaning in a character class.
785 .
786 .
788 .rs
789 .sp
790 Outside a character class, the escape sequence \eC matches any one byte, both
791 in and out of UTF-8 mode. Unlike a dot, it always matches any line-ending
792 characters. The feature is provided in Perl in order to match individual bytes
793 in UTF-8 mode. Because it breaks up UTF-8 characters into individual bytes,
794 what remains in the string may be a malformed UTF-8 string. For this reason,
795 the \eC escape sequence is best avoided.
796 .P
797 PCRE does not allow \eC to appear in lookbehind assertions
798 .\" HTML <a href="#lookbehind">
799 .\" </a>
800 (described below),
801 .\"
802 because in UTF-8 mode this would make it impossible to calculate the length of
803 the lookbehind.
804 .
805 .
806 .\" HTML <a name="characterclass"></a>
808 .rs
809 .sp
810 An opening square bracket introduces a character class, terminated by a closing
811 square bracket. A closing square bracket on its own is not special. If a
812 closing square bracket is required as a member of the class, it should be the
813 first data character in the class (after an initial circumflex, if present) or
814 escaped with a backslash.
815 .P
816 A character class matches a single character in the subject. In UTF-8 mode, the
817 character may occupy more than one byte. A matched character must be in the set
818 of characters defined by the class, unless the first character in the class
819 definition is a circumflex, in which case the subject character must not be in
820 the set defined by the class. If a circumflex is actually required as a member
821 of the class, ensure it is not the first character, or escape it with a
822 backslash.
823 .P
824 For example, the character class [aeiou] matches any lower case vowel, while
825 [^aeiou] matches any character that is not a lower case vowel. Note that a
826 circumflex is just a convenient notation for specifying the characters that
827 are in the class by enumerating those that are not. A class that starts with a
828 circumflex is not an assertion: it still consumes a character from the subject
829 string, and therefore it fails if the current pointer is at the end of the
830 string.
831 .P
832 In UTF-8 mode, characters with values greater than 255 can be included in a
833 class as a literal string of bytes, or by using the \ex{ escaping mechanism.
834 .P
835 When caseless matching is set, any letters in a class represent both their
836 upper case and lower case versions, so for example, a caseless [aeiou] matches
837 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
838 caseful version would. In UTF-8 mode, PCRE always understands the concept of
839 case for characters whose values are less than 128, so caseless matching is
840 always possible. For characters with higher values, the concept of case is
841 supported if PCRE is compiled with Unicode property support, but not otherwise.
842 If you want to use caseless matching for characters 128 and above, you must
843 ensure that PCRE is compiled with Unicode property support as well as with
844 UTF-8 support.
845 .P
846 Characters that might indicate line breaks are never treated in any special way
847 when matching character classes, whatever line-ending sequence is in use, and
848 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
849 such as [^a] always matches one of these characters.
850 .P
851 The minus (hyphen) character can be used to specify a range of characters in a
852 character class. For example, [d-m] matches any letter between d and m,
853 inclusive. If a minus character is required in a class, it must be escaped with
854 a backslash or appear in a position where it cannot be interpreted as
855 indicating a range, typically as the first or last character in the class.
856 .P
857 It is not possible to have the literal character "]" as the end character of a
858 range. A pattern such as [W-]46] is interpreted as a class of two characters
859 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
860 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
861 the end of range, so [W-\e]46] is interpreted as a class containing a range
862 followed by two other characters. The octal or hexadecimal representation of
863 "]" can also be used to end a range.
864 .P
865 Ranges operate in the collating sequence of character values. They can also be
866 used for characters specified numerically, for example [\e000-\e037]. In UTF-8
867 mode, ranges can include characters whose values are greater than 255, for
868 example [\ex{100}-\ex{2ff}].
869 .P
870 If a range that includes letters is used when caseless matching is set, it
871 matches the letters in either case. For example, [W-c] is equivalent to
872 [][\e\e^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
873 tables for a French locale are in use, [\exc8-\excb] matches accented E
874 characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
875 characters with values greater than 128 only when it is compiled with Unicode
876 property support.
877 .P
878 The character types \ed, \eD, \ep, \eP, \es, \eS, \ew, and \eW may also appear
879 in a character class, and add the characters that they match to the class. For
880 example, [\edABCDEF] matches any hexadecimal digit. A circumflex can
881 conveniently be used with the upper case character types to specify a more
882 restricted set of characters than the matching lower case type. For example,
883 the class [^\eW_] matches any letter or digit, but not underscore.
884 .P
885 The only metacharacters that are recognized in character classes are backslash,
886 hyphen (only where it can be interpreted as specifying a range), circumflex
887 (only at the start), opening square bracket (only when it can be interpreted as
888 introducing a POSIX class name - see the next section), and the terminating
889 closing square bracket. However, escaping other non-alphanumeric characters
890 does no harm.
891 .
892 .
894 .rs
895 .sp
896 Perl supports the POSIX notation for character classes. This uses names
897 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
898 this notation. For example,
899 .sp
900 [01[:alpha:]%]
901 .sp
902 matches "0", "1", any alphabetic character, or "%". The supported class names
903 are
904 .sp
905 alnum letters and digits
906 alpha letters
907 ascii character codes 0 - 127
908 blank space or tab only
909 cntrl control characters
910 digit decimal digits (same as \ed)
911 graph printing characters, excluding space
912 lower lower case letters
913 print printing characters, including space
914 punct printing characters, excluding letters and digits
915 space white space (not quite the same as \es)
916 upper upper case letters
917 word "word" characters (same as \ew)
918 xdigit hexadecimal digits
919 .sp
920 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
921 space (32). Notice that this list includes the VT character (code 11). This
922 makes "space" different to \es, which does not include VT (for Perl
923 compatibility).
924 .P
925 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
926 5.8. Another Perl extension is negation, which is indicated by a ^ character
927 after the colon. For example,
928 .sp
929 [12[:^digit:]]
930 .sp
931 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
932 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
933 supported, and an error is given if they are encountered.
934 .P
935 In UTF-8 mode, characters with values greater than 128 do not match any of
936 the POSIX character classes.
937 .
938 .
940 .rs
941 .sp
942 Vertical bar characters are used to separate alternative patterns. For example,
943 the pattern
944 .sp
945 gilbert|sullivan
946 .sp
947 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
948 and an empty alternative is permitted (matching the empty string). The matching
949 process tries each alternative in turn, from left to right, and the first one
950 that succeeds is used. If the alternatives are within a subpattern
951 .\" HTML <a href="#subpattern">
952 .\" </a>
953 (defined below),
954 .\"
955 "succeeds" means matching the rest of the main pattern as well as the
956 alternative in the subpattern.
957 .
958 .
960 .rs
961 .sp
963 PCRE_EXTENDED options can be changed from within the pattern by a sequence of
964 Perl option letters enclosed between "(?" and ")". The option letters are
965 .sp
968 s for PCRE_DOTALL
970 .sp
971 For example, (?im) sets caseless, multiline matching. It is also possible to
972 unset these options by preceding the letter with a hyphen, and a combined
973 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
974 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
975 permitted. If a letter appears both before and after the hyphen, the option is
976 unset.
