ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1341 - (show annotations)
Fri Jun 28 15:07:57 2013 UTC (6 years, 7 months ago) by ph10
File size: 130606 byte(s)
Fix typo.
1 .TH PCREPATTERN 3 "26 April 2013" "PCRE 8.33"
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
84 .sp
85 (*UCP)
86 .sp
87 This has the same effect as setting the PCRE_UCP option: it causes sequences
88 such as \ed and \ew to use Unicode properties to determine character types,
89 instead of recognizing only characters with codes less than 128 via a lookup
90 table.
91 .
92 .
93 .SS "Disabling start-up optimizations"
94 .rs
95 .sp
96 If a pattern starts with (*NO_START_OPT), it has the same effect as setting the
97 PCRE_NO_START_OPTIMIZE option either at compile or matching time.
98 .
99 .
100 .\" HTML <a name="newlines"></a>
101 .SS "Newline conventions"
102 .rs
103 .sp
104 PCRE supports five different conventions for indicating line breaks in
105 strings: a single CR (carriage return) character, a single LF (linefeed)
106 character, the two-character sequence CRLF, any of the three preceding, or any
107 Unicode newline sequence. The
108 .\" HREF
109 \fBpcreapi\fP
110 .\"
111 page has
112 .\" HTML <a href="pcreapi.html#newlines">
113 .\" </a>
114 further discussion
115 .\"
116 about newlines, and shows how to set the newline convention in the
117 \fIoptions\fP arguments for the compiling and matching functions.
118 .P
119 It is also possible to specify a newline convention by starting a pattern
120 string with one of the following five sequences:
121 .sp
122 (*CR) carriage return
123 (*LF) linefeed
124 (*CRLF) carriage return, followed by linefeed
125 (*ANYCRLF) any of the three above
126 (*ANY) all Unicode newline sequences
127 .sp
128 These override the default and the options given to the compiling function. For
129 example, on a Unix system where LF is the default newline sequence, the pattern
130 .sp
131 (*CR)a.b
132 .sp
133 changes the convention to CR. That pattern matches "a\enb" because LF is no
134 longer a newline. If more than one of these settings is present, the last one
135 is used.
136 .P
137 The newline convention affects where the circumflex and dollar assertions are
138 true. It also affects the interpretation of the dot metacharacter when
139 PCRE_DOTALL is not set, and the behaviour of \eN. However, it does not affect
140 what the \eR escape sequence matches. By default, this is any Unicode newline
141 sequence, for Perl compatibility. However, this can be changed; see the
142 description of \eR in the section entitled
143 .\" HTML <a href="#newlineseq">
144 .\" </a>
145 "Newline sequences"
146 .\"
147 below. A change of \eR setting can be combined with a change of newline
148 convention.
149 .
150 .
151 .SS "Setting match and recursion limits"
152 .rs
153 .sp
154 The caller of \fBpcre_exec()\fP can set a limit on the number of times the
155 internal \fBmatch()\fP function is called and on the maximum depth of
156 recursive calls. These facilities are provided to catch runaway matches that
157 are provoked by patterns with huge matching trees (a typical example is a
158 pattern with nested unlimited repeats) and to avoid running out of system stack
159 by too much recursion. When one of these limits is reached, \fBpcre_exec()\fP
160 gives an error return. The limits can also be set by items at the start of the
161 pattern of the form
162 .sp
163 (*LIMIT_MATCH=d)
165 .sp
166 where d is any number of decimal digits. However, the value of the setting must
167 be less than the value set by the caller of \fBpcre_exec()\fP for it to have
168 any effect. In other words, the pattern writer can lower the limit set by the
169 programmer, but not raise it. If there is more than one setting of one of these
170 limits, the lower value is used.
171 .
172 .
174 .rs
175 .sp
176 PCRE can be compiled to run in an environment that uses EBCDIC as its character
177 code rather than ASCII or Unicode (typically a mainframe system). In the
178 sections below, character code values are ASCII or Unicode; in an EBCDIC
179 environment these characters may have different code values, and there are no
180 code points greater than 255.
181 .
182 .
184 .rs
185 .sp
186 A regular expression is a pattern that is matched against a subject string from
187 left to right. Most characters stand for themselves in a pattern, and match the
188 corresponding characters in the subject. As a trivial example, the pattern
189 .sp
190 The quick brown fox
191 .sp
192 matches a portion of a subject string that is identical to itself. When
193 caseless matching is specified (the PCRE_CASELESS option), letters are matched
194 independently of case. In a UTF mode, PCRE always understands the concept of
195 case for characters whose values are less than 128, so caseless matching is
196 always possible. For characters with higher values, the concept of case is
197 supported if PCRE is compiled with Unicode property support, but not otherwise.
198 If you want to use caseless matching for characters 128 and above, you must
199 ensure that PCRE is compiled with Unicode property support as well as with
200 UTF support.
201 .P
202 The power of regular expressions comes from the ability to include alternatives
203 and repetitions in the pattern. These are encoded in the pattern by the use of
204 \fImetacharacters\fP, which do not stand for themselves but instead are
205 interpreted in some special way.
206 .P
207 There are two different sets of metacharacters: those that are recognized
208 anywhere in the pattern except within square brackets, and those that are
209 recognized within square brackets. Outside square brackets, the metacharacters
210 are as follows:
211 .sp
212 \e general escape character with several uses
213 ^ assert start of string (or line, in multiline mode)
214 $ assert end of string (or line, in multiline mode)
215 . match any character except newline (by default)
216 [ start character class definition
217 | start of alternative branch
218 ( start subpattern
219 ) end subpattern
220 ? extends the meaning of (
221 also 0 or 1 quantifier
222 also quantifier minimizer
223 * 0 or more quantifier
224 + 1 or more quantifier
225 also "possessive quantifier"
226 { start min/max quantifier
227 .sp
228 Part of a pattern that is in square brackets is called a "character class". In
229 a character class the only metacharacters are:
230 .sp
231 \e general escape character
232 ^ negate the class, but only if the first character
233 - indicates character range
234 .\" JOIN
235 [ POSIX character class (only if followed by POSIX
236 syntax)
237 ] terminates the character class
238 .sp
239 The following sections describe the use of each of the metacharacters.
240 .
241 .
243 .rs
244 .sp
245 The backslash character has several uses. Firstly, if it is followed by a
246 character that is not a number or a letter, it takes away any special meaning
247 that character may have. This use of backslash as an escape character applies
248 both inside and outside character classes.
249 .P
250 For example, if you want to match a * character, you write \e* in the pattern.
251 This escaping action applies whether or not the following character would
252 otherwise be interpreted as a metacharacter, so it is always safe to precede a
253 non-alphanumeric with backslash to specify that it stands for itself. In
254 particular, if you want to match a backslash, you write \e\e.
255 .P
256 In a UTF mode, only ASCII numbers and letters have any special meaning after a
257 backslash. All other characters (in particular, those whose codepoints are
258 greater than 127) are treated as literals.
259 .P
260 If a pattern is compiled with the PCRE_EXTENDED option, white space in the
261 pattern (other than in a character class) and characters between a # outside
262 a character class and the next newline are ignored. An escaping backslash can
263 be used to include a white space or # character as part of the pattern.
264 .P
265 If you want to remove the special meaning from a sequence of characters, you
266 can do so by putting them between \eQ and \eE. This is different from Perl in
267 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
268 Perl, $ and @ cause variable interpolation. Note the following examples:
269 .sp
270 Pattern PCRE matches Perl matches
271 .sp
272 .\" JOIN
273 \eQabc$xyz\eE abc$xyz abc followed by the
274 contents of $xyz
275 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
276 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
277 .sp
278 The \eQ...\eE sequence is recognized both inside and outside character classes.
279 An isolated \eE that is not preceded by \eQ is ignored. If \eQ is not followed
280 by \eE later in the pattern, the literal interpretation continues to the end of
281 the pattern (that is, \eE is assumed at the end). If the isolated \eQ is inside
282 a character class, this causes an error, because the character class is not
283 terminated.
284 .
285 .
286 .\" HTML <a name="digitsafterbackslash"></a>
287 .SS "Non-printing characters"
288 .rs
289 .sp
290 A second use of backslash provides a way of encoding non-printing characters
291 in patterns in a visible manner. There is no restriction on the appearance of
292 non-printing characters, apart from the binary zero that terminates a pattern,
293 but when a pattern is being prepared by text editing, it is often easier to use
294 one of the following escape sequences than the binary character it represents:
295 .sp
296 \ea alarm, that is, the BEL character (hex 07)
297 \ecx "control-x", where x is any ASCII character
298 \ee escape (hex 1B)
299 \ef form feed (hex 0C)
300 \en linefeed (hex 0A)
301 \er carriage return (hex 0D)
302 \et tab (hex 09)
303 \eddd character with octal code ddd, or back reference
304 \exhh character with hex code hh
305 \ex{hhh..} character with hex code hhh.. (non-JavaScript mode)
306 \euhhhh character with hex code hhhh (JavaScript mode only)
307 .sp
308 The precise effect of \ecx on ASCII characters is as follows: if x is a lower
309 case letter, it is converted to upper case. Then bit 6 of the character (hex
310 40) is inverted. Thus \ecA to \ecZ become hex 01 to hex 1A (A is 41, Z is 5A),
311 but \ec{ becomes hex 3B ({ is 7B), and \ec; becomes hex 7B (; is 3B). If the
312 data item (byte or 16-bit value) following \ec has a value greater than 127, a
313 compile-time error occurs. This locks out non-ASCII characters in all modes.
314 .P
315 The \ec facility was designed for use with ASCII characters, but with the
316 extension to Unicode it is even less useful than it once was. It is, however,
317 recognized when PCRE is compiled in EBCDIC mode, where data items are always
318 bytes. In this mode, all values are valid after \ec. If the next character is a
319 lower case letter, it is converted to upper case. Then the 0xc0 bits of the
320 byte are inverted. Thus \ecA becomes hex 01, as in ASCII (A is C1), but because
321 the EBCDIC letters are disjoint, \ecZ becomes hex 29 (Z is E9), and other
322 characters also generate different values.
323 .P
324 By default, after \ex, from zero to two hexadecimal digits are read (letters
325 can be in upper or lower case). Any number of hexadecimal digits may appear
326 between \ex{ and }, but the character code is constrained as follows:
327 .sp
328 8-bit non-UTF mode less than 0x100
329 8-bit UTF-8 mode less than 0x10ffff and a valid codepoint
330 16-bit non-UTF mode less than 0x10000
331 16-bit UTF-16 mode less than 0x10ffff and a valid codepoint
332 32-bit non-UTF mode less than 0x80000000
333 32-bit UTF-32 mode less than 0x10ffff and a valid codepoint
334 .sp
335 Invalid Unicode codepoints are the range 0xd800 to 0xdfff (the so-called
336 "surrogate" codepoints), and 0xffef.
337 .P
338 If characters other than hexadecimal digits appear between \ex{ and }, or if
339 there is no terminating }, this form of escape is not recognized. Instead, the
340 initial \ex will be interpreted as a basic hexadecimal escape, with no
341 following digits, giving a character whose value is zero.
342 .P
343 If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \ex is
344 as just described only when it is followed by two hexadecimal digits.
