ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1459 - (show annotations)
Tue Mar 4 10:45:15 2014 UTC (5 years, 6 months ago) by ph10
File size: 135252 byte(s)
Preparations for next release.
1 .TH PCREPATTERN 3 "08 January 2014" "PCRE 8.35"
3 PCRE - Perl-compatible regular expressions
5 .rs
6 .sp
7 The syntax and semantics of the regular expressions that are supported by PCRE
8 are described in detail below. There is a quick-reference syntax summary in the
9 .\" HREF
10 \fBpcresyntax\fP
11 .\"
12 page. PCRE tries to match Perl syntax and semantics as closely as it can. PCRE
13 also supports some alternative regular expression syntax (which does not
14 conflict with the Perl syntax) in order to provide some compatibility with
15 regular expressions in Python, .NET, and Oniguruma.
16 .P
17 Perl's regular expressions are described in its own documentation, and
18 regular expressions in general are covered in a number of books, some of which
19 have copious examples. Jeffrey Friedl's "Mastering Regular Expressions",
20 published by O'Reilly, covers regular expressions in great detail. This
21 description of PCRE's regular expressions is intended as reference material.
22 .P
23 This document discusses the patterns that are supported by PCRE when one its
24 main matching functions, \fBpcre_exec()\fP (8-bit) or \fBpcre[16|32]_exec()\fP
25 (16- or 32-bit), is used. PCRE also has alternative matching functions,
26 \fBpcre_dfa_exec()\fP and \fBpcre[16|32_dfa_exec()\fP, which match using a
27 different algorithm that is not Perl-compatible. Some of the features discussed
28 below are not available when DFA matching is used. The advantages and
29 disadvantages of the alternative functions, and how they differ from the normal
30 functions, are discussed in the
31 .\" HREF
32 \fBpcrematching\fP
33 .\"
34 page.
35 .
36 .
38 .rs
39 .sp
40 A number of options that can be passed to \fBpcre_compile()\fP can also be set
41 by special items at the start of a pattern. These are not Perl-compatible, but
42 are provided to make these options accessible to pattern writers who are not
43 able to change the program that processes the pattern. Any number of these
44 items may appear, but they must all be together right at the start of the
45 pattern string, and the letters must be in upper case.
46 .
47 .
48 .SS "UTF support"
49 .rs
50 .sp
51 The original operation of PCRE was on strings of one-byte characters. However,
52 there is now also support for UTF-8 strings in the original library, an
53 extra library that supports 16-bit and UTF-16 character strings, and a
54 third library that supports 32-bit and UTF-32 character strings. To use these
55 features, PCRE must be built to include appropriate support. When using UTF
56 strings you must either call the compiling function with the PCRE_UTF8,
57 PCRE_UTF16, or PCRE_UTF32 option, or the pattern must start with one of
58 these special sequences:
59 .sp
60 (*UTF8)
61 (*UTF16)
62 (*UTF32)
63 (*UTF)
64 .sp
65 (*UTF) is a generic sequence that can be used with any of the libraries.
66 Starting a pattern with such a sequence is equivalent to setting the relevant
67 option. How setting a UTF mode affects pattern matching is mentioned in several
68 places below. There is also a summary of features in the
69 .\" HREF
70 \fBpcreunicode\fP
71 .\"
72 page.
73 .P
74 Some applications that allow their users to supply patterns may wish to
75 restrict them to non-UTF data for security reasons. If the PCRE_NEVER_UTF
76 option is set at compile time, (*UTF) etc. are not allowed, and their
77 appearance causes an error.
78 .
79 .
80 .SS "Unicode property support"
81 .rs
82 .sp
83 Another special sequence that may appear at the start of a pattern is (*UCP).
84 This has the same effect as setting the PCRE_UCP option: it causes sequences
85 such as \ed and \ew to use Unicode properties to determine character types,
86 instead of recognizing only characters with codes less than 128 via a lookup
87 table.
88 .
89 .
90 .SS "Disabling auto-possessification"
91 .rs
92 .sp
93 If a pattern starts with (*NO_AUTO_POSSESS), it has the same effect as setting
94 the PCRE_NO_AUTO_POSSESS option at compile time. This stops PCRE from making
95 quantifiers possessive when what follows cannot match the repeated item. For
96 example, by default a+b is treated as a++b. For more details, see the
97 .\" HREF
98 \fBpcreapi\fP
99 .\"
100 documentation.
101 .
102 .
103 .SS "Disabling start-up optimizations"
104 .rs
105 .sp
106 If a pattern starts with (*NO_START_OPT), it has the same effect as setting the
107 PCRE_NO_START_OPTIMIZE option either at compile or matching time. This disables
108 several optimizations for quickly reaching "no match" results. For more
109 details, see the
110 .\" HREF
111 \fBpcreapi\fP
112 .\"
113 documentation.
114 .
115 .
116 .\" HTML <a name="newlines"></a>
117 .SS "Newline conventions"
118 .rs
119 .sp
120 PCRE supports five different conventions for indicating line breaks in
121 strings: a single CR (carriage return) character, a single LF (linefeed)
122 character, the two-character sequence CRLF, any of the three preceding, or any
123 Unicode newline sequence. The
124 .\" HREF
125 \fBpcreapi\fP
126 .\"
127 page has
128 .\" HTML <a href="pcreapi.html#newlines">
129 .\" </a>
130 further discussion
131 .\"
132 about newlines, and shows how to set the newline convention in the
133 \fIoptions\fP arguments for the compiling and matching functions.
134 .P
135 It is also possible to specify a newline convention by starting a pattern
136 string with one of the following five sequences:
137 .sp
138 (*CR) carriage return
139 (*LF) linefeed
140 (*CRLF) carriage return, followed by linefeed
141 (*ANYCRLF) any of the three above
142 (*ANY) all Unicode newline sequences
143 .sp
144 These override the default and the options given to the compiling function. For
145 example, on a Unix system where LF is the default newline sequence, the pattern
146 .sp
147 (*CR)a.b
148 .sp
149 changes the convention to CR. That pattern matches "a\enb" because LF is no
150 longer a newline. If more than one of these settings is present, the last one
151 is used.
152 .P
153 The newline convention affects where the circumflex and dollar assertions are
154 true. It also affects the interpretation of the dot metacharacter when
155 PCRE_DOTALL is not set, and the behaviour of \eN. However, it does not affect
156 what the \eR escape sequence matches. By default, this is any Unicode newline
157 sequence, for Perl compatibility. However, this can be changed; see the
158 description of \eR in the section entitled
159 .\" HTML <a href="#newlineseq">
160 .\" </a>
161 "Newline sequences"
162 .\"
163 below. A change of \eR setting can be combined with a change of newline
164 convention.
165 .
166 .
167 .SS "Setting match and recursion limits"
168 .rs
169 .sp
170 The caller of \fBpcre_exec()\fP can set a limit on the number of times the
171 internal \fBmatch()\fP function is called and on the maximum depth of
172 recursive calls. These facilities are provided to catch runaway matches that
173 are provoked by patterns with huge matching trees (a typical example is a
174 pattern with nested unlimited repeats) and to avoid running out of system stack
175 by too much recursion. When one of these limits is reached, \fBpcre_exec()\fP
176 gives an error return. The limits can also be set by items at the start of the
177 pattern of the form
178 .sp
179 (*LIMIT_MATCH=d)
181 .sp
182 where d is any number of decimal digits. However, the value of the setting must
183 be less than the value set (or defaulted) by the caller of \fBpcre_exec()\fP
184 for it to have any effect. In other words, the pattern writer can lower the
185 limits set by the programmer, but not raise them. If there is more than one
186 setting of one of these limits, the lower value is used.
187 .
188 .
190 .rs
191 .sp
192 PCRE can be compiled to run in an environment that uses EBCDIC as its character
193 code rather than ASCII or Unicode (typically a mainframe system). In the
194 sections below, character code values are ASCII or Unicode; in an EBCDIC
195 environment these characters may have different code values, and there are no
196 code points greater than 255.
197 .
198 .
200 .rs
201 .sp
202 A regular expression is a pattern that is matched against a subject string from
203 left to right. Most characters stand for themselves in a pattern, and match the
204 corresponding characters in the subject. As a trivial example, the pattern
205 .sp
206 The quick brown fox
207 .sp
208 matches a portion of a subject string that is identical to itself. When
209 caseless matching is specified (the PCRE_CASELESS option), letters are matched
210 independently of case. In a UTF mode, PCRE always understands the concept of
211 case for characters whose values are less than 128, so caseless matching is
212 always possible. For characters with higher values, the concept of case is
213 supported if PCRE is compiled with Unicode property support, but not otherwise.
214 If you want to use caseless matching for characters 128 and above, you must
215 ensure that PCRE is compiled with Unicode property support as well as with
216 UTF support.
217 .P
218 The power of regular expressions comes from the ability to include alternatives
219 and repetitions in the pattern. These are encoded in the pattern by the use of
220 \fImetacharacters\fP, which do not stand for themselves but instead are
221 interpreted in some special way.
222 .P
223 There are two different sets of metacharacters: those that are recognized
224 anywhere in the pattern except within square brackets, and those that are
225 recognized within square brackets. Outside square brackets, the metacharacters
226 are as follows:
227 .sp
228 \e general escape character with several uses
229 ^ assert start of string (or line, in multiline mode)
230 $ assert end of string (or line, in multiline mode)
231 . match any character except newline (by default)
232 [ start character class definition
233 | start of alternative branch
234 ( start subpattern
235 ) end subpattern
236 ? extends the meaning of (
237 also 0 or 1 quantifier
238 also quantifier minimizer
239 * 0 or more quantifier
240 + 1 or more quantifier
241 also "possessive quantifier"
242 { start min/max quantifier
243 .sp
244 Part of a pattern that is in square brackets is called a "character class". In
245 a character class the only metacharacters are:
246 .sp
247 \e general escape character
248 ^ negate the class, but only if the first character
249 - indicates character range
250 .\" JOIN
251 [ POSIX character class (only if followed by POSIX
252 syntax)
253 ] terminates the character class
254 .sp
255 The following sections describe the use of each of the metacharacters.
256 .
257 .
259 .rs
260 .sp
261 The backslash character has several uses. Firstly, if it is followed by a
262 character that is not a number or a letter, it takes away any special meaning
263 that character may have. This use of backslash as an escape character applies
264 both inside and outside character classes.
265 .P
266 For example, if you want to match a * character, you write \e* in the pattern.
267 This escaping action applies whether or not the following character would
268 otherwise be interpreted as a metacharacter, so it is always safe to precede a
269 non-alphanumeric with backslash to specify that it stands for itself. In
270 particular, if you want to match a backslash, you write \e\e.
271 .P
272 In a UTF mode, only ASCII numbers and letters have any special meaning after a
273 backslash. All other characters (in particular, those whose codepoints are
274 greater than 127) are treated as literals.
275 .P
276 If a pattern is compiled with the PCRE_EXTENDED option, most white space in the
277 pattern (other than in a character class), and characters between a # outside a
278 character class and the next newline, inclusive, are ignored. An escaping
279 backslash can be used to include a white space or # character as part of the
280 pattern.
281 .P
282 If you want to remove the special meaning from a sequence of characters, you
283 can do so by putting them between \eQ and \eE. This is different from Perl in
284 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
285 Perl, $ and @ cause variable interpolation. Note the following examples:
286 .sp
287 Pattern PCRE matches Perl matches
288 .sp
289 .\" JOIN
290 \eQabc$xyz\eE abc$xyz abc followed by the
291 contents of $xyz
292 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
293 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
294 .sp
295 The \eQ...\eE sequence is recognized both inside and outside character classes.
296 An isolated \eE that is not preceded by \eQ is ignored. If \eQ is not followed
297 by \eE later in the pattern, the literal interpretation continues to the end of
298 the pattern (that is, \eE is assumed at the end). If the isolated \eQ is inside
299 a character class, this causes an error, because the character class is not
300 terminated.
301 .
302 .
303 .\" HTML <a name="digitsafterbackslash"></a>
304 .SS "Non-printing characters"
305 .rs
306 .sp
307 A second use of backslash provides a way of encoding non-printing characters
308 in patterns in a visible manner. There is no restriction on the appearance of
309 non-printing characters, apart from the binary zero that terminates a pattern,
310 but when a pattern is being prepared by text editing, it is often easier to use
311 one of the following escape sequences than the binary character it represents:
312 .sp
313 \ea alarm, that is, the BEL character (hex 07)
314 \ecx "control-x", where x is any ASCII character
315 \ee escape (hex 1B)
316 \ef form feed (hex 0C)
317 \en linefeed (hex 0A)
318 \er carriage return (hex 0D)
319 \et tab (hex 09)
320 \e0dd character with octal code 0dd
321 \eddd character with octal code ddd, or back reference
322 \eo{ddd..} character with octal code ddd..
323 \exhh character with hex code hh
324 \ex{hhh..} character with hex code hhh.. (non-JavaScript mode)
325 \euhhhh character with hex code hhhh (JavaScript mode only)
326 .sp
327 The precise effect of \ecx on ASCII characters is as follows: if x is a lower
328 case letter, it is converted to upper case. Then bit 6 of the character (hex
329 40) is inverted. Thus \ecA to \ecZ become hex 01 to hex 1A (A is 41, Z is 5A),
330 but \ec{ becomes hex 3B ({ is 7B), and \ec; becomes hex 7B (; is 3B). If the
331 data item (byte or 16-bit value) following \ec has a value greater than 127, a
332 compile-time error occurs. This locks out non-ASCII characters in all modes.
333 .P
334 The \ec facility was designed for use with ASCII characters, but with the
335 extension to Unicode it is even less useful than it once was. It is, however,
336 recognized when PCRE is compiled in EBCDIC mode, where data items are always
337 bytes. In this mode, all values are valid after \ec. If the next character is a
338 lower case letter, it is converted to upper case. Then the 0xc0 bits of the
339 byte are inverted. Thus \ecA becomes hex 01, as in ASCII (A is C1), but because
340 the EBCDIC letters are disjoint, \ecZ becomes hex 29 (Z is E9), and other
341 characters also generate different values.
342 .P
343 After \e0 up to two further octal digits are read. If there are fewer than two
344 digits, just those that are present are used. Thus the sequence \e0\ex\e07
345 specifies two binary zeros followed by a BEL character (code value 7). Make
346 sure you supply two digits after the initial zero if the pattern character that
347 follows is itself an octal digit.
348 .P
349 The escape \eo must be followed by a sequence of octal digits, enclosed in
350 braces. An error occurs if this is not the case. This escape is a recent
351 addition to Perl; it provides way of specifying character code points as octal
352 numbers greater than 0777, and it also allows octal numbers and back references
353 to be unambiguously specified.
354 .P
355 For greater clarity and unambiguity, it is best to avoid following \e by a
356 digit greater than zero. Instead, use \eo{} or \ex{} to specify character
357 numbers, and \eg{} to specify back references. The following paragraphs
358 describe the old, ambiguous syntax.
359 .P
360 The handling of a backslash followed by a digit other than 0 is complicated,
361 and Perl has changed in recent releases, causing PCRE also to change. Outside a
362 character class, PCRE reads the digit and any following digits as a decimal
363 number. If the number is less than 8, or if there have been at least that many
364 previous capturing left parentheses in the expression, the entire sequence is
365 taken as a \fIback reference\fP. A description of how this works is given
366 .\" HTML <a href="#backreferences">
367 .\" </a>
368 later,
369 .\"
370 following the discussion of
371 .\" HTML <a href="#subpattern">
372 .\" </a>
373 parenthesized subpatterns.
374 .\"
375 .P
376 Inside a character class, or if the decimal number following \e is greater than
377 7 and there have not been that many capturing subpatterns, PCRE handles \e8 and
378 \e9 as the literal characters "8" and "9", and otherwise re-reads up to three
379 octal digits following the backslash, using them to generate a data character.
380 Any subsequent digits stand for themselves. For example:
381 .sp
382 \e040 is another way of writing an ASCII space
383 .\" JOIN
384 \e40 is the same, provided there are fewer than 40
385 previous capturing subpatterns
386 \e7 is always a back reference
387 .\" JOIN
388 \e11 might be a back reference, or another way of
389 writing a tab
390 \e011 is always a tab
391 \e0113 is a tab followed by the character "3"
392 .\" JOIN
393 \e113 might be a back reference, otherwise the
394 character with octal code 113
395 .\" JOIN
396 \e377 might be a back reference, otherwise
397 the value 255 (decimal)
398 .\" JOIN
399 \e81 is either a back reference, or the two
400 characters "8" and "1"
401 .sp
402 Note that octal values of 100 or greater that are specified using this syntax
403 must not be introduced by a leading zero, because no more than three octal
404 digits are ever read.