977 .P
978 When an option change occurs at top level (that is, not inside subpattern
979 parentheses), the change applies to the remainder of the pattern that follows.
980 If the change is placed right at the start of a pattern, PCRE extracts it into
981 the global options (and it will therefore show up in data extracted by the
982 \fBpcre_fullinfo()\fP function).
983 .P
984 An option change within a subpattern (see below for a description of
985 subpatterns) affects only that part of the current pattern that follows it, so
986 .sp
987 (a(?i)b)c
988 .sp
989 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
990 By this means, options can be made to have different settings in different
991 parts of the pattern. Any changes made in one alternative do carry on
992 into subsequent branches within the same subpattern. For example,
993 .sp
994 (a(?i)b|c)
995 .sp
996 matches "ab", "aB", "c", and "C", even though when matching "C" the first
997 branch is abandoned before the option setting. This is because the effects of
998 option settings happen at compile time. There would be some very weird
999 behaviour otherwise.
1000 .P
1001 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1002 changed in the same way as the Perl-compatible options by using the characters
1003 J, U and X respectively.
1004 .
1005 .
1006 .\" HTML <a name="subpattern"></a>
1008 .rs
1009 .sp
1010 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1011 Turning part of a pattern into a subpattern does two things:
1012 .sp
1013 1. It localizes a set of alternatives. For example, the pattern
1014 .sp
1015 cat(aract|erpillar|)
1016 .sp
1017 matches one of the words "cat", "cataract", or "caterpillar". Without the
1018 parentheses, it would match "cataract", "erpillar" or an empty string.
1019 .sp
1020 2. It sets up the subpattern as a capturing subpattern. This means that, when
1021 the whole pattern matches, that portion of the subject string that matched the
1022 subpattern is passed back to the caller via the \fIovector\fP argument of
1023 \fBpcre_exec()\fP. Opening parentheses are counted from left to right (starting
1024 from 1) to obtain numbers for the capturing subpatterns.
1025 .P
1026 For example, if the string "the red king" is matched against the pattern
1027 .sp
1028 the ((red|white) (king|queen))
1029 .sp
1030 the captured substrings are "red king", "red", and "king", and are numbered 1,
1031 2, and 3, respectively.
1032 .P
1033 The fact that plain parentheses fulfil two functions is not always helpful.
1034 There are often times when a grouping subpattern is required without a
1035 capturing requirement. If an opening parenthesis is followed by a question mark
1036 and a colon, the subpattern does not do any capturing, and is not counted when
1037 computing the number of any subsequent capturing subpatterns. For example, if
1038 the string "the white queen" is matched against the pattern
1039 .sp
1040 the ((?:red|white) (king|queen))
1041 .sp
1042 the captured substrings are "white queen" and "queen", and are numbered 1 and
1043 2. The maximum number of capturing subpatterns is 65535.
1044 .P
1045 As a convenient shorthand, if any option settings are required at the start of
1046 a non-capturing subpattern, the option letters may appear between the "?" and
1047 the ":". Thus the two patterns
1048 .sp
1049 (?i:saturday|sunday)
1050 (?:(?i)saturday|sunday)
1051 .sp
1052 match exactly the same set of strings. Because alternative branches are tried
1053 from left to right, and options are not reset until the end of the subpattern
1054 is reached, an option setting in one branch does affect subsequent branches, so
1055 the above patterns match "SUNDAY" as well as "Saturday".
1056 .
1057 .
1059 .rs
1060 .sp
1061 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1062 the same numbers for its capturing parentheses. Such a subpattern starts with
1063 (?| and is itself a non-capturing subpattern. For example, consider this
1064 pattern:
1065 .sp
1066 (?|(Sat)ur|(Sun))day
1067 .sp
1068 Because the two alternatives are inside a (?| group, both sets of capturing
1069 parentheses are numbered one. Thus, when the pattern matches, you can look
1070 at captured substring number one, whichever alternative matched. This construct
1071 is useful when you want to capture part, but not all, of one of a number of
1072 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1073 number is reset at the start of each branch. The numbers of any capturing
1074 buffers that follow the subpattern start after the highest number used in any
1075 branch. The following example is taken from the Perl documentation.
1076 The numbers underneath show in which buffer the captured content will be
1077 stored.
1078 .sp
1079 # before ---------------branch-reset----------- after
1080 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1081 # 1 2 2 3 2 3 4
1082 .sp
1083 A backreference or a recursive call to a numbered subpattern always refers to
1084 the first one in the pattern with the given number.
1085 .P
1086 An alternative approach to using this "branch reset" feature is to use
1087 duplicate named subpatterns, as described in the next section.
1088 .
1089 .
1091 .rs
1092 .sp
1093 Identifying capturing parentheses by number is simple, but it can be very hard
1094 to keep track of the numbers in complicated regular expressions. Furthermore,
1095 if an expression is modified, the numbers may change. To help with this
1096 difficulty, PCRE supports the naming of subpatterns. This feature was not
1097 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1098 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1099 the Perl and the Python syntax.
1100 .P
1101 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1102 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1103 parentheses from other parts of the pattern, such as
1104 .\" HTML <a href="#backreferences">
1105 .\" </a>
1106 backreferences,
1107 .\"
1108 .\" HTML <a href="#recursion">
1109 .\" </a>
1110 recursion,
1111 .\"
1112 and
1113 .\" HTML <a href="#conditions">
1114 .\" </a>
1115 conditions,
1116 .\"
1117 can be made by name as well as by number.
1118 .P
1119 Names consist of up to 32 alphanumeric characters and underscores. Named
1120 capturing parentheses are still allocated numbers as well as names, exactly as
1121 if the names were not present. The PCRE API provides function calls for
1122 extracting the name-to-number translation table from a compiled pattern. There
1123 is also a convenience function for extracting a captured substring by name.
1124 .P
1125 By default, a name must be unique within a pattern, but it is possible to relax
1126 this constraint by setting the PCRE_DUPNAMES option at compile time. This can
1127 be useful for patterns where only one instance of the named parentheses can
1128 match. Suppose you want to match the name of a weekday, either as a 3-letter
1129 abbreviation or as the full name, and in both cases you want to extract the
1130 abbreviation. This pattern (ignoring the line breaks) does the job:
1131 .sp
1132 (?<DN>Mon|Fri|Sun)(?:day)?|
1133 (?<DN>Tue)(?:sday)?|
1134 (?<DN>Wed)(?:nesday)?|
1135 (?<DN>Thu)(?:rsday)?|
1136 (?<DN>Sat)(?:urday)?
1137 .sp
1138 There are five capturing substrings, but only one is ever set after a match.
1139 (An alternative way of solving this problem is to use a "branch reset"
1140 subpattern, as described in the previous section.)