345 Otherwise, it matches a literal "x" character. In JavaScript mode, support for
346 code points greater than 256 is provided by \eu, which must be followed by
347 four hexadecimal digits; otherwise it matches a literal "u" character.
348 Character codes specified by \eu in JavaScript mode are constrained in the same
349 was as those specified by \ex in non-JavaScript mode.
350 .P
351 Characters whose value is less than 256 can be defined by either of the two
352 syntaxes for \ex (or by \eu in JavaScript mode). There is no difference in the
353 way they are handled. For example, \exdc is exactly the same as \ex{dc} (or
354 \eu00dc in JavaScript mode).
355 .P
356 After \e0 up to two further octal digits are read. If there are fewer than two
357 digits, just those that are present are used. Thus the sequence \e0\ex\e07
358 specifies two binary zeros followed by a BEL character (code value 7). Make
359 sure you supply two digits after the initial zero if the pattern character that
360 follows is itself an octal digit.
361 .P
362 The handling of a backslash followed by a digit other than 0 is complicated.
363 Outside a character class, PCRE reads it and any following digits as a decimal
364 number. If the number is less than 10, or if there have been at least that many
365 previous capturing left parentheses in the expression, the entire sequence is
366 taken as a \fIback reference\fP. A description of how this works is given
367 .\" HTML <a href="#backreferences">
368 .\" </a>
369 later,
370 .\"
371 following the discussion of
372 .\" HTML <a href="#subpattern">
373 .\" </a>
374 parenthesized subpatterns.
375 .\"
376 .P
377 Inside a character class, or if the decimal number is greater than 9 and there
378 have not been that many capturing subpatterns, PCRE re-reads up to three octal
379 digits following the backslash, and uses them to generate a data character. Any
380 subsequent digits stand for themselves. The value of the character is
381 constrained in the same way as characters specified in hexadecimal.
382 For example:
383 .sp
384 \e040 is another way of writing an ASCII space
385 .\" JOIN
386 \e40 is the same, provided there are fewer than 40
387 previous capturing subpatterns
388 \e7 is always a back reference
389 .\" JOIN
390 \e11 might be a back reference, or another way of
391 writing a tab
392 \e011 is always a tab
393 \e0113 is a tab followed by the character "3"
394 .\" JOIN
395 \e113 might be a back reference, otherwise the
396 character with octal code 113
397 .\" JOIN
398 \e377 might be a back reference, otherwise
399 the value 255 (decimal)
400 .\" JOIN
401 \e81 is either a back reference, or a binary zero
402 followed by the two characters "8" and "1"
403 .sp
404 Note that octal values of 100 or greater must not be introduced by a leading
405 zero, because no more than three octal digits are ever read.
406 .P
407 All the sequences that define a single character value can be used both inside
408 and outside character classes. In addition, inside a character class, \eb is
409 interpreted as the backspace character (hex 08).
410 .P
411 \eN is not allowed in a character class. \eB, \eR, and \eX are not special
412 inside a character class. Like other unrecognized escape sequences, they are
413 treated as the literal characters "B", "R", and "X" by default, but cause an
414 error if the PCRE_EXTRA option is set. Outside a character class, these
415 sequences have different meanings.
416 .
417 .
418 .SS "Unsupported escape sequences"
419 .rs
420 .sp
421 In Perl, the sequences \el, \eL, \eu, and \eU are recognized by its string
422 handler and used to modify the case of following characters. By default, PCRE
423 does not support these escape sequences. However, if the PCRE_JAVASCRIPT_COMPAT
424 option is set, \eU matches a "U" character, and \eu can be used to define a
425 character by code point, as described in the previous section.
426 .
427 .
428 .SS "Absolute and relative back references"
429 .rs
430 .sp
431 The sequence \eg followed by an unsigned or a negative number, optionally
432 enclosed in braces, is an absolute or relative back reference. A named back
433 reference can be coded as \eg{name}. Back references are discussed
434 .\" HTML <a href="#backreferences">
435 .\" </a>
436 later,
437 .\"
438 following the discussion of
439 .\" HTML <a href="#subpattern">
440 .\" </a>
441 parenthesized subpatterns.
442 .\"
443 .
444 .
445 .SS "Absolute and relative subroutine calls"
446 .rs
447 .sp
448 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
449 a number enclosed either in angle brackets or single quotes, is an alternative
450 syntax for referencing a subpattern as a "subroutine". Details are discussed
451 .\" HTML <a href="#onigurumasubroutines">
452 .\" </a>
453 later.
454 .\"
455 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
456 synonymous. The former is a back reference; the latter is a
457 .\" HTML <a href="#subpatternsassubroutines">
458 .\" </a>
459 subroutine
460 .\"
461 call.
462 .
463 .
464 .\" HTML <a name="genericchartypes"></a>
465 .SS "Generic character types"
466 .rs
467 .sp
468 Another use of backslash is for specifying generic character types:
469 .sp
470 \ed any decimal digit
471 \eD any character that is not a decimal digit
472 \eh any horizontal white space character
473 \eH any character that is not a horizontal white space character
474 \es any white space character
475 \eS any character that is not a white space character
476 \ev any vertical white space character
477 \eV any character that is not a vertical white space character
478 \ew any "word" character
479 \eW any "non-word" character
480 .sp
481 There is also the single sequence \eN, which matches a non-newline character.
482 This is the same as
483 .\" HTML <a href="#fullstopdot">
484 .\" </a>
485 the "." metacharacter
486 .\"
487 when PCRE_DOTALL is not set. Perl also uses \eN to match characters by name;
488 PCRE does not support this.
489 .P
490 Each pair of lower and upper case escape sequences partitions the complete set
491 of characters into two disjoint sets. Any given character matches one, and only
492 one, of each pair. The sequences can appear both inside and outside character
493 classes. They each match one character of the appropriate type. If the current
494 matching point is at the end of the subject string, all of them fail, because
495 there is no character to match.
496 .P
497 For compatibility with Perl, \es does not match the VT character (code 11).
498 This makes it different from the the POSIX "space" class. The \es characters
499 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
500 included in a Perl script, \es may match the VT character. In PCRE, it never
501 does.
502 .P
503 A "word" character is an underscore or any character that is a letter or digit.
504 By default, the definition of letters and digits is controlled by PCRE's
505 low-valued character tables, and may vary if locale-specific matching is taking
506 place (see
507 .\" HTML <a href="pcreapi.html#localesupport">
508 .\" </a>
509 "Locale support"
510 .\"
511 in the
512 .\" HREF
513 \fBpcreapi\fP
514 .\"
515 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
516 or "french" in Windows, some character codes greater than 128 are used for
517 accented letters, and these are then matched by \ew. The use of locales with
518 Unicode is discouraged.
519 .P
520 By default, in a UTF mode, characters with values greater than 128 never match
521 \ed, \es, or \ew, and always match \eD, \eS, and \eW. These sequences retain
522 their original meanings from before UTF support was available, mainly for
523 efficiency reasons. However, if PCRE is compiled with Unicode property support,
524 and the PCRE_UCP option is set, the behaviour is changed so that Unicode
525 properties are used to determine character types, as follows:
526 .sp
527 \ed any character that \ep{Nd} matches (decimal digit)
528 \es any character that \ep{Z} matches, plus HT, LF, FF, CR
529 \ew any character that \ep{L} or \ep{N} matches, plus underscore
530 .sp
531 The upper case escapes match the inverse sets of characters. Note that \ed
532 matches only decimal digits, whereas \ew matches any Unicode digit, as well as
533 any Unicode letter, and underscore. Note also that PCRE_UCP affects \eb, and
534 \eB because they are defined in terms of \ew and \eW. Matching these sequences
535 is noticeably slower when PCRE_UCP is set.
536 .P
537 The sequences \eh, \eH, \ev, and \eV are features that were added to Perl at
538 release 5.10. In contrast to the other sequences, which match only ASCII
539 characters by default, these always match certain high-valued codepoints,
540 whether or not PCRE_UCP is set. The horizontal space characters are:
541 .sp
542 U+0009 Horizontal tab (HT)
543 U+0020 Space
544 U+00A0 Non-break space
545 U+1680 Ogham space mark
546 U+180E Mongolian vowel separator
547 U+2000 En quad
548 U+2001 Em quad
549 U+2002 En space
550 U+2003 Em space
551 U+2004 Three-per-em space
552 U+2005 Four-per-em space
553 U+2006 Six-per-em space
554 U+2007 Figure space
555 U+2008 Punctuation space
556 U+2009 Thin space
557 U+200A Hair space
558 U+202F Narrow no-break space
559 U+205F Medium mathematical space
560 U+3000 Ideographic space
561 .sp
562 The vertical space characters are:
563 .sp
564 U+000A Linefeed (LF)
565 U+000B Vertical tab (VT)
566 U+000C Form feed (FF)
567 U+000D Carriage return (CR)
568 U+0085 Next line (NEL)
569 U+2028 Line separator
570 U+2029 Paragraph separator
571 .sp
572 In 8-bit, non-UTF-8 mode, only the characters with codepoints less than 256 are
573 relevant.
574 .
575 .
576 .\" HTML <a name="newlineseq"></a>
577 .SS "Newline sequences"
578 .rs
579 .sp
580 Outside a character class, by default, the escape sequence \eR matches any
581 Unicode newline sequence. In 8-bit non-UTF-8 mode \eR is equivalent to the
582 following:
583 .sp
584 (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
585 .sp
586 This is an example of an "atomic group", details of which are given
587 .\" HTML <a href="#atomicgroup">
588 .\" </a>
589 below.
590 .\"
591 This particular group matches either the two-character sequence CR followed by
592 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
593 U+000B), FF (form feed, U+000C), CR (carriage return, U+000D), or NEL (next
594 line, U+0085). The two-character sequence is treated as a single unit that
595 cannot be split.
596 .P
597 In other modes, two additional characters whose codepoints are greater than 255
598 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
599 Unicode character property support is not needed for these characters to be
600 recognized.
601 .P
602 It is possible to restrict \eR to match only CR, LF, or CRLF (instead of the
603 complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
604 either at compile time or when the pattern is matched. (BSR is an abbrevation
605 for "backslash R".) This can be made the default when PCRE is built; if this is
606 the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
607 It is also possible to specify these settings by starting a pattern string with
608 one of the following sequences:
609 .sp
610 (*BSR_ANYCRLF) CR, LF, or CRLF only
611 (*BSR_UNICODE) any Unicode newline sequence
612 .sp
613 These override the default and the options given to the compiling function, but
614 they can themselves be overridden by options given to a matching function. Note
615 that these special settings, which are not Perl-compatible, are recognized only
616 at the very start of a pattern, and that they must be in upper case. If more
617 than one of them is present, the last one is used. They can be combined with a
618 change of newline convention; for example, a pattern can start with:
619 .sp
621 .sp
622 They can also be combined with the (*UTF8), (*UTF16), (*UTF32), (*UTF) or
623 (*UCP) special sequences. Inside a character class, \eR is treated as an
624 unrecognized escape sequence, and so matches the letter "R" by default, but
625 causes an error if PCRE_EXTRA is set.
626 .
627 .
628 .\" HTML <a name="uniextseq"></a>
629 .SS Unicode character properties
630 .rs
631 .sp
632 When PCRE is built with Unicode character property support, three additional
633 escape sequences that match characters with specific properties are available.
