/[pcre]/code/trunk/doc/pcrepattern.3
ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log


Revision 211 - (show annotations) (download)
Thu Aug 9 09:52:43 2007 UTC (6 years, 8 months ago) by ph10
File size: 86179 byte(s)
Update UTF-8 validity check and documentation.

1 .TH PCREPATTERN 3
2 .SH NAME
3 PCRE - Perl-compatible regular expressions
4 .SH "PCRE REGULAR EXPRESSION DETAILS"
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 .
47 .SH "CHARACTERS AND METACHARACTERS"
48 .rs
49 .sp
50 A regular expression is a pattern that is matched against a subject string from
51 left to right. Most characters stand for themselves in a pattern, and match the
52 corresponding characters in the subject. As a trivial example, the pattern
53 .sp
54 The quick brown fox
55 .sp
56 matches a portion of a subject string that is identical to itself. When
57 caseless matching is specified (the PCRE_CASELESS option), letters are matched
58 independently of case. In UTF-8 mode, PCRE always understands the concept of
59 case for characters whose values are less than 128, so caseless matching is
60 always possible. For characters with higher values, the concept of case is
61 supported if PCRE is compiled with Unicode property support, but not otherwise.
62 If you want to use caseless matching for characters 128 and above, you must
63 ensure that PCRE is compiled with Unicode property support as well as with
64 UTF-8 support.
65 .P
66 The power of regular expressions comes from the ability to include alternatives
67 and repetitions in the pattern. These are encoded in the pattern by the use of
68 \fImetacharacters\fP, which do not stand for themselves but instead are
69 interpreted in some special way.
70 .P
71 There are two different sets of metacharacters: those that are recognized
72 anywhere in the pattern except within square brackets, and those that are
73 recognized within square brackets. Outside square brackets, the metacharacters
74 are as follows:
75 .sp
76 \e general escape character with several uses
77 ^ assert start of string (or line, in multiline mode)
78 $ assert end of string (or line, in multiline mode)
79 . match any character except newline (by default)
80 [ start character class definition
81 | start of alternative branch
82 ( start subpattern
83 ) end subpattern
84 ? extends the meaning of (
85 also 0 or 1 quantifier
86 also quantifier minimizer
87 * 0 or more quantifier
88 + 1 or more quantifier
89 also "possessive quantifier"
90 { start min/max quantifier
91 .sp
92 Part of a pattern that is in square brackets is called a "character class". In
93 a character class the only metacharacters are:
94 .sp
95 \e general escape character
96 ^ negate the class, but only if the first character
97 - indicates character range
98 .\" JOIN
99 [ POSIX character class (only if followed by POSIX
100 syntax)
101 ] terminates the character class
102 .sp
103 The following sections describe the use of each of the metacharacters.
104 .
105 .
106 .SH BACKSLASH
107 .rs
108 .sp
109 The backslash character has several uses. Firstly, if it is followed by a
110 non-alphanumeric character, it takes away any special meaning that character
111 may have. This use of backslash as an escape character applies both inside and
112 outside character classes.
113 .P
114 For example, if you want to match a * character, you write \e* in the pattern.
115 This escaping action applies whether or not the following character would
116 otherwise be interpreted as a metacharacter, so it is always safe to precede a
117 non-alphanumeric with backslash to specify that it stands for itself. In
118 particular, if you want to match a backslash, you write \e\e.
119 .P
120 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
121 pattern (other than in a character class) and characters between a # outside
122 a character class and the next newline are ignored. An escaping backslash can
123 be used to include a whitespace or # character as part of the pattern.
124 .P
125 If you want to remove the special meaning from a sequence of characters, you
126 can do so by putting them between \eQ and \eE. This is different from Perl in
127 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
128 Perl, $ and @ cause variable interpolation. Note the following examples:
129 .sp
130 Pattern PCRE matches Perl matches
131 .sp
132 .\" JOIN
133 \eQabc$xyz\eE abc$xyz abc followed by the
134 contents of $xyz
135 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
136 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
137 .sp
138 The \eQ...\eE sequence is recognized both inside and outside character classes.
139 .
140 .
141 .\" HTML <a name="digitsafterbackslash"></a>
142 .SS "Non-printing characters"
143 .rs
144 .sp
145 A second use of backslash provides a way of encoding non-printing characters
146 in patterns in a visible manner. There is no restriction on the appearance of
147 non-printing characters, apart from the binary zero that terminates a pattern,
148 but when a pattern is being prepared by text editing, it is usually easier to
149 use one of the following escape sequences than the binary character it
150 represents:
151 .sp
152 \ea alarm, that is, the BEL character (hex 07)
153 \ecx "control-x", where x is any character
154 \ee escape (hex 1B)
155 \ef formfeed (hex 0C)
156 \en newline (hex 0A)
157 \er carriage return (hex 0D)
158 \et tab (hex 09)
159 \eddd character with octal code ddd, or backreference
160 \exhh character with hex code hh
161 \ex{hhh..} character with hex code hhh..
162 .sp
163 The precise effect of \ecx is as follows: if x is a lower case letter, it
164 is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
165 Thus \ecz becomes hex 1A, but \ec{ becomes hex 3B, while \ec; becomes hex
166 7B.
167 .P
168 After \ex, from zero to two hexadecimal digits are read (letters can be in
169 upper or lower case). Any number of hexadecimal digits may appear between \ex{
170 and }, but the value of the character code must be less than 256 in non-UTF-8
171 mode, and less than 2**31 in UTF-8 mode. That is, the maximum value in
172 hexadecimal is 7FFFFFFF. Note that this is bigger than the largest Unicode code
173 point, which is 10FFFF.
174 .P
175 If characters other than hexadecimal digits appear between \ex{ and }, or if
176 there is no terminating }, this form of escape is not recognized. Instead, the
177 initial \ex will be interpreted as a basic hexadecimal escape, with no
178 following digits, giving a character whose value is zero.
179 .P
180 Characters whose value is less than 256 can be defined by either of the two
181 syntaxes for \ex. There is no difference in the way they are handled. For
182 example, \exdc is exactly the same as \ex{dc}.
183 .P
184 After \e0 up to two further octal digits are read. If there are fewer than two
185 digits, just those that are present are used. Thus the sequence \e0\ex\e07
186 specifies two binary zeros followed by a BEL character (code value 7). Make
187 sure you supply two digits after the initial zero if the pattern character that
188 follows is itself an octal digit.
189 .P
190 The handling of a backslash followed by a digit other than 0 is complicated.
191 Outside a character class, PCRE reads it and any following digits as a decimal
192 number. If the number is less than 10, or if there have been at least that many
193 previous capturing left parentheses in the expression, the entire sequence is
194 taken as a \fIback reference\fP. A description of how this works is given
195 .\" HTML <a href="#backreferences">
196 .\" </a>
197 later,
198 .\"
199 following the discussion of
200 .\" HTML <a href="#subpattern">
201 .\" </a>
202 parenthesized subpatterns.
203 .\"
204 .P
205 Inside a character class, or if the decimal number is greater than 9 and there
206 have not been that many capturing subpatterns, PCRE re-reads up to three octal
207 digits following the backslash, and uses them to generate a data character. Any
208 subsequent digits stand for themselves. In non-UTF-8 mode, the value of a
209 character specified in octal must be less than \e400. In UTF-8 mode, values up
210 to \e777 are permitted. For example:
211 .sp
212 \e040 is another way of writing a space
213 .\" JOIN
214 \e40 is the same, provided there are fewer than 40
215 previous capturing subpatterns
216 \e7 is always a back reference
217 .\" JOIN
218 \e11 might be a back reference, or another way of
219 writing a tab
220 \e011 is always a tab
221 \e0113 is a tab followed by the character "3"
222 .\" JOIN
223 \e113 might be a back reference, otherwise the
224 character with octal code 113
225 .\" JOIN
226 \e377 might be a back reference, otherwise
227 the byte consisting entirely of 1 bits
228 .\" JOIN
229 \e81 is either a back reference, or a binary zero
230 followed by the two characters "8" and "1"
231 .sp
232 Note that octal values of 100 or greater must not be introduced by a leading
233 zero, because no more than three octal digits are ever read.
234 .P
235 All the sequences that define a single character value can be used both inside
236 and outside character classes. In addition, inside a character class, the
237 sequence \eb is interpreted as the backspace character (hex 08), and the
238 sequences \eR and \eX are interpreted as the characters "R" and "X",
239 respectively. Outside a character class, these sequences have different
240 meanings
241 .\" HTML <a href="#uniextseq">
242 .\" </a>
243 (see below).
244 .\"
245 .
246 .
247 .SS "Absolute and relative back references"
248 .rs
249 .sp
250 The sequence \eg followed by an unsigned or a negative number, optionally
251 enclosed in braces, is an absolute or relative back reference. A named back
252 reference can be coded as \eg{name}. Back references are discussed
253 .\" HTML <a href="#backreferences">
254 .\" </a>
255 later,
256 .\"
257 following the discussion of
258 .\" HTML <a href="#subpattern">
259 .\" </a>
260 parenthesized subpatterns.
261 .\"
262 .
263 .
264 .SS "Generic character types"
265 .rs
266 .sp
267 Another use of backslash is for specifying generic character types. The
268 following are always recognized:
269 .sp
270 \ed any decimal digit
271 \eD any character that is not a decimal digit
272 \eh any horizontal whitespace character
273 \eH any character that is not a horizontal whitespace character
274 \es any whitespace character
275 \eS any character that is not a whitespace character
276 \ev any vertical whitespace character
277 \eV any character that is not a vertical whitespace character
278 \ew any "word" character
279 \eW any "non-word" character
280 .sp
281 Each pair of escape sequences partitions the complete set of characters into
282 two disjoint sets. Any given character matches one, and only one, of each pair.
283 .P
284 These character type sequences can appear both inside and outside character
285 classes. They each match one character of the appropriate type. If the current
286 matching point is at the end of the subject string, all of them fail, since
287 there is no character to match.
288 .P
289 For compatibility with Perl, \es does not match the VT character (code 11).
290 This makes it different from the the POSIX "space" class. The \es characters
291 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
292 included in a Perl script, \es may match the VT character. In PCRE, it never
293 does.