405 .P
406 By default, after \ex that is not followed by {, from zero to two hexadecimal
407 digits are read (letters can be in upper or lower case). Any number of
408 hexadecimal digits may appear between \ex{ and }. If a character other than
409 a hexadecimal digit appears between \ex{ and }, or if there is no terminating
410 }, an error occurs.
411 .P
412 If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \ex is
413 as just described only when it is followed by two hexadecimal digits.
414 Otherwise, it matches a literal "x" character. In JavaScript mode, support for
415 code points greater than 256 is provided by \eu, which must be followed by
416 four hexadecimal digits; otherwise it matches a literal "u" character.
417 .P
418 Characters whose value is less than 256 can be defined by either of the two
419 syntaxes for \ex (or by \eu in JavaScript mode). There is no difference in the
420 way they are handled. For example, \exdc is exactly the same as \ex{dc} (or
421 \eu00dc in JavaScript mode).
422 .
423 .
424 .SS "Constraints on character values"
425 .rs
426 .sp
427 Characters that are specified using octal or hexadecimal numbers are
428 limited to certain values, as follows:
429 .sp
430 8-bit non-UTF mode less than 0x100
431 8-bit UTF-8 mode less than 0x10ffff and a valid codepoint
432 16-bit non-UTF mode less than 0x10000
433 16-bit UTF-16 mode less than 0x10ffff and a valid codepoint
434 32-bit non-UTF mode less than 0x100000000
435 32-bit UTF-32 mode less than 0x10ffff and a valid codepoint
436 .sp
437 Invalid Unicode codepoints are the range 0xd800 to 0xdfff (the so-called
438 "surrogate" codepoints), and 0xffef.
439 .
440 .
441 .SS "Escape sequences in character classes"
442 .rs
443 .sp
444 All the sequences that define a single character value can be used both inside
445 and outside character classes. In addition, inside a character class, \eb is
446 interpreted as the backspace character (hex 08).
447 .P
448 \eN is not allowed in a character class. \eB, \eR, and \eX are not special
449 inside a character class. Like other unrecognized escape sequences, they are
450 treated as the literal characters "B", "R", and "X" by default, but cause an
451 error if the PCRE_EXTRA option is set. Outside a character class, these
452 sequences have different meanings.
453 .
454 .
455 .SS "Unsupported escape sequences"
456 .rs
457 .sp
458 In Perl, the sequences \el, \eL, \eu, and \eU are recognized by its string
459 handler and used to modify the case of following characters. By default, PCRE
460 does not support these escape sequences. However, if the PCRE_JAVASCRIPT_COMPAT
461 option is set, \eU matches a "U" character, and \eu can be used to define a
462 character by code point, as described in the previous section.
463 .
464 .
465 .SS "Absolute and relative back references"
466 .rs
467 .sp
468 The sequence \eg followed by an unsigned or a negative number, optionally
469 enclosed in braces, is an absolute or relative back reference. A named back
470 reference can be coded as \eg{name}. Back references are discussed
471 .\" HTML <a href="#backreferences">
472 .\" </a>
473 later,
474 .\"
475 following the discussion of
476 .\" HTML <a href="#subpattern">
477 .\" </a>
478 parenthesized subpatterns.
479 .\"
480 .
481 .
482 .SS "Absolute and relative subroutine calls"
483 .rs
484 .sp
485 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
486 a number enclosed either in angle brackets or single quotes, is an alternative
487 syntax for referencing a subpattern as a "subroutine". Details are discussed
488 .\" HTML <a href="#onigurumasubroutines">
489 .\" </a>
490 later.
491 .\"
492 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
493 synonymous. The former is a back reference; the latter is a
494 .\" HTML <a href="#subpatternsassubroutines">
495 .\" </a>
496 subroutine
497 .\"
498 call.
499 .
500 .
501 .\" HTML <a name="genericchartypes"></a>
502 .SS "Generic character types"
503 .rs
504 .sp
505 Another use of backslash is for specifying generic character types:
506 .sp
507 \ed any decimal digit
508 \eD any character that is not a decimal digit
509 \eh any horizontal white space character
510 \eH any character that is not a horizontal white space character
511 \es any white space character
512 \eS any character that is not a white space character
513 \ev any vertical white space character
514 \eV any character that is not a vertical white space character
515 \ew any "word" character
516 \eW any "non-word" character
517 .sp
518 There is also the single sequence \eN, which matches a non-newline character.
519 This is the same as
520 .\" HTML <a href="#fullstopdot">
521 .\" </a>
522 the "." metacharacter
523 .\"
524 when PCRE_DOTALL is not set. Perl also uses \eN to match characters by name;
525 PCRE does not support this.
526 .P
527 Each pair of lower and upper case escape sequences partitions the complete set
528 of characters into two disjoint sets. Any given character matches one, and only
529 one, of each pair. The sequences can appear both inside and outside character
530 classes. They each match one character of the appropriate type. If the current
531 matching point is at the end of the subject string, all of them fail, because
532 there is no character to match.
533 .P
534 For compatibility with Perl, \es did not used to match the VT character (code
535 11), which made it different from the the POSIX "space" class. However, Perl
536 added VT at release 5.18, and PCRE followed suit at release 8.34. The default
537 \es characters are now HT (9), LF (10), VT (11), FF (12), CR (13), and space
538 (32), which are defined as white space in the "C" locale. This list may vary if
539 locale-specific matching is taking place. For example, in some locales the
540 "non-breaking space" character (\exA0) is recognized as white space, and in
541 others the VT character is not.
542 .P
543 A "word" character is an underscore or any character that is a letter or digit.
544 By default, the definition of letters and digits is controlled by PCRE's
545 low-valued character tables, and may vary if locale-specific matching is taking
546 place (see
547 .\" HTML <a href="pcreapi.html#localesupport">
548 .\" </a>
549 "Locale support"
550 .\"
551 in the
552 .\" HREF
553 \fBpcreapi\fP
554 .\"
555 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
556 or "french" in Windows, some character codes greater than 127 are used for
557 accented letters, and these are then matched by \ew. The use of locales with
558 Unicode is discouraged.
559 .P
560 By default, characters whose code points are greater than 127 never match \ed,
561 \es, or \ew, and always match \eD, \eS, and \eW, although this may vary for
562 characters in the range 128-255 when locale-specific matching is happening.
563 These escape sequences retain their original meanings from before Unicode
564 support was available, mainly for efficiency reasons. If PCRE is compiled with
565 Unicode property support, and the PCRE_UCP option is set, the behaviour is
566 changed so that Unicode properties are used to determine character types, as
567 follows:
568 .sp
569 \ed any character that matches \ep{Nd} (decimal digit)
570 \es any character that matches \ep{Z} or \eh or \ev
571 \ew any character that matches \ep{L} or \ep{N}, plus underscore
572 .sp
573 The upper case escapes match the inverse sets of characters. Note that \ed
574 matches only decimal digits, whereas \ew matches any Unicode digit, as well as
575 any Unicode letter, and underscore. Note also that PCRE_UCP affects \eb, and
576 \eB because they are defined in terms of \ew and \eW. Matching these sequences
577 is noticeably slower when PCRE_UCP is set.
578 .P
579 The sequences \eh, \eH, \ev, and \eV are features that were added to Perl at
580 release 5.10. In contrast to the other sequences, which match only ASCII
581 characters by default, these always match certain high-valued code points,
582 whether or not PCRE_UCP is set. The horizontal space characters are:
583 .sp
584 U+0009 Horizontal tab (HT)
585 U+0020 Space
586 U+00A0 Non-break space
587 U+1680 Ogham space mark
588 U+180E Mongolian vowel separator
589 U+2000 En quad
590 U+2001 Em quad
591 U+2002 En space
592 U+2003 Em space
593 U+2004 Three-per-em space
594 U+2005 Four-per-em space
595 U+2006 Six-per-em space
596 U+2007 Figure space
597 U+2008 Punctuation space
598 U+2009 Thin space
599 U+200A Hair space
600 U+202F Narrow no-break space
601 U+205F Medium mathematical space
602 U+3000 Ideographic space
603 .sp
604 The vertical space characters are:
605 .sp
606 U+000A Linefeed (LF)
607 U+000B Vertical tab (VT)
608 U+000C Form feed (FF)
609 U+000D Carriage return (CR)
610 U+0085 Next line (NEL)
611 U+2028 Line separator
612 U+2029 Paragraph separator
613 .sp
614 In 8-bit, non-UTF-8 mode, only the characters with codepoints less than 256 are
615 relevant.
616 .
617 .
618 .\" HTML <a name="newlineseq"></a>
619 .SS "Newline sequences"
620 .rs
621 .sp
622 Outside a character class, by default, the escape sequence \eR matches any
623 Unicode newline sequence. In 8-bit non-UTF-8 mode \eR is equivalent to the
624 following:
625 .sp
626 (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
627 .sp
628 This is an example of an "atomic group", details of which are given
629 .\" HTML <a href="#atomicgroup">
630 .\" </a>
631 below.
632 .\"
633 This particular group matches either the two-character sequence CR followed by
634 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
635 U+000B), FF (form feed, U+000C), CR (carriage return, U+000D), or NEL (next
636 line, U+0085). The two-character sequence is treated as a single unit that
637 cannot be split.
638 .P
639 In other modes, two additional characters whose codepoints are greater than 255
640 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
641 Unicode character property support is not needed for these characters to be
642 recognized.
643 .P
644 It is possible to restrict \eR to match only CR, LF, or CRLF (instead of the
645 complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
646 either at compile time or when the pattern is matched. (BSR is an abbrevation
647 for "backslash R".) This can be made the default when PCRE is built; if this is
648 the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
649 It is also possible to specify these settings by starting a pattern string with
650 one of the following sequences:
651 .sp
652 (*BSR_ANYCRLF) CR, LF, or CRLF only
653 (*BSR_UNICODE) any Unicode newline sequence
654 .sp
655 These override the default and the options given to the compiling function, but
656 they can themselves be overridden by options given to a matching function. Note
657 that these special settings, which are not Perl-compatible, are recognized only
658 at the very start of a pattern, and that they must be in upper case. If more
659 than one of them is present, the last one is used. They can be combined with a
660 change of newline convention; for example, a pattern can start with:
661 .sp
663 .sp
664 They can also be combined with the (*UTF8), (*UTF16), (*UTF32), (*UTF) or
665 (*UCP) special sequences. Inside a character class, \eR is treated as an
666 unrecognized escape sequence, and so matches the letter "R" by default, but
667 causes an error if PCRE_EXTRA is set.
668 .
669 .
670 .\" HTML <a name="uniextseq"></a>
671 .SS Unicode character properties
672 .rs
673 .sp
674 When PCRE is built with Unicode character property support, three additional
675 escape sequences that match characters with specific properties are available.
676 When in 8-bit non-UTF-8 mode, these sequences are of course limited to testing
677 characters whose codepoints are less than 256, but they do work in this mode.
678 The extra escape sequences are:
679 .sp
680 \ep{\fIxx\fP} a character with the \fIxx\fP property
681 \eP{\fIxx\fP} a character without the \fIxx\fP property
682 \eX a Unicode extended grapheme cluster
683 .sp
684 The property names represented by \fIxx\fP above are limited to the Unicode
685 script names, the general category properties, "Any", which matches any
686 character (including newline), and some special PCRE properties (described
687 in the
688 .\" HTML <a href="#extraprops">
689 .\" </a>
690 next section).
691 .\"
692 Other Perl properties such as "InMusicalSymbols" are not currently supported by
693 PCRE. Note that \eP{Any} does not match any characters, so always causes a
694 match failure.
695 .P
696 Sets of Unicode characters are defined as belonging to certain scripts. A
697 character from one of these sets can be matched using a script name. For
698 example:
699 .sp
700 \ep{Greek}
701 \eP{Han}
702 .sp
703 Those that are not part of an identified script are lumped together as
704 "Common". The current list of scripts is:
705 .P
706 Arabic,
707 Armenian,
708 Avestan,
709 Balinese,
710 Bamum,
711 Batak,
712 Bengali,
713 Bopomofo,
714 Brahmi,
715 Braille,
716 Buginese,
717 Buhid,
718 Canadian_Aboriginal,
719 Carian,
720 Chakma,
721 Cham,
722 Cherokee,
723 Common,
724 Coptic,
725 Cuneiform,
726 Cypriot,
727 Cyrillic,
728 Deseret,
729 Devanagari,
730 Egyptian_Hieroglyphs,
731 Ethiopic,
732 Georgian,
733 Glagolitic,
734 Gothic,
735 Greek,
736 Gujarati,
737 Gurmukhi,
738 Han,
739 Hangul,
740 Hanunoo,
741 Hebrew,
742 Hiragana,
743 Imperial_Aramaic,
744 Inherited,
745 Inscriptional_Pahlavi,
746 Inscriptional_Parthian,
747 Javanese,
748 Kaithi,
749 Kannada,
750 Katakana,
751 Kayah_Li,
752 Kharoshthi,
753 Khmer,
754 Lao,
755 Latin,
756 Lepcha,
757 Limbu,
758 Linear_B,
759 Lisu,
760 Lycian,
761 Lydian,
762 Malayalam,
763 Mandaic,
764 Meetei_Mayek,
765 Meroitic_Cursive,
766 Meroitic_Hieroglyphs,
767 Miao,
768 Mongolian,
769 Myanmar,
770 New_Tai_Lue,
771 Nko,
772 Ogham,
773 Old_Italic,
774 Old_Persian,
775 Old_South_Arabian,
776 Old_Turkic,
777 Ol_Chiki,
778 Oriya,
779 Osmanya,
780 Phags_Pa,
781 Phoenician,
782 Rejang,
783 Runic,
784 Samaritan,
785 Saurashtra,
786 Sharada,
787 Shavian,
788 Sinhala,
789 Sora_Sompeng,
790 Sundanese,
791 Syloti_Nagri,
792 Syriac,
793 Tagalog,
794 Tagbanwa,
795 Tai_Le,
796 Tai_Tham,
797 Tai_Viet,
798 Takri,
799 Tamil,
800 Telugu,
801 Thaana,
802 Thai,
803 Tibetan,
804 Tifinagh,
805 Ugaritic,
806 Vai,
807 Yi.
808 .P
809 Each character has exactly one Unicode general category property, specified by
810 a two-letter abbreviation. For compatibility with Perl, negation can be
811 specified by including a circumflex between the opening brace and the property
812 name. For example, \ep{^Lu} is the same as \eP{Lu}.
813 .P
814 If only one letter is specified with \ep or \eP, it includes all the general
815 category properties that start with that letter. In this case, in the absence
816 of negation, the curly brackets in the escape sequence are optional; these two
817 examples have the same effect:
818 .sp
819 \ep{L}
820 \epL
821 .sp
822 The following general category property codes are supported:
823 .sp
824 C Other
825 Cc Control
826 Cf Format
827 Cn Unassigned
828 Co Private use
829 Cs Surrogate
830 .sp
831 L Letter
832 Ll Lower case letter
833 Lm Modifier letter
834 Lo Other letter
835 Lt Title case letter
836 Lu Upper case letter
837 .sp
838 M Mark
839 Mc Spacing mark
840 Me Enclosing mark
841 Mn Non-spacing mark
842 .sp
843 N Number
844 Nd Decimal number
845 Nl Letter number
846 No Other number
847 .sp
848 P Punctuation
849 Pc Connector punctuation
850 Pd Dash punctuation
851 Pe Close punctuation
852 Pf Final punctuation
853 Pi Initial punctuation
854 Po Other punctuation
855 Ps Open punctuation
856 .sp
857 S Symbol
858 Sc Currency symbol
859 Sk Modifier symbol
860 Sm Mathematical symbol
861 So Other symbol
862 .sp
863 Z Separator
864 Zl Line separator
865 Zp Paragraph separator
866 Zs Space separator
867 .sp
868 The special property L& is also supported: it matches a character that has
869 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
870 a modifier or "other".
871 .P
872 The Cs (Surrogate) property applies only to characters in the range U+D800 to
873 U+DFFF. Such characters are not valid in Unicode strings and so
874 cannot be tested by PCRE, unless UTF validity checking has been turned off
875 (see the discussion of PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK and
876 PCRE_NO_UTF32_CHECK in the
877 .\" HREF
878 \fBpcreapi\fP
879 .\"
880 page). Perl does not support the Cs property.
881 .P
882 The long synonyms for property names that Perl supports (such as \ep{Letter})
883 are not supported by PCRE, nor is it permitted to prefix any of these
884 properties with "Is".