634 When in 8-bit non-UTF-8 mode, these sequences are of course limited to testing
635 characters whose codepoints are less than 256, but they do work in this mode.
636 The extra escape sequences are:
637 .sp
638 \ep{\fIxx\fP} a character with the \fIxx\fP property
639 \eP{\fIxx\fP} a character without the \fIxx\fP property
640 \eX a Unicode extended grapheme cluster
641 .sp
642 The property names represented by \fIxx\fP above are limited to the Unicode
643 script names, the general category properties, "Any", which matches any
644 character (including newline), and some special PCRE properties (described
645 in the
646 .\" HTML <a href="#extraprops">
647 .\" </a>
648 next section).
649 .\"
650 Other Perl properties such as "InMusicalSymbols" are not currently supported by
651 PCRE. Note that \eP{Any} does not match any characters, so always causes a
652 match failure.
653 .P
654 Sets of Unicode characters are defined as belonging to certain scripts. A
655 character from one of these sets can be matched using a script name. For
656 example:
657 .sp
658 \ep{Greek}
659 \eP{Han}
660 .sp
661 Those that are not part of an identified script are lumped together as
662 "Common". The current list of scripts is:
663 .P
664 Arabic,
665 Armenian,
666 Avestan,
667 Balinese,
668 Bamum,
669 Batak,
670 Bengali,
671 Bopomofo,
672 Brahmi,
673 Braille,
674 Buginese,
675 Buhid,
676 Canadian_Aboriginal,
677 Carian,
678 Chakma,
679 Cham,
680 Cherokee,
681 Common,
682 Coptic,
683 Cuneiform,
684 Cypriot,
685 Cyrillic,
686 Deseret,
687 Devanagari,
688 Egyptian_Hieroglyphs,
689 Ethiopic,
690 Georgian,
691 Glagolitic,
692 Gothic,
693 Greek,
694 Gujarati,
695 Gurmukhi,
696 Han,
697 Hangul,
698 Hanunoo,
699 Hebrew,
700 Hiragana,
701 Imperial_Aramaic,
702 Inherited,
703 Inscriptional_Pahlavi,
704 Inscriptional_Parthian,
705 Javanese,
706 Kaithi,
707 Kannada,
708 Katakana,
709 Kayah_Li,
710 Kharoshthi,
711 Khmer,
712 Lao,
713 Latin,
714 Lepcha,
715 Limbu,
716 Linear_B,
717 Lisu,
718 Lycian,
719 Lydian,
720 Malayalam,
721 Mandaic,
722 Meetei_Mayek,
723 Meroitic_Cursive,
724 Meroitic_Hieroglyphs,
725 Miao,
726 Mongolian,
727 Myanmar,
728 New_Tai_Lue,
729 Nko,
730 Ogham,
731 Old_Italic,
732 Old_Persian,
733 Old_South_Arabian,
734 Old_Turkic,
735 Ol_Chiki,
736 Oriya,
737 Osmanya,
738 Phags_Pa,
739 Phoenician,
740 Rejang,
741 Runic,
742 Samaritan,
743 Saurashtra,
744 Sharada,
745 Shavian,
746 Sinhala,
747 Sora_Sompeng,
748 Sundanese,
749 Syloti_Nagri,
750 Syriac,
751 Tagalog,
752 Tagbanwa,
753 Tai_Le,
754 Tai_Tham,
755 Tai_Viet,
756 Takri,
757 Tamil,
758 Telugu,
759 Thaana,
760 Thai,
761 Tibetan,
762 Tifinagh,
763 Ugaritic,
764 Vai,
765 Yi.
766 .P
767 Each character has exactly one Unicode general category property, specified by
768 a two-letter abbreviation. For compatibility with Perl, negation can be
769 specified by including a circumflex between the opening brace and the property
770 name. For example, \ep{^Lu} is the same as \eP{Lu}.
771 .P
772 If only one letter is specified with \ep or \eP, it includes all the general
773 category properties that start with that letter. In this case, in the absence
774 of negation, the curly brackets in the escape sequence are optional; these two
775 examples have the same effect:
776 .sp
777 \ep{L}
778 \epL
779 .sp
780 The following general category property codes are supported:
781 .sp
782 C Other
783 Cc Control
784 Cf Format
785 Cn Unassigned
786 Co Private use
787 Cs Surrogate
788 .sp
789 L Letter
790 Ll Lower case letter
791 Lm Modifier letter
792 Lo Other letter
793 Lt Title case letter
794 Lu Upper case letter
795 .sp
796 M Mark
797 Mc Spacing mark
798 Me Enclosing mark
799 Mn Non-spacing mark
800 .sp
801 N Number
802 Nd Decimal number
803 Nl Letter number
804 No Other number
805 .sp
806 P Punctuation
807 Pc Connector punctuation
808 Pd Dash punctuation
809 Pe Close punctuation
810 Pf Final punctuation
811 Pi Initial punctuation
812 Po Other punctuation
813 Ps Open punctuation
814 .sp
815 S Symbol
816 Sc Currency symbol
817 Sk Modifier symbol
818 Sm Mathematical symbol
819 So Other symbol
820 .sp
821 Z Separator
822 Zl Line separator
823 Zp Paragraph separator
824 Zs Space separator
825 .sp
826 The special property L& is also supported: it matches a character that has
827 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
828 a modifier or "other".
829 .P
830 The Cs (Surrogate) property applies only to characters in the range U+D800 to
831 U+DFFF. Such characters are not valid in Unicode strings and so
832 cannot be tested by PCRE, unless UTF validity checking has been turned off
833 (see the discussion of PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK and
834 PCRE_NO_UTF32_CHECK in the
835 .\" HREF
836 \fBpcreapi\fP
837 .\"
838 page). Perl does not support the Cs property.
839 .P
840 The long synonyms for property names that Perl supports (such as \ep{Letter})
841 are not supported by PCRE, nor is it permitted to prefix any of these
842 properties with "Is".
843 .P
844 No character that is in the Unicode table has the Cn (unassigned) property.
845 Instead, this property is assumed for any code point that is not in the
846 Unicode table.
847 .P
848 Specifying caseless matching does not affect these escape sequences. For
849 example, \ep{Lu} always matches only upper case letters. This is different from
850 the behaviour of current versions of Perl.
851 .P
852 Matching characters by Unicode property is not fast, because PCRE has to do a
853 multistage table lookup in order to find a character's property. That is why
854 the traditional escape sequences such as \ed and \ew do not use Unicode
855 properties in PCRE by default, though you can make them do so by setting the
856 PCRE_UCP option or by starting the pattern with (*UCP).
857 .
858 .
859 .SS Extended grapheme clusters
860 .rs
861 .sp
862 The \eX escape matches any number of Unicode characters that form an "extended
863 grapheme cluster", and treats the sequence as an atomic group
864 .\" HTML <a href="#atomicgroup">
865 .\" </a>
866 (see below).
867 .\"
868 Up to and including release 8.31, PCRE matched an earlier, simpler definition
869 that was equivalent to
870 .sp
871 (?>\ePM\epM*)
872 .sp
873 That is, it matched a character without the "mark" property, followed by zero
874 or more characters with the "mark" property. Characters with the "mark"
875 property are typically non-spacing accents that affect the preceding character.
876 .P
877 This simple definition was extended in Unicode to include more complicated
878 kinds of composite character by giving each character a grapheme breaking
879 property, and creating rules that use these properties to define the boundaries
880 of extended grapheme clusters. In releases of PCRE later than 8.31, \eX matches
881 one of these clusters.
882 .P
883 \eX always matches at least one character. Then it decides whether to add
884 additional characters according to the following rules for ending a cluster:
885 .P
886 1. End at the end of the subject string.
887 .P
888 2. Do not end between CR and LF; otherwise end after any control character.
889 .P
890 3. Do not break Hangul (a Korean script) syllable sequences. Hangul characters
891 are of five types: L, V, T, LV, and LVT. An L character may be followed by an
892 L, V, LV, or LVT character; an LV or V character may be followed by a V or T
893 character; an LVT or T character may be follwed only by a T character.
894 .P
895 4. Do not end before extending characters or spacing marks. Characters with
896 the "mark" property always have the "extend" grapheme breaking property.
897 .P
898 5. Do not end after prepend characters.
899 .P
900 6. Otherwise, end the cluster.
901 .
902 .
903 .\" HTML <a name="extraprops"></a>
904 .SS PCRE's additional properties
905 .rs
906 .sp
907 As well as the standard Unicode properties described above, PCRE supports four
908 more that make it possible to convert traditional escape sequences such as \ew
909 and \es and POSIX character classes to use Unicode properties. PCRE uses these
910 non-standard, non-Perl properties internally when PCRE_UCP is set. However,
911 they may also be used explicitly. These properties are:
912 .sp
913 Xan Any alphanumeric character
914 Xps Any POSIX space character
915 Xsp Any Perl space character
916 Xwd Any Perl "word" character
917 .sp
918 Xan matches characters that have either the L (letter) or the N (number)
919 property. Xps matches the characters tab, linefeed, vertical tab, form feed, or
920 carriage return, and any other character that has the Z (separator) property.
921 Xsp is the same as Xps, except that vertical tab is excluded. Xwd matches the
922 same characters as Xan, plus underscore.
923 .P
924 There is another non-standard property, Xuc, which matches any character that
925 can be represented by a Universal Character Name in C++ and other programming
926 languages. These are the characters $, @, ` (grave accent), and all characters
927 with Unicode code points greater than or equal to U+00A0, except for the
928 surrogates U+D800 to U+DFFF. Note that most base (ASCII) characters are
929 excluded. (Universal Character Names are of the form \euHHHH or \eUHHHHHHHH
930 where H is a hexadecimal digit. Note that the Xuc property does not match these
931 sequences but the characters that they represent.)
932 .
933 .
934 .\" HTML <a name="resetmatchstart"></a>
935 .SS "Resetting the match start"
936 .rs
937 .sp
938 The escape sequence \eK causes any previously matched characters not to be
939 included in the final matched sequence. For example, the pattern:
940 .sp
941 foo\eKbar
942 .sp
943 matches "foobar", but reports that it has matched "bar". This feature is
944 similar to a lookbehind assertion
945 .\" HTML <a href="#lookbehind">
946 .\" </a>
947 (described below).
948 .\"
949 However, in this case, the part of the subject before the real match does not
950 have to be of fixed length, as lookbehind assertions do. The use of \eK does
951 not interfere with the setting of
952 .\" HTML <a href="#subpattern">
953 .\" </a>
954 captured substrings.
955 .\"
956 For example, when the pattern
957 .sp
958 (foo)\eKbar
959 .sp
960 matches "foobar", the first substring is still set to "foo".
961 .P
962 Perl documents that the use of \eK within assertions is "not well defined". In
963 PCRE, \eK is acted upon when it occurs inside positive assertions, but is
964 ignored in negative assertions.
965 .
966 .
967 .\" HTML <a name="smallassertions"></a>
968 .SS "Simple assertions"
969 .rs
970 .sp
971 The final use of backslash is for certain simple assertions. An assertion
972 specifies a condition that has to be met at a particular point in a match,
973 without consuming any characters from the subject string. The use of
974 subpatterns for more complicated assertions is described
975 .\" HTML <a href="#bigassertions">
976 .\" </a>
977 below.