294 .P
295 In UTF-8 mode, characters with values greater than 128 never match \ed, \es, or
296 \ew, and always match \eD, \eS, and \eW. This is true even when Unicode
297 character property support is available. These sequences retain their original
298 meanings from before UTF-8 support was available, mainly for efficiency
299 reasons.
300 .P
301 The sequences \eh, \eH, \ev, and \eV are Perl 5.10 features. In contrast to the
302 other sequences, these do match certain high-valued codepoints in UTF-8 mode.
303 The horizontal space characters are:
304 .sp
305 U+0009 Horizontal tab
306 U+0020 Space
307 U+00A0 Non-break space
308 U+1680 Ogham space mark
309 U+180E Mongolian vowel separator
310 U+2000 En quad
311 U+2001 Em quad
312 U+2002 En space
313 U+2003 Em space
314 U+2004 Three-per-em space
315 U+2005 Four-per-em space
316 U+2006 Six-per-em space
317 U+2007 Figure space
318 U+2008 Punctuation space
319 U+2009 Thin space
320 U+200A Hair space
321 U+202F Narrow no-break space
322 U+205F Medium mathematical space
323 U+3000 Ideographic space
324 .sp
325 The vertical space characters are:
326 .sp
327 U+000A Linefeed
328 U+000B Vertical tab
329 U+000C Formfeed
330 U+000D Carriage return
331 U+0085 Next line
332 U+2028 Line separator
333 U+2029 Paragraph separator
334 .P
335 A "word" character is an underscore or any character less than 256 that is a
336 letter or digit. The definition of letters and digits is controlled by PCRE's
337 low-valued character tables, and may vary if locale-specific matching is taking
338 place (see
339 .\" HTML <a href="pcreapi.html#localesupport">
340 .\" </a>
341 "Locale support"
342 .\"
343 in the
344 .\" HREF
345 \fBpcreapi\fP
346 .\"
347 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
348 or "french" in Windows, some character codes greater than 128 are used for
349 accented letters, and these are matched by \ew. The use of locales with Unicode
350 is discouraged.
351 .
352 .
353 .SS "Newline sequences"
354 .rs
355 .sp
356 Outside a character class, the escape sequence \eR matches any Unicode newline
357 sequence. This is a Perl 5.10 feature. In non-UTF-8 mode \eR is equivalent to
358 the following:
359 .sp
360 (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
361 .sp
362 This is an example of an "atomic group", details of which are given
363 .\" HTML <a href="#atomicgroup">
364 .\" </a>
365 below.
366 .\"
367 This particular group matches either the two-character sequence CR followed by
368 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
369 U+000B), FF (formfeed, U+000C), CR (carriage return, U+000D), or NEL (next
370 line, U+0085). The two-character sequence is treated as a single unit that
371 cannot be split.
372 .P
373 In UTF-8 mode, two additional characters whose codepoints are greater than 255
374 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
375 Unicode character property support is not needed for these characters to be
376 recognized.
377 .P
378 Inside a character class, \eR matches the letter "R".
379 .
380 .
381 .\" HTML <a name="uniextseq"></a>
382 .SS Unicode character properties
383 .rs
384 .sp
385 When PCRE is built with Unicode character property support, three additional
386 escape sequences that match characters with specific properties are available.
387 When not in UTF-8 mode, these sequences are of course limited to testing
388 characters whose codepoints are less than 256, but they do work in this mode.
389 The extra escape sequences are:
390 .sp
391 \ep{\fIxx\fP} a character with the \fIxx\fP property
392 \eP{\fIxx\fP} a character without the \fIxx\fP property
393 \eX an extended Unicode sequence
394 .sp
395 The property names represented by \fIxx\fP above are limited to the Unicode
396 script names, the general category properties, and "Any", which matches any
397 character (including newline). Other properties such as "InMusicalSymbols" are
398 not currently supported by PCRE. Note that \eP{Any} does not match any
399 characters, so always causes a match failure.
400 .P
401 Sets of Unicode characters are defined as belonging to certain scripts. A
402 character from one of these sets can be matched using a script name. For
403 example:
404 .sp
405 \ep{Greek}
406 \eP{Han}
407 .sp
408 Those that are not part of an identified script are lumped together as
409 "Common". The current list of scripts is:
410 .P
411 Arabic,
412 Armenian,
413 Balinese,
414 Bengali,
415 Bopomofo,
416 Braille,
417 Buginese,
418 Buhid,
419 Canadian_Aboriginal,
420 Cherokee,
421 Common,
422 Coptic,
423 Cuneiform,
424 Cypriot,
425 Cyrillic,
426 Deseret,
427 Devanagari,
428 Ethiopic,
429 Georgian,
430 Glagolitic,
431 Gothic,
432 Greek,
433 Gujarati,
434 Gurmukhi,
435 Han,
436 Hangul,
437 Hanunoo,
438 Hebrew,
439 Hiragana,
440 Inherited,
441 Kannada,
442 Katakana,
443 Kharoshthi,
444 Khmer,
445 Lao,
446 Latin,
447 Limbu,
448 Linear_B,
449 Malayalam,
450 Mongolian,
451 Myanmar,
452 New_Tai_Lue,
453 Nko,
454 Ogham,
455 Old_Italic,
456 Old_Persian,
457 Oriya,
458 Osmanya,
459 Phags_Pa,
460 Phoenician,
461 Runic,
462 Shavian,
463 Sinhala,
464 Syloti_Nagri,
465 Syriac,
466 Tagalog,
467 Tagbanwa,
468 Tai_Le,
469 Tamil,
470 Telugu,
471 Thaana,
472 Thai,
473 Tibetan,
474 Tifinagh,
475 Ugaritic,
476 Yi.
477 .P
478 Each character has exactly one general category property, specified by a
479 two-letter abbreviation. For compatibility with Perl, negation can be specified
480 by including a circumflex between the opening brace and the property name. For
481 example, \ep{^Lu} is the same as \eP{Lu}.
482 .P
483 If only one letter is specified with \ep or \eP, it includes all the general
484 category properties that start with that letter. In this case, in the absence
485 of negation, the curly brackets in the escape sequence are optional; these two
486 examples have the same effect:
487 .sp
488 \ep{L}
489 \epL
490 .sp
491 The following general category property codes are supported:
492 .sp
493 C Other
494 Cc Control
495 Cf Format
496 Cn Unassigned
497 Co Private use
498 Cs Surrogate
499 .sp
500 L Letter
501 Ll Lower case letter
502 Lm Modifier letter
503 Lo Other letter
504 Lt Title case letter
505 Lu Upper case letter
506 .sp
507 M Mark
508 Mc Spacing mark
509 Me Enclosing mark
510 Mn Non-spacing mark
511 .sp
512 N Number
513 Nd Decimal number
514 Nl Letter number
515 No Other number
516 .sp
517 P Punctuation
518 Pc Connector punctuation
519 Pd Dash punctuation
520 Pe Close punctuation
521 Pf Final punctuation
522 Pi Initial punctuation
523 Po Other punctuation
524 Ps Open punctuation
525 .sp
526 S Symbol
527 Sc Currency symbol
528 Sk Modifier symbol
529 Sm Mathematical symbol
530 So Other symbol
531 .sp
532 Z Separator
533 Zl Line separator
534 Zp Paragraph separator
535 Zs Space separator
536 .sp
537 The special property L& is also supported: it matches a character that has
538 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
539 a modifier or "other".
540 .P
541 The Cs (Surrogate) property applies only to characters in the range U+D800 to
542 U+DFFF. Such characters are not valid in UTF-8 strings (see RFC 3629) and so
543 cannot be tested by PCRE, unless UTF-8 validity checking has been turned off
544 (see the discussion of PCRE_NO_UTF8_CHECK in the
545 .\" HREF
546 \fBpcreapi\fP
547 .\"
548 page).
549 .P
550 The long synonyms for these properties that Perl supports (such as \ep{Letter})
551 are not supported by PCRE, nor is it permitted to prefix any of these
552 properties with "Is".
553 .P
554 No character that is in the Unicode table has the Cn (unassigned) property.
555 Instead, this property is assumed for any code point that is not in the
556 Unicode table.
557 .P
558 Specifying caseless matching does not affect these escape sequences. For
559 example, \ep{Lu} always matches only upper case letters.
560 .P
561 The \eX escape matches any number of Unicode characters that form an extended
562 Unicode sequence. \eX is equivalent to
563 .sp
564 (?>\ePM\epM*)
565 .sp
566 That is, it matches a character without the "mark" property, followed by zero
567 or more characters with the "mark" property, and treats the sequence as an
568 atomic group
569 .\" HTML <a href="#atomicgroup">
570 .\" </a>
571 (see below).
572 .\"
573 Characters with the "mark" property are typically accents that affect the
574 preceding character. None of them have codepoints less than 256, so in
575 non-UTF-8 mode \eX matches any one character.
576 .P
577 Matching characters by Unicode property is not fast, because PCRE has to search
578 a structure that contains data for over fifteen thousand characters. That is
579 why the traditional escape sequences such as \ed and \ew do not use Unicode
580 properties in PCRE.
581 .
582 .
583 .\" HTML <a name="resetmatchstart"></a>
584 .SS "Resetting the match start"
585 .rs
586 .sp
587 The escape sequence \eK, which is a Perl 5.10 feature, causes any previously
588 matched characters not to be included in the final matched sequence. For
589 example, the pattern:
590 .sp
591 foo\eKbar
592 .sp
593 matches "foobar", but reports that it has matched "bar". This feature is
594 similar to a lookbehind assertion
595 .\" HTML <a href="#lookbehind">
596 .\" </a>
597 (described below).
598 .\"
599 However, in this case, the part of the subject before the real match does not
600 have to be of fixed length, as lookbehind assertions do. The use of \eK does
601 not interfere with the setting of
602 .\" HTML <a href="#subpattern">
603 .\" </a>
604 captured substrings.
605 .\"
606 For example, when the pattern
607 .sp
608 (foo)\eKbar
609 .sp
610 matches "foobar", the first substring is still set to "foo".
611 .
612 .
613 .\" HTML <a name="smallassertions"></a>
614 .SS "Simple assertions"
615 .rs
616 .sp
617 The final use of backslash is for certain simple assertions. An assertion
618 specifies a condition that has to be met at a particular point in a match,
619 without consuming any characters from the subject string. The use of
620 subpatterns for more complicated assertions is described
621 .\" HTML <a href="#bigassertions">
622 .\" </a>
623 below.