885 .P
886 No character that is in the Unicode table has the Cn (unassigned) property.
887 Instead, this property is assumed for any code point that is not in the
888 Unicode table.
889 .P
890 Specifying caseless matching does not affect these escape sequences. For
891 example, \ep{Lu} always matches only upper case letters. This is different from
892 the behaviour of current versions of Perl.
893 .P
894 Matching characters by Unicode property is not fast, because PCRE has to do a
895 multistage table lookup in order to find a character's property. That is why
896 the traditional escape sequences such as \ed and \ew do not use Unicode
897 properties in PCRE by default, though you can make them do so by setting the
898 PCRE_UCP option or by starting the pattern with (*UCP).
899 .
900 .
901 .SS Extended grapheme clusters
902 .rs
903 .sp
904 The \eX escape matches any number of Unicode characters that form an "extended
905 grapheme cluster", and treats the sequence as an atomic group
906 .\" HTML <a href="#atomicgroup">
907 .\" </a>
908 (see below).
909 .\"
910 Up to and including release 8.31, PCRE matched an earlier, simpler definition
911 that was equivalent to
912 .sp
913 (?>\ePM\epM*)
914 .sp
915 That is, it matched a character without the "mark" property, followed by zero
916 or more characters with the "mark" property. Characters with the "mark"
917 property are typically non-spacing accents that affect the preceding character.
918 .P
919 This simple definition was extended in Unicode to include more complicated
920 kinds of composite character by giving each character a grapheme breaking
921 property, and creating rules that use these properties to define the boundaries
922 of extended grapheme clusters. In releases of PCRE later than 8.31, \eX matches
923 one of these clusters.
924 .P
925 \eX always matches at least one character. Then it decides whether to add
926 additional characters according to the following rules for ending a cluster:
927 .P
928 1. End at the end of the subject string.
929 .P
930 2. Do not end between CR and LF; otherwise end after any control character.
931 .P
932 3. Do not break Hangul (a Korean script) syllable sequences. Hangul characters
933 are of five types: L, V, T, LV, and LVT. An L character may be followed by an
934 L, V, LV, or LVT character; an LV or V character may be followed by a V or T
935 character; an LVT or T character may be follwed only by a T character.
936 .P
937 4. Do not end before extending characters or spacing marks. Characters with
938 the "mark" property always have the "extend" grapheme breaking property.
939 .P
940 5. Do not end after prepend characters.
941 .P
942 6. Otherwise, end the cluster.
943 .
944 .
945 .\" HTML <a name="extraprops"></a>
946 .SS PCRE's additional properties
947 .rs
948 .sp
949 As well as the standard Unicode properties described above, PCRE supports four
950 more that make it possible to convert traditional escape sequences such as \ew
951 and \es to use Unicode properties. PCRE uses these non-standard, non-Perl
952 properties internally when PCRE_UCP is set. However, they may also be used
953 explicitly. These properties are:
954 .sp
955 Xan Any alphanumeric character
956 Xps Any POSIX space character
957 Xsp Any Perl space character
958 Xwd Any Perl "word" character
959 .sp
960 Xan matches characters that have either the L (letter) or the N (number)
961 property. Xps matches the characters tab, linefeed, vertical tab, form feed, or
962 carriage return, and any other character that has the Z (separator) property.
963 Xsp is the same as Xps; it used to exclude vertical tab, for Perl
964 compatibility, but Perl changed, and so PCRE followed at release 8.34. Xwd
965 matches the same characters as Xan, plus underscore.
966 .P
967 There is another non-standard property, Xuc, which matches any character that
968 can be represented by a Universal Character Name in C++ and other programming
969 languages. These are the characters $, @, ` (grave accent), and all characters
970 with Unicode code points greater than or equal to U+00A0, except for the
971 surrogates U+D800 to U+DFFF. Note that most base (ASCII) characters are
972 excluded. (Universal Character Names are of the form \euHHHH or \eUHHHHHHHH
973 where H is a hexadecimal digit. Note that the Xuc property does not match these
974 sequences but the characters that they represent.)
975 .
976 .
977 .\" HTML <a name="resetmatchstart"></a>
978 .SS "Resetting the match start"
979 .rs
980 .sp
981 The escape sequence \eK causes any previously matched characters not to be
982 included in the final matched sequence. For example, the pattern:
983 .sp
984 foo\eKbar
985 .sp
986 matches "foobar", but reports that it has matched "bar". This feature is
987 similar to a lookbehind assertion
988 .\" HTML <a href="#lookbehind">
989 .\" </a>
990 (described below).
991 .\"
992 However, in this case, the part of the subject before the real match does not
993 have to be of fixed length, as lookbehind assertions do. The use of \eK does
994 not interfere with the setting of
995 .\" HTML <a href="#subpattern">
996 .\" </a>
997 captured substrings.
998 .\"
999 For example, when the pattern
1000 .sp
1001 (foo)\eKbar
1002 .sp
1003 matches "foobar", the first substring is still set to "foo".
1004 .P
1005 Perl documents that the use of \eK within assertions is "not well defined". In
1006 PCRE, \eK is acted upon when it occurs inside positive assertions, but is
1007 ignored in negative assertions. Note that when a pattern such as (?=ab\eK)
1008 matches, the reported start of the match can be greater than the end of the
1009 match.
1010 .
1011 .
1012 .\" HTML <a name="smallassertions"></a>
1013 .SS "Simple assertions"
1014 .rs
1015 .sp
1016 The final use of backslash is for certain simple assertions. An assertion
1017 specifies a condition that has to be met at a particular point in a match,
1018 without consuming any characters from the subject string. The use of
1019 subpatterns for more complicated assertions is described
1020 .\" HTML <a href="#bigassertions">
1021 .\" </a>
1022 below.
1023 .\"
1024 The backslashed assertions are:
1025 .sp
1026 \eb matches at a word boundary
1027 \eB matches when not at a word boundary
1028 \eA matches at the start of the subject
1029 \eZ matches at the end of the subject
1030 also matches before a newline at the end of the subject
1031 \ez matches only at the end of the subject
1032 \eG matches at the first matching position in the subject
1033 .sp
1034 Inside a character class, \eb has a different meaning; it matches the backspace
1035 character. If any other of these assertions appears in a character class, by
1036 default it matches the corresponding literal character (for example, \eB
1037 matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
1038 escape sequence" error is generated instead.
1039 .P
1040 A word boundary is a position in the subject string where the current character
1041 and the previous character do not both match \ew or \eW (i.e. one matches
1042 \ew and the other matches \eW), or the start or end of the string if the
1043 first or last character matches \ew, respectively. In a UTF mode, the meanings
1044 of \ew and \eW can be changed by setting the PCRE_UCP option. When this is
1045 done, it also affects \eb and \eB. Neither PCRE nor Perl has a separate "start
1046 of word" or "end of word" metasequence. However, whatever follows \eb normally
1047 determines which it is. For example, the fragment \eba matches "a" at the start
1048 of a word.
1049 .P
1050 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
1051 dollar (described in the next section) in that they only ever match at the very
1052 start and end of the subject string, whatever options are set. Thus, they are
1053 independent of multiline mode. These three assertions are not affected by the
1054 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
1055 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
1056 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
1057 at a point other than the beginning of the subject, \eA can never match. The
1058 difference between \eZ and \ez is that \eZ matches before a newline at the end
1059 of the string as well as at the very end, whereas \ez matches only at the end.
1060 .P
1061 The \eG assertion is true only when the current matching position is at the
1062 start point of the match, as specified by the \fIstartoffset\fP argument of
1063 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
1064 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
1065 arguments, you can mimic Perl's /g option, and it is in this kind of
1066 implementation where \eG can be useful.
1067 .P
1068 Note, however, that PCRE's interpretation of \eG, as the start of the current
1069 match, is subtly different from Perl's, which defines it as the end of the
1070 previous match. In Perl, these can be different when the previously matched
1071 string was empty. Because PCRE does just one match at a time, it cannot
1072 reproduce this behaviour.
1073 .P
1074 If all the alternatives of a pattern begin with \eG, the expression is anchored
1075 to the starting match position, and the "anchored" flag is set in the compiled
1076 regular expression.
1077 .
1078 .
1080 .rs
1081 .sp
1082 The circumflex and dollar metacharacters are zero-width assertions. That is,
1083 they test for a particular condition being true without consuming any
1084 characters from the subject string.
1085 .P
1086 Outside a character class, in the default matching mode, the circumflex
1087 character is an assertion that is true only if the current matching point is at
1088 the start of the subject string. If the \fIstartoffset\fP argument of
1089 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
1090 option is unset. Inside a character class, circumflex has an entirely different
1091 meaning
1092 .\" HTML <a href="#characterclass">
1093 .\" </a>
1094 (see below).
1095 .\"
1096 .P
1097 Circumflex need not be the first character of the pattern if a number of
1098 alternatives are involved, but it should be the first thing in each alternative
1099 in which it appears if the pattern is ever to match that branch. If all
1100 possible alternatives start with a circumflex, that is, if the pattern is
1101 constrained to match only at the start of the subject, it is said to be an
1102 "anchored" pattern. (There are also other constructs that can cause a pattern
1103 to be anchored.)
1104 .P
1105 The dollar character is an assertion that is true only if the current matching
1106 point is at the end of the subject string, or immediately before a newline at
1107 the end of the string (by default). Note, however, that it does not actually
1108 match the newline. Dollar need not be the last character of the pattern if a
1109 number of alternatives are involved, but it should be the last item in any
1110 branch in which it appears. Dollar has no special meaning in a character class.
1111 .P
1112 The meaning of dollar can be changed so that it matches only at the very end of
1113 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
1114 does not affect the \eZ assertion.
1115 .P
1116 The meanings of the circumflex and dollar characters are changed if the
1117 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
1118 immediately after internal newlines as well as at the start of the subject
1119 string. It does not match after a newline that ends the string. A dollar
1120 matches before any newlines in the string, as well as at the very end, when
1121 PCRE_MULTILINE is set. When newline is specified as the two-character
1122 sequence CRLF, isolated CR and LF characters do not indicate newlines.
1123 .P
1124 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
1125 \en represents a newline) in multiline mode, but not otherwise. Consequently,
1126 patterns that are anchored in single line mode because all branches start with
1127 ^ are not anchored in multiline mode, and a match for circumflex is possible
1128 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
1129 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
1130 .P
1131 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
1132 end of the subject in both modes, and if all branches of a pattern start with
1133 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
1134 .
1135 .
1136 .\" HTML <a name="fullstopdot"></a>
1138 .rs
1139 .sp
1140 Outside a character class, a dot in the pattern matches any one character in
1141 the subject string except (by default) a character that signifies the end of a
1142 line.
1143 .P
1144 When a line ending is defined as a single character, dot never matches that
1145 character; when the two-character sequence CRLF is used, dot does not match CR
1146 if it is immediately followed by LF, but otherwise it matches all characters
1147 (including isolated CRs and LFs). When any Unicode line endings are being
1148 recognized, dot does not match CR or LF or any of the other line ending
1149 characters.
1150 .P
1151 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
1152 option is set, a dot matches any one character, without exception. If the
1153 two-character sequence CRLF is present in the subject string, it takes two dots
1154 to match it.
1155 .P
1156 The handling of dot is entirely independent of the handling of circumflex and
1157 dollar, the only relationship being that they both involve newlines. Dot has no
1158 special meaning in a character class.
1159 .P
1160 The escape sequence \eN behaves like a dot, except that it is not affected by
1161 the PCRE_DOTALL option. In other words, it matches any character except one
1162 that signifies the end of a line. Perl also uses \eN to match characters by
1163 name; PCRE does not support this.
1164 .
1165 .
1167 .rs
1168 .sp
1169 Outside a character class, the escape sequence \eC matches any one data unit,
1170 whether or not a UTF mode is set. In the 8-bit library, one data unit is one
1171 byte; in the 16-bit library it is a 16-bit unit; in the 32-bit library it is
1172 a 32-bit unit. Unlike a dot, \eC always
1173 matches line-ending characters. The feature is provided in Perl in order to
1174 match individual bytes in UTF-8 mode, but it is unclear how it can usefully be
1175 used. Because \eC breaks up characters into individual data units, matching one
1176 unit with \eC in a UTF mode means that the rest of the string may start with a
1177 malformed UTF character. This has undefined results, because PCRE assumes that
1178 it is dealing with valid UTF strings (and by default it checks this at the
1179 start of processing unless the PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK or
1180 PCRE_NO_UTF32_CHECK option is used).
1181 .P
1182 PCRE does not allow \eC to appear in lookbehind assertions
1183 .\" HTML <a href="#lookbehind">
1184 .\" </a>
1185 (described below)
1186 .\"
1187 in a UTF mode, because this would make it impossible to calculate the length of
1188 the lookbehind.
1189 .P
1190 In general, the \eC escape sequence is best avoided. However, one
1191 way of using it that avoids the problem of malformed UTF characters is to use a
1192 lookahead to check the length of the next character, as in this pattern, which
1193 could be used with a UTF-8 string (ignore white space and line breaks):
1194 .sp
1195 (?| (?=[\ex00-\ex7f])(\eC) |
1196 (?=[\ex80-\ex{7ff}])(\eC)(\eC) |
1197 (?=[\ex{800}-\ex{ffff}])(\eC)(\eC)(\eC) |
1198 (?=[\ex{10000}-\ex{1fffff}])(\eC)(\eC)(\eC)(\eC))
1199 .sp
1200 A group that starts with (?| resets the capturing parentheses numbers in each
1201 alternative (see
1202 .\" HTML <a href="#dupsubpatternnumber">
1203 .\" </a>
1204 "Duplicate Subpattern Numbers"
1205 .\"
1206 below). The assertions at the start of each branch check the next UTF-8
1207 character for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The
1208 character's individual bytes are then captured by the appropriate number of
1209 groups.
1210 .
1211 .
1212 .\" HTML <a name="characterclass"></a>
1214 .rs
1215 .sp
1216 An opening square bracket introduces a character class, terminated by a closing
1217 square bracket. A closing square bracket on its own is not special by default.
1218 However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
1219 bracket causes a compile-time error. If a closing square bracket is required as
1220 a member of the class, it should be the first data character in the class
1221 (after an initial circumflex, if present) or escaped with a backslash.
1222 .P
1223 A character class matches a single character in the subject. In a UTF mode, the
1224 character may be more than one data unit long. A matched character must be in
1225 the set of characters defined by the class, unless the first character in the
1226 class definition is a circumflex, in which case the subject character must not
1227 be in the set defined by the class. If a circumflex is actually required as a
1228 member of the class, ensure it is not the first character, or escape it with a
1229 backslash.
1230 .P
1231 For example, the character class [aeiou] matches any lower case vowel, while
1232 [^aeiou] matches any character that is not a lower case vowel. Note that a
1233 circumflex is just a convenient notation for specifying the characters that
1234 are in the class by enumerating those that are not. A class that starts with a
1235 circumflex is not an assertion; it still consumes a character from the subject
1236 string, and therefore it fails if the current pointer is at the end of the
1237 string.
1238 .P
1239 In UTF-8 (UTF-16, UTF-32) mode, characters with values greater than 255 (0xffff)
1240 can be included in a class as a literal string of data units, or by using the
1241 \ex{ escaping mechanism.
1242 .P
1243 When caseless matching is set, any letters in a class represent both their
1244 upper case and lower case versions, so for example, a caseless [aeiou] matches
1245 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1246 caseful version would. In a UTF mode, PCRE always understands the concept of
1247 case for characters whose values are less than 128, so caseless matching is
1248 always possible. For characters with higher values, the concept of case is
1249 supported if PCRE is compiled with Unicode property support, but not otherwise.
1250 If you want to use caseless matching in a UTF mode for characters 128 and
1251 above, you must ensure that PCRE is compiled with Unicode property support as
1252 well as with UTF support.
1253 .P
1254 Characters that might indicate line breaks are never treated in any special way
1255 when matching character classes, whatever line-ending sequence is in use, and
1256 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1257 such as [^a] always matches one of these characters.
1258 .P
1259 The minus (hyphen) character can be used to specify a range of characters in a
1260 character class. For example, [d-m] matches any letter between d and m,
1261 inclusive. If a minus character is required in a class, it must be escaped with
1262 a backslash or appear in a position where it cannot be interpreted as
1263 indicating a range, typically as the first or last character in the class, or
1264 immediately after a range. For example, [b-d-z] matches letters in the range b
1265 to d, a hyphen character, or z.