1141 .P
1142 The convenience function for extracting the data by name returns the substring
1143 for the first (and in this example, the only) subpattern of that name that
1144 matched. This saves searching to find which numbered subpattern it was. If you
1145 make a reference to a non-unique named subpattern from elsewhere in the
1146 pattern, the one that corresponds to the lowest number is used. For further
1147 details of the interfaces for handling named subpatterns, see the
1148 .\" HREF
1149 \fBpcreapi\fP
1150 .\"
1151 documentation.
1152 .
1153 .
1155 .rs
1156 .sp
1157 Repetition is specified by quantifiers, which can follow any of the following
1158 items:
1159 .sp
1160 a literal data character
1161 the dot metacharacter
1162 the \eC escape sequence
1163 the \eX escape sequence (in UTF-8 mode with Unicode properties)
1164 the \eR escape sequence
1165 an escape such as \ed that matches a single character
1166 a character class
1167 a back reference (see next section)
1168 a parenthesized subpattern (unless it is an assertion)
1169 .sp
1170 The general repetition quantifier specifies a minimum and maximum number of
1171 permitted matches, by giving the two numbers in curly brackets (braces),
1172 separated by a comma. The numbers must be less than 65536, and the first must
1173 be less than or equal to the second. For example:
1174 .sp
1175 z{2,4}
1176 .sp
1177 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1178 character. If the second number is omitted, but the comma is present, there is
1179 no upper limit; if the second number and the comma are both omitted, the
1180 quantifier specifies an exact number of required matches. Thus
1181 .sp
1182 [aeiou]{3,}
1183 .sp
1184 matches at least 3 successive vowels, but may match many more, while
1185 .sp
1186 \ed{8}
1187 .sp
1188 matches exactly 8 digits. An opening curly bracket that appears in a position
1189 where a quantifier is not allowed, or one that does not match the syntax of a
1190 quantifier, is taken as a literal character. For example, {,6} is not a
1191 quantifier, but a literal string of four characters.
1192 .P
1193 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
1194 bytes. Thus, for example, \ex{100}{2} matches two UTF-8 characters, each of
1195 which is represented by a two-byte sequence. Similarly, when Unicode property
1196 support is available, \eX{3} matches three Unicode extended sequences, each of
1197 which may be several bytes long (and they may be of different lengths).
1198 .P
1199 The quantifier {0} is permitted, causing the expression to behave as if the
1200 previous item and the quantifier were not present.
1201 .P
1202 For convenience, the three most common quantifiers have single-character
1203 abbreviations:
1204 .sp
1205 * is equivalent to {0,}
1206 + is equivalent to {1,}
1207 ? is equivalent to {0,1}
1208 .sp
1209 It is possible to construct infinite loops by following a subpattern that can
1210 match no characters with a quantifier that has no upper limit, for example:
1211 .sp
1212 (a?)*
1213 .sp
1214 Earlier versions of Perl and PCRE used to give an error at compile time for
1215 such patterns. However, because there are cases where this can be useful, such
1216 patterns are now accepted, but if any repetition of the subpattern does in fact
1217 match no characters, the loop is forcibly broken.
1218 .P
1219 By default, the quantifiers are "greedy", that is, they match as much as
1220 possible (up to the maximum number of permitted times), without causing the
1221 rest of the pattern to fail. The classic example of where this gives problems
1222 is in trying to match comments in C programs. These appear between /* and */
1223 and within the comment, individual * and / characters may appear. An attempt to
1224 match C comments by applying the pattern
1225 .sp
1226 /\e*.*\e*/
1227 .sp
1228 to the string
1229 .sp
1230 /* first comment */ not comment /* second comment */
1231 .sp
1232 fails, because it matches the entire string owing to the greediness of the .*
1233 item.
1234 .P
1235 However, if a quantifier is followed by a question mark, it ceases to be
1236 greedy, and instead matches the minimum number of times possible, so the
1237 pattern
1238 .sp
1239 /\e*.*?\e*/
1240 .sp
1241 does the right thing with the C comments. The meaning of the various
1242 quantifiers is not otherwise changed, just the preferred number of matches.
1243 Do not confuse this use of question mark with its use as a quantifier in its
1244 own right. Because it has two uses, it can sometimes appear doubled, as in
1245 .sp
1246 \ed??\ed
1247 .sp
1248 which matches one digit by preference, but can match two if that is the only
1249 way the rest of the pattern matches.
1250 .P
1251 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1252 the quantifiers are not greedy by default, but individual ones can be made
1253 greedy by following them with a question mark. In other words, it inverts the
1254 default behaviour.
1255 .P
1256 When a parenthesized subpattern is quantified with a minimum repeat count that
1257 is greater than 1 or with a limited maximum, more memory is required for the
1258 compiled pattern, in proportion to the size of the minimum or maximum.
1259 .P
1260 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1261 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1262 implicitly anchored, because whatever follows will be tried against every
1263 character position in the subject string, so there is no point in retrying the
1264 overall match at any position after the first. PCRE normally treats such a
1265 pattern as though it were preceded by \eA.
1266 .P
1267 In cases where it is known that the subject string contains no newlines, it is
1268 worth setting PCRE_DOTALL in order to obtain this optimization, or
1269 alternatively using ^ to indicate anchoring explicitly.
1270 .P
1271 However, there is one situation where the optimization cannot be used. When .*
1272 is inside capturing parentheses that are the subject of a backreference
1273 elsewhere in the pattern, a match at the start may fail where a later one
1274 succeeds. Consider, for example:
1275 .sp
1276 (.*)abc\e1
1277 .sp
1278 If the subject is "xyz123abc123" the match point is the fourth character. For
1279 this reason, such a pattern is not implicitly anchored.
1280 .P
1281 When a capturing subpattern is repeated, the value captured is the substring
1282 that matched the final iteration. For example, after
1283 .sp
1284 (tweedle[dume]{3}\es*)+
1285 .sp
1286 has matched "tweedledum tweedledee" the value of the captured substring is
1287 "tweedledee". However, if there are nested capturing subpatterns, the
1288 corresponding captured values may have been set in previous iterations. For
1289 example, after
1290 .sp
1291 /(a|(b))+/
1292 .sp
1293 matches "aba" the value of the second captured substring is "b".
1294 .
1295 .
1296 .\" HTML <a name="atomicgroup"></a>
1298 .rs
1299 .sp
1300 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1301 repetition, failure of what follows normally causes the repeated item to be
1302 re-evaluated to see if a different number of repeats allows the rest of the
1303 pattern to match. Sometimes it is useful to prevent this, either to change the
1304 nature of the match, or to cause it fail earlier than it otherwise might, when
1305 the author of the pattern knows there is no point in carrying on.
1306 .P
1307 Consider, for example, the pattern \ed+foo when applied to the subject line
1308 .sp
1309 123456bar
1310 .sp
1311 After matching all 6 digits and then failing to match "foo", the normal
1312 action of the matcher is to try again with only 5 digits matching the \ed+
1313 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1314 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1315 that once a subpattern has matched, it is not to be re-evaluated in this way.