624 .\"
625 The backslashed assertions are:
626 .sp
627 \eb matches at a word boundary
628 \eB matches when not at a word boundary
629 \eA matches at the start of the subject
630 \eZ matches at the end of the subject
631 also matches before a newline at the end of the subject
632 \ez matches only at the end of the subject
633 \eG matches at the first matching position in the subject
634 .sp
635 These assertions may not appear in character classes (but note that \eb has a
636 different meaning, namely the backspace character, inside a character class).
637 .P
638 A word boundary is a position in the subject string where the current character
639 and the previous character do not both match \ew or \eW (i.e. one matches
640 \ew and the other matches \eW), or the start or end of the string if the
641 first or last character matches \ew, respectively.
642 .P
643 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
644 dollar (described in the next section) in that they only ever match at the very
645 start and end of the subject string, whatever options are set. Thus, they are
646 independent of multiline mode. These three assertions are not affected by the
647 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
648 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
649 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
650 at a point other than the beginning of the subject, \eA can never match. The
651 difference between \eZ and \ez is that \eZ matches before a newline at the end
652 of the string as well as at the very end, whereas \ez matches only at the end.
653 .P
654 The \eG assertion is true only when the current matching position is at the
655 start point of the match, as specified by the \fIstartoffset\fP argument of
656 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
657 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
658 arguments, you can mimic Perl's /g option, and it is in this kind of
659 implementation where \eG can be useful.
660 .P
661 Note, however, that PCRE's interpretation of \eG, as the start of the current
662 match, is subtly different from Perl's, which defines it as the end of the
663 previous match. In Perl, these can be different when the previously matched
664 string was empty. Because PCRE does just one match at a time, it cannot
665 reproduce this behaviour.
666 .P
667 If all the alternatives of a pattern begin with \eG, the expression is anchored
668 to the starting match position, and the "anchored" flag is set in the compiled
669 regular expression.
670 .
671 .
672 .SH "CIRCUMFLEX AND DOLLAR"
673 .rs
674 .sp
675 Outside a character class, in the default matching mode, the circumflex
676 character is an assertion that is true only if the current matching point is
677 at the start of the subject string. If the \fIstartoffset\fP argument of
678 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
679 option is unset. Inside a character class, circumflex has an entirely different
680 meaning
681 .\" HTML <a href="#characterclass">
682 .\" </a>
683 (see below).
684 .\"
685 .P
686 Circumflex need not be the first character of the pattern if a number of
687 alternatives are involved, but it should be the first thing in each alternative
688 in which it appears if the pattern is ever to match that branch. If all
689 possible alternatives start with a circumflex, that is, if the pattern is
690 constrained to match only at the start of the subject, it is said to be an
691 "anchored" pattern. (There are also other constructs that can cause a pattern
692 to be anchored.)
693 .P
694 A dollar character is an assertion that is true only if the current matching
695 point is at the end of the subject string, or immediately before a newline
696 at the end of the string (by default). Dollar need not be the last character of
697 the pattern if a number of alternatives are involved, but it should be the last
698 item in any branch in which it appears. Dollar has no special meaning in a
699 character class.
700 .P
701 The meaning of dollar can be changed so that it matches only at the very end of
702 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
703 does not affect the \eZ assertion.
704 .P
705 The meanings of the circumflex and dollar characters are changed if the
706 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
707 immediately after internal newlines as well as at the start of the subject
708 string. It does not match after a newline that ends the string. A dollar
709 matches before any newlines in the string, as well as at the very end, when
710 PCRE_MULTILINE is set. When newline is specified as the two-character
711 sequence CRLF, isolated CR and LF characters do not indicate newlines.
712 .P
713 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
714 \en represents a newline) in multiline mode, but not otherwise. Consequently,
715 patterns that are anchored in single line mode because all branches start with
716 ^ are not anchored in multiline mode, and a match for circumflex is possible
717 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
718 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
719 .P
720 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
721 end of the subject in both modes, and if all branches of a pattern start with
722 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
723 .
724 .
725 .SH "FULL STOP (PERIOD, DOT)"
726 .rs
727 .sp
728 Outside a character class, a dot in the pattern matches any one character in
729 the subject string except (by default) a character that signifies the end of a
730 line. In UTF-8 mode, the matched character may be more than one byte long.
731 .P
732 When a line ending is defined as a single character, dot never matches that
733 character; when the two-character sequence CRLF is used, dot does not match CR
734 if it is immediately followed by LF, but otherwise it matches all characters
735 (including isolated CRs and LFs). When any Unicode line endings are being
736 recognized, dot does not match CR or LF or any of the other line ending
737 characters.
738 .P
739 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
740 option is set, a dot matches any one character, without exception. If the
741 two-character sequence CRLF is present in the subject string, it takes two dots
742 to match it.
743 .P
744 The handling of dot is entirely independent of the handling of circumflex and
745 dollar, the only relationship being that they both involve newlines. Dot has no
746 special meaning in a character class.
747 .
748 .
749 .SH "MATCHING A SINGLE BYTE"
750 .rs
751 .sp
752 Outside a character class, the escape sequence \eC matches any one byte, both
753 in and out of UTF-8 mode. Unlike a dot, it always matches any line-ending
754 characters. The feature is provided in Perl in order to match individual bytes
755 in UTF-8 mode. Because it breaks up UTF-8 characters into individual bytes,
756 what remains in the string may be a malformed UTF-8 string. For this reason,
757 the \eC escape sequence is best avoided.
758 .P
759 PCRE does not allow \eC to appear in lookbehind assertions
760 .\" HTML <a href="#lookbehind">
761 .\" </a>
762 (described below),
763 .\"
764 because in UTF-8 mode this would make it impossible to calculate the length of
765 the lookbehind.
766 .
767 .
768 .\" HTML <a name="characterclass"></a>
769 .SH "SQUARE BRACKETS AND CHARACTER CLASSES"
770 .rs
771 .sp
772 An opening square bracket introduces a character class, terminated by a closing
773 square bracket. A closing square bracket on its own is not special. If a
774 closing square bracket is required as a member of the class, it should be the
775 first data character in the class (after an initial circumflex, if present) or
776 escaped with a backslash.
777 .P
778 A character class matches a single character in the subject. In UTF-8 mode, the
779 character may occupy more than one byte. A matched character must be in the set
780 of characters defined by the class, unless the first character in the class
781 definition is a circumflex, in which case the subject character must not be in
782 the set defined by the class. If a circumflex is actually required as a member
783 of the class, ensure it is not the first character, or escape it with a
784 backslash.
785 .P
786 For example, the character class [aeiou] matches any lower case vowel, while
787 [^aeiou] matches any character that is not a lower case vowel. Note that a
788 circumflex is just a convenient notation for specifying the characters that
789 are in the class by enumerating those that are not. A class that starts with a
790 circumflex is not an assertion: it still consumes a character from the subject
791 string, and therefore it fails if the current pointer is at the end of the
792 string.
793 .P
794 In UTF-8 mode, characters with values greater than 255 can be included in a
795 class as a literal string of bytes, or by using the \ex{ escaping mechanism.
796 .P
797 When caseless matching is set, any letters in a class represent both their
798 upper case and lower case versions, so for example, a caseless [aeiou] matches
799 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
800 caseful version would. In UTF-8 mode, PCRE always understands the concept of
801 case for characters whose values are less than 128, so caseless matching is
802 always possible. For characters with higher values, the concept of case is
803 supported if PCRE is compiled with Unicode property support, but not otherwise.
804 If you want to use caseless matching for characters 128 and above, you must
805 ensure that PCRE is compiled with Unicode property support as well as with
806 UTF-8 support.
807 .P
808 Characters that might indicate line breaks are never treated in any special way
809 when matching character classes, whatever line-ending sequence is in use, and
810 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
811 such as [^a] always matches one of these characters.
812 .P
813 The minus (hyphen) character can be used to specify a range of characters in a
814 character class. For example, [d-m] matches any letter between d and m,
815 inclusive. If a minus character is required in a class, it must be escaped with
816 a backslash or appear in a position where it cannot be interpreted as
817 indicating a range, typically as the first or last character in the class.
818 .P
819 It is not possible to have the literal character "]" as the end character of a
820 range. A pattern such as [W-]46] is interpreted as a class of two characters
821 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
822 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
823 the end of range, so [W-\e]46] is interpreted as a class containing a range
824 followed by two other characters. The octal or hexadecimal representation of
825 "]" can also be used to end a range.
826 .P
827 Ranges operate in the collating sequence of character values. They can also be
828 used for characters specified numerically, for example [\e000-\e037]. In UTF-8
829 mode, ranges can include characters whose values are greater than 255, for
830 example [\ex{100}-\ex{2ff}].
831 .P
832 If a range that includes letters is used when caseless matching is set, it
833 matches the letters in either case. For example, [W-c] is equivalent to
834 [][\e\e^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
835 tables for a French locale are in use, [\exc8-\excb] matches accented E
836 characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
837 characters with values greater than 128 only when it is compiled with Unicode
838 property support.
839 .P
840 The character types \ed, \eD, \ep, \eP, \es, \eS, \ew, and \eW may also appear
841 in a character class, and add the characters that they match to the class. For
842 example, [\edABCDEF] matches any hexadecimal digit. A circumflex can
843 conveniently be used with the upper case character types to specify a more
844 restricted set of characters than the matching lower case type. For example,
845 the class [^\eW_] matches any letter or digit, but not underscore.
846 .P
847 The only metacharacters that are recognized in character classes are backslash,
848 hyphen (only where it can be interpreted as specifying a range), circumflex
849 (only at the start), opening square bracket (only when it can be interpreted as
850 introducing a POSIX class name - see the next section), and the terminating
851 closing square bracket. However, escaping other non-alphanumeric characters
852 does no harm.
853 .
854 .