1266 .P
1267 It is not possible to have the literal character "]" as the end character of a
1268 range. A pattern such as [W-]46] is interpreted as a class of two characters
1269 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1270 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1271 the end of range, so [W-\e]46] is interpreted as a class containing a range
1272 followed by two other characters. The octal or hexadecimal representation of
1273 "]" can also be used to end a range.
1274 .P
1275 An error is generated if a POSIX character class (see below) or an escape
1276 sequence other than one that defines a single character appears at a point
1277 where a range ending character is expected. For example, [z-\exff] is valid,
1278 but [A-\ed] and [A-[:digit:]] are not.
1279 .P
1280 Ranges operate in the collating sequence of character values. They can also be
1281 used for characters specified numerically, for example [\e000-\e037]. Ranges
1282 can include any characters that are valid for the current mode.
1283 .P
1284 If a range that includes letters is used when caseless matching is set, it
1285 matches the letters in either case. For example, [W-c] is equivalent to
1286 [][\e\e^_`wxyzabc], matched caselessly, and in a non-UTF mode, if character
1287 tables for a French locale are in use, [\exc8-\excb] matches accented E
1288 characters in both cases. In UTF modes, PCRE supports the concept of case for
1289 characters with values greater than 128 only when it is compiled with Unicode
1290 property support.
1291 .P
1292 The character escape sequences \ed, \eD, \eh, \eH, \ep, \eP, \es, \eS, \ev,
1293 \eV, \ew, and \eW may appear in a character class, and add the characters that
1294 they match to the class. For example, [\edABCDEF] matches any hexadecimal
1295 digit. In UTF modes, the PCRE_UCP option affects the meanings of \ed, \es, \ew
1296 and their upper case partners, just as it does when they appear outside a
1297 character class, as described in the section entitled
1298 .\" HTML <a href="#genericchartypes">
1299 .\" </a>
1300 "Generic character types"
1301 .\"
1302 above. The escape sequence \eb has a different meaning inside a character
1303 class; it matches the backspace character. The sequences \eB, \eN, \eR, and \eX
1304 are not special inside a character class. Like any other unrecognized escape
1305 sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1306 default, but cause an error if the PCRE_EXTRA option is set.
1307 .P
1308 A circumflex can conveniently be used with the upper case character types to
1309 specify a more restricted set of characters than the matching lower case type.
1310 For example, the class [^\eW_] matches any letter or digit, but not underscore,
1311 whereas [\ew] includes underscore. A positive character class should be read as
1312 "something OR something OR ..." and a negative class as "NOT something AND NOT
1313 something AND NOT ...".
1314 .P
1315 The only metacharacters that are recognized in character classes are backslash,
1316 hyphen (only where it can be interpreted as specifying a range), circumflex
1317 (only at the start), opening square bracket (only when it can be interpreted as
1318 introducing a POSIX class name, or for a special compatibility feature - see
1319 the next two sections), and the terminating closing square bracket. However,
1320 escaping other non-alphanumeric characters does no harm.
1321 .
1322 .
1324 .rs
1325 .sp
1326 Perl supports the POSIX notation for character classes. This uses names
1327 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1328 this notation. For example,
1329 .sp
1330 [01[:alpha:]%]
1331 .sp
1332 matches "0", "1", any alphabetic character, or "%". The supported class names
1333 are:
1334 .sp
1335 alnum letters and digits
1336 alpha letters
1337 ascii character codes 0 - 127
1338 blank space or tab only
1339 cntrl control characters
1340 digit decimal digits (same as \ed)
1341 graph printing characters, excluding space
1342 lower lower case letters
1343 print printing characters, including space
1344 punct printing characters, excluding letters and digits and space
1345 space white space (the same as \es from PCRE 8.34)
1346 upper upper case letters
1347 word "word" characters (same as \ew)
1348 xdigit hexadecimal digits
1349 .sp
1350 The default "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13),
1351 and space (32). If locale-specific matching is taking place, the list of space
1352 characters may be different; there may be fewer or more of them. "Space" used
1353 to be different to \es, which did not include VT, for Perl compatibility.
1354 However, Perl changed at release 5.18, and PCRE followed at release 8.34.
1355 "Space" and \es now match the same set of characters.
1356 .P
1357 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1358 5.8. Another Perl extension is negation, which is indicated by a ^ character
1359 after the colon. For example,
1360 .sp
1361 [12[:^digit:]]
1362 .sp
1363 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1364 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1365 supported, and an error is given if they are encountered.
1366 .P
1367 By default, characters with values greater than 128 do not match any of the
1368 POSIX character classes. However, if the PCRE_UCP option is passed to
1369 \fBpcre_compile()\fP, some of the classes are changed so that Unicode character
1370 properties are used. This is achieved by replacing certain POSIX classes by
1371 other sequences, as follows:
1372 .sp
1373 [:alnum:] becomes \ep{Xan}
1374 [:alpha:] becomes \ep{L}
1375 [:blank:] becomes \eh
1376 [:digit:] becomes \ep{Nd}
1377 [:lower:] becomes \ep{Ll}
1378 [:space:] becomes \ep{Xps}
1379 [:upper:] becomes \ep{Lu}
1380 [:word:] becomes \ep{Xwd}
1381 .sp
1382 Negated versions, such as [:^alpha:] use \eP instead of \ep. Three other POSIX
1383 classes are handled specially in UCP mode:
1384 .TP 10
1385 [:graph:]
1386 This matches characters that have glyphs that mark the page when printed. In
1387 Unicode property terms, it matches all characters with the L, M, N, P, S, or Cf
1388 properties, except for:
1389 .sp
1390 U+061C Arabic Letter Mark
1391 U+180E Mongolian Vowel Separator
1392 U+2066 - U+2069 Various "isolate"s
1393 .sp
1394 .TP 10
1395 [:print:]
1396 This matches the same characters as [:graph:] plus space characters that are
1397 not controls, that is, characters with the Zs property.
1398 .TP 10
1399 [:punct:]
1400 This matches all characters that have the Unicode P (punctuation) property,
1401 plus those characters whose code points are less than 128 that have the S
1402 (Symbol) property.
1403 .P
1404 The other POSIX classes are unchanged, and match only characters with code
1405 points less than 128.
1406 .
1407 .
1409 .rs
1410 .sp
1411 In the POSIX.2 compliant library that was included in 4.4BSD Unix, the ugly
1412 syntax [[:<:]] and [[:>:]] is used for matching "start of word" and "end of
1413 word". PCRE treats these items as follows:
1414 .sp
1415 [[:<:]] is converted to \eb(?=\ew)
1416 [[:>:]] is converted to \eb(?<=\ew)
1417 .sp
1418 Only these exact character sequences are recognized. A sequence such as
1419 [a[:<:]b] provokes error for an unrecognized POSIX class name. This support is
1420 not compatible with Perl. It is provided to help migrations from other
1421 environments, and is best not used in any new patterns. Note that \eb matches
1422 at the start and the end of a word (see
1423 .\" HTML <a href="#smallassertions">
1424 .\" </a>
1425 "Simple assertions"
1426 .\"
1427 above), and in a Perl-style pattern the preceding or following character
1428 normally shows which is wanted, without the need for the assertions that are
1429 used above in order to give exactly the POSIX behaviour.
1430 .
1431 .
1433 .rs
1434 .sp
1435 Vertical bar characters are used to separate alternative patterns. For example,
1436 the pattern
1437 .sp
1438 gilbert|sullivan
1439 .sp
1440 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1441 and an empty alternative is permitted (matching the empty string). The matching
1442 process tries each alternative in turn, from left to right, and the first one
1443 that succeeds is used. If the alternatives are within a subpattern
1444 .\" HTML <a href="#subpattern">
1445 .\" </a>
1446 (defined below),
1447 .\"
1448 "succeeds" means matching the rest of the main pattern as well as the
1449 alternative in the subpattern.
1450 .
1451 .
1453 .rs
1454 .sp
1455 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1456 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1457 the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1458 The option letters are
1459 .sp
1460 i for PCRE_CASELESS
1462 s for PCRE_DOTALL
1463 x for PCRE_EXTENDED
1464 .sp
1465 For example, (?im) sets caseless, multiline matching. It is also possible to
1466 unset these options by preceding the letter with a hyphen, and a combined
1467 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1468 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1469 permitted. If a letter appears both before and after the hyphen, the option is
1470 unset.
1471 .P
1472 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1473 changed in the same way as the Perl-compatible options by using the characters
1474 J, U and X respectively.
1475 .P
1476 When one of these option changes occurs at top level (that is, not inside
1477 subpattern parentheses), the change applies to the remainder of the pattern
1478 that follows. If the change is placed right at the start of a pattern, PCRE
1479 extracts it into the global options (and it will therefore show up in data
1480 extracted by the \fBpcre_fullinfo()\fP function).
1481 .P
1482 An option change within a subpattern (see below for a description of
1483 subpatterns) affects only that part of the subpattern that follows it, so
1484 .sp
1485 (a(?i)b)c
1486 .sp
1487 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1488 By this means, options can be made to have different settings in different
1489 parts of the pattern. Any changes made in one alternative do carry on
1490 into subsequent branches within the same subpattern. For example,
1491 .sp
1492 (a(?i)b|c)
1493 .sp
1494 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1495 branch is abandoned before the option setting. This is because the effects of
1496 option settings happen at compile time. There would be some very weird
1497 behaviour otherwise.
1498 .P
1499 \fBNote:\fP There are other PCRE-specific options that can be set by the
1500 application when the compiling or matching functions are called. In some cases
1501 the pattern can contain special leading sequences such as (*CRLF) to override
1502 what the application has set or what has been defaulted. Details are given in
1503 the section entitled
1504 .\" HTML <a href="#newlineseq">
1505 .\" </a>
1506 "Newline sequences"
1507 .\"
1508 above. There are also the (*UTF8), (*UTF16),(*UTF32), and (*UCP) leading
1509 sequences that can be used to set UTF and Unicode property modes; they are
1510 equivalent to setting the PCRE_UTF8, PCRE_UTF16, PCRE_UTF32 and the PCRE_UCP
1511 options, respectively. The (*UTF) sequence is a generic version that can be
1512 used with any of the libraries. However, the application can set the
1513 PCRE_NEVER_UTF option, which locks out the use of the (*UTF) sequences.
1514 .
1515 .
1516 .\" HTML <a name="subpattern"></a>
1518 .rs
1519 .sp
1520 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1521 Turning part of a pattern into a subpattern does two things:
1522 .sp
1523 1. It localizes a set of alternatives. For example, the pattern
1524 .sp
1525 cat(aract|erpillar|)
1526 .sp
1527 matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1528 match "cataract", "erpillar" or an empty string.
1529 .sp
1530 2. It sets up the subpattern as a capturing subpattern. This means that, when
1531 the whole pattern matches, that portion of the subject string that matched the
1532 subpattern is passed back to the caller via the \fIovector\fP argument of the
1533 matching function. (This applies only to the traditional matching functions;
1534 the DFA matching functions do not support capturing.)
1535 .P
1536 Opening parentheses are counted from left to right (starting from 1) to obtain
1537 numbers for the capturing subpatterns. For example, if the string "the red
1538 king" is matched against the pattern
1539 .sp
1540 the ((red|white) (king|queen))
1541 .sp
1542 the captured substrings are "red king", "red", and "king", and are numbered 1,
1543 2, and 3, respectively.
1544 .P
1545 The fact that plain parentheses fulfil two functions is not always helpful.
1546 There are often times when a grouping subpattern is required without a
1547 capturing requirement. If an opening parenthesis is followed by a question mark
1548 and a colon, the subpattern does not do any capturing, and is not counted when
1549 computing the number of any subsequent capturing subpatterns. For example, if
1550 the string "the white queen" is matched against the pattern
1551 .sp
1552 the ((?:red|white) (king|queen))
1553 .sp
1554 the captured substrings are "white queen" and "queen", and are numbered 1 and
1555 2. The maximum number of capturing subpatterns is 65535.
1556 .P
1557 As a convenient shorthand, if any option settings are required at the start of
1558 a non-capturing subpattern, the option letters may appear between the "?" and
1559 the ":". Thus the two patterns
1560 .sp
1561 (?i:saturday|sunday)
1562 (?:(?i)saturday|sunday)
1563 .sp
1564 match exactly the same set of strings. Because alternative branches are tried
1565 from left to right, and options are not reset until the end of the subpattern
1566 is reached, an option setting in one branch does affect subsequent branches, so
1567 the above patterns match "SUNDAY" as well as "Saturday".
1568 .
1569 .
1570 .\" HTML <a name="dupsubpatternnumber"></a>
1572 .rs
1573 .sp
1574 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1575 the same numbers for its capturing parentheses. Such a subpattern starts with
1576 (?| and is itself a non-capturing subpattern. For example, consider this
1577 pattern:
1578 .sp
1579 (?|(Sat)ur|(Sun))day
1580 .sp
1581 Because the two alternatives are inside a (?| group, both sets of capturing
1582 parentheses are numbered one. Thus, when the pattern matches, you can look
1583 at captured substring number one, whichever alternative matched. This construct
1584 is useful when you want to capture part, but not all, of one of a number of
1585 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1586 number is reset at the start of each branch. The numbers of any capturing
1587 parentheses that follow the subpattern start after the highest number used in
1588 any branch. The following example is taken from the Perl documentation. The
1589 numbers underneath show in which buffer the captured content will be stored.
1590 .sp
1591 # before ---------------branch-reset----------- after
1592 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1593 # 1 2 2 3 2 3 4
1594 .sp
1595 A back reference to a numbered subpattern uses the most recent value that is
1596 set for that number by any subpattern. The following pattern matches "abcabc"
1597 or "defdef":
1598 .sp
1599 /(?|(abc)|(def))\e1/
1600 .sp
1601 In contrast, a subroutine call to a numbered subpattern always refers to the
1602 first one in the pattern with the given number. The following pattern matches
1603 "abcabc" or "defabc":
1604 .sp
1605 /(?|(abc)|(def))(?1)/
1606 .sp
1607 If a
1608 .\" HTML <a href="#conditions">
1609 .\" </a>
1610 condition test
1611 .\"
1612 for a subpattern's having matched refers to a non-unique number, the test is
1613 true if any of the subpatterns of that number have matched.
1614 .P
1615 An alternative approach to using this "branch reset" feature is to use
1616 duplicate named subpatterns, as described in the next section.
1617 .
1618 .
1620 .rs
1621 .sp
1622 Identifying capturing parentheses by number is simple, but it can be very hard
1623 to keep track of the numbers in complicated regular expressions. Furthermore,
1624 if an expression is modified, the numbers may change. To help with this
1625 difficulty, PCRE supports the naming of subpatterns. This feature was not
1626 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1627 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1628 the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1629 have different names, but PCRE does not.
1630 .P
1631 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1632 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1633 parentheses from other parts of the pattern, such as
1634 .\" HTML <a href="#backreferences">
1635 .\" </a>
1636 back references,
1637 .\"
1638 .\" HTML <a href="#recursion">
1639 .\" </a>
1640 recursion,
1641 .\"
1642 and
1643 .\" HTML <a href="#conditions">
1644 .\" </a>
1645 conditions,
1646 .\"
1647 can be made by name as well as by number.
1648 .P
1649 Names consist of up to 32 alphanumeric characters and underscores, but must
1650 start with a non-digit. Named capturing parentheses are still allocated numbers
1651 as well as names, exactly as if the names were not present. The PCRE API
1652 provides function calls for extracting the name-to-number translation table
1653 from a compiled pattern. There is also a convenience function for extracting a
1654 captured substring by name.
1655 .P
1656 By default, a name must be unique within a pattern, but it is possible to relax
1657 this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1658 names are also always permitted for subpatterns with the same number, set up as
1659 described in the previous section.) Duplicate names can be useful for patterns
1660 where only one instance of the named parentheses can match. Suppose you want to
1661 match the name of a weekday, either as a 3-letter abbreviation or as the full
1662 name, and in both cases you want to extract the abbreviation. This pattern
1663 (ignoring the line breaks) does the job:
1664 .sp
1665 (?<DN>Mon|Fri|Sun)(?:day)?|
1666 (?<DN>Tue)(?:sday)?|
1667 (?<DN>Wed)(?:nesday)?|
1668 (?<DN>Thu)(?:rsday)?|
1669 (?<DN>Sat)(?:urday)?
1670 .sp
1671 There are five capturing substrings, but only one is ever set after a match.
1672 (An alternative way of solving this problem is to use a "branch reset"
1673 subpattern, as described in the previous section.)
1674 .P
1675 The convenience function for extracting the data by name returns the substring
1676 for the first (and in this example, the only) subpattern of that name that
1677 matched. This saves searching to find which numbered subpattern it was.