1316 .P
1317 If we use atomic grouping for the previous example, the matcher gives up
1318 immediately on failing to match "foo" the first time. The notation is a kind of
1319 special parenthesis, starting with (?> as in this example:
1320 .sp
1321 (?>\ed+)foo
1322 .sp
1323 This kind of parenthesis "locks up" the part of the pattern it contains once
1324 it has matched, and a failure further into the pattern is prevented from
1325 backtracking into it. Backtracking past it to previous items, however, works as
1326 normal.
1327 .P
1328 An alternative description is that a subpattern of this type matches the string
1329 of characters that an identical standalone pattern would match, if anchored at
1330 the current point in the subject string.
1331 .P
1332 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1333 the above example can be thought of as a maximizing repeat that must swallow
1334 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1335 number of digits they match in order to make the rest of the pattern match,
1336 (?>\ed+) can only match an entire sequence of digits.
1337 .P
1338 Atomic groups in general can of course contain arbitrarily complicated
1339 subpatterns, and can be nested. However, when the subpattern for an atomic
1340 group is just a single repeated item, as in the example above, a simpler
1341 notation, called a "possessive quantifier" can be used. This consists of an
1342 additional + character following a quantifier. Using this notation, the
1343 previous example can be rewritten as
1344 .sp
1345 \ed++foo
1346 .sp
1347 Note that a possessive quantifier can be used with an entire group, for
1348 example:
1349 .sp
1350 (abc|xyz){2,3}+
1351 .sp
1352 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1353 option is ignored. They are a convenient notation for the simpler forms of
1354 atomic group. However, there is no difference in the meaning of a possessive
1355 quantifier and the equivalent atomic group, though there may be a performance
1356 difference; possessive quantifiers should be slightly faster.
1357 .P
1358 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1359 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1360 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1361 package, and PCRE copied it from there. It ultimately found its way into Perl
1362 at release 5.10.
1363 .P
1364 PCRE has an optimization that automatically "possessifies" certain simple
1365 pattern constructs. For example, the sequence A+B is treated as A++B because
1366 there is no point in backtracking into a sequence of A's when B must follow.
1367 .P
1368 When a pattern contains an unlimited repeat inside a subpattern that can itself
1369 be repeated an unlimited number of times, the use of an atomic group is the
1370 only way to avoid some failing matches taking a very long time indeed. The
1371 pattern
1372 .sp
1373 (\eD+|<\ed+>)*[!?]
1374 .sp
1375 matches an unlimited number of substrings that either consist of non-digits, or
1376 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1377 quickly. However, if it is applied to
1378 .sp
1379 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1380 .sp
1381 it takes a long time before reporting failure. This is because the string can
1382 be divided between the internal \eD+ repeat and the external * repeat in a
1383 large number of ways, and all have to be tried. (The example uses [!?] rather
1384 than a single character at the end, because both PCRE and Perl have an
1385 optimization that allows for fast failure when a single character is used. They
1386 remember the last single character that is required for a match, and fail early
1387 if it is not present in the string.) If the pattern is changed so that it uses
1388 an atomic group, like this:
1389 .sp
1390 ((?>\eD+)|<\ed+>)*[!?]
1391 .sp
1392 sequences of non-digits cannot be broken, and failure happens quickly.
1393 .
1394 .
1395 .\" HTML <a name="backreferences"></a>
1397 .rs
1398 .sp
1399 Outside a character class, a backslash followed by a digit greater than 0 (and
1400 possibly further digits) is a back reference to a capturing subpattern earlier
1401 (that is, to its left) in the pattern, provided there have been that many
1402 previous capturing left parentheses.
1403 .P
1404 However, if the decimal number following the backslash is less than 10, it is
1405 always taken as a back reference, and causes an error only if there are not
1406 that many capturing left parentheses in the entire pattern. In other words, the
1407 parentheses that are referenced need not be to the left of the reference for
1408 numbers less than 10. A "forward back reference" of this type can make sense
1409 when a repetition is involved and the subpattern to the right has participated
1410 in an earlier iteration.
1411 .P
1412 It is not possible to have a numerical "forward back reference" to a subpattern
1413 whose number is 10 or more using this syntax because a sequence such as \e50 is
1414 interpreted as a character defined in octal. See the subsection entitled
1415 "Non-printing characters"
1416 .\" HTML <a href="#digitsafterbackslash">
1417 .\" </a>
1418 above
1419 .\"
1420 for further details of the handling of digits following a backslash. There is
1421 no such problem when named parentheses are used. A back reference to any
1422 subpattern is possible using named parentheses (see below).
1423 .P
1424 Another way of avoiding the ambiguity inherent in the use of digits following a
1425 backslash is to use the \eg escape sequence, which is a feature introduced in
1426 Perl 5.10. This escape must be followed by an unsigned number or a negative
1427 number, optionally enclosed in braces. These examples are all identical:
1428 .sp
1429 (ring), \e1
1430 (ring), \eg1
1431 (ring), \eg{1}
1432 .sp
1433 An unsigned number specifies an absolute reference without the ambiguity that
1434 is present in the older syntax. It is also useful when literal digits follow
1435 the reference. A negative number is a relative reference. Consider this
1436 example:
1437 .sp
1438 (abc(def)ghi)\eg{-1}
1439 .sp
1440 The sequence \eg{-1} is a reference to the most recently started capturing
1441 subpattern before \eg, that is, is it equivalent to \e2. Similarly, \eg{-2}
1442 would be equivalent to \e1. The use of relative references can be helpful in
1443 long patterns, and also in patterns that are created by joining together
1444 fragments that contain references within themselves.
1445 .P
1446 A back reference matches whatever actually matched the capturing subpattern in
1447 the current subject string, rather than anything matching the subpattern
1448 itself (see
1449 .\" HTML <a href="#subpatternsassubroutines">
1450 .\" </a>
1451 "Subpatterns as subroutines"
1452 .\"
1453 below for a way of doing that). So the pattern
1454 .sp
1455 (sens|respons)e and \e1ibility
1456 .sp
1457 matches "sense and sensibility" and "response and responsibility", but not
1458 "sense and responsibility". If caseful matching is in force at the time of the
1459 back reference, the case of letters is relevant. For example,
1460 .sp
1461 ((?i)rah)\es+\e1
1462 .sp
1463 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1464 capturing subpattern is matched caselessly.
1465 .P
1466 There are several different ways of writing back references to named
1467 subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
1468 \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1469 back reference syntax, in which \eg can be used for both numeric and named
1470 references, is also supported. We could rewrite the above example in any of
1471 the following ways:
1472 .sp
1473 (?<p1>(?i)rah)\es+\ek<p1>
1474 (?'p1'(?i)rah)\es+\ek{p1}
1475 (?P<p1>(?i)rah)\es+(?P=p1)
1476 (?<p1>(?i)rah)\es+\eg{p1}
1477 .sp
1478 A subpattern that is referenced by name may appear in the pattern before or
1479 after the reference.