855 .SH "POSIX CHARACTER CLASSES"
856 .rs
857 .sp
858 Perl supports the POSIX notation for character classes. This uses names
859 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
860 this notation. For example,
861 .sp
862 [01[:alpha:]%]
863 .sp
864 matches "0", "1", any alphabetic character, or "%". The supported class names
865 are
866 .sp
867 alnum letters and digits
868 alpha letters
869 ascii character codes 0 - 127
870 blank space or tab only
871 cntrl control characters
872 digit decimal digits (same as \ed)
873 graph printing characters, excluding space
874 lower lower case letters
875 print printing characters, including space
876 punct printing characters, excluding letters and digits
877 space white space (not quite the same as \es)
878 upper upper case letters
879 word "word" characters (same as \ew)
880 xdigit hexadecimal digits
881 .sp
882 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
883 space (32). Notice that this list includes the VT character (code 11). This
884 makes "space" different to \es, which does not include VT (for Perl
885 compatibility).
886 .P
887 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
888 5.8. Another Perl extension is negation, which is indicated by a ^ character
889 after the colon. For example,
890 .sp
891 [12[:^digit:]]
892 .sp
893 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
894 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
895 supported, and an error is given if they are encountered.
896 .P
897 In UTF-8 mode, characters with values greater than 128 do not match any of
898 the POSIX character classes.
899 .
900 .
901 .SH "VERTICAL BAR"
902 .rs
903 .sp
904 Vertical bar characters are used to separate alternative patterns. For example,
905 the pattern
906 .sp
907 gilbert|sullivan
908 .sp
909 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
910 and an empty alternative is permitted (matching the empty string). The matching
911 process tries each alternative in turn, from left to right, and the first one
912 that succeeds is used. If the alternatives are within a subpattern
913 .\" HTML <a href="#subpattern">
914 .\" </a>
915 (defined below),
916 .\"
917 "succeeds" means matching the rest of the main pattern as well as the
918 alternative in the subpattern.
919 .
920 .
921 .SH "INTERNAL OPTION SETTING"
922 .rs
923 .sp
924 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
925 PCRE_EXTENDED options can be changed from within the pattern by a sequence of
926 Perl option letters enclosed between "(?" and ")". The option letters are
927 .sp
928 i for PCRE_CASELESS
929 m for PCRE_MULTILINE
930 s for PCRE_DOTALL
931 x for PCRE_EXTENDED
932 .sp
933 For example, (?im) sets caseless, multiline matching. It is also possible to
934 unset these options by preceding the letter with a hyphen, and a combined
935 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
936 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
937 permitted. If a letter appears both before and after the hyphen, the option is
938 unset.
939 .P
940 When an option change occurs at top level (that is, not inside subpattern
941 parentheses), the change applies to the remainder of the pattern that follows.
942 If the change is placed right at the start of a pattern, PCRE extracts it into
943 the global options (and it will therefore show up in data extracted by the
944 \fBpcre_fullinfo()\fP function).
945 .P
946 An option change within a subpattern (see below for a description of
947 subpatterns) affects only that part of the current pattern that follows it, so
948 .sp
949 (a(?i)b)c
950 .sp
951 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
952 By this means, options can be made to have different settings in different
953 parts of the pattern. Any changes made in one alternative do carry on
954 into subsequent branches within the same subpattern. For example,
955 .sp
956 (a(?i)b|c)
957 .sp
958 matches "ab", "aB", "c", and "C", even though when matching "C" the first
959 branch is abandoned before the option setting. This is because the effects of
960 option settings happen at compile time. There would be some very weird
961 behaviour otherwise.
962 .P
963 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
964 changed in the same way as the Perl-compatible options by using the characters
965 J, U and X respectively.
966 .
967 .
968 .\" HTML <a name="subpattern"></a>
969 .SH SUBPATTERNS
970 .rs
971 .sp
972 Subpatterns are delimited by parentheses (round brackets), which can be nested.
973 Turning part of a pattern into a subpattern does two things:
974 .sp
975 1. It localizes a set of alternatives. For example, the pattern
976 .sp
977 cat(aract|erpillar|)
978 .sp
979 matches one of the words "cat", "cataract", or "caterpillar". Without the
980 parentheses, it would match "cataract", "erpillar" or an empty string.
981 .sp
982 2. It sets up the subpattern as a capturing subpattern. This means that, when
983 the whole pattern matches, that portion of the subject string that matched the
984 subpattern is passed back to the caller via the \fIovector\fP argument of
985 \fBpcre_exec()\fP. Opening parentheses are counted from left to right (starting
986 from 1) to obtain numbers for the capturing subpatterns.
987 .P
988 For example, if the string "the red king" is matched against the pattern
989 .sp
990 the ((red|white) (king|queen))
991 .sp
992 the captured substrings are "red king", "red", and "king", and are numbered 1,
993 2, and 3, respectively.
994 .P
995 The fact that plain parentheses fulfil two functions is not always helpful.
996 There are often times when a grouping subpattern is required without a
997 capturing requirement. If an opening parenthesis is followed by a question mark
998 and a colon, the subpattern does not do any capturing, and is not counted when
999 computing the number of any subsequent capturing subpatterns. For example, if
1000 the string "the white queen" is matched against the pattern
1001 .sp
1002 the ((?:red|white) (king|queen))
1003 .sp
1004 the captured substrings are "white queen" and "queen", and are numbered 1 and
1005 2. The maximum number of capturing subpatterns is 65535.
1006 .P
1007 As a convenient shorthand, if any option settings are required at the start of
1008 a non-capturing subpattern, the option letters may appear between the "?" and
1009 the ":". Thus the two patterns
1010 .sp
1011 (?i:saturday|sunday)
1012 (?:(?i)saturday|sunday)
1013 .sp
1014 match exactly the same set of strings. Because alternative branches are tried
1015 from left to right, and options are not reset until the end of the subpattern
1016 is reached, an option setting in one branch does affect subsequent branches, so
1017 the above patterns match "SUNDAY" as well as "Saturday".
1018 .
1019 .
1020 .SH "DUPLICATE SUBPATTERN NUMBERS"
1021 .rs
1022 .sp
1023 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1024 the same numbers for its capturing parentheses. Such a subpattern starts with
1025 (?| and is itself a non-capturing subpattern. For example, consider this
1026 pattern:
1027 .sp
1028 (?|(Sat)ur|(Sun))day
1029 .sp
1030 Because the two alternatives are inside a (?| group, both sets of capturing
1031 parentheses are numbered one. Thus, when the pattern matches, you can look
1032 at captured substring number one, whichever alternative matched. This construct
1033 is useful when you want to capture part, but not all, of one of a number of
1034 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1035 number is reset at the start of each branch. The numbers of any capturing
1036 buffers that follow the subpattern start after the highest number used in any
1037 branch. The following example is taken from the Perl documentation.
1038 The numbers underneath show in which buffer the captured content will be
1039 stored.
1040 .sp
1041 # before ---------------branch-reset----------- after
1042 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1043 # 1 2 2 3 2 3 4
1044 .sp
1045 A backreference or a recursive call to a numbered subpattern always refers to
1046 the first one in the pattern with the given number.
1047 .P
1048 An alternative approach to using this "branch reset" feature is to use
1049 duplicate named subpatterns, as described in the next section.
1050 .
1051 .
1052 .SH "NAMED SUBPATTERNS"
1053 .rs
1054 .sp
1055 Identifying capturing parentheses by number is simple, but it can be very hard
1056 to keep track of the numbers in complicated regular expressions. Furthermore,
1057 if an expression is modified, the numbers may change. To help with this
1058 difficulty, PCRE supports the naming of subpatterns. This feature was not
1059 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1060 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1061 the Perl and the Python syntax.
1062 .P
1063 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1064 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1065 parentheses from other parts of the pattern, such as
1066 .\" HTML <a href="#backreferences">
1067 .\" </a>
1068 backreferences,
1069 .\"
1070 .\" HTML <a href="#recursion">
1071 .\" </a>
1072 recursion,
1073 .\"
1074 and
1075 .\" HTML <a href="#conditions">
1076 .\" </a>
1077 conditions,
1078 .\"
1079 can be made by name as well as by number.
1080 .P
1081 Names consist of up to 32 alphanumeric characters and underscores. Named
1082 capturing parentheses are still allocated numbers as well as names, exactly as
1083 if the names were not present. The PCRE API provides function calls for
1084 extracting the name-to-number translation table from a compiled pattern. There
1085 is also a convenience function for extracting a captured substring by name.
1086 .P
1087 By default, a name must be unique within a pattern, but it is possible to relax
1088 this constraint by setting the PCRE_DUPNAMES option at compile time. This can
1089 be useful for patterns where only one instance of the named parentheses can
1090 match. Suppose you want to match the name of a weekday, either as a 3-letter
1091 abbreviation or as the full name, and in both cases you want to extract the
1092 abbreviation. This pattern (ignoring the line breaks) does the job:
1093 .sp
1094 (?<DN>Mon|Fri|Sun)(?:day)?|
1095 (?<DN>Tue)(?:sday)?|
1096 (?<DN>Wed)(?:nesday)?|
1097 (?<DN>Thu)(?:rsday)?|
1098 (?<DN>Sat)(?:urday)?
1099 .sp
1100 There are five capturing substrings, but only one is ever set after a match.
1101 (An alternative way of solving this problem is to use a "branch reset"
1102 subpattern, as described in the previous section.)
1103 .P
1104 The convenience function for extracting the data by name returns the substring
1105 for the first (and in this example, the only) subpattern of that name that
1106 matched. This saves searching to find which numbered subpattern it was. If you
1107 make a reference to a non-unique named subpattern from elsewhere in the
1108 pattern, the one that corresponds to the lowest number is used. For further
1109 details of the interfaces for handling named subpatterns, see the
1110 .\" HREF
1111 \fBpcreapi\fP
1112 .\"
1113 documentation.
1114 .
1115 .