1678 .P
1679 If you make a back reference to a non-unique named subpattern from elsewhere in
1680 the pattern, the subpatterns to which the name refers are checked in the order
1681 in which they appear in the overall pattern. The first one that is set is used
1682 for the reference. For example, this pattern matches both "foofoo" and
1683 "barbar" but not "foobar" or "barfoo":
1684 .sp
1685 (?:(?<n>foo)|(?<n>bar))\ek<n>
1686 .sp
1687 .P
1688 If you make a subroutine call to a non-unique named subpattern, the one that
1689 corresponds to the first occurrence of the name is used. In the absence of
1690 duplicate numbers (see the previous section) this is the one with the lowest
1691 number.
1692 .P
1693 If you use a named reference in a condition
1694 test (see the
1695 .\"
1696 .\" HTML <a href="#conditions">
1697 .\" </a>
1698 section about conditions
1699 .\"
1700 below), either to check whether a subpattern has matched, or to check for
1701 recursion, all subpatterns with the same name are tested. If the condition is
1702 true for any one of them, the overall condition is true. This is the same
1703 behaviour as testing by number. For further details of the interfaces for
1704 handling named subpatterns, see the
1705 .\" HREF
1706 \fBpcreapi\fP
1707 .\"
1708 documentation.
1709 .P
1710 \fBWarning:\fP You cannot use different names to distinguish between two
1711 subpatterns with the same number because PCRE uses only the numbers when
1712 matching. For this reason, an error is given at compile time if different names
1713 are given to subpatterns with the same number. However, you can always give the
1714 same name to subpatterns with the same number, even when PCRE_DUPNAMES is not
1715 set.
1716 .
1717 .
1719 .rs
1720 .sp
1721 Repetition is specified by quantifiers, which can follow any of the following
1722 items:
1723 .sp
1724 a literal data character
1725 the dot metacharacter
1726 the \eC escape sequence
1727 the \eX escape sequence
1728 the \eR escape sequence
1729 an escape such as \ed or \epL that matches a single character
1730 a character class
1731 a back reference (see next section)
1732 a parenthesized subpattern (including assertions)
1733 a subroutine call to a subpattern (recursive or otherwise)
1734 .sp
1735 The general repetition quantifier specifies a minimum and maximum number of
1736 permitted matches, by giving the two numbers in curly brackets (braces),
1737 separated by a comma. The numbers must be less than 65536, and the first must
1738 be less than or equal to the second. For example:
1739 .sp
1740 z{2,4}
1741 .sp
1742 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1743 character. If the second number is omitted, but the comma is present, there is
1744 no upper limit; if the second number and the comma are both omitted, the
1745 quantifier specifies an exact number of required matches. Thus
1746 .sp
1747 [aeiou]{3,}
1748 .sp
1749 matches at least 3 successive vowels, but may match many more, while
1750 .sp
1751 \ed{8}
1752 .sp
1753 matches exactly 8 digits. An opening curly bracket that appears in a position
1754 where a quantifier is not allowed, or one that does not match the syntax of a
1755 quantifier, is taken as a literal character. For example, {,6} is not a
1756 quantifier, but a literal string of four characters.
1757 .P
1758 In UTF modes, quantifiers apply to characters rather than to individual data
1759 units. Thus, for example, \ex{100}{2} matches two characters, each of
1760 which is represented by a two-byte sequence in a UTF-8 string. Similarly,
1761 \eX{3} matches three Unicode extended grapheme clusters, each of which may be
1762 several data units long (and they may be of different lengths).
1763 .P
1764 The quantifier {0} is permitted, causing the expression to behave as if the
1765 previous item and the quantifier were not present. This may be useful for
1766 subpatterns that are referenced as
1767 .\" HTML <a href="#subpatternsassubroutines">
1768 .\" </a>
1769 subroutines
1770 .\"
1771 from elsewhere in the pattern (but see also the section entitled
1772 .\" HTML <a href="#subdefine">
1773 .\" </a>
1774 "Defining subpatterns for use by reference only"
1775 .\"
1776 below). Items other than subpatterns that have a {0} quantifier are omitted
1777 from the compiled pattern.
1778 .P
1779 For convenience, the three most common quantifiers have single-character
1780 abbreviations:
1781 .sp
1782 * is equivalent to {0,}
1783 + is equivalent to {1,}
1784 ? is equivalent to {0,1}
1785 .sp
1786 It is possible to construct infinite loops by following a subpattern that can
1787 match no characters with a quantifier that has no upper limit, for example:
1788 .sp
1789 (a?)*
1790 .sp
1791 Earlier versions of Perl and PCRE used to give an error at compile time for
1792 such patterns. However, because there are cases where this can be useful, such
1793 patterns are now accepted, but if any repetition of the subpattern does in fact
1794 match no characters, the loop is forcibly broken.
1795 .P
1796 By default, the quantifiers are "greedy", that is, they match as much as
1797 possible (up to the maximum number of permitted times), without causing the
1798 rest of the pattern to fail. The classic example of where this gives problems
1799 is in trying to match comments in C programs. These appear between /* and */
1800 and within the comment, individual * and / characters may appear. An attempt to
1801 match C comments by applying the pattern
1802 .sp
1803 /\e*.*\e*/
1804 .sp
1805 to the string
1806 .sp
1807 /* first comment */ not comment /* second comment */
1808 .sp
1809 fails, because it matches the entire string owing to the greediness of the .*
1810 item.
1811 .P
1812 However, if a quantifier is followed by a question mark, it ceases to be
1813 greedy, and instead matches the minimum number of times possible, so the
1814 pattern
1815 .sp
1816 /\e*.*?\e*/
1817 .sp
1818 does the right thing with the C comments. The meaning of the various
1819 quantifiers is not otherwise changed, just the preferred number of matches.
1820 Do not confuse this use of question mark with its use as a quantifier in its
1821 own right. Because it has two uses, it can sometimes appear doubled, as in
1822 .sp
1823 \ed??\ed
1824 .sp
1825 which matches one digit by preference, but can match two if that is the only
1826 way the rest of the pattern matches.
1827 .P
1828 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1829 the quantifiers are not greedy by default, but individual ones can be made
1830 greedy by following them with a question mark. In other words, it inverts the
1831 default behaviour.
1832 .P
1833 When a parenthesized subpattern is quantified with a minimum repeat count that
1834 is greater than 1 or with a limited maximum, more memory is required for the
1835 compiled pattern, in proportion to the size of the minimum or maximum.
1836 .P
1837 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1838 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1839 implicitly anchored, because whatever follows will be tried against every
1840 character position in the subject string, so there is no point in retrying the
1841 overall match at any position after the first. PCRE normally treats such a
1842 pattern as though it were preceded by \eA.
1843 .P
1844 In cases where it is known that the subject string contains no newlines, it is
1845 worth setting PCRE_DOTALL in order to obtain this optimization, or
1846 alternatively using ^ to indicate anchoring explicitly.
1847 .P
1848 However, there are some cases where the optimization cannot be used. When .*
1849 is inside capturing parentheses that are the subject of a back reference
1850 elsewhere in the pattern, a match at the start may fail where a later one
1851 succeeds. Consider, for example:
1852 .sp
1853 (.*)abc\e1
1854 .sp
1855 If the subject is "xyz123abc123" the match point is the fourth character. For
1856 this reason, such a pattern is not implicitly anchored.
1857 .P
1858 Another case where implicit anchoring is not applied is when the leading .* is
1859 inside an atomic group. Once again, a match at the start may fail where a later
1860 one succeeds. Consider this pattern:
1861 .sp
1862 (?>.*?a)b
1863 .sp
1864 It matches "ab" in the subject "aab". The use of the backtracking control verbs
1865 (*PRUNE) and (*SKIP) also disable this optimization.
1866 .P
1867 When a capturing subpattern is repeated, the value captured is the substring
1868 that matched the final iteration. For example, after
1869 .sp
1870 (tweedle[dume]{3}\es*)+
1871 .sp
1872 has matched "tweedledum tweedledee" the value of the captured substring is
1873 "tweedledee". However, if there are nested capturing subpatterns, the
1874 corresponding captured values may have been set in previous iterations. For
1875 example, after
1876 .sp
1877 /(a|(b))+/
1878 .sp
1879 matches "aba" the value of the second captured substring is "b".
1880 .
1881 .
1882 .\" HTML <a name="atomicgroup"></a>
1884 .rs
1885 .sp
1886 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1887 repetition, failure of what follows normally causes the repeated item to be
1888 re-evaluated to see if a different number of repeats allows the rest of the
1889 pattern to match. Sometimes it is useful to prevent this, either to change the
1890 nature of the match, or to cause it fail earlier than it otherwise might, when
1891 the author of the pattern knows there is no point in carrying on.
1892 .P
1893 Consider, for example, the pattern \ed+foo when applied to the subject line
1894 .sp
1895 123456bar
1896 .sp
1897 After matching all 6 digits and then failing to match "foo", the normal
1898 action of the matcher is to try again with only 5 digits matching the \ed+
1899 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1900 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1901 that once a subpattern has matched, it is not to be re-evaluated in this way.
1902 .P
1903 If we use atomic grouping for the previous example, the matcher gives up
1904 immediately on failing to match "foo" the first time. The notation is a kind of
1905 special parenthesis, starting with (?> as in this example:
1906 .sp
1907 (?>\ed+)foo
1908 .sp
1909 This kind of parenthesis "locks up" the part of the pattern it contains once
1910 it has matched, and a failure further into the pattern is prevented from
1911 backtracking into it. Backtracking past it to previous items, however, works as
1912 normal.
1913 .P
1914 An alternative description is that a subpattern of this type matches the string
1915 of characters that an identical standalone pattern would match, if anchored at
1916 the current point in the subject string.
1917 .P
1918 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1919 the above example can be thought of as a maximizing repeat that must swallow
1920 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1921 number of digits they match in order to make the rest of the pattern match,
1922 (?>\ed+) can only match an entire sequence of digits.
1923 .P
1924 Atomic groups in general can of course contain arbitrarily complicated
1925 subpatterns, and can be nested. However, when the subpattern for an atomic
1926 group is just a single repeated item, as in the example above, a simpler
1927 notation, called a "possessive quantifier" can be used. This consists of an
1928 additional + character following a quantifier. Using this notation, the
1929 previous example can be rewritten as
1930 .sp
1931 \ed++foo
1932 .sp
1933 Note that a possessive quantifier can be used with an entire group, for
1934 example:
1935 .sp
1936 (abc|xyz){2,3}+
1937 .sp
1938 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1939 option is ignored. They are a convenient notation for the simpler forms of
1940 atomic group. However, there is no difference in the meaning of a possessive
1941 quantifier and the equivalent atomic group, though there may be a performance
1942 difference; possessive quantifiers should be slightly faster.
1943 .P
1944 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1945 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1946 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1947 package, and PCRE copied it from there. It ultimately found its way into Perl
1948 at release 5.10.
1949 .P
1950 PCRE has an optimization that automatically "possessifies" certain simple
1951 pattern constructs. For example, the sequence A+B is treated as A++B because
1952 there is no point in backtracking into a sequence of A's when B must follow.
1953 .P
1954 When a pattern contains an unlimited repeat inside a subpattern that can itself
1955 be repeated an unlimited number of times, the use of an atomic group is the
1956 only way to avoid some failing matches taking a very long time indeed. The
1957 pattern
1958 .sp
1959 (\eD+|<\ed+>)*[!?]
1960 .sp
1961 matches an unlimited number of substrings that either consist of non-digits, or
1962 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1963 quickly. However, if it is applied to
1964 .sp
1965 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1966 .sp
1967 it takes a long time before reporting failure. This is because the string can
1968 be divided between the internal \eD+ repeat and the external * repeat in a
1969 large number of ways, and all have to be tried. (The example uses [!?] rather
1970 than a single character at the end, because both PCRE and Perl have an
1971 optimization that allows for fast failure when a single character is used. They
1972 remember the last single character that is required for a match, and fail early
1973 if it is not present in the string.) If the pattern is changed so that it uses
1974 an atomic group, like this:
1975 .sp
1976 ((?>\eD+)|<\ed+>)*[!?]
1977 .sp
1978 sequences of non-digits cannot be broken, and failure happens quickly.
1979 .
1980 .
1981 .\" HTML <a name="backreferences"></a>
1983 .rs
1984 .sp
1985 Outside a character class, a backslash followed by a digit greater than 0 (and
1986 possibly further digits) is a back reference to a capturing subpattern earlier
1987 (that is, to its left) in the pattern, provided there have been that many
1988 previous capturing left parentheses.
1989 .P
1990 However, if the decimal number following the backslash is less than 10, it is
1991 always taken as a back reference, and causes an error only if there are not
1992 that many capturing left parentheses in the entire pattern. In other words, the
1993 parentheses that are referenced need not be to the left of the reference for
1994 numbers less than 10. A "forward back reference" of this type can make sense
1995 when a repetition is involved and the subpattern to the right has participated
1996 in an earlier iteration.
1997 .P
1998 It is not possible to have a numerical "forward back reference" to a subpattern
1999 whose number is 10 or more using this syntax because a sequence such as \e50 is
2000 interpreted as a character defined in octal. See the subsection entitled
2001 "Non-printing characters"
2002 .\" HTML <a href="#digitsafterbackslash">
2003 .\" </a>
2004 above
2005 .\"
2006 for further details of the handling of digits following a backslash. There is
2007 no such problem when named parentheses are used. A back reference to any
2008 subpattern is possible using named parentheses (see below).
2009 .P
2010 Another way of avoiding the ambiguity inherent in the use of digits following a
2011 backslash is to use the \eg escape sequence. This escape must be followed by an
2012 unsigned number or a negative number, optionally enclosed in braces. These
2013 examples are all identical:
2014 .sp
2015 (ring), \e1
2016 (ring), \eg1
2017 (ring), \eg{1}
2018 .sp
2019 An unsigned number specifies an absolute reference without the ambiguity that
2020 is present in the older syntax. It is also useful when literal digits follow
2021 the reference. A negative number is a relative reference. Consider this
2022 example:
2023 .sp
2024 (abc(def)ghi)\eg{-1}
2025 .sp
2026 The sequence \eg{-1} is a reference to the most recently started capturing
2027 subpattern before \eg, that is, is it equivalent to \e2 in this example.
2028 Similarly, \eg{-2} would be equivalent to \e1. The use of relative references
2029 can be helpful in long patterns, and also in patterns that are created by
2030 joining together fragments that contain references within themselves.
2031 .P
2032 A back reference matches whatever actually matched the capturing subpattern in
2033 the current subject string, rather than anything matching the subpattern
2034 itself (see
2035 .\" HTML <a href="#subpatternsassubroutines">
2036 .\" </a>
2037 "Subpatterns as subroutines"
2038 .\"
2039 below for a way of doing that). So the pattern
2040 .sp
2041 (sens|respons)e and \e1ibility
2042 .sp
2043 matches "sense and sensibility" and "response and responsibility", but not
2044 "sense and responsibility". If caseful matching is in force at the time of the
2045 back reference, the case of letters is relevant. For example,
2046 .sp
2047 ((?i)rah)\es+\e1
2048 .sp
2049 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
2050 capturing subpattern is matched caselessly.
2051 .P
2052 There are several different ways of writing back references to named
2053 subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
2054 \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
2055 back reference syntax, in which \eg can be used for both numeric and named
2056 references, is also supported. We could rewrite the above example in any of
2057 the following ways:
2058 .sp
2059 (?<p1>(?i)rah)\es+\ek<p1>
2060 (?'p1'(?i)rah)\es+\ek{p1}
2061 (?P<p1>(?i)rah)\es+(?P=p1)
2062 (?<p1>(?i)rah)\es+\eg{p1}
2063 .sp
2064 A subpattern that is referenced by name may appear in the pattern before or
2065 after the reference.
2066 .P
2067 There may be more than one back reference to the same subpattern. If a
2068 subpattern has not actually been used in a particular match, any back
2069 references to it always fail by default. For example, the pattern
2070 .sp
2071 (a|(bc))\e2
2072 .sp
2073 always fails if it starts to match "a" rather than "bc". However, if the
2074 PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
2075 unset value matches an empty string.
2076 .P
2077 Because there may be many capturing parentheses in a pattern, all digits
2078 following a backslash are taken as part of a potential back reference number.
2079 If the pattern continues with a digit character, some delimiter must be used to
2080 terminate the back reference. If the PCRE_EXTENDED option is set, this can be
2081 white space. Otherwise, the \eg{ syntax or an empty comment (see
2082 .\" HTML <a href="#comments">
2083 .\" </a>
2084 "Comments"
2085 .\"
2086 below) can be used.