978 .\"
979 The backslashed assertions are:
980 .sp
981 \eb matches at a word boundary
982 \eB matches when not at a word boundary
983 \eA matches at the start of the subject
984 \eZ matches at the end of the subject
985 also matches before a newline at the end of the subject
986 \ez matches only at the end of the subject
987 \eG matches at the first matching position in the subject
988 .sp
989 Inside a character class, \eb has a different meaning; it matches the backspace
990 character. If any other of these assertions appears in a character class, by
991 default it matches the corresponding literal character (for example, \eB
992 matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
993 escape sequence" error is generated instead.
994 .P
995 A word boundary is a position in the subject string where the current character
996 and the previous character do not both match \ew or \eW (i.e. one matches
997 \ew and the other matches \eW), or the start or end of the string if the
998 first or last character matches \ew, respectively. In a UTF mode, the meanings
999 of \ew and \eW can be changed by setting the PCRE_UCP option. When this is
1000 done, it also affects \eb and \eB. Neither PCRE nor Perl has a separate "start
1001 of word" or "end of word" metasequence. However, whatever follows \eb normally
1002 determines which it is. For example, the fragment \eba matches "a" at the start
1003 of a word.
1004 .P
1005 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
1006 dollar (described in the next section) in that they only ever match at the very
1007 start and end of the subject string, whatever options are set. Thus, they are
1008 independent of multiline mode. These three assertions are not affected by the
1009 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
1010 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
1011 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
1012 at a point other than the beginning of the subject, \eA can never match. The
1013 difference between \eZ and \ez is that \eZ matches before a newline at the end
1014 of the string as well as at the very end, whereas \ez matches only at the end.
1015 .P
1016 The \eG assertion is true only when the current matching position is at the
1017 start point of the match, as specified by the \fIstartoffset\fP argument of
1018 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
1019 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
1020 arguments, you can mimic Perl's /g option, and it is in this kind of
1021 implementation where \eG can be useful.
1022 .P
1023 Note, however, that PCRE's interpretation of \eG, as the start of the current
1024 match, is subtly different from Perl's, which defines it as the end of the
1025 previous match. In Perl, these can be different when the previously matched
1026 string was empty. Because PCRE does just one match at a time, it cannot
1027 reproduce this behaviour.
1028 .P
1029 If all the alternatives of a pattern begin with \eG, the expression is anchored
1030 to the starting match position, and the "anchored" flag is set in the compiled
1031 regular expression.
1032 .
1033 .
1035 .rs
1036 .sp
1037 The circumflex and dollar metacharacters are zero-width assertions. That is,
1038 they test for a particular condition being true without consuming any
1039 characters from the subject string.
1040 .P
1041 Outside a character class, in the default matching mode, the circumflex
1042 character is an assertion that is true only if the current matching point is at
1043 the start of the subject string. If the \fIstartoffset\fP argument of
1044 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
1045 option is unset. Inside a character class, circumflex has an entirely different
1046 meaning
1047 .\" HTML <a href="#characterclass">
1048 .\" </a>
1049 (see below).
1050 .\"
1051 .P
1052 Circumflex need not be the first character of the pattern if a number of
1053 alternatives are involved, but it should be the first thing in each alternative
1054 in which it appears if the pattern is ever to match that branch. If all
1055 possible alternatives start with a circumflex, that is, if the pattern is
1056 constrained to match only at the start of the subject, it is said to be an
1057 "anchored" pattern. (There are also other constructs that can cause a pattern
1058 to be anchored.)
1059 .P
1060 The dollar character is an assertion that is true only if the current matching
1061 point is at the end of the subject string, or immediately before a newline at
1062 the end of the string (by default). Note, however, that it does not actually
1063 match the newline. Dollar need not be the last character of the pattern if a
1064 number of alternatives are involved, but it should be the last item in any
1065 branch in which it appears. Dollar has no special meaning in a character class.
1066 .P
1067 The meaning of dollar can be changed so that it matches only at the very end of
1068 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
1069 does not affect the \eZ assertion.
1070 .P
1071 The meanings of the circumflex and dollar characters are changed if the
1072 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
1073 immediately after internal newlines as well as at the start of the subject
1074 string. It does not match after a newline that ends the string. A dollar
1075 matches before any newlines in the string, as well as at the very end, when
1076 PCRE_MULTILINE is set. When newline is specified as the two-character
1077 sequence CRLF, isolated CR and LF characters do not indicate newlines.
1078 .P
1079 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
1080 \en represents a newline) in multiline mode, but not otherwise. Consequently,
1081 patterns that are anchored in single line mode because all branches start with
1082 ^ are not anchored in multiline mode, and a match for circumflex is possible
1083 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
1084 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
1085 .P
1086 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
1087 end of the subject in both modes, and if all branches of a pattern start with
1088 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
1089 .
1090 .
1091 .\" HTML <a name="fullstopdot"></a>
1093 .rs
1094 .sp
1095 Outside a character class, a dot in the pattern matches any one character in
1096 the subject string except (by default) a character that signifies the end of a
1097 line.
1098 .P
1099 When a line ending is defined as a single character, dot never matches that
1100 character; when the two-character sequence CRLF is used, dot does not match CR
1101 if it is immediately followed by LF, but otherwise it matches all characters
1102 (including isolated CRs and LFs). When any Unicode line endings are being
1103 recognized, dot does not match CR or LF or any of the other line ending
1104 characters.
1105 .P
1106 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
1107 option is set, a dot matches any one character, without exception. If the
1108 two-character sequence CRLF is present in the subject string, it takes two dots
1109 to match it.
1110 .P
1111 The handling of dot is entirely independent of the handling of circumflex and
1112 dollar, the only relationship being that they both involve newlines. Dot has no
1113 special meaning in a character class.
1114 .P
1115 The escape sequence \eN behaves like a dot, except that it is not affected by
1116 the PCRE_DOTALL option. In other words, it matches any character except one
1117 that signifies the end of a line. Perl also uses \eN to match characters by
1118 name; PCRE does not support this.
1119 .
1120 .
1122 .rs
1123 .sp
1124 Outside a character class, the escape sequence \eC matches any one data unit,
1125 whether or not a UTF mode is set. In the 8-bit library, one data unit is one
1126 byte; in the 16-bit library it is a 16-bit unit; in the 32-bit library it is
1127 a 32-bit unit. Unlike a dot, \eC always
1128 matches line-ending characters. The feature is provided in Perl in order to
1129 match individual bytes in UTF-8 mode, but it is unclear how it can usefully be
1130 used. Because \eC breaks up characters into individual data units, matching one
1131 unit with \eC in a UTF mode means that the rest of the string may start with a
1132 malformed UTF character. This has undefined results, because PCRE assumes that
1133 it is dealing with valid UTF strings (and by default it checks this at the
1134 start of processing unless the PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK or
1135 PCRE_NO_UTF32_CHECK option is used).
1136 .P
1137 PCRE does not allow \eC to appear in lookbehind assertions
1138 .\" HTML <a href="#lookbehind">
1139 .\" </a>
1140 (described below)
1141 .\"
1142 in a UTF mode, because this would make it impossible to calculate the length of
1143 the lookbehind.
1144 .P
1145 In general, the \eC escape sequence is best avoided. However, one
1146 way of using it that avoids the problem of malformed UTF characters is to use a
1147 lookahead to check the length of the next character, as in this pattern, which
1148 could be used with a UTF-8 string (ignore white space and line breaks):
1149 .sp
1150 (?| (?=[\ex00-\ex7f])(\eC) |
1151 (?=[\ex80-\ex{7ff}])(\eC)(\eC) |
1152 (?=[\ex{800}-\ex{ffff}])(\eC)(\eC)(\eC) |
1153 (?=[\ex{10000}-\ex{1fffff}])(\eC)(\eC)(\eC)(\eC))
1154 .sp
1155 A group that starts with (?| resets the capturing parentheses numbers in each
1156 alternative (see
1157 .\" HTML <a href="#dupsubpatternnumber">
1158 .\" </a>
1159 "Duplicate Subpattern Numbers"
1160 .\"
1161 below). The assertions at the start of each branch check the next UTF-8
1162 character for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The
1163 character's individual bytes are then captured by the appropriate number of
1164 groups.
1165 .
1166 .
1167 .\" HTML <a name="characterclass"></a>
1169 .rs
1170 .sp
1171 An opening square bracket introduces a character class, terminated by a closing
1172 square bracket. A closing square bracket on its own is not special by default.
1173 However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
1174 bracket causes a compile-time error. If a closing square bracket is required as
1175 a member of the class, it should be the first data character in the class
1176 (after an initial circumflex, if present) or escaped with a backslash.
1177 .P
1178 A character class matches a single character in the subject. In a UTF mode, the
1179 character may be more than one data unit long. A matched character must be in
1180 the set of characters defined by the class, unless the first character in the
1181 class definition is a circumflex, in which case the subject character must not
1182 be in the set defined by the class. If a circumflex is actually required as a
1183 member of the class, ensure it is not the first character, or escape it with a
1184 backslash.
1185 .P
1186 For example, the character class [aeiou] matches any lower case vowel, while
1187 [^aeiou] matches any character that is not a lower case vowel. Note that a
1188 circumflex is just a convenient notation for specifying the characters that
1189 are in the class by enumerating those that are not. A class that starts with a
1190 circumflex is not an assertion; it still consumes a character from the subject
1191 string, and therefore it fails if the current pointer is at the end of the
1192 string.
1193 .P
1194 In UTF-8 (UTF-16, UTF-32) mode, characters with values greater than 255 (0xffff)
1195 can be included in a class as a literal string of data units, or by using the
1196 \ex{ escaping mechanism.
1197 .P
1198 When caseless matching is set, any letters in a class represent both their
1199 upper case and lower case versions, so for example, a caseless [aeiou] matches
1200 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1201 caseful version would. In a UTF mode, PCRE always understands the concept of
1202 case for characters whose values are less than 128, so caseless matching is
1203 always possible. For characters with higher values, the concept of case is
1204 supported if PCRE is compiled with Unicode property support, but not otherwise.
1205 If you want to use caseless matching in a UTF mode for characters 128 and
1206 above, you must ensure that PCRE is compiled with Unicode property support as
1207 well as with UTF support.
1208 .P
1209 Characters that might indicate line breaks are never treated in any special way
1210 when matching character classes, whatever line-ending sequence is in use, and
1211 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1212 such as [^a] always matches one of these characters.
1213 .P
1214 The minus (hyphen) character can be used to specify a range of characters in a
1215 character class. For example, [d-m] matches any letter between d and m,
1216 inclusive. If a minus character is required in a class, it must be escaped with
1217 a backslash or appear in a position where it cannot be interpreted as
1218 indicating a range, typically as the first or last character in the class.
1219 .P
1220 It is not possible to have the literal character "]" as the end character of a
1221 range. A pattern such as [W-]46] is interpreted as a class of two characters
1222 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1223 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1224 the end of range, so [W-\e]46] is interpreted as a class containing a range
1225 followed by two other characters. The octal or hexadecimal representation of
1226 "]" can also be used to end a range.