1116 .SH REPETITION
1117 .rs
1118 .sp
1119 Repetition is specified by quantifiers, which can follow any of the following
1120 items:
1121 .sp
1122 a literal data character
1123 the dot metacharacter
1124 the \eC escape sequence
1125 the \eX escape sequence (in UTF-8 mode with Unicode properties)
1126 the \eR escape sequence
1127 an escape such as \ed that matches a single character
1128 a character class
1129 a back reference (see next section)
1130 a parenthesized subpattern (unless it is an assertion)
1131 .sp
1132 The general repetition quantifier specifies a minimum and maximum number of
1133 permitted matches, by giving the two numbers in curly brackets (braces),
1134 separated by a comma. The numbers must be less than 65536, and the first must
1135 be less than or equal to the second. For example:
1136 .sp
1137 z{2,4}
1138 .sp
1139 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1140 character. If the second number is omitted, but the comma is present, there is
1141 no upper limit; if the second number and the comma are both omitted, the
1142 quantifier specifies an exact number of required matches. Thus
1143 .sp
1144 [aeiou]{3,}
1145 .sp
1146 matches at least 3 successive vowels, but may match many more, while
1147 .sp
1148 \ed{8}
1149 .sp
1150 matches exactly 8 digits. An opening curly bracket that appears in a position
1151 where a quantifier is not allowed, or one that does not match the syntax of a
1152 quantifier, is taken as a literal character. For example, {,6} is not a
1153 quantifier, but a literal string of four characters.
1154 .P
1155 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
1156 bytes. Thus, for example, \ex{100}{2} matches two UTF-8 characters, each of
1157 which is represented by a two-byte sequence. Similarly, when Unicode property
1158 support is available, \eX{3} matches three Unicode extended sequences, each of
1159 which may be several bytes long (and they may be of different lengths).
1160 .P
1161 The quantifier {0} is permitted, causing the expression to behave as if the
1162 previous item and the quantifier were not present.
1163 .P
1164 For convenience, the three most common quantifiers have single-character
1165 abbreviations:
1166 .sp
1167 * is equivalent to {0,}
1168 + is equivalent to {1,}
1169 ? is equivalent to {0,1}
1170 .sp
1171 It is possible to construct infinite loops by following a subpattern that can
1172 match no characters with a quantifier that has no upper limit, for example:
1173 .sp
1174 (a?)*
1175 .sp
1176 Earlier versions of Perl and PCRE used to give an error at compile time for
1177 such patterns. However, because there are cases where this can be useful, such
1178 patterns are now accepted, but if any repetition of the subpattern does in fact
1179 match no characters, the loop is forcibly broken.
1180 .P
1181 By default, the quantifiers are "greedy", that is, they match as much as
1182 possible (up to the maximum number of permitted times), without causing the
1183 rest of the pattern to fail. The classic example of where this gives problems
1184 is in trying to match comments in C programs. These appear between /* and */
1185 and within the comment, individual * and / characters may appear. An attempt to
1186 match C comments by applying the pattern
1187 .sp
1188 /\e*.*\e*/
1189 .sp
1190 to the string
1191 .sp
1192 /* first comment */ not comment /* second comment */
1193 .sp
1194 fails, because it matches the entire string owing to the greediness of the .*
1195 item.
1196 .P
1197 However, if a quantifier is followed by a question mark, it ceases to be
1198 greedy, and instead matches the minimum number of times possible, so the
1199 pattern
1200 .sp
1201 /\e*.*?\e*/
1202 .sp
1203 does the right thing with the C comments. The meaning of the various
1204 quantifiers is not otherwise changed, just the preferred number of matches.
1205 Do not confuse this use of question mark with its use as a quantifier in its
1206 own right. Because it has two uses, it can sometimes appear doubled, as in
1207 .sp
1208 \ed??\ed
1209 .sp
1210 which matches one digit by preference, but can match two if that is the only
1211 way the rest of the pattern matches.
1212 .P
1213 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1214 the quantifiers are not greedy by default, but individual ones can be made
1215 greedy by following them with a question mark. In other words, it inverts the
1216 default behaviour.
1217 .P
1218 When a parenthesized subpattern is quantified with a minimum repeat count that
1219 is greater than 1 or with a limited maximum, more memory is required for the
1220 compiled pattern, in proportion to the size of the minimum or maximum.
1221 .P
1222 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1223 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1224 implicitly anchored, because whatever follows will be tried against every
1225 character position in the subject string, so there is no point in retrying the
1226 overall match at any position after the first. PCRE normally treats such a
1227 pattern as though it were preceded by \eA.
1228 .P
1229 In cases where it is known that the subject string contains no newlines, it is
1230 worth setting PCRE_DOTALL in order to obtain this optimization, or
1231 alternatively using ^ to indicate anchoring explicitly.
1232 .P
1233 However, there is one situation where the optimization cannot be used. When .*
1234 is inside capturing parentheses that are the subject of a backreference
1235 elsewhere in the pattern, a match at the start may fail where a later one
1236 succeeds. Consider, for example:
1237 .sp
1238 (.*)abc\e1
1239 .sp
1240 If the subject is "xyz123abc123" the match point is the fourth character. For
1241 this reason, such a pattern is not implicitly anchored.
1242 .P
1243 When a capturing subpattern is repeated, the value captured is the substring
1244 that matched the final iteration. For example, after
1245 .sp
1246 (tweedle[dume]{3}\es*)+
1247 .sp
1248 has matched "tweedledum tweedledee" the value of the captured substring is
1249 "tweedledee". However, if there are nested capturing subpatterns, the
1250 corresponding captured values may have been set in previous iterations. For
1251 example, after
1252 .sp
1253 /(a|(b))+/
1254 .sp
1255 matches "aba" the value of the second captured substring is "b".
1256 .
1257 .
1258 .\" HTML <a name="atomicgroup"></a>
1259 .SH "ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS"
1260 .rs
1261 .sp
1262 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1263 repetition, failure of what follows normally causes the repeated item to be
1264 re-evaluated to see if a different number of repeats allows the rest of the
1265 pattern to match. Sometimes it is useful to prevent this, either to change the
1266 nature of the match, or to cause it fail earlier than it otherwise might, when
1267 the author of the pattern knows there is no point in carrying on.
1268 .P
1269 Consider, for example, the pattern \ed+foo when applied to the subject line
1270 .sp
1271 123456bar
1272 .sp
1273 After matching all 6 digits and then failing to match "foo", the normal
1274 action of the matcher is to try again with only 5 digits matching the \ed+
1275 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1276 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1277 that once a subpattern has matched, it is not to be re-evaluated in this way.
1278 .P
1279 If we use atomic grouping for the previous example, the matcher gives up
1280 immediately on failing to match "foo" the first time. The notation is a kind of
1281 special parenthesis, starting with (?> as in this example:
1282 .sp
1283 (?>\ed+)foo
1284 .sp
1285 This kind of parenthesis "locks up" the part of the pattern it contains once
1286 it has matched, and a failure further into the pattern is prevented from
1287 backtracking into it. Backtracking past it to previous items, however, works as
1288 normal.
1289 .P
1290 An alternative description is that a subpattern of this type matches the string
1291 of characters that an identical standalone pattern would match, if anchored at
1292 the current point in the subject string.
1293 .P
1294 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1295 the above example can be thought of as a maximizing repeat that must swallow
1296 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1297 number of digits they match in order to make the rest of the pattern match,
1298 (?>\ed+) can only match an entire sequence of digits.
1299 .P
1300 Atomic groups in general can of course contain arbitrarily complicated
1301 subpatterns, and can be nested. However, when the subpattern for an atomic
1302 group is just a single repeated item, as in the example above, a simpler
1303 notation, called a "possessive quantifier" can be used. This consists of an
1304 additional + character following a quantifier. Using this notation, the
1305 previous example can be rewritten as
1306 .sp
1307 \ed++foo
1308 .sp
1309 Note that a possessive quantifier can be used with an entire group, for
1310 example:
1311 .sp
1312 (abc|xyz){2,3}+
1313 .sp
1314 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1315 option is ignored. They are a convenient notation for the simpler forms of
1316 atomic group. However, there is no difference in the meaning of a possessive
1317 quantifier and the equivalent atomic group, though there may be a performance
1318 difference; possessive quantifiers should be slightly faster.
1319 .P
1320 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1321 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1322 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1323 package, and PCRE copied it from there. It ultimately found its way into Perl
1324 at release 5.10.
1325 .P
1326 PCRE has an optimization that automatically "possessifies" certain simple
1327 pattern constructs. For example, the sequence A+B is treated as A++B because
1328 there is no point in backtracking into a sequence of A's when B must follow.
1329 .P
1330 When a pattern contains an unlimited repeat inside a subpattern that can itself
1331 be repeated an unlimited number of times, the use of an atomic group is the
1332 only way to avoid some failing matches taking a very long time indeed. The
1333 pattern
1334 .sp
1335 (\eD+|<\ed+>)*[!?]
1336 .sp
1337 matches an unlimited number of substrings that either consist of non-digits, or
1338 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1339 quickly. However, if it is applied to
1340 .sp
1341 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1342 .sp
1343 it takes a long time before reporting failure. This is because the string can
1344 be divided between the internal \eD+ repeat and the external * repeat in a
1345 large number of ways, and all have to be tried. (The example uses [!?] rather
1346 than a single character at the end, because both PCRE and Perl have an
1347 optimization that allows for fast failure when a single character is used. They
1348 remember the last single character that is required for a match, and fail early
1349 if it is not present in the string.) If the pattern is changed so that it uses
1350 an atomic group, like this:
1351 .sp
1352 ((?>\eD+)|<\ed+>)*[!?]
1353 .sp
1354 sequences of non-digits cannot be broken, and failure happens quickly.
1355 .
1356 .
1357 .\" HTML <a name="backreferences"></a>
1358 .SH "BACK REFERENCES"
1359 .rs
1360 .sp
1361 Outside a character class, a backslash followed by a digit greater than 0 (and
1362 possibly further digits) is a back reference to a capturing subpattern earlier
1363 (that is, to its left) in the pattern, provided there have been that many
1364 previous capturing left parentheses.
1365 .P
1366 However, if the decimal number following the backslash is less than 10, it is
1367 always taken as a back reference, and causes an error only if there are not
1368 that many capturing left parentheses in the entire pattern. In other words, the
1369 parentheses that are referenced need not be to the left of the reference for
1370 numbers less than 10. A "forward back reference" of this type can make sense
1371 when a repetition is involved and the subpattern to the right has participated
1372 in an earlier iteration.
1373 .P
1374 It is not possible to have a numerical "forward back reference" to a subpattern
1375 whose number is 10 or more using this syntax because a sequence such as \e50 is
1376 interpreted as a character defined in octal. See the subsection entitled
1377 "Non-printing characters"
1378 .\" HTML <a href="#digitsafterbackslash">
1379 .\" </a>
1380 above
1381 .\"
1382 for further details of the handling of digits following a backslash. There is
1383 no such problem when named parentheses are used. A back reference to any
1384 subpattern is possible using named parentheses (see below).