2087 .
2088 .SS "Recursive back references"
2089 .rs
2090 .sp
2091 A back reference that occurs inside the parentheses to which it refers fails
2092 when the subpattern is first used, so, for example, (a\e1) never matches.
2093 However, such references can be useful inside repeated subpatterns. For
2094 example, the pattern
2095 .sp
2096 (a|b\e1)+
2097 .sp
2098 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
2099 the subpattern, the back reference matches the character string corresponding
2100 to the previous iteration. In order for this to work, the pattern must be such
2101 that the first iteration does not need to match the back reference. This can be
2102 done using alternation, as in the example above, or by a quantifier with a
2103 minimum of zero.
2104 .P
2105 Back references of this type cause the group that they reference to be treated
2106 as an
2107 .\" HTML <a href="#atomicgroup">
2108 .\" </a>
2109 atomic group.
2110 .\"
2111 Once the whole group has been matched, a subsequent matching failure cannot
2112 cause backtracking into the middle of the group.
2113 .
2114 .
2115 .\" HTML <a name="bigassertions"></a>
2117 .rs
2118 .sp
2119 An assertion is a test on the characters following or preceding the current
2120 matching point that does not actually consume any characters. The simple
2121 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
2122 .\" HTML <a href="#smallassertions">
2123 .\" </a>
2124 above.
2125 .\"
2126 .P
2127 More complicated assertions are coded as subpatterns. There are two kinds:
2128 those that look ahead of the current position in the subject string, and those
2129 that look behind it. An assertion subpattern is matched in the normal way,
2130 except that it does not cause the current matching position to be changed.
2131 .P
2132 Assertion subpatterns are not capturing subpatterns. If such an assertion
2133 contains capturing subpatterns within it, these are counted for the purposes of
2134 numbering the capturing subpatterns in the whole pattern. However, substring
2135 capturing is carried out only for positive assertions. (Perl sometimes, but not
2136 always, does do capturing in negative assertions.)
2137 .P
2138 For compatibility with Perl, assertion subpatterns may be repeated; though
2139 it makes no sense to assert the same thing several times, the side effect of
2140 capturing parentheses may occasionally be useful. In practice, there only three
2141 cases:
2142 .sp
2143 (1) If the quantifier is {0}, the assertion is never obeyed during matching.
2144 However, it may contain internal capturing parenthesized groups that are called
2145 from elsewhere via the
2146 .\" HTML <a href="#subpatternsassubroutines">
2147 .\" </a>
2148 subroutine mechanism.
2149 .\"
2150 .sp
2151 (2) If quantifier is {0,n} where n is greater than zero, it is treated as if it
2152 were {0,1}. At run time, the rest of the pattern match is tried with and
2153 without the assertion, the order depending on the greediness of the quantifier.
2154 .sp
2155 (3) If the minimum repetition is greater than zero, the quantifier is ignored.
2156 The assertion is obeyed just once when encountered during matching.
2157 .
2158 .
2159 .SS "Lookahead assertions"
2160 .rs
2161 .sp
2162 Lookahead assertions start with (?= for positive assertions and (?! for
2163 negative assertions. For example,
2164 .sp
2165 \ew+(?=;)
2166 .sp
2167 matches a word followed by a semicolon, but does not include the semicolon in
2168 the match, and
2169 .sp
2170 foo(?!bar)
2171 .sp
2172 matches any occurrence of "foo" that is not followed by "bar". Note that the
2173 apparently similar pattern
2174 .sp
2175 (?!foo)bar
2176 .sp
2177 does not find an occurrence of "bar" that is preceded by something other than
2178 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
2179 (?!foo) is always true when the next three characters are "bar". A
2180 lookbehind assertion is needed to achieve the other effect.
2181 .P
2182 If you want to force a matching failure at some point in a pattern, the most
2183 convenient way to do it is with (?!) because an empty string always matches, so
2184 an assertion that requires there not to be an empty string must always fail.
2185 The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
2186 .
2187 .
2188 .\" HTML <a name="lookbehind"></a>
2189 .SS "Lookbehind assertions"
2190 .rs
2191 .sp
2192 Lookbehind assertions start with (?<= for positive assertions and (?<! for
2193 negative assertions. For example,
2194 .sp
2195 (?<!foo)bar
2196 .sp
2197 does find an occurrence of "bar" that is not preceded by "foo". The contents of
2198 a lookbehind assertion are restricted such that all the strings it matches must
2199 have a fixed length. However, if there are several top-level alternatives, they
2200 do not all have to have the same fixed length. Thus
2201 .sp
2202 (?<=bullock|donkey)
2203 .sp
2204 is permitted, but
2205 .sp
2206 (?<!dogs?|cats?)
2207 .sp
2208 causes an error at compile time. Branches that match different length strings
2209 are permitted only at the top level of a lookbehind assertion. This is an
2210 extension compared with Perl, which requires all branches to match the same
2211 length of string. An assertion such as
2212 .sp
2213 (?<=ab(c|de))
2214 .sp
2215 is not permitted, because its single top-level branch can match two different
2216 lengths, but it is acceptable to PCRE if rewritten to use two top-level
2217 branches:
2218 .sp
2219 (?<=abc|abde)
2220 .sp
2221 In some cases, the escape sequence \eK
2222 .\" HTML <a href="#resetmatchstart">
2223 .\" </a>
2224 (see above)
2225 .\"
2226 can be used instead of a lookbehind assertion to get round the fixed-length
2227 restriction.
2228 .P
2229 The implementation of lookbehind assertions is, for each alternative, to
2230 temporarily move the current position back by the fixed length and then try to
2231 match. If there are insufficient characters before the current position, the
2232 assertion fails.
2233 .P
2234 In a UTF mode, PCRE does not allow the \eC escape (which matches a single data
2235 unit even in a UTF mode) to appear in lookbehind assertions, because it makes
2236 it impossible to calculate the length of the lookbehind. The \eX and \eR
2237 escapes, which can match different numbers of data units, are also not
2238 permitted.
2239 .P
2240 .\" HTML <a href="#subpatternsassubroutines">
2241 .\" </a>
2242 "Subroutine"
2243 .\"
2244 calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
2245 as the subpattern matches a fixed-length string.
2246 .\" HTML <a href="#recursion">
2247 .\" </a>
2248 Recursion,
2249 .\"
2250 however, is not supported.
2251 .P
2252 Possessive quantifiers can be used in conjunction with lookbehind assertions to
2253 specify efficient matching of fixed-length strings at the end of subject
2254 strings. Consider a simple pattern such as
2255 .sp
2256 abcd$
2257 .sp
2258 when applied to a long string that does not match. Because matching proceeds
2259 from left to right, PCRE will look for each "a" in the subject and then see if
2260 what follows matches the rest of the pattern. If the pattern is specified as
2261 .sp
2262 ^.*abcd$
2263 .sp
2264 the initial .* matches the entire string at first, but when this fails (because
2265 there is no following "a"), it backtracks to match all but the last character,
2266 then all but the last two characters, and so on. Once again the search for "a"
2267 covers the entire string, from right to left, so we are no better off. However,
2268 if the pattern is written as
2269 .sp
2270 ^.*+(?<=abcd)
2271 .sp
2272 there can be no backtracking for the .*+ item; it can match only the entire
2273 string. The subsequent lookbehind assertion does a single test on the last four
2274 characters. If it fails, the match fails immediately. For long strings, this
2275 approach makes a significant difference to the processing time.
2276 .
2277 .
2278 .SS "Using multiple assertions"
2279 .rs
2280 .sp
2281 Several assertions (of any sort) may occur in succession. For example,
2282 .sp
2283 (?<=\ed{3})(?<!999)foo
2284 .sp
2285 matches "foo" preceded by three digits that are not "999". Notice that each of
2286 the assertions is applied independently at the same point in the subject
2287 string. First there is a check that the previous three characters are all
2288 digits, and then there is a check that the same three characters are not "999".
2289 This pattern does \fInot\fP match "foo" preceded by six characters, the first
2290 of which are digits and the last three of which are not "999". For example, it
2291 doesn't match "123abcfoo". A pattern to do that is
2292 .sp
2293 (?<=\ed{3}...)(?<!999)foo
2294 .sp
2295 This time the first assertion looks at the preceding six characters, checking
2296 that the first three are digits, and then the second assertion checks that the
2297 preceding three characters are not "999".
2298 .P
2299 Assertions can be nested in any combination. For example,
2300 .sp
2301 (?<=(?<!foo)bar)baz
2302 .sp
2303 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
2304 preceded by "foo", while
2305 .sp
2306 (?<=\ed{3}(?!999)...)foo
2307 .sp
2308 is another pattern that matches "foo" preceded by three digits and any three
2309 characters that are not "999".
2310 .
2311 .
2312 .\" HTML <a name="conditions"></a>
2314 .rs
2315 .sp
2316 It is possible to cause the matching process to obey a subpattern
2317 conditionally or to choose between two alternative subpatterns, depending on
2318 the result of an assertion, or whether a specific capturing subpattern has
2319 already been matched. The two possible forms of conditional subpattern are:
2320 .sp
2321 (?(condition)yes-pattern)
2322 (?(condition)yes-pattern|no-pattern)
2323 .sp
2324 If the condition is satisfied, the yes-pattern is used; otherwise the
2325 no-pattern (if present) is used. If there are more than two alternatives in the
2326 subpattern, a compile-time error occurs. Each of the two alternatives may
2327 itself contain nested subpatterns of any form, including conditional
2328 subpatterns; the restriction to two alternatives applies only at the level of
2329 the condition. This pattern fragment is an example where the alternatives are
2330 complex:
2331 .sp
2332 (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
2333 .sp
2334 .P
2335 There are four kinds of condition: references to subpatterns, references to
2336 recursion, a pseudo-condition called DEFINE, and assertions.
2337 .
2338 .SS "Checking for a used subpattern by number"
2339 .rs
2340 .sp
2341 If the text between the parentheses consists of a sequence of digits, the
2342 condition is true if a capturing subpattern of that number has previously
2343 matched. If there is more than one capturing subpattern with the same number
2344 (see the earlier
2345 .\"
2346 .\" HTML <a href="#recursion">
2347 .\" </a>
2348 section about duplicate subpattern numbers),
2349 .\"
2350 the condition is true if any of them have matched. An alternative notation is
2351 to precede the digits with a plus or minus sign. In this case, the subpattern
2352 number is relative rather than absolute. The most recently opened parentheses
2353 can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
2354 loops it can also make sense to refer to subsequent groups. The next
2355 parentheses to be opened can be referenced as (?(+1), and so on. (The value
2356 zero in any of these forms is not used; it provokes a compile-time error.)
2357 .P
2358 Consider the following pattern, which contains non-significant white space to
2359 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2360 three parts for ease of discussion:
2361 .sp
2362 ( \e( )? [^()]+ (?(1) \e) )
2363 .sp
2364 The first part matches an optional opening parenthesis, and if that
2365 character is present, sets it as the first captured substring. The second part
2366 matches one or more characters that are not parentheses. The third part is a
2367 conditional subpattern that tests whether or not the first set of parentheses
2368 matched. If they did, that is, if subject started with an opening parenthesis,
2369 the condition is true, and so the yes-pattern is executed and a closing
2370 parenthesis is required. Otherwise, since no-pattern is not present, the
2371 subpattern matches nothing. In other words, this pattern matches a sequence of
2372 non-parentheses, optionally enclosed in parentheses.
2373 .P
2374 If you were embedding this pattern in a larger one, you could use a relative
2375 reference:
2376 .sp
2377 ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
2378 .sp
2379 This makes the fragment independent of the parentheses in the larger pattern.
2380 .
2381 .SS "Checking for a used subpattern by name"
2382 .rs
2383 .sp
2384 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
2385 subpattern by name. For compatibility with earlier versions of PCRE, which had
2386 this facility before Perl, the syntax (?(name)...) is also recognized.
2387 .P
2388 Rewriting the above example to use a named subpattern gives this:
2389 .sp
2390 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
2391 .sp
2392 If the name used in a condition of this kind is a duplicate, the test is
2393 applied to all subpatterns of the same name, and is true if any one of them has
2394 matched.
2395 .
2396 .SS "Checking for pattern recursion"
2397 .rs
2398 .sp
2399 If the condition is the string (R), and there is no subpattern with the name R,
2400 the condition is true if a recursive call to the whole pattern or any
2401 subpattern has been made. If digits or a name preceded by ampersand follow the
2402 letter R, for example:
2403 .sp
2404 (?(R3)...) or (?(R&name)...)
2405 .sp
2406 the condition is true if the most recent recursion is into a subpattern whose
2407 number or name is given. This condition does not check the entire recursion
2408 stack. If the name used in a condition of this kind is a duplicate, the test is
2409 applied to all subpatterns of the same name, and is true if any one of them is
2410 the most recent recursion.
2411 .P
2412 At "top level", all these recursion test conditions are false.
2413 .\" HTML <a href="#recursion">
2414 .\" </a>
2415 The syntax for recursive patterns
2416 .\"
2417 is described below.
2418 .
2419 .\" HTML <a name="subdefine"></a>
2420 .SS "Defining subpatterns for use by reference only"
2421 .rs
2422 .sp
2423 If the condition is the string (DEFINE), and there is no subpattern with the
2424 name DEFINE, the condition is always false. In this case, there may be only one
2425 alternative in the subpattern. It is always skipped if control reaches this
2426 point in the pattern; the idea of DEFINE is that it can be used to define
2427 subroutines that can be referenced from elsewhere. (The use of
2428 .\" HTML <a href="#subpatternsassubroutines">
2429 .\" </a>
2430 subroutines
2431 .\"
2432 is described below.) For example, a pattern to match an IPv4 address such as
2433 "" could be written like this (ignore white space and line
2434 breaks):
2435 .sp
2436 (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
2437 \eb (?&byte) (\e.(?&byte)){3} \eb
2438 .sp
2439 The first part of the pattern is a DEFINE group inside which a another group
2440 named "byte" is defined. This matches an individual component of an IPv4
2441 address (a number less than 256). When matching takes place, this part of the
2442 pattern is skipped because DEFINE acts like a false condition. The rest of the
2443 pattern uses references to the named group to match the four dot-separated
2444 components of an IPv4 address, insisting on a word boundary at each end.
2445 .
2446 .SS "Assertion conditions"
2447 .rs
2448 .sp
2449 If the condition is not in any of the above formats, it must be an assertion.
2450 This may be a positive or negative lookahead or lookbehind assertion. Consider
2451 this pattern, again containing non-significant white space, and with the two
2452 alternatives on the second line:
2453 .sp
2454 (?(?=[^a-z]*[a-z])
2455 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
2456 .sp
2457 The condition is a positive lookahead assertion that matches an optional
2458 sequence of non-letters followed by a letter. In other words, it tests for the
2459 presence of at least one letter in the subject. If a letter is found, the
2460 subject is matched against the first alternative; otherwise it is matched
2461 against the second. This pattern matches strings in one of the two forms
2462 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2463 .
2464 .
2465 .\" HTML <a name="comments"></a>
2467 .rs
2468 .sp
2469 There are two ways of including comments in patterns that are processed by
2470 PCRE. In both cases, the start of the comment must not be in a character class,
2471 nor in the middle of any other sequence of related characters such as (?: or a
2472 subpattern name or number. The characters that make up a comment play no part
2473 in the pattern matching.
2474 .P
2475 The sequence (?# marks the start of a comment that continues up to the next
2476 closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2477 option is set, an unescaped # character also introduces a comment, which in
2478 this case continues to immediately after the next newline character or
2479 character sequence in the pattern. Which characters are interpreted as newlines
2480 is controlled by the options passed to a compiling function or by a special
2481 sequence at the start of the pattern, as described in the section entitled
2482 .\" HTML <a href="#newlines">
2483 .\" </a>
2484 "Newline conventions"
2485 .\"
2486 above. Note that the end of this type of comment is a literal newline sequence
2487 in the pattern; escape sequences that happen to represent a newline do not
2488 count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2489 default newline convention is in force:
2490 .sp
2491 abc #comment \en still comment
2492 .sp
2493 On encountering the # character, \fBpcre_compile()\fP skips along, looking for
2494 a newline in the pattern. The sequence \en is still literal at this stage, so
2495 it does not terminate the comment. Only an actual character with the code value
2496 0x0a (the default newline) does so.
2497 .
2498 .