1227 .P
1228 Ranges operate in the collating sequence of character values. They can also be
1229 used for characters specified numerically, for example [\e000-\e037]. Ranges
1230 can include any characters that are valid for the current mode.
1231 .P
1232 If a range that includes letters is used when caseless matching is set, it
1233 matches the letters in either case. For example, [W-c] is equivalent to
1234 [][\e\e^_`wxyzabc], matched caselessly, and in a non-UTF mode, if character
1235 tables for a French locale are in use, [\exc8-\excb] matches accented E
1236 characters in both cases. In UTF modes, PCRE supports the concept of case for
1237 characters with values greater than 128 only when it is compiled with Unicode
1238 property support.
1239 .P
1240 The character escape sequences \ed, \eD, \eh, \eH, \ep, \eP, \es, \eS, \ev,
1241 \eV, \ew, and \eW may appear in a character class, and add the characters that
1242 they match to the class. For example, [\edABCDEF] matches any hexadecimal
1243 digit. In UTF modes, the PCRE_UCP option affects the meanings of \ed, \es, \ew
1244 and their upper case partners, just as it does when they appear outside a
1245 character class, as described in the section entitled
1246 .\" HTML <a href="#genericchartypes">
1247 .\" </a>
1248 "Generic character types"
1249 .\"
1250 above. The escape sequence \eb has a different meaning inside a character
1251 class; it matches the backspace character. The sequences \eB, \eN, \eR, and \eX
1252 are not special inside a character class. Like any other unrecognized escape
1253 sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1254 default, but cause an error if the PCRE_EXTRA option is set.
1255 .P
1256 A circumflex can conveniently be used with the upper case character types to
1257 specify a more restricted set of characters than the matching lower case type.
1258 For example, the class [^\eW_] matches any letter or digit, but not underscore,
1259 whereas [\ew] includes underscore. A positive character class should be read as
1260 "something OR something OR ..." and a negative class as "NOT something AND NOT
1261 something AND NOT ...".
1262 .P
1263 The only metacharacters that are recognized in character classes are backslash,
1264 hyphen (only where it can be interpreted as specifying a range), circumflex
1265 (only at the start), opening square bracket (only when it can be interpreted as
1266 introducing a POSIX class name - see the next section), and the terminating
1267 closing square bracket. However, escaping other non-alphanumeric characters
1268 does no harm.
1269 .
1270 .
1272 .rs
1273 .sp
1274 Perl supports the POSIX notation for character classes. This uses names
1275 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1276 this notation. For example,
1277 .sp
1278 [01[:alpha:]%]
1279 .sp
1280 matches "0", "1", any alphabetic character, or "%". The supported class names
1281 are:
1282 .sp
1283 alnum letters and digits
1284 alpha letters
1285 ascii character codes 0 - 127
1286 blank space or tab only
1287 cntrl control characters
1288 digit decimal digits (same as \ed)
1289 graph printing characters, excluding space
1290 lower lower case letters
1291 print printing characters, including space
1292 punct printing characters, excluding letters and digits and space
1293 space white space (not quite the same as \es)
1294 upper upper case letters
1295 word "word" characters (same as \ew)
1296 xdigit hexadecimal digits
1297 .sp
1298 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
1299 space (32). Notice that this list includes the VT character (code 11). This
1300 makes "space" different to \es, which does not include VT (for Perl
1301 compatibility).
1302 .P
1303 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1304 5.8. Another Perl extension is negation, which is indicated by a ^ character
1305 after the colon. For example,
1306 .sp
1307 [12[:^digit:]]
1308 .sp
1309 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1310 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1311 supported, and an error is given if they are encountered.
1312 .P
1313 By default, in UTF modes, characters with values greater than 128 do not match
1314 any of the POSIX character classes. However, if the PCRE_UCP option is passed
1315 to \fBpcre_compile()\fP, some of the classes are changed so that Unicode
1316 character properties are used. This is achieved by replacing the POSIX classes
1317 by other sequences, as follows:
1318 .sp
1319 [:alnum:] becomes \ep{Xan}
1320 [:alpha:] becomes \ep{L}
1321 [:blank:] becomes \eh
1322 [:digit:] becomes \ep{Nd}
1323 [:lower:] becomes \ep{Ll}
1324 [:space:] becomes \ep{Xps}
1325 [:upper:] becomes \ep{Lu}
1326 [:word:] becomes \ep{Xwd}
1327 .sp
1328 Negated versions, such as [:^alpha:] use \eP instead of \ep. The other POSIX
1329 classes are unchanged, and match only characters with code points less than
1330 128.
1331 .
1332 .
1334 .rs
1335 .sp
1336 Vertical bar characters are used to separate alternative patterns. For example,
1337 the pattern
1338 .sp
1339 gilbert|sullivan
1340 .sp
1341 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1342 and an empty alternative is permitted (matching the empty string). The matching
1343 process tries each alternative in turn, from left to right, and the first one
1344 that succeeds is used. If the alternatives are within a subpattern
1345 .\" HTML <a href="#subpattern">
1346 .\" </a>
1347 (defined below),
1348 .\"
1349 "succeeds" means matching the rest of the main pattern as well as the
1350 alternative in the subpattern.
1351 .
1352 .
1354 .rs
1355 .sp
1356 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1357 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1358 the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1359 The option letters are
1360 .sp
1361 i for PCRE_CASELESS
1363 s for PCRE_DOTALL
1364 x for PCRE_EXTENDED
1365 .sp
1366 For example, (?im) sets caseless, multiline matching. It is also possible to
1367 unset these options by preceding the letter with a hyphen, and a combined
1368 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1369 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1370 permitted. If a letter appears both before and after the hyphen, the option is
1371 unset.
1372 .P
1373 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1374 changed in the same way as the Perl-compatible options by using the characters
1375 J, U and X respectively.
1376 .P
1377 When one of these option changes occurs at top level (that is, not inside
1378 subpattern parentheses), the change applies to the remainder of the pattern
1379 that follows. If the change is placed right at the start of a pattern, PCRE
1380 extracts it into the global options (and it will therefore show up in data
1381 extracted by the \fBpcre_fullinfo()\fP function).
1382 .P
1383 An option change within a subpattern (see below for a description of
1384 subpatterns) affects only that part of the subpattern that follows it, so
1385 .sp
1386 (a(?i)b)c
1387 .sp
1388 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1389 By this means, options can be made to have different settings in different
1390 parts of the pattern. Any changes made in one alternative do carry on
1391 into subsequent branches within the same subpattern. For example,
1392 .sp
1393 (a(?i)b|c)
1394 .sp
1395 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1396 branch is abandoned before the option setting. This is because the effects of
1397 option settings happen at compile time. There would be some very weird
1398 behaviour otherwise.
1399 .P
1400 \fBNote:\fP There are other PCRE-specific options that can be set by the
1401 application when the compiling or matching functions are called. In some cases
1402 the pattern can contain special leading sequences such as (*CRLF) to override
1403 what the application has set or what has been defaulted. Details are given in
1404 the section entitled
1405 .\" HTML <a href="#newlineseq">
1406 .\" </a>
1407 "Newline sequences"
1408 .\"
1409 above. There are also the (*UTF8), (*UTF16),(*UTF32), and (*UCP) leading
1410 sequences that can be used to set UTF and Unicode property modes; they are
1411 equivalent to setting the PCRE_UTF8, PCRE_UTF16, PCRE_UTF32 and the PCRE_UCP
1412 options, respectively. The (*UTF) sequence is a generic version that can be
1413 used with any of the libraries. However, the application can set the
1414 PCRE_NEVER_UTF option, which locks out the use of the (*UTF) sequences.
1415 .
1416 .
1417 .\" HTML <a name="subpattern"></a>
1419 .rs
1420 .sp
1421 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1422 Turning part of a pattern into a subpattern does two things:
1423 .sp
1424 1. It localizes a set of alternatives. For example, the pattern
1425 .sp
1426 cat(aract|erpillar|)
1427 .sp
1428 matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1429 match "cataract", "erpillar" or an empty string.
1430 .sp
1431 2. It sets up the subpattern as a capturing subpattern. This means that, when
1432 the whole pattern matches, that portion of the subject string that matched the
1433 subpattern is passed back to the caller via the \fIovector\fP argument of the
1434 matching function. (This applies only to the traditional matching functions;
1435 the DFA matching functions do not support capturing.)
1436 .P
1437 Opening parentheses are counted from left to right (starting from 1) to obtain
1438 numbers for the capturing subpatterns. For example, if the string "the red
1439 king" is matched against the pattern
1440 .sp
1441 the ((red|white) (king|queen))
1442 .sp
1443 the captured substrings are "red king", "red", and "king", and are numbered 1,
1444 2, and 3, respectively.
1445 .P
1446 The fact that plain parentheses fulfil two functions is not always helpful.
1447 There are often times when a grouping subpattern is required without a
1448 capturing requirement. If an opening parenthesis is followed by a question mark
1449 and a colon, the subpattern does not do any capturing, and is not counted when
1450 computing the number of any subsequent capturing subpatterns. For example, if
1451 the string "the white queen" is matched against the pattern
1452 .sp
1453 the ((?:red|white) (king|queen))
1454 .sp
1455 the captured substrings are "white queen" and "queen", and are numbered 1 and
1456 2. The maximum number of capturing subpatterns is 65535.
1457 .P
1458 As a convenient shorthand, if any option settings are required at the start of
1459 a non-capturing subpattern, the option letters may appear between the "?" and
1460 the ":". Thus the two patterns
1461 .sp
1462 (?i:saturday|sunday)
1463 (?:(?i)saturday|sunday)
1464 .sp
1465 match exactly the same set of strings. Because alternative branches are tried
1466 from left to right, and options are not reset until the end of the subpattern
1467 is reached, an option setting in one branch does affect subsequent branches, so
1468 the above patterns match "SUNDAY" as well as "Saturday".
1469 .
1470 .
1471 .\" HTML <a name="dupsubpatternnumber"></a>
1473 .rs
1474 .sp
1475 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1476 the same numbers for its capturing parentheses. Such a subpattern starts with
1477 (?| and is itself a non-capturing subpattern. For example, consider this
1478 pattern:
1479 .sp
1480 (?|(Sat)ur|(Sun))day
1481 .sp
1482 Because the two alternatives are inside a (?| group, both sets of capturing
1483 parentheses are numbered one. Thus, when the pattern matches, you can look
1484 at captured substring number one, whichever alternative matched. This construct
1485 is useful when you want to capture part, but not all, of one of a number of
1486 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1487 number is reset at the start of each branch. The numbers of any capturing
1488 parentheses that follow the subpattern start after the highest number used in
1489 any branch. The following example is taken from the Perl documentation. The
1490 numbers underneath show in which buffer the captured content will be stored.
1491 .sp
1492 # before ---------------branch-reset----------- after
1493 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1494 # 1 2 2 3 2 3 4
1495 .sp
1496 A back reference to a numbered subpattern uses the most recent value that is
1497 set for that number by any subpattern. The following pattern matches "abcabc"
1498 or "defdef":
1499 .sp
1500 /(?|(abc)|(def))\e1/
1501 .sp
1502 In contrast, a subroutine call to a numbered subpattern always refers to the
1503 first one in the pattern with the given number. The following pattern matches
1504 "abcabc" or "defabc":
1505 .sp
1506 /(?|(abc)|(def))(?1)/
1507 .sp
1508 If a
1509 .\" HTML <a href="#conditions">
1510 .\" </a>
1511 condition test
1512 .\"
1513 for a subpattern's having matched refers to a non-unique number, the test is
1514 true if any of the subpatterns of that number have matched.