1480 .P
1481 There may be more than one back reference to the same subpattern. If a
1482 subpattern has not actually been used in a particular match, any back
1483 references to it always fail. For example, the pattern
1484 .sp
1485 (a|(bc))\e2
1486 .sp
1487 always fails if it starts to match "a" rather than "bc". Because there may be
1488 many capturing parentheses in a pattern, all digits following the backslash are
1489 taken as part of a potential back reference number. If the pattern continues
1490 with a digit character, some delimiter must be used to terminate the back
1491 reference. If the PCRE_EXTENDED option is set, this can be whitespace.
1492 Otherwise an empty comment (see
1493 .\" HTML <a href="#comments">
1494 .\" </a>
1495 "Comments"
1496 .\"
1497 below) can be used.
1498 .P
1499 A back reference that occurs inside the parentheses to which it refers fails
1500 when the subpattern is first used, so, for example, (a\e1) never matches.
1501 However, such references can be useful inside repeated subpatterns. For
1502 example, the pattern
1503 .sp
1504 (a|b\e1)+
1505 .sp
1506 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1507 the subpattern, the back reference matches the character string corresponding
1508 to the previous iteration. In order for this to work, the pattern must be such
1509 that the first iteration does not need to match the back reference. This can be
1510 done using alternation, as in the example above, or by a quantifier with a
1511 minimum of zero.
1512 .
1513 .
1514 .\" HTML <a name="bigassertions"></a>
1516 .rs
1517 .sp
1518 An assertion is a test on the characters following or preceding the current
1519 matching point that does not actually consume any characters. The simple
1520 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
1521 .\" HTML <a href="#smallassertions">
1522 .\" </a>
1523 above.
1524 .\"
1525 .P
1526 More complicated assertions are coded as subpatterns. There are two kinds:
1527 those that look ahead of the current position in the subject string, and those
1528 that look behind it. An assertion subpattern is matched in the normal way,
1529 except that it does not cause the current matching position to be changed.
1530 .P
1531 Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1532 because it makes no sense to assert the same thing several times. If any kind
1533 of assertion contains capturing subpatterns within it, these are counted for
1534 the purposes of numbering the capturing subpatterns in the whole pattern.
1535 However, substring capturing is carried out only for positive assertions,
1536 because it does not make sense for negative assertions.
1537 .
1538 .
1539 .SS "Lookahead assertions"
1540 .rs
1541 .sp
1542 Lookahead assertions start with (?= for positive assertions and (?! for
1543 negative assertions. For example,
1544 .sp
1545 \ew+(?=;)
1546 .sp
1547 matches a word followed by a semicolon, but does not include the semicolon in
1548 the match, and
1549 .sp
1550 foo(?!bar)
1551 .sp
1552 matches any occurrence of "foo" that is not followed by "bar". Note that the
1553 apparently similar pattern
1554 .sp
1555 (?!foo)bar
1556 .sp
1557 does not find an occurrence of "bar" that is preceded by something other than
1558 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1559 (?!foo) is always true when the next three characters are "bar". A
1560 lookbehind assertion is needed to achieve the other effect.
1561 .P
1562 If you want to force a matching failure at some point in a pattern, the most
1563 convenient way to do it is with (?!) because an empty string always matches, so
1564 an assertion that requires there not to be an empty string must always fail.
1565 .
1566 .
1567 .\" HTML <a name="lookbehind"></a>
1568 .SS "Lookbehind assertions"
1569 .rs
1570 .sp
1571 Lookbehind assertions start with (?<= for positive assertions and (?<! for
1572 negative assertions. For example,
1573 .sp
1574 (?<!foo)bar
1575 .sp
1576 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1577 a lookbehind assertion are restricted such that all the strings it matches must
1578 have a fixed length. However, if there are several top-level alternatives, they
1579 do not all have to have the same fixed length. Thus
1580 .sp
1581 (?<=bullock|donkey)
1582 .sp
1583 is permitted, but
1584 .sp
1585 (?<!dogs?|cats?)
1586 .sp
1587 causes an error at compile time. Branches that match different length strings
1588 are permitted only at the top level of a lookbehind assertion. This is an
1589 extension compared with Perl (at least for 5.8), which requires all branches to
1590 match the same length of string. An assertion such as
1591 .sp
1592 (?<=ab(c|de))
1593 .sp
1594 is not permitted, because its single top-level branch can match two different
1595 lengths, but it is acceptable if rewritten to use two top-level branches:
1596 .sp
1597 (?<=abc|abde)
1598 .sp
1599 In some cases, the Perl 5.10 escape sequence \eK
1600 .\" HTML <a href="#resetmatchstart">
1601 .\" </a>
1602 (see above)
1603 .\"
1604 can be used instead of a lookbehind assertion; this is not restricted to a
1605 fixed-length.
1606 .P
1607 The implementation of lookbehind assertions is, for each alternative, to
1608 temporarily move the current position back by the fixed length and then try to
1609 match. If there are insufficient characters before the current position, the
1610 assertion fails.
1611 .P
1612 PCRE does not allow the \eC escape (which matches a single byte in UTF-8 mode)
1613 to appear in lookbehind assertions, because it makes it impossible to calculate
1614 the length of the lookbehind. The \eX and \eR escapes, which can match
1615 different numbers of bytes, are also not permitted.
1616 .P
1617 Possessive quantifiers can be used in conjunction with lookbehind assertions to
1618 specify efficient matching at the end of the subject string. Consider a simple
1619 pattern such as
1620 .sp
1621 abcd$
1622 .sp
1623 when applied to a long string that does not match. Because matching proceeds
1624 from left to right, PCRE will look for each "a" in the subject and then see if
1625 what follows matches the rest of the pattern. If the pattern is specified as
1626 .sp
1627 ^.*abcd$
1628 .sp
1629 the initial .* matches the entire string at first, but when this fails (because
1630 there is no following "a"), it backtracks to match all but the last character,
1631 then all but the last two characters, and so on. Once again the search for "a"
1632 covers the entire string, from right to left, so we are no better off. However,
1633 if the pattern is written as
1634 .sp
1635 ^.*+(?<=abcd)
1636 .sp
1637 there can be no backtracking for the .*+ item; it can match only the entire
1638 string. The subsequent lookbehind assertion does a single test on the last four
1639 characters. If it fails, the match fails immediately. For long strings, this
1640 approach makes a significant difference to the processing time.
1641 .
1642 .
1643 .SS "Using multiple assertions"
1644 .rs
1645 .sp
1646 Several assertions (of any sort) may occur in succession. For example,
1647 .sp
1648 (?<=\ed{3})(?<!999)foo
1649 .sp
1650 matches "foo" preceded by three digits that are not "999". Notice that each of
1651 the assertions is applied independently at the same point in the subject
1652 string. First there is a check that the previous three characters are all
1653 digits, and then there is a check that the same three characters are not "999".