2499 .\" HTML <a name="recursion"></a>
2501 .rs
2502 .sp
2503 Consider the problem of matching a string in parentheses, allowing for
2504 unlimited nested parentheses. Without the use of recursion, the best that can
2505 be done is to use a pattern that matches up to some fixed depth of nesting. It
2506 is not possible to handle an arbitrary nesting depth.
2507 .P
2508 For some time, Perl has provided a facility that allows regular expressions to
2509 recurse (amongst other things). It does this by interpolating Perl code in the
2510 expression at run time, and the code can refer to the expression itself. A Perl
2511 pattern using code interpolation to solve the parentheses problem can be
2512 created like this:
2513 .sp
2514 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
2515 .sp
2516 The (?p{...}) item interpolates Perl code at run time, and in this case refers
2517 recursively to the pattern in which it appears.
2518 .P
2519 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2520 supports special syntax for recursion of the entire pattern, and also for
2521 individual subpattern recursion. After its introduction in PCRE and Python,
2522 this kind of recursion was subsequently introduced into Perl at release 5.10.
2523 .P
2524 A special item that consists of (? followed by a number greater than zero and a
2525 closing parenthesis is a recursive subroutine call of the subpattern of the
2526 given number, provided that it occurs inside that subpattern. (If not, it is a
2527 .\" HTML <a href="#subpatternsassubroutines">
2528 .\" </a>
2529 non-recursive subroutine
2530 .\"
2531 call, which is described in the next section.) The special item (?R) or (?0) is
2532 a recursive call of the entire regular expression.
2533 .P
2534 This PCRE pattern solves the nested parentheses problem (assume the
2535 PCRE_EXTENDED option is set so that white space is ignored):
2536 .sp
2537 \e( ( [^()]++ | (?R) )* \e)
2538 .sp
2539 First it matches an opening parenthesis. Then it matches any number of
2540 substrings which can either be a sequence of non-parentheses, or a recursive
2541 match of the pattern itself (that is, a correctly parenthesized substring).
2542 Finally there is a closing parenthesis. Note the use of a possessive quantifier
2543 to avoid backtracking into sequences of non-parentheses.
2544 .P
2545 If this were part of a larger pattern, you would not want to recurse the entire
2546 pattern, so instead you could use this:
2547 .sp
2548 ( \e( ( [^()]++ | (?1) )* \e) )
2549 .sp
2550 We have put the pattern into parentheses, and caused the recursion to refer to
2551 them instead of the whole pattern.
2552 .P
2553 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2554 is made easier by the use of relative references. Instead of (?1) in the
2555 pattern above you can write (?-2) to refer to the second most recently opened
2556 parentheses preceding the recursion. In other words, a negative number counts
2557 capturing parentheses leftwards from the point at which it is encountered.
2558 .P
2559 It is also possible to refer to subsequently opened parentheses, by writing
2560 references such as (?+2). However, these cannot be recursive because the
2561 reference is not inside the parentheses that are referenced. They are always
2562 .\" HTML <a href="#subpatternsassubroutines">
2563 .\" </a>
2564 non-recursive subroutine
2565 .\"
2566 calls, as described in the next section.
2567 .P
2568 An alternative approach is to use named parentheses instead. The Perl syntax
2569 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
2570 could rewrite the above example as follows:
2571 .sp
2572 (?<pn> \e( ( [^()]++ | (?&pn) )* \e) )
2573 .sp
2574 If there is more than one subpattern with the same name, the earliest one is
2575 used.
2576 .P
2577 This particular example pattern that we have been looking at contains nested
2578 unlimited repeats, and so the use of a possessive quantifier for matching
2579 strings of non-parentheses is important when applying the pattern to strings
2580 that do not match. For example, when this pattern is applied to
2581 .sp
2582 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2583 .sp
2584 it yields "no match" quickly. However, if a possessive quantifier is not used,
2585 the match runs for a very long time indeed because there are so many different
2586 ways the + and * repeats can carve up the subject, and all have to be tested
2587 before failure can be reported.
2588 .P
2589 At the end of a match, the values of capturing parentheses are those from
2590 the outermost level. If you want to obtain intermediate values, a callout
2591 function can be used (see below and the
2592 .\" HREF
2593 \fBpcrecallout\fP
2594 .\"
2595 documentation). If the pattern above is matched against
2596 .sp
2597 (ab(cd)ef)
2598 .sp
2599 the value for the inner capturing parentheses (numbered 2) is "ef", which is
2600 the last value taken on at the top level. If a capturing subpattern is not
2601 matched at the top level, its final captured value is unset, even if it was
2602 (temporarily) set at a deeper level during the matching process.
2603 .P
2604 If there are more than 15 capturing parentheses in a pattern, PCRE has to
2605 obtain extra memory to store data during a recursion, which it does by using
2606 \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no memory can
2607 be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2608 .P
2609 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2610 Consider this pattern, which matches text in angle brackets, allowing for
2611 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2612 recursing), whereas any characters are permitted at the outer level.
2613 .sp
2614 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
2615 .sp
2616 In this pattern, (?(R) is the start of a conditional subpattern, with two
2617 different alternatives for the recursive and non-recursive cases. The (?R) item
2618 is the actual recursive call.
2619 .
2620 .
2621 .\" HTML <a name="recursiondifference"></a>
2622 .SS "Differences in recursion processing between PCRE and Perl"
2623 .rs
2624 .sp
2625 Recursion processing in PCRE differs from Perl in two important ways. In PCRE
2626 (like Python, but unlike Perl), a recursive subpattern call is always treated
2627 as an atomic group. That is, once it has matched some of the subject string, it
2628 is never re-entered, even if it contains untried alternatives and there is a
2629 subsequent matching failure. This can be illustrated by the following pattern,
2630 which purports to match a palindromic string that contains an odd number of
2631 characters (for example, "a", "aba", "abcba", "abcdcba"):
2632 .sp
2633 ^(.|(.)(?1)\e2)$
2634 .sp
2635 The idea is that it either matches a single character, or two identical
2636 characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2637 it does not if the pattern is longer than three characters. Consider the
2638 subject string "abcba":
2639 .P
2640 At the top level, the first character is matched, but as it is not at the end
2641 of the string, the first alternative fails; the second alternative is taken
2642 and the recursion kicks in. The recursive call to subpattern 1 successfully
2643 matches the next character ("b"). (Note that the beginning and end of line
2644 tests are not part of the recursion).
2645 .P
2646 Back at the top level, the next character ("c") is compared with what
2647 subpattern 2 matched, which was "a". This fails. Because the recursion is
2648 treated as an atomic group, there are now no backtracking points, and so the
2649 entire match fails. (Perl is able, at this point, to re-enter the recursion and
2650 try the second alternative.) However, if the pattern is written with the
2651 alternatives in the other order, things are different:
2652 .sp
2653 ^((.)(?1)\e2|.)$
2654 .sp
2655 This time, the recursing alternative is tried first, and continues to recurse
2656 until it runs out of characters, at which point the recursion fails. But this
2657 time we do have another alternative to try at the higher level. That is the big
2658 difference: in the previous case the remaining alternative is at a deeper
2659 recursion level, which PCRE cannot use.
2660 .P
2661 To change the pattern so that it matches all palindromic strings, not just
2662 those with an odd number of characters, it is tempting to change the pattern to
2663 this:
2664 .sp
2665 ^((.)(?1)\e2|.?)$
2666 .sp
2667 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2668 deeper recursion has matched a single character, it cannot be entered again in
2669 order to match an empty string. The solution is to separate the two cases, and
2670 write out the odd and even cases as alternatives at the higher level:
2671 .sp
2672 ^(?:((.)(?1)\e2|)|((.)(?3)\e4|.))
2673 .sp
2674 If you want to match typical palindromic phrases, the pattern has to ignore all
2675 non-word characters, which can be done like this:
2676 .sp
2677 ^\eW*+(?:((.)\eW*+(?1)\eW*+\e2|)|((.)\eW*+(?3)\eW*+\e4|\eW*+.\eW*+))\eW*+$
2678 .sp
2679 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2680 man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2681 the use of the possessive quantifier *+ to avoid backtracking into sequences of
2682 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2683 more) to match typical phrases, and Perl takes so long that you think it has
2684 gone into a loop.
2685 .P
2686 \fBWARNING\fP: The palindrome-matching patterns above work only if the subject
2687 string does not start with a palindrome that is shorter than the entire string.
2688 For example, although "abcba" is correctly matched, if the subject is "ababa",
2689 PCRE finds the palindrome "aba" at the start, then fails at top level because
2690 the end of the string does not follow. Once again, it cannot jump back into the
2691 recursion to try other alternatives, so the entire match fails.
2692 .P
2693 The second way in which PCRE and Perl differ in their recursion processing is
2694 in the handling of captured values. In Perl, when a subpattern is called
2695 recursively or as a subpattern (see the next section), it has no access to any
2696 values that were captured outside the recursion, whereas in PCRE these values
2697 can be referenced. Consider this pattern:
2698 .sp
2699 ^(.)(\e1|a(?2))
2700 .sp
2701 In PCRE, this pattern matches "bab". The first capturing parentheses match "b",
2702 then in the second group, when the back reference \e1 fails to match "b", the
2703 second alternative matches "a" and then recurses. In the recursion, \e1 does
2704 now match "b" and so the whole match succeeds. In Perl, the pattern fails to
2705 match because inside the recursive call \e1 cannot access the externally set
2706 value.
2707 .
2708 .
2709 .\" HTML <a name="subpatternsassubroutines"></a>
2711 .rs
2712 .sp
2713 If the syntax for a recursive subpattern call (either by number or by
2714 name) is used outside the parentheses to which it refers, it operates like a
2715 subroutine in a programming language. The called subpattern may be defined
2716 before or after the reference. A numbered reference can be absolute or
2717 relative, as in these examples:
2718 .sp
2719 (...(absolute)...)...(?2)...
2720 (...(relative)...)...(?-1)...
2721 (...(?+1)...(relative)...
2722 .sp
2723 An earlier example pointed out that the pattern
2724 .sp
2725 (sens|respons)e and \e1ibility
2726 .sp
2727 matches "sense and sensibility" and "response and responsibility", but not
2728 "sense and responsibility". If instead the pattern
2729 .sp
2730 (sens|respons)e and (?1)ibility
2731 .sp
2732 is used, it does match "sense and responsibility" as well as the other two
2733 strings. Another example is given in the discussion of DEFINE above.
2734 .P
2735 All subroutine calls, whether recursive or not, are always treated as atomic
2736 groups. That is, once a subroutine has matched some of the subject string, it
2737 is never re-entered, even if it contains untried alternatives and there is a
2738 subsequent matching failure. Any capturing parentheses that are set during the
2739 subroutine call revert to their previous values afterwards.
2740 .P
2741 Processing options such as case-independence are fixed when a subpattern is
2742 defined, so if it is used as a subroutine, such options cannot be changed for
2743 different calls. For example, consider this pattern:
2744 .sp
2745 (abc)(?i:(?-1))
2746 .sp
2747 It matches "abcabc". It does not match "abcABC" because the change of
2748 processing option does not affect the called subpattern.
2749 .
2750 .
2751 .\" HTML <a name="onigurumasubroutines"></a>
2753 .rs
2754 .sp
2755 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
2756 a number enclosed either in angle brackets or single quotes, is an alternative
2757 syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2758 are two of the examples used above, rewritten using this syntax:
2759 .sp
2760 (?<pn> \e( ( (?>[^()]+) | \eg<pn> )* \e) )
2761 (sens|respons)e and \eg'1'ibility
2762 .sp
2763 PCRE supports an extension to Oniguruma: if a number is preceded by a
2764 plus or a minus sign it is taken as a relative reference. For example:
2765 .sp
2766 (abc)(?i:\eg<-1>)
2767 .sp
2768 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
2769 synonymous. The former is a back reference; the latter is a subroutine call.
2770 .
2771 .
2773 .rs
2774 .sp
2775 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2776 code to be obeyed in the middle of matching a regular expression. This makes it
2777 possible, amongst other things, to extract different substrings that match the
2778 same pair of parentheses when there is a repetition.
2779 .P
2780 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2781 code. The feature is called "callout". The caller of PCRE provides an external
2782 function by putting its entry point in the global variable \fIpcre_callout\fP
2783 (8-bit library) or \fIpcre[16|32]_callout\fP (16-bit or 32-bit library).
2784 By default, this variable contains NULL, which disables all calling out.
2785 .P
2786 Within a regular expression, (?C) indicates the points at which the external
2787 function is to be called. If you want to identify different callout points, you
2788 can put a number less than 256 after the letter C. The default value is zero.
2789 For example, this pattern has two callout points:
2790 .sp
2791 (?C1)abc(?C2)def
2792 .sp
2793 If the PCRE_AUTO_CALLOUT flag is passed to a compiling function, callouts are
2794 automatically installed before each item in the pattern. They are all numbered
2795 255. If there is a conditional group in the pattern whose condition is an
2796 assertion, an additional callout is inserted just before the condition. An
2797 explicit callout may also be set at this position, as in this example:
2798 .sp
2799 (?(?C9)(?=a)abc|def)
2800 .sp
2801 Note that this applies only to assertion conditions, not to other types of
2802 condition.
2803 .P
2804 During matching, when PCRE reaches a callout point, the external function is
2805 called. It is provided with the number of the callout, the position in the
2806 pattern, and, optionally, one item of data originally supplied by the caller of
2807 the matching function. The callout function may cause matching to proceed, to
2808 backtrack, or to fail altogether.
2809 .P
2810 By default, PCRE implements a number of optimizations at compile time and
2811 matching time, and one side-effect is that sometimes callouts are skipped. If
2812 you need all possible callouts to happen, you need to set options that disable
2813 the relevant optimizations. More details, and a complete description of the
2814 interface to the callout function, are given in the
2815 .\" HREF
2816 \fBpcrecallout\fP
2817 .\"
2818 documentation.
2819 .
2820 .
2821 .\" HTML <a name="backtrackcontrol"></a>
2823 .rs
2824 .sp
2825 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2826 are still described in the Perl documentation as "experimental and subject to
2827 change or removal in a future version of Perl". It goes on to say: "Their usage
2828 in production code should be noted to avoid problems during upgrades." The same
2829 remarks apply to the PCRE features described in this section.
2830 .P
2831 The new verbs make use of what was previously invalid syntax: an opening
2832 parenthesis followed by an asterisk. They are generally of the form
2833 (*VERB) or (*VERB:NAME). Some may take either form, possibly behaving
2834 differently depending on whether or not a name is present. A name is any
2835 sequence of characters that does not include a closing parenthesis. The maximum
2836 length of name is 255 in the 8-bit library and 65535 in the 16-bit and 32-bit
2837 libraries. If the name is empty, that is, if the closing parenthesis
2838 immediately follows the colon, the effect is as if the colon were not there.
2839 Any number of these verbs may occur in a pattern.
2840 .P
2841 Since these verbs are specifically related to backtracking, most of them can be
2842 used only when the pattern is to be matched using one of the traditional
2843 matching functions, because these use a backtracking algorithm. With the
2844 exception of (*FAIL), which behaves like a failing negative assertion, the
2845 backtracking control verbs cause an error if encountered by a DFA matching
2846 function.
2847 .P
2848 The behaviour of these verbs in
2849 .\" HTML <a href="#btrepeat">
2850 .\" </a>
2851 repeated groups,
2852 .\"
2853 .\" HTML <a href="#btassert">
2854 .\" </a>
2855 assertions,
2856 .\"
2857 and in
2858 .\" HTML <a href="#btsub">
2859 .\" </a>
2860 subpatterns called as subroutines
2861 .\"
2862 (whether or not recursively) is documented below.
2863 .
2864 .
2865 .\" HTML <a name="nooptimize"></a>
2866 .SS "Optimizations that affect backtracking verbs"
2867 .rs
2868 .sp
2869 PCRE contains some optimizations that are used to speed up matching by running
2870 some checks at the start of each match attempt. For example, it may know the
2871 minimum length of matching subject, or that a particular character must be
2872 present. When one of these optimizations bypasses the running of a match, any
2873 included backtracking verbs will not, of course, be processed. You can suppress
2874 the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2875 when calling \fBpcre_compile()\fP or \fBpcre_exec()\fP, or by starting the
2876 pattern with (*NO_START_OPT). There is more discussion of this option in the
2877 section entitled
2878 .\" HTML <a href="pcreapi.html#execoptions">
2879 .\" </a>
2880 "Option bits for \fBpcre_exec()\fP"
2881 .\"
2882 in the
2883 .\" HREF
2884 \fBpcreapi\fP
2885 .\"
2886 documentation.