1515 .P
1516 An alternative approach to using this "branch reset" feature is to use
1517 duplicate named subpatterns, as described in the next section.
1518 .
1519 .
1521 .rs
1522 .sp
1523 Identifying capturing parentheses by number is simple, but it can be very hard
1524 to keep track of the numbers in complicated regular expressions. Furthermore,
1525 if an expression is modified, the numbers may change. To help with this
1526 difficulty, PCRE supports the naming of subpatterns. This feature was not
1527 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1528 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1529 the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1530 have different names, but PCRE does not.
1531 .P
1532 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1533 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1534 parentheses from other parts of the pattern, such as
1535 .\" HTML <a href="#backreferences">
1536 .\" </a>
1537 back references,
1538 .\"
1539 .\" HTML <a href="#recursion">
1540 .\" </a>
1541 recursion,
1542 .\"
1543 and
1544 .\" HTML <a href="#conditions">
1545 .\" </a>
1546 conditions,
1547 .\"
1548 can be made by name as well as by number.
1549 .P
1550 Names consist of up to 32 alphanumeric characters and underscores. Named
1551 capturing parentheses are still allocated numbers as well as names, exactly as
1552 if the names were not present. The PCRE API provides function calls for
1553 extracting the name-to-number translation table from a compiled pattern. There
1554 is also a convenience function for extracting a captured substring by name.
1555 .P
1556 By default, a name must be unique within a pattern, but it is possible to relax
1557 this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1558 names are also always permitted for subpatterns with the same number, set up as
1559 described in the previous section.) Duplicate names can be useful for patterns
1560 where only one instance of the named parentheses can match. Suppose you want to
1561 match the name of a weekday, either as a 3-letter abbreviation or as the full
1562 name, and in both cases you want to extract the abbreviation. This pattern
1563 (ignoring the line breaks) does the job:
1564 .sp
1565 (?<DN>Mon|Fri|Sun)(?:day)?|
1566 (?<DN>Tue)(?:sday)?|
1567 (?<DN>Wed)(?:nesday)?|
1568 (?<DN>Thu)(?:rsday)?|
1569 (?<DN>Sat)(?:urday)?
1570 .sp
1571 There are five capturing substrings, but only one is ever set after a match.
1572 (An alternative way of solving this problem is to use a "branch reset"
1573 subpattern, as described in the previous section.)
1574 .P
1575 The convenience function for extracting the data by name returns the substring
1576 for the first (and in this example, the only) subpattern of that name that
1577 matched. This saves searching to find which numbered subpattern it was.
1578 .P
1579 If you make a back reference to a non-unique named subpattern from elsewhere in
1580 the pattern, the one that corresponds to the first occurrence of the name is
1581 used. In the absence of duplicate numbers (see the previous section) this is
1582 the one with the lowest number. If you use a named reference in a condition
1583 test (see the
1584 .\"
1585 .\" HTML <a href="#conditions">
1586 .\" </a>
1587 section about conditions
1588 .\"
1589 below), either to check whether a subpattern has matched, or to check for
1590 recursion, all subpatterns with the same name are tested. If the condition is
1591 true for any one of them, the overall condition is true. This is the same
1592 behaviour as testing by number. For further details of the interfaces for
1593 handling named subpatterns, see the
1594 .\" HREF
1595 \fBpcreapi\fP
1596 .\"
1597 documentation.
1598 .P
1599 \fBWarning:\fP You cannot use different names to distinguish between two
1600 subpatterns with the same number because PCRE uses only the numbers when
1601 matching. For this reason, an error is given at compile time if different names
1602 are given to subpatterns with the same number. However, you can give the same
1603 name to subpatterns with the same number, even when PCRE_DUPNAMES is not set.
1604 .
1605 .
1607 .rs
1608 .sp
1609 Repetition is specified by quantifiers, which can follow any of the following
1610 items:
1611 .sp
1612 a literal data character
1613 the dot metacharacter
1614 the \eC escape sequence
1615 the \eX escape sequence
1616 the \eR escape sequence
1617 an escape such as \ed or \epL that matches a single character
1618 a character class
1619 a back reference (see next section)
1620 a parenthesized subpattern (including assertions)
1621 a subroutine call to a subpattern (recursive or otherwise)
1622 .sp
1623 The general repetition quantifier specifies a minimum and maximum number of
1624 permitted matches, by giving the two numbers in curly brackets (braces),
1625 separated by a comma. The numbers must be less than 65536, and the first must
1626 be less than or equal to the second. For example:
1627 .sp
1628 z{2,4}
1629 .sp
1630 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1631 character. If the second number is omitted, but the comma is present, there is
1632 no upper limit; if the second number and the comma are both omitted, the
1633 quantifier specifies an exact number of required matches. Thus
1634 .sp
1635 [aeiou]{3,}
1636 .sp
1637 matches at least 3 successive vowels, but may match many more, while
1638 .sp
1639 \ed{8}
1640 .sp
1641 matches exactly 8 digits. An opening curly bracket that appears in a position
1642 where a quantifier is not allowed, or one that does not match the syntax of a
1643 quantifier, is taken as a literal character. For example, {,6} is not a
1644 quantifier, but a literal string of four characters.
1645 .P
1646 In UTF modes, quantifiers apply to characters rather than to individual data
1647 units. Thus, for example, \ex{100}{2} matches two characters, each of
1648 which is represented by a two-byte sequence in a UTF-8 string. Similarly,
1649 \eX{3} matches three Unicode extended grapheme clusters, each of which may be
1650 several data units long (and they may be of different lengths).
1651 .P
1652 The quantifier {0} is permitted, causing the expression to behave as if the
1653 previous item and the quantifier were not present. This may be useful for
1654 subpatterns that are referenced as
1655 .\" HTML <a href="#subpatternsassubroutines">
1656 .\" </a>
1657 subroutines
1658 .\"
1659 from elsewhere in the pattern (but see also the section entitled
1660 .\" HTML <a href="#subdefine">
1661 .\" </a>
1662 "Defining subpatterns for use by reference only"
1663 .\"
1664 below). Items other than subpatterns that have a {0} quantifier are omitted
1665 from the compiled pattern.
1666 .P
1667 For convenience, the three most common quantifiers have single-character
1668 abbreviations:
1669 .sp
1670 * is equivalent to {0,}
1671 + is equivalent to {1,}
1672 ? is equivalent to {0,1}
1673 .sp
1674 It is possible to construct infinite loops by following a subpattern that can
1675 match no characters with a quantifier that has no upper limit, for example:
1676 .sp
1677 (a?)*
1678 .sp
1679 Earlier versions of Perl and PCRE used to give an error at compile time for
1680 such patterns. However, because there are cases where this can be useful, such
1681 patterns are now accepted, but if any repetition of the subpattern does in fact
1682 match no characters, the loop is forcibly broken.
1683 .P
1684 By default, the quantifiers are "greedy", that is, they match as much as
1685 possible (up to the maximum number of permitted times), without causing the
1686 rest of the pattern to fail. The classic example of where this gives problems
1687 is in trying to match comments in C programs. These appear between /* and */
1688 and within the comment, individual * and / characters may appear. An attempt to
1689 match C comments by applying the pattern
1690 .sp
1691 /\e*.*\e*/
1692 .sp
1693 to the string
1694 .sp
1695 /* first comment */ not comment /* second comment */
1696 .sp
1697 fails, because it matches the entire string owing to the greediness of the .*
1698 item.
1699 .P
1700 However, if a quantifier is followed by a question mark, it ceases to be
1701 greedy, and instead matches the minimum number of times possible, so the
1702 pattern
1703 .sp
1704 /\e*.*?\e*/
1705 .sp
1706 does the right thing with the C comments. The meaning of the various
1707 quantifiers is not otherwise changed, just the preferred number of matches.
1708 Do not confuse this use of question mark with its use as a quantifier in its
1709 own right. Because it has two uses, it can sometimes appear doubled, as in
1710 .sp
1711 \ed??\ed
1712 .sp
1713 which matches one digit by preference, but can match two if that is the only
1714 way the rest of the pattern matches.
1715 .P
1716 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1717 the quantifiers are not greedy by default, but individual ones can be made
1718 greedy by following them with a question mark. In other words, it inverts the
1719 default behaviour.
1720 .P
1721 When a parenthesized subpattern is quantified with a minimum repeat count that
1722 is greater than 1 or with a limited maximum, more memory is required for the
1723 compiled pattern, in proportion to the size of the minimum or maximum.
1724 .P
1725 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1726 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1727 implicitly anchored, because whatever follows will be tried against every
1728 character position in the subject string, so there is no point in retrying the
1729 overall match at any position after the first. PCRE normally treats such a
1730 pattern as though it were preceded by \eA.
1731 .P
1732 In cases where it is known that the subject string contains no newlines, it is
1733 worth setting PCRE_DOTALL in order to obtain this optimization, or
1734 alternatively using ^ to indicate anchoring explicitly.
1735 .P
1736 However, there are some cases where the optimization cannot be used. When .*
1737 is inside capturing parentheses that are the subject of a back reference
1738 elsewhere in the pattern, a match at the start may fail where a later one
1739 succeeds. Consider, for example:
1740 .sp
1741 (.*)abc\e1
1742 .sp
1743 If the subject is "xyz123abc123" the match point is the fourth character. For
1744 this reason, such a pattern is not implicitly anchored.
1745 .P
1746 Another case where implicit anchoring is not applied is when the leading .* is
1747 inside an atomic group. Once again, a match at the start may fail where a later
1748 one succeeds. Consider this pattern:
1749 .sp
1750 (?>.*?a)b
1751 .sp
1752 It matches "ab" in the subject "aab". The use of the backtracking control verbs
1753 (*PRUNE) and (*SKIP) also disable this optimization.
1754 .P
1755 When a capturing subpattern is repeated, the value captured is the substring
1756 that matched the final iteration. For example, after
1757 .sp
1758 (tweedle[dume]{3}\es*)+
1759 .sp
1760 has matched "tweedledum tweedledee" the value of the captured substring is
1761 "tweedledee". However, if there are nested capturing subpatterns, the
1762 corresponding captured values may have been set in previous iterations. For
1763 example, after
1764 .sp
1765 /(a|(b))+/
1766 .sp
1767 matches "aba" the value of the second captured substring is "b".
1768 .
1769 .
1770 .\" HTML <a name="atomicgroup"></a>
1772 .rs
1773 .sp
1774 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1775 repetition, failure of what follows normally causes the repeated item to be
1776 re-evaluated to see if a different number of repeats allows the rest of the
1777 pattern to match. Sometimes it is useful to prevent this, either to change the
1778 nature of the match, or to cause it fail earlier than it otherwise might, when
1779 the author of the pattern knows there is no point in carrying on.
1780 .P
1781 Consider, for example, the pattern \ed+foo when applied to the subject line
1782 .sp
1783 123456bar
1784 .sp
1785 After matching all 6 digits and then failing to match "foo", the normal
1786 action of the matcher is to try again with only 5 digits matching the \ed+
1787 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1788 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1789 that once a subpattern has matched, it is not to be re-evaluated in this way.