1385 .P
1386 Another way of avoiding the ambiguity inherent in the use of digits following a
1387 backslash is to use the \eg escape sequence, which is a feature introduced in
1388 Perl 5.10. This escape must be followed by an unsigned number or a negative
1389 number, optionally enclosed in braces. These examples are all identical:
1390 .sp
1391 (ring), \e1
1392 (ring), \eg1
1393 (ring), \eg{1}
1394 .sp
1395 An unsigned number specifies an absolute reference without the ambiguity that
1396 is present in the older syntax. It is also useful when literal digits follow
1397 the reference. A negative number is a relative reference. Consider this
1398 example:
1399 .sp
1400 (abc(def)ghi)\eg{-1}
1401 .sp
1402 The sequence \eg{-1} is a reference to the most recently started capturing
1403 subpattern before \eg, that is, is it equivalent to \e2. Similarly, \eg{-2}
1404 would be equivalent to \e1. The use of relative references can be helpful in
1405 long patterns, and also in patterns that are created by joining together
1406 fragments that contain references within themselves.
1407 .P
1408 A back reference matches whatever actually matched the capturing subpattern in
1409 the current subject string, rather than anything matching the subpattern
1410 itself (see
1411 .\" HTML <a href="#subpatternsassubroutines">
1412 .\" </a>
1413 "Subpatterns as subroutines"
1414 .\"
1415 below for a way of doing that). So the pattern
1416 .sp
1417 (sens|respons)e and \e1ibility
1418 .sp
1419 matches "sense and sensibility" and "response and responsibility", but not
1420 "sense and responsibility". If caseful matching is in force at the time of the
1421 back reference, the case of letters is relevant. For example,
1422 .sp
1423 ((?i)rah)\es+\e1
1424 .sp
1425 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1426 capturing subpattern is matched caselessly.
1427 .P
1428 There are several different ways of writing back references to named
1429 subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
1430 \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1431 back reference syntax, in which \eg can be used for both numeric and named
1432 references, is also supported. We could rewrite the above example in any of
1433 the following ways:
1434 .sp
1435 (?<p1>(?i)rah)\es+\ek<p1>
1436 (?'p1'(?i)rah)\es+\ek{p1}
1437 (?P<p1>(?i)rah)\es+(?P=p1)
1438 (?<p1>(?i)rah)\es+\eg{p1}
1439 .sp
1440 A subpattern that is referenced by name may appear in the pattern before or
1441 after the reference.
1442 .P
1443 There may be more than one back reference to the same subpattern. If a
1444 subpattern has not actually been used in a particular match, any back
1445 references to it always fail. For example, the pattern
1446 .sp
1447 (a|(bc))\e2
1448 .sp
1449 always fails if it starts to match "a" rather than "bc". Because there may be
1450 many capturing parentheses in a pattern, all digits following the backslash are
1451 taken as part of a potential back reference number. If the pattern continues
1452 with a digit character, some delimiter must be used to terminate the back
1453 reference. If the PCRE_EXTENDED option is set, this can be whitespace.
1454 Otherwise an empty comment (see
1455 .\" HTML <a href="#comments">
1456 .\" </a>
1457 "Comments"
1458 .\"
1459 below) can be used.
1460 .P
1461 A back reference that occurs inside the parentheses to which it refers fails
1462 when the subpattern is first used, so, for example, (a\e1) never matches.
1463 However, such references can be useful inside repeated subpatterns. For
1464 example, the pattern
1465 .sp
1466 (a|b\e1)+
1467 .sp
1468 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1469 the subpattern, the back reference matches the character string corresponding
1470 to the previous iteration. In order for this to work, the pattern must be such
1471 that the first iteration does not need to match the back reference. This can be
1472 done using alternation, as in the example above, or by a quantifier with a
1473 minimum of zero.
1474 .
1475 .
1476 .\" HTML <a name="bigassertions"></a>
1477 .SH ASSERTIONS
1478 .rs
1479 .sp
1480 An assertion is a test on the characters following or preceding the current
1481 matching point that does not actually consume any characters. The simple
1482 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
1483 .\" HTML <a href="#smallassertions">
1484 .\" </a>
1485 above.
1486 .\"
1487 .P
1488 More complicated assertions are coded as subpatterns. There are two kinds:
1489 those that look ahead of the current position in the subject string, and those
1490 that look behind it. An assertion subpattern is matched in the normal way,
1491 except that it does not cause the current matching position to be changed.
1492 .P
1493 Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1494 because it makes no sense to assert the same thing several times. If any kind
1495 of assertion contains capturing subpatterns within it, these are counted for
1496 the purposes of numbering the capturing subpatterns in the whole pattern.
1497 However, substring capturing is carried out only for positive assertions,
1498 because it does not make sense for negative assertions.
1499 .
1500 .
1501 .SS "Lookahead assertions"
1502 .rs
1503 .sp
1504 Lookahead assertions start with (?= for positive assertions and (?! for
1505 negative assertions. For example,
1506 .sp
1507 \ew+(?=;)
1508 .sp
1509 matches a word followed by a semicolon, but does not include the semicolon in
1510 the match, and
1511 .sp
1512 foo(?!bar)
1513 .sp
1514 matches any occurrence of "foo" that is not followed by "bar". Note that the
1515 apparently similar pattern
1516 .sp
1517 (?!foo)bar
1518 .sp
1519 does not find an occurrence of "bar" that is preceded by something other than
1520 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1521 (?!foo) is always true when the next three characters are "bar". A
1522 lookbehind assertion is needed to achieve the other effect.
1523 .P
1524 If you want to force a matching failure at some point in a pattern, the most
1525 convenient way to do it is with (?!) because an empty string always matches, so
1526 an assertion that requires there not to be an empty string must always fail.
1527 .
1528 .
1529 .\" HTML <a name="lookbehind"></a>
1530 .SS "Lookbehind assertions"
1531 .rs
1532 .sp
1533 Lookbehind assertions start with (?<= for positive assertions and (?<! for
1534 negative assertions. For example,
1535 .sp
1536 (?<!foo)bar
1537 .sp
1538 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1539 a lookbehind assertion are restricted such that all the strings it matches must
1540 have a fixed length. However, if there are several top-level alternatives, they
1541 do not all have to have the same fixed length. Thus
1542 .sp
1543 (?<=bullock|donkey)
1544 .sp
1545 is permitted, but
1546 .sp
1547 (?<!dogs?|cats?)
1548 .sp
1549 causes an error at compile time. Branches that match different length strings
1550 are permitted only at the top level of a lookbehind assertion. This is an
1551 extension compared with Perl (at least for 5.8), which requires all branches to
1552 match the same length of string. An assertion such as
1553 .sp
1554 (?<=ab(c|de))
1555 .sp
1556 is not permitted, because its single top-level branch can match two different
1557 lengths, but it is acceptable if rewritten to use two top-level branches:
1558 .sp
1559 (?<=abc|abde)
1560 .sp
1561 In some cases, the Perl 5.10 escape sequence \eK
1562 .\" HTML <a href="#resetmatchstart">
1563 .\" </a>
1564 (see above)
1565 .\"
1566 can be used instead of a lookbehind assertion; this is not restricted to a
1567 fixed-length.
1568 .P
1569 The implementation of lookbehind assertions is, for each alternative, to
1570 temporarily move the current position back by the fixed length and then try to
1571 match. If there are insufficient characters before the current position, the
1572 assertion fails.
1573 .P
1574 PCRE does not allow the \eC escape (which matches a single byte in UTF-8 mode)
1575 to appear in lookbehind assertions, because it makes it impossible to calculate
1576 the length of the lookbehind. The \eX and \eR escapes, which can match
1577 different numbers of bytes, are also not permitted.
1578 .P
1579 Possessive quantifiers can be used in conjunction with lookbehind assertions to
1580 specify efficient matching at the end of the subject string. Consider a simple
1581 pattern such as
1582 .sp
1583 abcd$
1584 .sp
1585 when applied to a long string that does not match. Because matching proceeds
1586 from left to right, PCRE will look for each "a" in the subject and then see if
1587 what follows matches the rest of the pattern. If the pattern is specified as
1588 .sp
1589 ^.*abcd$
1590 .sp
1591 the initial .* matches the entire string at first, but when this fails (because
1592 there is no following "a"), it backtracks to match all but the last character,
1593 then all but the last two characters, and so on. Once again the search for "a"
1594 covers the entire string, from right to left, so we are no better off. However,
1595 if the pattern is written as
1596 .sp
1597 ^.*+(?<=abcd)
1598 .sp
1599 there can be no backtracking for the .*+ item; it can match only the entire
1600 string. The subsequent lookbehind assertion does a single test on the last four
1601 characters. If it fails, the match fails immediately. For long strings, this
1602 approach makes a significant difference to the processing time.
1603 .
1604 .
1605 .SS "Using multiple assertions"
1606 .rs
1607 .sp
1608 Several assertions (of any sort) may occur in succession. For example,
1609 .sp
1610 (?<=\ed{3})(?<!999)foo
1611 .sp
1612 matches "foo" preceded by three digits that are not "999". Notice that each of
1613 the assertions is applied independently at the same point in the subject
1614 string. First there is a check that the previous three characters are all
1615 digits, and then there is a check that the same three characters are not "999".
1616 This pattern does \fInot\fP match "foo" preceded by six characters, the first
1617 of which are digits and the last three of which are not "999". For example, it
1618 doesn't match "123abcfoo". A pattern to do that is
1619 .sp
1620 (?<=\ed{3}...)(?<!999)foo
1621 .sp
1622 This time the first assertion looks at the preceding six characters, checking
1623 that the first three are digits, and then the second assertion checks that the
1624 preceding three characters are not "999".
1625 .P
1626 Assertions can be nested in any combination. For example,
1627 .sp
1628 (?<=(?<!foo)bar)baz
1629 .sp
1630 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1631 preceded by "foo", while
1632 .sp
1633 (?<=\ed{3}(?!999)...)foo
1634 .sp
1635 is another pattern that matches "foo" preceded by three digits and any three
1636 characters that are not "999".