2887 .P
2888 Experiments with Perl suggest that it too has similar optimizations, sometimes
2889 leading to anomalous results.
2890 .
2891 .
2892 .SS "Verbs that act immediately"
2893 .rs
2894 .sp
2895 The following verbs act as soon as they are encountered. They may not be
2896 followed by a name.
2897 .sp
2898 (*ACCEPT)
2899 .sp
2900 This verb causes the match to end successfully, skipping the remainder of the
2901 pattern. However, when it is inside a subpattern that is called as a
2902 subroutine, only that subpattern is ended successfully. Matching then continues
2903 at the outer level. If (*ACCEPT) in triggered in a positive assertion, the
2904 assertion succeeds; in a negative assertion, the assertion fails.
2905 .P
2906 If (*ACCEPT) is inside capturing parentheses, the data so far is captured. For
2907 example:
2908 .sp
2909 A((?:A|B(*ACCEPT)|C)D)
2910 .sp
2911 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2912 the outer parentheses.
2913 .sp
2914 (*FAIL) or (*F)
2915 .sp
2916 This verb causes a matching failure, forcing backtracking to occur. It is
2917 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2918 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2919 Perl features that are not present in PCRE. The nearest equivalent is the
2920 callout feature, as for example in this pattern:
2921 .sp
2922 a+(?C)(*FAIL)
2923 .sp
2924 A match with the string "aaaa" always fails, but the callout is taken before
2925 each backtrack happens (in this example, 10 times).
2926 .
2927 .
2928 .SS "Recording which path was taken"
2929 .rs
2930 .sp
2931 There is one verb whose main purpose is to track how a match was arrived at,
2932 though it also has a secondary use in conjunction with advancing the match
2933 starting point (see (*SKIP) below).
2934 .sp
2935 (*MARK:NAME) or (*:NAME)
2936 .sp
2937 A name is always required with this verb. There may be as many instances of
2938 (*MARK) as you like in a pattern, and their names do not have to be unique.
2939 .P
2940 When a match succeeds, the name of the last-encountered (*MARK:NAME),
2941 (*PRUNE:NAME), or (*THEN:NAME) on the matching path is passed back to the
2942 caller as described in the section entitled
2943 .\" HTML <a href="pcreapi.html#extradata">
2944 .\" </a>
2945 "Extra data for \fBpcre_exec()\fP"
2946 .\"
2947 in the
2948 .\" HREF
2949 \fBpcreapi\fP
2950 .\"
2951 documentation. Here is an example of \fBpcretest\fP output, where the /K
2952 modifier requests the retrieval and outputting of (*MARK) data:
2953 .sp
2954 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2955 data> XY
2956 0: XY
2957 MK: A
2958 XZ
2959 0: XZ
2960 MK: B
2961 .sp
2962 The (*MARK) name is tagged with "MK:" in this output, and in this example it
2963 indicates which of the two alternatives matched. This is a more efficient way
2964 of obtaining this information than putting each alternative in its own
2965 capturing parentheses.
2966 .P
2967 If a verb with a name is encountered in a positive assertion that is true, the
2968 name is recorded and passed back if it is the last-encountered. This does not
2969 happen for negative assertions or failing positive assertions.
2970 .P
2971 After a partial match or a failed match, the last encountered name in the
2972 entire match process is returned. For example:
2973 .sp
2974 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2975 data> XP
2976 No match, mark = B
2977 .sp
2978 Note that in this unanchored example the mark is retained from the match
2979 attempt that started at the letter "X" in the subject. Subsequent match
2980 attempts starting at "P" and then with an empty string do not get as far as the
2981 (*MARK) item, but nevertheless do not reset it.
2982 .P
2983 If you are interested in (*MARK) values after failed matches, you should
2984 probably set the PCRE_NO_START_OPTIMIZE option
2985 .\" HTML <a href="#nooptimize">
2986 .\" </a>
2987 (see above)
2988 .\"
2989 to ensure that the match is always attempted.
2990 .
2991 .
2992 .SS "Verbs that act after backtracking"
2993 .rs
2994 .sp
2995 The following verbs do nothing when they are encountered. Matching continues
2996 with what follows, but if there is no subsequent match, causing a backtrack to
2997 the verb, a failure is forced. That is, backtracking cannot pass to the left of
2998 the verb. However, when one of these verbs appears inside an atomic group or an
2999 assertion that is true, its effect is confined to that group, because once the
3000 group has been matched, there is never any backtracking into it. In this
3001 situation, backtracking can "jump back" to the left of the entire atomic group
3002 or assertion. (Remember also, as stated above, that this localization also
3003 applies in subroutine calls.)
3004 .P
3005 These verbs differ in exactly what kind of failure occurs when backtracking
3006 reaches them. The behaviour described below is what happens when the verb is
3007 not in a subroutine or an assertion. Subsequent sections cover these special
3008 cases.
3009 .sp
3010 (*COMMIT)
3011 .sp
3012 This verb, which may not be followed by a name, causes the whole match to fail
3013 outright if there is a later matching failure that causes backtracking to reach
3014 it. Even if the pattern is unanchored, no further attempts to find a match by
3015 advancing the starting point take place. If (*COMMIT) is the only backtracking
3016 verb that is encountered, once it has been passed \fBpcre_exec()\fP is
3017 committed to finding a match at the current starting point, or not at all. For
3018 example:
3019 .sp
3020 a+(*COMMIT)b
3021 .sp
3022 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
3023 dynamic anchor, or "I've started, so I must finish." The name of the most
3024 recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
3025 match failure.
3026 .P
3027 If there is more than one backtracking verb in a pattern, a different one that
3028 follows (*COMMIT) may be triggered first, so merely passing (*COMMIT) during a
3029 match does not always guarantee that a match must be at this starting point.
3030 .P
3031 Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
3032 unless PCRE's start-of-match optimizations are turned off, as shown in this
3033 output from \fBpcretest\fP:
3034 .sp
3035 re> /(*COMMIT)abc/
3036 data> xyzabc
3037 0: abc
3038 data> xyzabc\eY
3039 No match
3040 .sp
3041 For this pattern, PCRE knows that any match must start with "a", so the
3042 optimization skips along the subject to "a" before applying the pattern to the
3043 first set of data. The match attempt then succeeds. In the second set of data,
3044 the escape sequence \eY is interpreted by the \fBpcretest\fP program. It causes
3045 the PCRE_NO_START_OPTIMIZE option to be set when \fBpcre_exec()\fP is called.
3046 This disables the optimization that skips along to the first character. The
3047 pattern is now applied starting at "x", and so the (*COMMIT) causes the match
3048 to fail without trying any other starting points.
3049 .sp
3050 (*PRUNE) or (*PRUNE:NAME)
3051 .sp
3052 This verb causes the match to fail at the current starting position in the
3053 subject if there is a later matching failure that causes backtracking to reach
3054 it. If the pattern is unanchored, the normal "bumpalong" advance to the next
3055 starting character then happens. Backtracking can occur as usual to the left of
3056 (*PRUNE), before it is reached, or when matching to the right of (*PRUNE), but
3057 if there is no match to the right, backtracking cannot cross (*PRUNE). In
3058 simple cases, the use of (*PRUNE) is just an alternative to an atomic group or
3059 possessive quantifier, but there are some uses of (*PRUNE) that cannot be
3060 expressed in any other way. In an anchored pattern (*PRUNE) has the same effect
3061 as (*COMMIT).
3062 .P
3063 The behaviour of (*PRUNE:NAME) is the not the same as (*MARK:NAME)(*PRUNE).
3064 It is like (*MARK:NAME) in that the name is remembered for passing back to the
3065 caller. However, (*SKIP:NAME) searches only for names set with (*MARK).
3066 .sp
3067 (*SKIP)
3068 .sp
3069 This verb, when given without a name, is like (*PRUNE), except that if the
3070 pattern is unanchored, the "bumpalong" advance is not to the next character,
3071 but to the position in the subject where (*SKIP) was encountered. (*SKIP)
3072 signifies that whatever text was matched leading up to it cannot be part of a
3073 successful match. Consider:
3074 .sp
3075 a+(*SKIP)b
3076 .sp
3077 If the subject is "aaaac...", after the first match attempt fails (starting at
3078 the first character in the string), the starting point skips on to start the
3079 next attempt at "c". Note that a possessive quantifer does not have the same
3080 effect as this example; although it would suppress backtracking during the
3081 first match attempt, the second attempt would start at the second character
3082 instead of skipping on to "c".
3083 .sp
3084 (*SKIP:NAME)
3085 .sp
3086 When (*SKIP) has an associated name, its behaviour is modified. When it is
3087 triggered, the previous path through the pattern is searched for the most
3088 recent (*MARK) that has the same name. If one is found, the "bumpalong" advance
3089 is to the subject position that corresponds to that (*MARK) instead of to where
3090 (*SKIP) was encountered. If no (*MARK) with a matching name is found, the
3091 (*SKIP) is ignored.
3092 .P
3093 Note that (*SKIP:NAME) searches only for names set by (*MARK:NAME). It ignores
3094 names that are set by (*PRUNE:NAME) or (*THEN:NAME).
3095 .sp
3096 (*THEN) or (*THEN:NAME)
3097 .sp
3098 This verb causes a skip to the next innermost alternative when backtracking
3099 reaches it. That is, it cancels any further backtracking within the current
3100 alternative. Its name comes from the observation that it can be used for a
3101 pattern-based if-then-else block:
3102 .sp
3103 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
3104 .sp
3105 If the COND1 pattern matches, FOO is tried (and possibly further items after
3106 the end of the group if FOO succeeds); on failure, the matcher skips to the
3107 second alternative and tries COND2, without backtracking into COND1. If that
3108 succeeds and BAR fails, COND3 is tried. If subsequently BAZ fails, there are no
3109 more alternatives, so there is a backtrack to whatever came before the entire
3110 group. If (*THEN) is not inside an alternation, it acts like (*PRUNE).
3111 .P
3112 The behaviour of (*THEN:NAME) is the not the same as (*MARK:NAME)(*THEN).
3113 It is like (*MARK:NAME) in that the name is remembered for passing back to the
3114 caller. However, (*SKIP:NAME) searches only for names set with (*MARK).
3115 .P
3116 A subpattern that does not contain a | character is just a part of the
3117 enclosing alternative; it is not a nested alternation with only one
3118 alternative. The effect of (*THEN) extends beyond such a subpattern to the
3119 enclosing alternative. Consider this pattern, where A, B, etc. are complex
3120 pattern fragments that do not contain any | characters at this level:
3121 .sp
3122 A (B(*THEN)C) | D
3123 .sp
3124 If A and B are matched, but there is a failure in C, matching does not
3125 backtrack into A; instead it moves to the next alternative, that is, D.
3126 However, if the subpattern containing (*THEN) is given an alternative, it
3127 behaves differently:
3128 .sp
3129 A (B(*THEN)C | (*FAIL)) | D
3130 .sp
3131 The effect of (*THEN) is now confined to the inner subpattern. After a failure
3132 in C, matching moves to (*FAIL), which causes the whole subpattern to fail
3133 because there are no more alternatives to try. In this case, matching does now
3134 backtrack into A.
3135 .P
3136 Note that a conditional subpattern is not considered as having two
3137 alternatives, because only one is ever used. In other words, the | character in
3138 a conditional subpattern has a different meaning. Ignoring white space,
3139 consider:
3140 .sp
3141 ^.*? (?(?=a) a | b(*THEN)c )
3142 .sp
3143 If the subject is "ba", this pattern does not match. Because .*? is ungreedy,
3144 it initially matches zero characters. The condition (?=a) then fails, the
3145 character "b" is matched, but "c" is not. At this point, matching does not
3146 backtrack to .*? as might perhaps be expected from the presence of the |
3147 character. The conditional subpattern is part of the single alternative that
3148 comprises the whole pattern, and so the match fails. (If there was a backtrack
3149 into .*?, allowing it to match "b", the match would succeed.)
3150 .P
3151 The verbs just described provide four different "strengths" of control when
3152 subsequent matching fails. (*THEN) is the weakest, carrying on the match at the
3153 next alternative. (*PRUNE) comes next, failing the match at the current
3154 starting position, but allowing an advance to the next character (for an
3155 unanchored pattern). (*SKIP) is similar, except that the advance may be more
3156 than one character. (*COMMIT) is the strongest, causing the entire match to
3157 fail.
3158 .
3159 .
3160 .SS "More than one backtracking verb"
3161 .rs
3162 .sp
3163 If more than one backtracking verb is present in a pattern, the one that is
3164 backtracked onto first acts. For example, consider this pattern, where A, B,
3165 etc. are complex pattern fragments:
3166 .sp
3168 .sp
3169 If A matches but B fails, the backtrack to (*COMMIT) causes the entire match to
3170 fail. However, if A and B match, but C fails, the backtrack to (*THEN) causes
3171 the next alternative (ABD) to be tried. This behaviour is consistent, but is
3172 not always the same as Perl's. It means that if two or more backtracking verbs
3173 appear in succession, all the the last of them has no effect. Consider this
3174 example:
3175 .sp
3176 ...(*COMMIT)(*PRUNE)...
3177 .sp
3178 If there is a matching failure to the right, backtracking onto (*PRUNE) causes
3179 it to be triggered, and its action is taken. There can never be a backtrack
3180 onto (*COMMIT).
3181 .
3182 .
3183 .\" HTML <a name="btrepeat"></a>
3184 .SS "Backtracking verbs in repeated groups"
3185 .rs
3186 .sp
3187 PCRE differs from Perl in its handling of backtracking verbs in repeated
3188 groups. For example, consider:
3189 .sp
3190 /(a(*COMMIT)b)+ac/
3191 .sp
3192 If the subject is "abac", Perl matches, but PCRE fails because the (*COMMIT) in
3193 the second repeat of the group acts.
3194 .
3195 .
3196 .\" HTML <a name="btassert"></a>
3197 .SS "Backtracking verbs in assertions"
3198 .rs
3199 .sp
3200 (*FAIL) in an assertion has its normal effect: it forces an immediate backtrack.
3201 .P
3202 (*ACCEPT) in a positive assertion causes the assertion to succeed without any
3203 further processing. In a negative assertion, (*ACCEPT) causes the assertion to
3204 fail without any further processing.
3205 .P
3206 The other backtracking verbs are not treated specially if they appear in a
3207 positive assertion. In particular, (*THEN) skips to the next alternative in the
3208 innermost enclosing group that has alternations, whether or not this is within
3209 the assertion.
3210 .P
3211 Negative assertions are, however, different, in order to ensure that changing a
3212 positive assertion into a negative assertion changes its result. Backtracking
3213 into (*COMMIT), (*SKIP), or (*PRUNE) causes a negative assertion to be true,
3214 without considering any further alternative branches in the assertion.
3215 Backtracking into (*THEN) causes it to skip to the next enclosing alternative
3216 within the assertion (the normal behaviour), but if the assertion does not have
3217 such an alternative, (*THEN) behaves like (*PRUNE).
3218 .
3219 .
3220 .\" HTML <a name="btsub"></a>
3221 .SS "Backtracking verbs in subroutines"
3222 .rs
3223 .sp
3224 These behaviours occur whether or not the subpattern is called recursively.
3225 Perl's treatment of subroutines is different in some cases.
3226 .P
3227 (*FAIL) in a subpattern called as a subroutine has its normal effect: it forces
3228 an immediate backtrack.
3229 .P
3230 (*ACCEPT) in a subpattern called as a subroutine causes the subroutine match to
3231 succeed without any further processing. Matching then continues after the
3232 subroutine call.
3233 .P
3234 (*COMMIT), (*SKIP), and (*PRUNE) in a subpattern called as a subroutine cause
3235 the subroutine match to fail.
3236 .P
3237 (*THEN) skips to the next alternative in the innermost enclosing group within
3238 the subpattern that has alternatives. If there is no such group within the
3239 subpattern, (*THEN) causes the subroutine match to fail.
3240 .
3241 .
3242 .SH "SEE ALSO"
3243 .rs
3244 .sp
3245 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3),
3246 \fBpcresyntax\fP(3), \fBpcre\fP(3), \fBpcre16(3)\fP, \fBpcre32(3)\fP.
3247 .
3248 .
3250 .rs
3251 .sp
3252 .nf
3253 Philip Hazel
3254 University Computing Service
3255 Cambridge CB2 3QH, England.
3256 .fi
3257 .
3258 .
3260 .rs
3261 .sp
3262 .nf
3263 Last updated: 08 January 2014
3264 Copyright (c) 1997-2014 University of Cambridge.
3265 .fi


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

  ViewVC Help
Powered by ViewVC 1.1.5