1654 This pattern does \fInot\fP match "foo" preceded by six characters, the first
1655 of which are digits and the last three of which are not "999". For example, it
1656 doesn't match "123abcfoo". A pattern to do that is
1657 .sp
1658 (?<=\ed{3}...)(?<!999)foo
1659 .sp
1660 This time the first assertion looks at the preceding six characters, checking
1661 that the first three are digits, and then the second assertion checks that the
1662 preceding three characters are not "999".
1663 .P
1664 Assertions can be nested in any combination. For example,
1665 .sp
1666 (?<=(?<!foo)bar)baz
1667 .sp
1668 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1669 preceded by "foo", while
1670 .sp
1671 (?<=\ed{3}(?!999)...)foo
1672 .sp
1673 is another pattern that matches "foo" preceded by three digits and any three
1674 characters that are not "999".
1675 .
1676 .
1677 .\" HTML <a name="conditions"></a>
1679 .rs
1680 .sp
1681 It is possible to cause the matching process to obey a subpattern
1682 conditionally or to choose between two alternative subpatterns, depending on
1683 the result of an assertion, or whether a previous capturing subpattern matched
1684 or not. The two possible forms of conditional subpattern are
1685 .sp
1686 (?(condition)yes-pattern)
1687 (?(condition)yes-pattern|no-pattern)
1688 .sp
1689 If the condition is satisfied, the yes-pattern is used; otherwise the
1690 no-pattern (if present) is used. If there are more than two alternatives in the
1691 subpattern, a compile-time error occurs.
1692 .P
1693 There are four kinds of condition: references to subpatterns, references to
1694 recursion, a pseudo-condition called DEFINE, and assertions.
1695 .
1696 .SS "Checking for a used subpattern by number"
1697 .rs
1698 .sp
1699 If the text between the parentheses consists of a sequence of digits, the
1700 condition is true if the capturing subpattern of that number has previously
1701 matched. An alternative notation is to precede the digits with a plus or minus
1702 sign. In this case, the subpattern number is relative rather than absolute.
1703 The most recently opened parentheses can be referenced by (?(-1), the next most
1704 recent by (?(-2), and so on. In looping constructs it can also make sense to
1705 refer to subsequent groups with constructs such as (?(+2).
1706 .P
1707 Consider the following pattern, which contains non-significant white space to
1708 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
1709 three parts for ease of discussion:
1710 .sp
1711 ( \e( )? [^()]+ (?(1) \e) )
1712 .sp
1713 The first part matches an optional opening parenthesis, and if that
1714 character is present, sets it as the first captured substring. The second part
1715 matches one or more characters that are not parentheses. The third part is a
1716 conditional subpattern that tests whether the first set of parentheses matched
1717 or not. If they did, that is, if subject started with an opening parenthesis,
1718 the condition is true, and so the yes-pattern is executed and a closing
1719 parenthesis is required. Otherwise, since no-pattern is not present, the
1720 subpattern matches nothing. In other words, this pattern matches a sequence of
1721 non-parentheses, optionally enclosed in parentheses.
1722 .P
1723 If you were embedding this pattern in a larger one, you could use a relative
1724 reference:
1725 .sp
1726 ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
1727 .sp
1728 This makes the fragment independent of the parentheses in the larger pattern.
1729 .
1730 .SS "Checking for a used subpattern by name"
1731 .rs
1732 .sp
1733 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
1734 subpattern by name. For compatibility with earlier versions of PCRE, which had
1735 this facility before Perl, the syntax (?(name)...) is also recognized. However,
1736 there is a possible ambiguity with this syntax, because subpattern names may
1737 consist entirely of digits. PCRE looks first for a named subpattern; if it
1738 cannot find one and the name consists entirely of digits, PCRE looks for a
1739 subpattern of that number, which must be greater than zero. Using subpattern
1740 names that consist entirely of digits is not recommended.
1741 .P
1742 Rewriting the above example to use a named subpattern gives this:
1743 .sp
1744 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
1745 .sp
1746 .
1747 .SS "Checking for pattern recursion"
1748 .rs
1749 .sp
1750 If the condition is the string (R), and there is no subpattern with the name R,
1751 the condition is true if a recursive call to the whole pattern or any
1752 subpattern has been made. If digits or a name preceded by ampersand follow the
1753 letter R, for example:
1754 .sp
1755 (?(R3)...) or (?(R&name)...)
1756 .sp
1757 the condition is true if the most recent recursion is into the subpattern whose
1758 number or name is given. This condition does not check the entire recursion
1759 stack.
1760 .P
1761 At "top level", all these recursion test conditions are false. Recursive
1762 patterns are described below.
1763 .
1764 .SS "Defining subpatterns for use by reference only"
1765 .rs
1766 .sp
1767 If the condition is the string (DEFINE), and there is no subpattern with the
1768 name DEFINE, the condition is always false. In this case, there may be only one
1769 alternative in the subpattern. It is always skipped if control reaches this
1770 point in the pattern; the idea of DEFINE is that it can be used to define
1771 "subroutines" that can be referenced from elsewhere. (The use of "subroutines"
1772 is described below.) For example, a pattern to match an IPv4 address could be
1773 written like this (ignore whitespace and line breaks):
1774 .sp
1775 (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
1776 \eb (?&byte) (\e.(?&byte)){3} \eb
1777 .sp
1778 The first part of the pattern is a DEFINE group inside which a another group
1779 named "byte" is defined. This matches an individual component of an IPv4
1780 address (a number less than 256). When matching takes place, this part of the
1781 pattern is skipped because DEFINE acts like a false condition.
1782 .P
1783 The rest of the pattern uses references to the named group to match the four
1784 dot-separated components of an IPv4 address, insisting on a word boundary at
1785 each end.
1786 .
1787 .SS "Assertion conditions"
1788 .rs
1789 .sp
1790 If the condition is not in any of the above formats, it must be an assertion.
1791 This may be a positive or negative lookahead or lookbehind assertion. Consider
1792 this pattern, again containing non-significant white space, and with the two
1793 alternatives on the second line:
1794 .sp
1795 (?(?=[^a-z]*[a-z])
1796 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
1797 .sp
1798 The condition is a positive lookahead assertion that matches an optional
1799 sequence of non-letters followed by a letter. In other words, it tests for the
1800 presence of at least one letter in the subject. If a letter is found, the
1801 subject is matched against the first alternative; otherwise it is matched
1802 against the second. This pattern matches strings in one of the two forms
1803 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
1804 .
1805 .
1806 .\" HTML <a name="comments"></a>
1808 .rs
1809 .sp
1810 The sequence (?# marks the start of a comment that continues up to the next
1811 closing parenthesis. Nested parentheses are not permitted. The characters
1812 that make up a comment play no part in the pattern matching at all.
1813 .P
1814 If the PCRE_EXTENDED option is set, an unescaped # character outside a
1815 character class introduces a comment that continues to immediately after the
1816 next newline in the pattern.
1817 .
1818 .