1637 .
1638 .
1639 .\" HTML <a name="conditions"></a>
1640 .SH "CONDITIONAL SUBPATTERNS"
1641 .rs
1642 .sp
1643 It is possible to cause the matching process to obey a subpattern
1644 conditionally or to choose between two alternative subpatterns, depending on
1645 the result of an assertion, or whether a previous capturing subpattern matched
1646 or not. The two possible forms of conditional subpattern are
1647 .sp
1648 (?(condition)yes-pattern)
1649 (?(condition)yes-pattern|no-pattern)
1650 .sp
1651 If the condition is satisfied, the yes-pattern is used; otherwise the
1652 no-pattern (if present) is used. If there are more than two alternatives in the
1653 subpattern, a compile-time error occurs.
1654 .P
1655 There are four kinds of condition: references to subpatterns, references to
1656 recursion, a pseudo-condition called DEFINE, and assertions.
1657 .
1658 .SS "Checking for a used subpattern by number"
1659 .rs
1660 .sp
1661 If the text between the parentheses consists of a sequence of digits, the
1662 condition is true if the capturing subpattern of that number has previously
1663 matched. An alternative notation is to precede the digits with a plus or minus
1664 sign. In this case, the subpattern number is relative rather than absolute.
1665 The most recently opened parentheses can be referenced by (?(-1), the next most
1666 recent by (?(-2), and so on. In looping constructs it can also make sense to
1667 refer to subsequent groups with constructs such as (?(+2).
1668 .P
1669 Consider the following pattern, which contains non-significant white space to
1670 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
1671 three parts for ease of discussion:
1672 .sp
1673 ( \e( )? [^()]+ (?(1) \e) )
1674 .sp
1675 The first part matches an optional opening parenthesis, and if that
1676 character is present, sets it as the first captured substring. The second part
1677 matches one or more characters that are not parentheses. The third part is a
1678 conditional subpattern that tests whether the first set of parentheses matched
1679 or not. If they did, that is, if subject started with an opening parenthesis,
1680 the condition is true, and so the yes-pattern is executed and a closing
1681 parenthesis is required. Otherwise, since no-pattern is not present, the
1682 subpattern matches nothing. In other words, this pattern matches a sequence of
1683 non-parentheses, optionally enclosed in parentheses.
1684 .P
1685 If you were embedding this pattern in a larger one, you could use a relative
1686 reference:
1687 .sp
1688 ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
1689 .sp
1690 This makes the fragment independent of the parentheses in the larger pattern.
1691 .
1692 .SS "Checking for a used subpattern by name"
1693 .rs
1694 .sp
1695 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
1696 subpattern by name. For compatibility with earlier versions of PCRE, which had
1697 this facility before Perl, the syntax (?(name)...) is also recognized. However,
1698 there is a possible ambiguity with this syntax, because subpattern names may
1699 consist entirely of digits. PCRE looks first for a named subpattern; if it
1700 cannot find one and the name consists entirely of digits, PCRE looks for a
1701 subpattern of that number, which must be greater than zero. Using subpattern
1702 names that consist entirely of digits is not recommended.
1703 .P
1704 Rewriting the above example to use a named subpattern gives this:
1705 .sp
1706 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
1707 .sp
1708 .
1709 .SS "Checking for pattern recursion"
1710 .rs
1711 .sp
1712 If the condition is the string (R), and there is no subpattern with the name R,
1713 the condition is true if a recursive call to the whole pattern or any
1714 subpattern has been made. If digits or a name preceded by ampersand follow the
1715 letter R, for example:
1716 .sp
1717 (?(R3)...) or (?(R&name)...)
1718 .sp
1719 the condition is true if the most recent recursion is into the subpattern whose
1720 number or name is given. This condition does not check the entire recursion
1721 stack.
1722 .P
1723 At "top level", all these recursion test conditions are false. Recursive
1724 patterns are described below.
1725 .
1726 .SS "Defining subpatterns for use by reference only"
1727 .rs
1728 .sp
1729 If the condition is the string (DEFINE), and there is no subpattern with the
1730 name DEFINE, the condition is always false. In this case, there may be only one
1731 alternative in the subpattern. It is always skipped if control reaches this
1732 point in the pattern; the idea of DEFINE is that it can be used to define
1733 "subroutines" that can be referenced from elsewhere. (The use of "subroutines"
1734 is described below.) For example, a pattern to match an IPv4 address could be
1735 written like this (ignore whitespace and line breaks):
1736 .sp
1737 (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
1738 \eb (?&byte) (\e.(?&byte)){3} \eb
1739 .sp
1740 The first part of the pattern is a DEFINE group inside which a another group
1741 named "byte" is defined. This matches an individual component of an IPv4
1742 address (a number less than 256). When matching takes place, this part of the
1743 pattern is skipped because DEFINE acts like a false condition.
1744 .P
1745 The rest of the pattern uses references to the named group to match the four
1746 dot-separated components of an IPv4 address, insisting on a word boundary at
1747 each end.
1748 .
1749 .SS "Assertion conditions"
1750 .rs
1751 .sp
1752 If the condition is not in any of the above formats, it must be an assertion.
1753 This may be a positive or negative lookahead or lookbehind assertion. Consider
1754 this pattern, again containing non-significant white space, and with the two
1755 alternatives on the second line:
1756 .sp
1757 (?(?=[^a-z]*[a-z])
1758 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
1759 .sp
1760 The condition is a positive lookahead assertion that matches an optional
1761 sequence of non-letters followed by a letter. In other words, it tests for the
1762 presence of at least one letter in the subject. If a letter is found, the
1763 subject is matched against the first alternative; otherwise it is matched
1764 against the second. This pattern matches strings in one of the two forms
1765 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
1766 .
1767 .
1768 .\" HTML <a name="comments"></a>
1769 .SH COMMENTS
1770 .rs
1771 .sp
1772 The sequence (?# marks the start of a comment that continues up to the next
1773 closing parenthesis. Nested parentheses are not permitted. The characters
1774 that make up a comment play no part in the pattern matching at all.
1775 .P
1776 If the PCRE_EXTENDED option is set, an unescaped # character outside a
1777 character class introduces a comment that continues to immediately after the
1778 next newline in the pattern.
1779 .
1780 .
1781 .\" HTML <a name="recursion"></a>
1782 .SH "RECURSIVE PATTERNS"
1783 .rs
1784 .sp
1785 Consider the problem of matching a string in parentheses, allowing for
1786 unlimited nested parentheses. Without the use of recursion, the best that can
1787 be done is to use a pattern that matches up to some fixed depth of nesting. It
1788 is not possible to handle an arbitrary nesting depth.
1789 .P
1790 For some time, Perl has provided a facility that allows regular expressions to
1791 recurse (amongst other things). It does this by interpolating Perl code in the
1792 expression at run time, and the code can refer to the expression itself. A Perl
1793 pattern using code interpolation to solve the parentheses problem can be
1794 created like this:
1795 .sp
1796 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
1797 .sp
1798 The (?p{...}) item interpolates Perl code at run time, and in this case refers
1799 recursively to the pattern in which it appears.
1800 .P
1801 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
1802 supports special syntax for recursion of the entire pattern, and also for
1803 individual subpattern recursion. After its introduction in PCRE and Python,
1804 this kind of recursion was introduced into Perl at release 5.10.
1805 .P
1806 A special item that consists of (? followed by a number greater than zero and a
1807 closing parenthesis is a recursive call of the subpattern of the given number,
1808 provided that it occurs inside that subpattern. (If not, it is a "subroutine"
1809 call, which is described in the next section.) The special item (?R) or (?0) is
1810 a recursive call of the entire regular expression.
1811 .P
1812 In PCRE (like Python, but unlike Perl), a recursive subpattern call is always
1813 treated as an atomic group. That is, once it has matched some of the subject
1814 string, it is never re-entered, even if it contains untried alternatives and
1815 there is a subsequent matching failure.
1816 .P
1817 This PCRE pattern solves the nested parentheses problem (assume the
1818 PCRE_EXTENDED option is set so that white space is ignored):
1819 .sp
1820 \e( ( (?>[^()]+) | (?R) )* \e)
1821 .sp
1822 First it matches an opening parenthesis. Then it matches any number of
1823 substrings which can either be a sequence of non-parentheses, or a recursive
1824 match of the pattern itself (that is, a correctly parenthesized substring).
1825 Finally there is a closing parenthesis.
1826 .P
1827 If this were part of a larger pattern, you would not want to recurse the entire
1828 pattern, so instead you could use this:
1829 .sp
1830 ( \e( ( (?>[^()]+) | (?1) )* \e) )
1831 .sp
1832 We have put the pattern into parentheses, and caused the recursion to refer to
1833 them instead of the whole pattern.
1834 .P
1835 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
1836 is made easier by the use of relative references. (A Perl 5.10 feature.)
1837 Instead of (?1) in the pattern above you can write (?-2) to refer to the second
1838 most recently opened parentheses preceding the recursion. In other words, a
1839 negative number counts capturing parentheses leftwards from the point at which
1840 it is encountered.
1841 .P
1842 It is also possible to refer to subsequently opened parentheses, by writing
1843 references such as (?+2). However, these cannot be recursive because the
1844 reference is not inside the parentheses that are referenced. They are always
1845 "subroutine" calls, as described in the next section.
1846 .P
1847 An alternative approach is to use named parentheses instead. The Perl syntax
1848 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
1849 could rewrite the above example as follows:
1850 .sp
1851 (?<pn> \e( ( (?>[^()]+) | (?&pn) )* \e) )
1852 .sp
1853 If there is more than one subpattern with the same name, the earliest one is
1854 used.
1855 .P
1856 This particular example pattern that we have been looking at contains nested
1857 unlimited repeats, and so the use of atomic grouping for matching strings of
1858 non-parentheses is important when applying the pattern to strings that do not
1859 match. For example, when this pattern is applied to
1860 .sp
1861 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
1862 .sp
1863 it yields "no match" quickly. However, if atomic grouping is not used,
1864 the match runs for a very long time indeed because there are so many different
1865 ways the + and * repeats can carve up the subject, and all have to be tested
1866 before failure can be reported.