1790 .P
1791 If we use atomic grouping for the previous example, the matcher gives up
1792 immediately on failing to match "foo" the first time. The notation is a kind of
1793 special parenthesis, starting with (?> as in this example:
1794 .sp
1795 (?>\ed+)foo
1796 .sp
1797 This kind of parenthesis "locks up" the part of the pattern it contains once
1798 it has matched, and a failure further into the pattern is prevented from
1799 backtracking into it. Backtracking past it to previous items, however, works as
1800 normal.
1801 .P
1802 An alternative description is that a subpattern of this type matches the string
1803 of characters that an identical standalone pattern would match, if anchored at
1804 the current point in the subject string.
1805 .P
1806 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1807 the above example can be thought of as a maximizing repeat that must swallow
1808 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1809 number of digits they match in order to make the rest of the pattern match,
1810 (?>\ed+) can only match an entire sequence of digits.
1811 .P
1812 Atomic groups in general can of course contain arbitrarily complicated
1813 subpatterns, and can be nested. However, when the subpattern for an atomic
1814 group is just a single repeated item, as in the example above, a simpler
1815 notation, called a "possessive quantifier" can be used. This consists of an
1816 additional + character following a quantifier. Using this notation, the
1817 previous example can be rewritten as
1818 .sp
1819 \ed++foo
1820 .sp
1821 Note that a possessive quantifier can be used with an entire group, for
1822 example:
1823 .sp
1824 (abc|xyz){2,3}+
1825 .sp
1826 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1827 option is ignored. They are a convenient notation for the simpler forms of
1828 atomic group. However, there is no difference in the meaning of a possessive
1829 quantifier and the equivalent atomic group, though there may be a performance
1830 difference; possessive quantifiers should be slightly faster.
1831 .P
1832 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1833 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1834 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1835 package, and PCRE copied it from there. It ultimately found its way into Perl
1836 at release 5.10.
1837 .P
1838 PCRE has an optimization that automatically "possessifies" certain simple
1839 pattern constructs. For example, the sequence A+B is treated as A++B because
1840 there is no point in backtracking into a sequence of A's when B must follow.
1841 .P
1842 When a pattern contains an unlimited repeat inside a subpattern that can itself
1843 be repeated an unlimited number of times, the use of an atomic group is the
1844 only way to avoid some failing matches taking a very long time indeed. The
1845 pattern
1846 .sp
1847 (\eD+|<\ed+>)*[!?]
1848 .sp
1849 matches an unlimited number of substrings that either consist of non-digits, or
1850 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1851 quickly. However, if it is applied to
1852 .sp
1853 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1854 .sp
1855 it takes a long time before reporting failure. This is because the string can
1856 be divided between the internal \eD+ repeat and the external * repeat in a
1857 large number of ways, and all have to be tried. (The example uses [!?] rather
1858 than a single character at the end, because both PCRE and Perl have an
1859 optimization that allows for fast failure when a single character is used. They
1860 remember the last single character that is required for a match, and fail early
1861 if it is not present in the string.) If the pattern is changed so that it uses
1862 an atomic group, like this:
1863 .sp
1864 ((?>\eD+)|<\ed+>)*[!?]
1865 .sp
1866 sequences of non-digits cannot be broken, and failure happens quickly.
1867 .
1868 .
1869 .\" HTML <a name="backreferences"></a>
1871 .rs
1872 .sp
1873 Outside a character class, a backslash followed by a digit greater than 0 (and
1874 possibly further digits) is a back reference to a capturing subpattern earlier
1875 (that is, to its left) in the pattern, provided there have been that many
1876 previous capturing left parentheses.
1877 .P
1878 However, if the decimal number following the backslash is less than 10, it is
1879 always taken as a back reference, and causes an error only if there are not
1880 that many capturing left parentheses in the entire pattern. In other words, the
1881 parentheses that are referenced need not be to the left of the reference for
1882 numbers less than 10. A "forward back reference" of this type can make sense
1883 when a repetition is involved and the subpattern to the right has participated
1884 in an earlier iteration.
1885 .P
1886 It is not possible to have a numerical "forward back reference" to a subpattern
1887 whose number is 10 or more using this syntax because a sequence such as \e50 is
1888 interpreted as a character defined in octal. See the subsection entitled
1889 "Non-printing characters"
1890 .\" HTML <a href="#digitsafterbackslash">
1891 .\" </a>
1892 above
1893 .\"
1894 for further details of the handling of digits following a backslash. There is
1895 no such problem when named parentheses are used. A back reference to any
1896 subpattern is possible using named parentheses (see below).
1897 .P
1898 Another way of avoiding the ambiguity inherent in the use of digits following a
1899 backslash is to use the \eg escape sequence. This escape must be followed by an
1900 unsigned number or a negative number, optionally enclosed in braces. These
1901 examples are all identical:
1902 .sp
1903 (ring), \e1
1904 (ring), \eg1
1905 (ring), \eg{1}
1906 .sp
1907 An unsigned number specifies an absolute reference without the ambiguity that
1908 is present in the older syntax. It is also useful when literal digits follow
1909 the reference. A negative number is a relative reference. Consider this
1910 example:
1911 .sp
1912 (abc(def)ghi)\eg{-1}
1913 .sp
1914 The sequence \eg{-1} is a reference to the most recently started capturing
1915 subpattern before \eg, that is, is it equivalent to \e2 in this example.
1916 Similarly, \eg{-2} would be equivalent to \e1. The use of relative references
1917 can be helpful in long patterns, and also in patterns that are created by
1918 joining together fragments that contain references within themselves.
1919 .P
1920 A back reference matches whatever actually matched the capturing subpattern in
1921 the current subject string, rather than anything matching the subpattern
1922 itself (see
1923 .\" HTML <a href="#subpatternsassubroutines">
1924 .\" </a>
1925 "Subpatterns as subroutines"
1926 .\"
1927 below for a way of doing that). So the pattern
1928 .sp
1929 (sens|respons)e and \e1ibility
1930 .sp
1931 matches "sense and sensibility" and "response and responsibility", but not
1932 "sense and responsibility". If caseful matching is in force at the time of the
1933 back reference, the case of letters is relevant. For example,
1934 .sp
1935 ((?i)rah)\es+\e1
1936 .sp
1937 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1938 capturing subpattern is matched caselessly.
1939 .P
1940 There are several different ways of writing back references to named
1941 subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
1942 \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1943 back reference syntax, in which \eg can be used for both numeric and named
1944 references, is also supported. We could rewrite the above example in any of
1945 the following ways:
1946 .sp
1947 (?<p1>(?i)rah)\es+\ek<p1>
1948 (?'p1'(?i)rah)\es+\ek{p1}
1949 (?P<p1>(?i)rah)\es+(?P=p1)
1950 (?<p1>(?i)rah)\es+\eg{p1}
1951 .sp
1952 A subpattern that is referenced by name may appear in the pattern before or
1953 after the reference.
1954 .P
1955 There may be more than one back reference to the same subpattern. If a
1956 subpattern has not actually been used in a particular match, any back
1957 references to it always fail by default. For example, the pattern
1958 .sp
1959 (a|(bc))\e2
1960 .sp
1961 always fails if it starts to match "a" rather than "bc". However, if the
1962 PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
1963 unset value matches an empty string.
1964 .P
1965 Because there may be many capturing parentheses in a pattern, all digits
1966 following a backslash are taken as part of a potential back reference number.
1967 If the pattern continues with a digit character, some delimiter must be used to
1968 terminate the back reference. If the PCRE_EXTENDED option is set, this can be
1969 white space. Otherwise, the \eg{ syntax or an empty comment (see
1970 .\" HTML <a href="#comments">
1971 .\" </a>
1972 "Comments"
1973 .\"
1974 below) can be used.
1975 .
1976 .SS "Recursive back references"
1977 .rs
1978 .sp
1979 A back reference that occurs inside the parentheses to which it refers fails
1980 when the subpattern is first used, so, for example, (a\e1) never matches.
1981 However, such references can be useful inside repeated subpatterns. For
1982 example, the pattern
1983 .sp
1984 (a|b\e1)+
1985 .sp
1986 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1987 the subpattern, the back reference matches the character string corresponding
1988 to the previous iteration. In order for this to work, the pattern must be such
1989 that the first iteration does not need to match the back reference. This can be
1990 done using alternation, as in the example above, or by a quantifier with a
1991 minimum of zero.
1992 .P
1993 Back references of this type cause the group that they reference to be treated
1994 as an
1995 .\" HTML <a href="#atomicgroup">
1996 .\" </a>
1997 atomic group.
1998 .\"
1999 Once the whole group has been matched, a subsequent matching failure cannot
2000 cause backtracking into the middle of the group.
2001 .
2002 .
2003 .\" HTML <a name="bigassertions"></a>
2005 .rs
2006 .sp
2007 An assertion is a test on the characters following or preceding the current
2008 matching point that does not actually consume any characters. The simple
2009 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
2010 .\" HTML <a href="#smallassertions">
2011 .\" </a>
2012 above.
2013 .\"
2014 .P
2015 More complicated assertions are coded as subpatterns. There are two kinds:
2016 those that look ahead of the current position in the subject string, and those
2017 that look behind it. An assertion subpattern is matched in the normal way,
2018 except that it does not cause the current matching position to be changed.
2019 .P
2020 Assertion subpatterns are not capturing subpatterns. If such an assertion
2021 contains capturing subpatterns within it, these are counted for the purposes of
2022 numbering the capturing subpatterns in the whole pattern. However, substring
2023 capturing is carried out only for positive assertions. (Perl sometimes, but not
2024 always, does do capturing in negative assertions.)
2025 .P
2026 For compatibility with Perl, assertion subpatterns may be repeated; though
2027 it makes no sense to assert the same thing several times, the side effect of
2028 capturing parentheses may occasionally be useful. In practice, there only three
2029 cases:
2030 .sp
2031 (1) If the quantifier is {0}, the assertion is never obeyed during matching.
2032 However, it may contain internal capturing parenthesized groups that are called
2033 from elsewhere via the
2034 .\" HTML <a href="#subpatternsassubroutines">
2035 .\" </a>
2036 subroutine mechanism.
2037 .\"
2038 .sp
2039 (2) If quantifier is {0,n} where n is greater than zero, it is treated as if it
2040 were {0,1}. At run time, the rest of the pattern match is tried with and
2041 without the assertion, the order depending on the greediness of the quantifier.
2042 .sp
2043 (3) If the minimum repetition is greater than zero, the quantifier is ignored.
2044 The assertion is obeyed just once when encountered during matching.
2045 .
2046 .
2047 .SS "Lookahead assertions"
2048 .rs
2049 .sp
2050 Lookahead assertions start with (?= for positive assertions and (?! for
2051 negative assertions. For example,
2052 .sp
2053 \ew+(?=;)
2054 .sp
2055 matches a word followed by a semicolon, but does not include the semicolon in
2056 the match, and
2057 .sp
2058 foo(?!bar)
2059 .sp
2060 matches any occurrence of "foo" that is not followed by "bar". Note that the
2061 apparently similar pattern
2062 .sp
2063 (?!foo)bar
2064 .sp
2065 does not find an occurrence of "bar" that is preceded by something other than
2066 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
2067 (?!foo) is always true when the next three characters are "bar". A
2068 lookbehind assertion is needed to achieve the other effect.