1819 .\" HTML <a name="recursion"></a>
1821 .rs
1822 .sp
1823 Consider the problem of matching a string in parentheses, allowing for
1824 unlimited nested parentheses. Without the use of recursion, the best that can
1825 be done is to use a pattern that matches up to some fixed depth of nesting. It
1826 is not possible to handle an arbitrary nesting depth.
1827 .P
1828 For some time, Perl has provided a facility that allows regular expressions to
1829 recurse (amongst other things). It does this by interpolating Perl code in the
1830 expression at run time, and the code can refer to the expression itself. A Perl
1831 pattern using code interpolation to solve the parentheses problem can be
1832 created like this:
1833 .sp
1834 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
1835 .sp
1836 The (?p{...}) item interpolates Perl code at run time, and in this case refers
1837 recursively to the pattern in which it appears.
1838 .P
1839 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
1840 supports special syntax for recursion of the entire pattern, and also for
1841 individual subpattern recursion. After its introduction in PCRE and Python,
1842 this kind of recursion was introduced into Perl at release 5.10.
1843 .P
1844 A special item that consists of (? followed by a number greater than zero and a
1845 closing parenthesis is a recursive call of the subpattern of the given number,
1846 provided that it occurs inside that subpattern. (If not, it is a "subroutine"
1847 call, which is described in the next section.) The special item (?R) or (?0) is
1848 a recursive call of the entire regular expression.
1849 .P
1850 In PCRE (like Python, but unlike Perl), a recursive subpattern call is always
1851 treated as an atomic group. That is, once it has matched some of the subject
1852 string, it is never re-entered, even if it contains untried alternatives and
1853 there is a subsequent matching failure.
1854 .P
1855 This PCRE pattern solves the nested parentheses problem (assume the
1856 PCRE_EXTENDED option is set so that white space is ignored):
1857 .sp
1858 \e( ( (?>[^()]+) | (?R) )* \e)
1859 .sp
1860 First it matches an opening parenthesis. Then it matches any number of
1861 substrings which can either be a sequence of non-parentheses, or a recursive
1862 match of the pattern itself (that is, a correctly parenthesized substring).
1863 Finally there is a closing parenthesis.
1864 .P
1865 If this were part of a larger pattern, you would not want to recurse the entire
1866 pattern, so instead you could use this:
1867 .sp
1868 ( \e( ( (?>[^()]+) | (?1) )* \e) )
1869 .sp
1870 We have put the pattern into parentheses, and caused the recursion to refer to
1871 them instead of the whole pattern.
1872 .P
1873 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
1874 is made easier by the use of relative references. (A Perl 5.10 feature.)
1875 Instead of (?1) in the pattern above you can write (?-2) to refer to the second
1876 most recently opened parentheses preceding the recursion. In other words, a
1877 negative number counts capturing parentheses leftwards from the point at which
1878 it is encountered.
1879 .P
1880 It is also possible to refer to subsequently opened parentheses, by writing
1881 references such as (?+2). However, these cannot be recursive because the
1882 reference is not inside the parentheses that are referenced. They are always
1883 "subroutine" calls, as described in the next section.
1884 .P
1885 An alternative approach is to use named parentheses instead. The Perl syntax
1886 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
1887 could rewrite the above example as follows:
1888 .sp
1889 (?<pn> \e( ( (?>[^()]+) | (?&pn) )* \e) )
1890 .sp
1891 If there is more than one subpattern with the same name, the earliest one is
1892 used.
1893 .P
1894 This particular example pattern that we have been looking at contains nested
1895 unlimited repeats, and so the use of atomic grouping for matching strings of
1896 non-parentheses is important when applying the pattern to strings that do not
1897 match. For example, when this pattern is applied to
1898 .sp
1899 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
1900 .sp
1901 it yields "no match" quickly. However, if atomic grouping is not used,
1902 the match runs for a very long time indeed because there are so many different
1903 ways the + and * repeats can carve up the subject, and all have to be tested
1904 before failure can be reported.
1905 .P
1906 At the end of a match, the values set for any capturing subpatterns are those
1907 from the outermost level of the recursion at which the subpattern value is set.
1908 If you want to obtain intermediate values, a callout function can be used (see
1909 below and the
1910 .\" HREF
1911 \fBpcrecallout\fP
1912 .\"
1913 documentation). If the pattern above is matched against
1914 .sp
1915 (ab(cd)ef)
1916 .sp
1917 the value for the capturing parentheses is "ef", which is the last value taken
1918 on at the top level. If additional parentheses are added, giving
1919 .sp
1920 \e( ( ( (?>[^()]+) | (?R) )* ) \e)
1921 ^ ^
1922 ^ ^
1923 .sp
1924 the string they capture is "ab(cd)ef", the contents of the top level
1925 parentheses. If there are more than 15 capturing parentheses in a pattern, PCRE
1926 has to obtain extra memory to store data during a recursion, which it does by
1927 using \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no
1928 memory can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
1929 .P
1930 Do not confuse the (?R) item with the condition (R), which tests for recursion.
1931 Consider this pattern, which matches text in angle brackets, allowing for
1932 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
1933 recursing), whereas any characters are permitted at the outer level.
1934 .sp
1935 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
1936 .sp
1937 In this pattern, (?(R) is the start of a conditional subpattern, with two
1938 different alternatives for the recursive and non-recursive cases. The (?R) item
1939 is the actual recursive call.
1940 .
1941 .
1942 .\" HTML <a name="subpatternsassubroutines"></a>
1944 .rs
1945 .sp
1946 If the syntax for a recursive subpattern reference (either by number or by
1947 name) is used outside the parentheses to which it refers, it operates like a
1948 subroutine in a programming language. The "called" subpattern may be defined
1949 before or after the reference. A numbered reference can be absolute or
1950 relative, as in these examples:
1951 .sp
1952 (...(absolute)...)...(?2)...
1953 (...(relative)...)...(?-1)...
1954 (...(?+1)...(relative)...
1955 .sp
1956 An earlier example pointed out that the pattern
1957 .sp
1958 (sens|respons)e and \e1ibility
1959 .sp
1960 matches "sense and sensibility" and "response and responsibility", but not
1961 "sense and responsibility". If instead the pattern
1962 .sp
1963 (sens|respons)e and (?1)ibility
1964 .sp
1965 is used, it does match "sense and responsibility" as well as the other two
1966 strings. Another example is given in the discussion of DEFINE above.
1967 .P
1968 Like recursive subpatterns, a "subroutine" call is always treated as an atomic
1969 group. That is, once it has matched some of the subject string, it is never
1970 re-entered, even if it contains untried alternatives and there is a subsequent
1971 matching failure.
1972 .P
1973 When a subpattern is used as a subroutine, processing options such as
1974 case-independence are fixed when the subpattern is defined. They cannot be
1975 changed for different calls. For example, consider this pattern:
1976 .sp
1977 (abc)(?i:(?-1))
1978 .sp
1979 It matches "abcabc". It does not match "abcABC" because the change of
1980 processing option does not affect the called subpattern.
1981 .
1982 .