1867 .P
1868 At the end of a match, the values set for any capturing subpatterns are those
1869 from the outermost level of the recursion at which the subpattern value is set.
1870 If you want to obtain intermediate values, a callout function can be used (see
1871 below and the
1872 .\" HREF
1873 \fBpcrecallout\fP
1874 .\"
1875 documentation). If the pattern above is matched against
1876 .sp
1877 (ab(cd)ef)
1878 .sp
1879 the value for the capturing parentheses is "ef", which is the last value taken
1880 on at the top level. If additional parentheses are added, giving
1881 .sp
1882 \e( ( ( (?>[^()]+) | (?R) )* ) \e)
1883 ^ ^
1884 ^ ^
1885 .sp
1886 the string they capture is "ab(cd)ef", the contents of the top level
1887 parentheses. If there are more than 15 capturing parentheses in a pattern, PCRE
1888 has to obtain extra memory to store data during a recursion, which it does by
1889 using \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no
1890 memory can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
1891 .P
1892 Do not confuse the (?R) item with the condition (R), which tests for recursion.
1893 Consider this pattern, which matches text in angle brackets, allowing for
1894 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
1895 recursing), whereas any characters are permitted at the outer level.
1896 .sp
1897 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
1898 .sp
1899 In this pattern, (?(R) is the start of a conditional subpattern, with two
1900 different alternatives for the recursive and non-recursive cases. The (?R) item
1901 is the actual recursive call.
1902 .
1903 .
1904 .\" HTML <a name="subpatternsassubroutines"></a>
1905 .SH "SUBPATTERNS AS SUBROUTINES"
1906 .rs
1907 .sp
1908 If the syntax for a recursive subpattern reference (either by number or by
1909 name) is used outside the parentheses to which it refers, it operates like a
1910 subroutine in a programming language. The "called" subpattern may be defined
1911 before or after the reference. A numbered reference can be absolute or
1912 relative, as in these examples:
1913 .sp
1914 (...(absolute)...)...(?2)...
1915 (...(relative)...)...(?-1)...
1916 (...(?+1)...(relative)...
1917 .sp
1918 An earlier example pointed out that the pattern
1919 .sp
1920 (sens|respons)e and \e1ibility
1921 .sp
1922 matches "sense and sensibility" and "response and responsibility", but not
1923 "sense and responsibility". If instead the pattern
1924 .sp
1925 (sens|respons)e and (?1)ibility
1926 .sp
1927 is used, it does match "sense and responsibility" as well as the other two
1928 strings. Another example is given in the discussion of DEFINE above.
1929 .P
1930 Like recursive subpatterns, a "subroutine" call is always treated as an atomic
1931 group. That is, once it has matched some of the subject string, it is never
1932 re-entered, even if it contains untried alternatives and there is a subsequent
1933 matching failure.
1934 .P
1935 When a subpattern is used as a subroutine, processing options such as
1936 case-independence are fixed when the subpattern is defined. They cannot be
1937 changed for different calls. For example, consider this pattern:
1938 .sp
1939 (abc)(?i:(?-1))
1940 .sp
1941 It matches "abcabc". It does not match "abcABC" because the change of
1942 processing option does not affect the called subpattern.
1943 .
1944 .
1945 .SH CALLOUTS
1946 .rs
1947 .sp
1948 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
1949 code to be obeyed in the middle of matching a regular expression. This makes it
1950 possible, amongst other things, to extract different substrings that match the
1951 same pair of parentheses when there is a repetition.
1952 .P
1953 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
1954 code. The feature is called "callout". The caller of PCRE provides an external
1955 function by putting its entry point in the global variable \fIpcre_callout\fP.
1956 By default, this variable contains NULL, which disables all calling out.
1957 .P
1958 Within a regular expression, (?C) indicates the points at which the external
1959 function is to be called. If you want to identify different callout points, you
1960 can put a number less than 256 after the letter C. The default value is zero.
1961 For example, this pattern has two callout points:
1962 .sp
1963 (?C1)abc(?C2)def
1964 .sp
1965 If the PCRE_AUTO_CALLOUT flag is passed to \fBpcre_compile()\fP, callouts are
1966 automatically installed before each item in the pattern. They are all numbered
1967 255.
1968 .P
1969 During matching, when PCRE reaches a callout point (and \fIpcre_callout\fP is
1970 set), the external function is called. It is provided with the number of the
1971 callout, the position in the pattern, and, optionally, one item of data
1972 originally supplied by the caller of \fBpcre_exec()\fP. The callout function
1973 may cause matching to proceed, to backtrack, or to fail altogether. A complete
1974 description of the interface to the callout function is given in the
1975 .\" HREF
1976 \fBpcrecallout\fP
1977 .\"
1978 documentation.
1979 .
1980 .
1981 .SH "BACTRACKING CONTROL"
1982 .rs
1983 .sp
1984 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
1985 are described in the Perl documentation as "experimental and subject to change
1986 or removal in a future version of Perl". It goes on to say: "Their usage in
1987 production code should be noted to avoid problems during upgrades." The same
1988 remarks apply to the PCRE features described in this section.
1989 .P
1990 Since these verbs are specifically related to backtracking, they can be used
1991 only when the pattern is to be matched using \fBpcre_exec()\fP, which uses a
1992 backtracking algorithm. They cause an error if encountered by
1993 \fBpcre_dfa_exec()\fP.
1994 .P
1995 The new verbs make use of what was previously invalid syntax: an opening
1996 parenthesis followed by an asterisk. In Perl, they are generally of the form
1997 (*VERB:ARG) but PCRE does not support the use of arguments, so its general
1998 form is just (*VERB). Any number of these verbs may occur in a pattern. There
1999 are two kinds:
2000 .
2001 .SS "Verbs that act immediately"
2002 .rs
2003 .sp
2004 The following verbs act as soon as they are encountered:
2005 .sp
2006 (*ACCEPT)
2007 .sp
2008 This verb causes the match to end successfully, skipping the remainder of the
2009 pattern. When inside a recursion, only the innermost pattern is ended
2010 immediately. PCRE differs from Perl in what happens if the (*ACCEPT) is inside
2011 capturing parentheses. In Perl, the data so far is captured: in PCRE no data is
2012 captured. For example:
2013 .sp
2014 A(A|B(*ACCEPT)|C)D
2015 .sp
2016 This matches "AB", "AAD", or "ACD", but when it matches "AB", no data is
2017 captured.
2018 .sp
2019 (*FAIL) or (*F)
2020 .sp
2021 This verb causes the match to fail, forcing backtracking to occur. It is
2022 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2023 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2024 Perl features that are not present in PCRE. The nearest equivalent is the
2025 callout feature, as for example in this pattern:
2026 .sp
2027 a+(?C)(*FAIL)
2028 .sp
2029 A match with the string "aaaa" always fails, but the callout is taken before
2030 each backtrack happens (in this example, 10 times).
2031 .
2032 .SS "Verbs that act after backtracking"
2033 .rs
2034 .sp
2035 The following verbs do nothing when they are encountered. Matching continues
2036 with what follows, but if there is no subsequent match, a failure is forced.
2037 The verbs differ in exactly what kind of failure occurs.
2038 .sp
2039 (*COMMIT)
2040 .sp
2041 This verb causes the whole match to fail outright if the rest of the pattern
2042 does not match. Even if the pattern is unanchored, no further attempts to find
2043 a match by advancing the start point take place. Once (*COMMIT) has been
2044 passed, \fBpcre_exec()\fP is committed to finding a match at the current
2045 starting point, or not at all. For example:
2046 .sp
2047 a+(*COMMIT)b
2048 .sp
2049 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2050 dynamic anchor, or "I've started, so I must finish."
2051 .sp
2052 (*PRUNE)
2053 .sp
2054 This verb causes the match to fail at the current position if the rest of the
2055 pattern does not match. If the pattern is unanchored, the normal "bumpalong"
2056 advance to the next starting character then happens. Backtracking can occur as
2057 usual to the left of (*PRUNE), or when matching to the right of (*PRUNE), but
2058 if there is no match to the right, backtracking cannot cross (*PRUNE).
2059 In simple cases, the use of (*PRUNE) is just an alternative to an atomic
2060 group or possessive quantifier, but there are some uses of (*PRUNE) that cannot
2061 be expressed in any other way.
2062 .sp
2063 (*SKIP)
2064 .sp
2065 This verb is like (*PRUNE), except that if the pattern is unanchored, the
2066 "bumpalong" advance is not to the next character, but to the position in the
2067 subject where (*SKIP) was encountered. (*SKIP) signifies that whatever text
2068 was matched leading up to it cannot be part of a successful match. Consider:
2069 .sp
2070 a+(*SKIP)b
2071 .sp
2072 If the subject is "aaaac...", after the first match attempt fails (starting at
2073 the first character in the string), the starting point skips on to start the
2074 next attempt at "c". Note that a possessive quantifer does not have the same
2075 effect in this example; although it would suppress backtracking during the
2076 first match attempt, the second attempt would start at the second character
2077 instead of skipping on to "c".
2078 .sp
2079 (*THEN)
2080 .sp
2081 This verb causes a skip to the next alternation if the rest of the pattern does
2082 not match. That is, it cancels pending backtracking, but only within the
2083 current alternation. Its name comes from the observation that it can be used
2084 for a pattern-based if-then-else block:
2085 .sp
2086 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2087 .sp
2088 If the COND1 pattern matches, FOO is tried (and possibly further items after
2089 the end of the group if FOO succeeds); on failure the matcher skips to the
2090 second alternative and tries COND2, without backtracking into COND1. If (*THEN)
2091 is used outside of any alternation, it acts exactly like (*PRUNE).
2092 .
2093 .
2094 .SH "SEE ALSO"
2095 .rs
2096 .sp
2097 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3), \fBpcre\fP(3).
2098 .
2099 .
2100 .SH AUTHOR
2101 .rs
2102 .sp
2103 .nf
2104 Philip Hazel
2105 University Computing Service
2106 Cambridge CB2 3QH, England.
2107 .fi
2108 .
2109 .
2110 .SH REVISION
2111 .rs
2112 .sp
2113 .nf
2114 Last updated: 09 August 2007
2115 Copyright (c) 1997-2007 University of Cambridge.
2116 .fi

Properties

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

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12