1984 .rs
1985 .sp
1986 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
1987 code to be obeyed in the middle of matching a regular expression. This makes it
1988 possible, amongst other things, to extract different substrings that match the
1989 same pair of parentheses when there is a repetition.
1990 .P
1991 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
1992 code. The feature is called "callout". The caller of PCRE provides an external
1993 function by putting its entry point in the global variable \fIpcre_callout\fP.
1994 By default, this variable contains NULL, which disables all calling out.
1995 .P
1996 Within a regular expression, (?C) indicates the points at which the external
1997 function is to be called. If you want to identify different callout points, you
1998 can put a number less than 256 after the letter C. The default value is zero.
1999 For example, this pattern has two callout points:
2000 .sp
2001 (?C1)abc(?C2)def
2002 .sp
2003 If the PCRE_AUTO_CALLOUT flag is passed to \fBpcre_compile()\fP, callouts are
2004 automatically installed before each item in the pattern. They are all numbered
2005 255.
2006 .P
2007 During matching, when PCRE reaches a callout point (and \fIpcre_callout\fP is
2008 set), the external function is called. It is provided with the number of the
2009 callout, the position in the pattern, and, optionally, one item of data
2010 originally supplied by the caller of \fBpcre_exec()\fP. The callout function
2011 may cause matching to proceed, to backtrack, or to fail altogether. A complete
2012 description of the interface to the callout function is given in the
2013 .\" HREF
2014 \fBpcrecallout\fP
2015 .\"
2016 documentation.
2017 .
2018 .
2020 .rs
2021 .sp
2022 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2023 are described in the Perl documentation as "experimental and subject to change
2024 or removal in a future version of Perl". It goes on to say: "Their usage in
2025 production code should be noted to avoid problems during upgrades." The same
2026 remarks apply to the PCRE features described in this section.
2027 .P
2028 Since these verbs are specifically related to backtracking, they can be used
2029 only when the pattern is to be matched using \fBpcre_exec()\fP, which uses a
2030 backtracking algorithm. They cause an error if encountered by
2031 \fBpcre_dfa_exec()\fP.
2032 .P
2033 The new verbs make use of what was previously invalid syntax: an opening
2034 parenthesis followed by an asterisk. In Perl, they are generally of the form
2035 (*VERB:ARG) but PCRE does not support the use of arguments, so its general
2036 form is just (*VERB). Any number of these verbs may occur in a pattern. There
2037 are two kinds:
2038 .
2039 .SS "Verbs that act immediately"
2040 .rs
2041 .sp
2042 The following verbs act as soon as they are encountered:
2043 .sp
2044 (*ACCEPT)
2045 .sp
2046 This verb causes the match to end successfully, skipping the remainder of the
2047 pattern. When inside a recursion, only the innermost pattern is ended
2048 immediately. PCRE differs from Perl in what happens if the (*ACCEPT) is inside
2049 capturing parentheses. In Perl, the data so far is captured: in PCRE no data is
2050 captured. For example:
2051 .sp
2052 A(A|B(*ACCEPT)|C)D
2053 .sp
2054 This matches "AB", "AAD", or "ACD", but when it matches "AB", no data is
2055 captured.
2056 .sp
2057 (*FAIL) or (*F)
2058 .sp
2059 This verb causes the match to fail, forcing backtracking to occur. It is
2060 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2061 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2062 Perl features that are not present in PCRE. The nearest equivalent is the
2063 callout feature, as for example in this pattern:
2064 .sp
2065 a+(?C)(*FAIL)
2066 .sp
2067 A match with the string "aaaa" always fails, but the callout is taken before
2068 each backtrack happens (in this example, 10 times).
2069 .
2070 .SS "Verbs that act after backtracking"
2071 .rs
2072 .sp
2073 The following verbs do nothing when they are encountered. Matching continues
2074 with what follows, but if there is no subsequent match, a failure is forced.
2075 The verbs differ in exactly what kind of failure occurs.
2076 .sp
2077 (*COMMIT)
2078 .sp
2079 This verb causes the whole match to fail outright if the rest of the pattern
2080 does not match. Even if the pattern is unanchored, no further attempts to find
2081 a match by advancing the start point take place. Once (*COMMIT) has been
2082 passed, \fBpcre_exec()\fP is committed to finding a match at the current
2083 starting point, or not at all. For example:
2084 .sp
2085 a+(*COMMIT)b
2086 .sp
2087 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2088 dynamic anchor, or "I've started, so I must finish."
2089 .sp
2090 (*PRUNE)
2091 .sp
2092 This verb causes the match to fail at the current position if the rest of the
2093 pattern does not match. If the pattern is unanchored, the normal "bumpalong"
2094 advance to the next starting character then happens. Backtracking can occur as
2095 usual to the left of (*PRUNE), or when matching to the right of (*PRUNE), but
2096 if there is no match to the right, backtracking cannot cross (*PRUNE).
2097 In simple cases, the use of (*PRUNE) is just an alternative to an atomic
2098 group or possessive quantifier, but there are some uses of (*PRUNE) that cannot
2099 be expressed in any other way.
2100 .sp
2101 (*SKIP)
2102 .sp
2103 This verb is like (*PRUNE), except that if the pattern is unanchored, the
2104 "bumpalong" advance is not to the next character, but to the position in the
2105 subject where (*SKIP) was encountered. (*SKIP) signifies that whatever text
2106 was matched leading up to it cannot be part of a successful match. Consider:
2107 .sp
2108 a+(*SKIP)b
2109 .sp
2110 If the subject is "aaaac...", after the first match attempt fails (starting at
2111 the first character in the string), the starting point skips on to start the
2112 next attempt at "c". Note that a possessive quantifer does not have the same
2113 effect in this example; although it would suppress backtracking during the
2114 first match attempt, the second attempt would start at the second character
2115 instead of skipping on to "c".
2116 .sp
2117 (*THEN)
2118 .sp
2119 This verb causes a skip to the next alternation if the rest of the pattern does
2120 not match. That is, it cancels pending backtracking, but only within the
2121 current alternation. Its name comes from the observation that it can be used
2122 for a pattern-based if-then-else block:
2123 .sp
2124 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2125 .sp
2126 If the COND1 pattern matches, FOO is tried (and possibly further items after
2127 the end of the group if FOO succeeds); on failure the matcher skips to the
2128 second alternative and tries COND2, without backtracking into COND1. If (*THEN)
2129 is used outside of any alternation, it acts exactly like (*PRUNE).
2130 .
2131 .
2132 .SH "SEE ALSO"
2133 .rs
2134 .sp
2135 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3), \fBpcre\fP(3).
2136 .
2137 .
2139 .rs
2140 .sp
2141 .nf
2142 Philip Hazel
2143 University Computing Service
2144 Cambridge CB2 3QH, England.
2145 .fi
2146 .
2147 .
2149 .rs
2150 .sp
2151 .nf
2152 Last updated: 21 August 2007
2153 Copyright (c) 1997-2007 University of Cambridge.
2154 .fi


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

  ViewVC Help
Powered by ViewVC 1.1.5