2069 .P
2070 If you want to force a matching failure at some point in a pattern, the most
2071 convenient way to do it is with (?!) because an empty string always matches, so
2072 an assertion that requires there not to be an empty string must always fail.
2073 The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
2074 .
2075 .
2076 .\" HTML <a name="lookbehind"></a>
2077 .SS "Lookbehind assertions"
2078 .rs
2079 .sp
2080 Lookbehind assertions start with (?<= for positive assertions and (?<! for
2081 negative assertions. For example,
2082 .sp
2083 (?<!foo)bar
2084 .sp
2085 does find an occurrence of "bar" that is not preceded by "foo". The contents of
2086 a lookbehind assertion are restricted such that all the strings it matches must
2087 have a fixed length. However, if there are several top-level alternatives, they
2088 do not all have to have the same fixed length. Thus
2089 .sp
2090 (?<=bullock|donkey)
2091 .sp
2092 is permitted, but
2093 .sp
2094 (?<!dogs?|cats?)
2095 .sp
2096 causes an error at compile time. Branches that match different length strings
2097 are permitted only at the top level of a lookbehind assertion. This is an
2098 extension compared with Perl, which requires all branches to match the same
2099 length of string. An assertion such as
2100 .sp
2101 (?<=ab(c|de))
2102 .sp
2103 is not permitted, because its single top-level branch can match two different
2104 lengths, but it is acceptable to PCRE if rewritten to use two top-level
2105 branches:
2106 .sp
2107 (?<=abc|abde)
2108 .sp
2109 In some cases, the escape sequence \eK
2110 .\" HTML <a href="#resetmatchstart">
2111 .\" </a>
2112 (see above)
2113 .\"
2114 can be used instead of a lookbehind assertion to get round the fixed-length
2115 restriction.
2116 .P
2117 The implementation of lookbehind assertions is, for each alternative, to
2118 temporarily move the current position back by the fixed length and then try to
2119 match. If there are insufficient characters before the current position, the
2120 assertion fails.
2121 .P
2122 In a UTF mode, PCRE does not allow the \eC escape (which matches a single data
2123 unit even in a UTF mode) to appear in lookbehind assertions, because it makes
2124 it impossible to calculate the length of the lookbehind. The \eX and \eR
2125 escapes, which can match different numbers of data units, are also not
2126 permitted.
2127 .P
2128 .\" HTML <a href="#subpatternsassubroutines">
2129 .\" </a>
2130 "Subroutine"
2131 .\"
2132 calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
2133 as the subpattern matches a fixed-length string.
2134 .\" HTML <a href="#recursion">
2135 .\" </a>
2136 Recursion,
2137 .\"
2138 however, is not supported.
2139 .P
2140 Possessive quantifiers can be used in conjunction with lookbehind assertions to
2141 specify efficient matching of fixed-length strings at the end of subject
2142 strings. Consider a simple pattern such as
2143 .sp
2144 abcd$
2145 .sp
2146 when applied to a long string that does not match. Because matching proceeds
2147 from left to right, PCRE will look for each "a" in the subject and then see if
2148 what follows matches the rest of the pattern. If the pattern is specified as
2149 .sp
2150 ^.*abcd$
2151 .sp
2152 the initial .* matches the entire string at first, but when this fails (because
2153 there is no following "a"), it backtracks to match all but the last character,
2154 then all but the last two characters, and so on. Once again the search for "a"
2155 covers the entire string, from right to left, so we are no better off. However,
2156 if the pattern is written as
2157 .sp
2158 ^.*+(?<=abcd)
2159 .sp
2160 there can be no backtracking for the .*+ item; it can match only the entire
2161 string. The subsequent lookbehind assertion does a single test on the last four
2162 characters. If it fails, the match fails immediately. For long strings, this
2163 approach makes a significant difference to the processing time.
2164 .
2165 .
2166 .SS "Using multiple assertions"
2167 .rs
2168 .sp
2169 Several assertions (of any sort) may occur in succession. For example,
2170 .sp
2171 (?<=\ed{3})(?<!999)foo
2172 .sp
2173 matches "foo" preceded by three digits that are not "999". Notice that each of
2174 the assertions is applied independently at the same point in the subject
2175 string. First there is a check that the previous three characters are all
2176 digits, and then there is a check that the same three characters are not "999".
2177 This pattern does \fInot\fP match "foo" preceded by six characters, the first
2178 of which are digits and the last three of which are not "999". For example, it
2179 doesn't match "123abcfoo". A pattern to do that is
2180 .sp
2181 (?<=\ed{3}...)(?<!999)foo
2182 .sp
2183 This time the first assertion looks at the preceding six characters, checking
2184 that the first three are digits, and then the second assertion checks that the
2185 preceding three characters are not "999".
2186 .P
2187 Assertions can be nested in any combination. For example,
2188 .sp
2189 (?<=(?<!foo)bar)baz
2190 .sp
2191 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
2192 preceded by "foo", while
2193 .sp
2194 (?<=\ed{3}(?!999)...)foo
2195 .sp
2196 is another pattern that matches "foo" preceded by three digits and any three
2197 characters that are not "999".
2198 .
2199 .
2200 .\" HTML <a name="conditions"></a>
2202 .rs
2203 .sp
2204 It is possible to cause the matching process to obey a subpattern
2205 conditionally or to choose between two alternative subpatterns, depending on
2206 the result of an assertion, or whether a specific capturing subpattern has
2207 already been matched. The two possible forms of conditional subpattern are:
2208 .sp
2209 (?(condition)yes-pattern)
2210 (?(condition)yes-pattern|no-pattern)
2211 .sp
2212 If the condition is satisfied, the yes-pattern is used; otherwise the
2213 no-pattern (if present) is used. If there are more than two alternatives in the
2214 subpattern, a compile-time error occurs. Each of the two alternatives may
2215 itself contain nested subpatterns of any form, including conditional
2216 subpatterns; the restriction to two alternatives applies only at the level of
2217 the condition. This pattern fragment is an example where the alternatives are
2218 complex:
2219 .sp
2220 (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
2221 .sp
2222 .P
2223 There are four kinds of condition: references to subpatterns, references to
2224 recursion, a pseudo-condition called DEFINE, and assertions.
2225 .
2226 .SS "Checking for a used subpattern by number"
2227 .rs
2228 .sp
2229 If the text between the parentheses consists of a sequence of digits, the
2230 condition is true if a capturing subpattern of that number has previously
2231 matched. If there is more than one capturing subpattern with the same number
2232 (see the earlier
2233 .\"
2234 .\" HTML <a href="#recursion">
2235 .\" </a>
2236 section about duplicate subpattern numbers),
2237 .\"
2238 the condition is true if any of them have matched. An alternative notation is
2239 to precede the digits with a plus or minus sign. In this case, the subpattern
2240 number is relative rather than absolute. The most recently opened parentheses
2241 can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
2242 loops it can also make sense to refer to subsequent groups. The next
2243 parentheses to be opened can be referenced as (?(+1), and so on. (The value
2244 zero in any of these forms is not used; it provokes a compile-time error.)
2245 .P
2246 Consider the following pattern, which contains non-significant white space to
2247 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2248 three parts for ease of discussion:
2249 .sp
2250 ( \e( )? [^()]+ (?(1) \e) )
2251 .sp
2252 The first part matches an optional opening parenthesis, and if that
2253 character is present, sets it as the first captured substring. The second part
2254 matches one or more characters that are not parentheses. The third part is a
2255 conditional subpattern that tests whether or not the first set of parentheses
2256 matched. If they did, that is, if subject started with an opening parenthesis,
2257 the condition is true, and so the yes-pattern is executed and a closing
2258 parenthesis is required. Otherwise, since no-pattern is not present, the
2259 subpattern matches nothing. In other words, this pattern matches a sequence of
2260 non-parentheses, optionally enclosed in parentheses.
2261 .P
2262 If you were embedding this pattern in a larger one, you could use a relative
2263 reference:
2264 .sp
2265 ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
2266 .sp
2267 This makes the fragment independent of the parentheses in the larger pattern.
2268 .
2269 .SS "Checking for a used subpattern by name"
2270 .rs
2271 .sp
2272 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
2273 subpattern by name. For compatibility with earlier versions of PCRE, which had
2274 this facility before Perl, the syntax (?(name)...) is also recognized. However,
2275 there is a possible ambiguity with this syntax, because subpattern names may
2276 consist entirely of digits. PCRE looks first for a named subpattern; if it
2277 cannot find one and the name consists entirely of digits, PCRE looks for a
2278 subpattern of that number, which must be greater than zero. Using subpattern
2279 names that consist entirely of digits is not recommended.
2280 .P
2281 Rewriting the above example to use a named subpattern gives this:
2282 .sp
2283 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
2284 .sp
2285 If the name used in a condition of this kind is a duplicate, the test is
2286 applied to all subpatterns of the same name, and is true if any one of them has
2287 matched.
2288 .
2289 .SS "Checking for pattern recursion"
2290 .rs
2291 .sp
2292 If the condition is the string (R), and there is no subpattern with the name R,
2293 the condition is true if a recursive call to the whole pattern or any
2294 subpattern has been made. If digits or a name preceded by ampersand follow the
2295 letter R, for example:
2296 .sp
2297 (?(R3)...) or (?(R&name)...)
2298 .sp
2299 the condition is true if the most recent recursion is into a subpattern whose
2300 number or name is given. This condition does not check the entire recursion
2301 stack. If the name used in a condition of this kind is a duplicate, the test is
2302 applied to all subpatterns of the same name, and is true if any one of them is
2303 the most recent recursion.
2304 .P
2305 At "top level", all these recursion test conditions are false.
2306 .\" HTML <a href="#recursion">
2307 .\" </a>
2308 The syntax for recursive patterns
2309 .\"
2310 is described below.
2311 .
2312 .\" HTML <a name="subdefine"></a>
2313 .SS "Defining subpatterns for use by reference only"
2314 .rs
2315 .sp
2316 If the condition is the string (DEFINE), and there is no subpattern with the
2317 name DEFINE, the condition is always false. In this case, there may be only one
2318 alternative in the subpattern. It is always skipped if control reaches this
2319 point in the pattern; the idea of DEFINE is that it can be used to define
2320 subroutines that can be referenced from elsewhere. (The use of
2321 .\" HTML <a href="#subpatternsassubroutines">
2322 .\" </a>
2323 subroutines
2324 .\"
2325 is described below.) For example, a pattern to match an IPv4 address such as
2326 "" could be written like this (ignore white space and line
2327 breaks):
2328 .sp
2329 (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
2330 \eb (?&byte) (\e.(?&byte)){3} \eb
2331 .sp
2332 The first part of the pattern is a DEFINE group inside which a another group
2333 named "byte" is defined. This matches an individual component of an IPv4
2334 address (a number less than 256). When matching takes place, this part of the
2335 pattern is skipped because DEFINE acts like a false condition. The rest of the
2336 pattern uses references to the named group to match the four dot-separated
2337 components of an IPv4 address, insisting on a word boundary at each end.
2338 .
2339 .SS "Assertion conditions"
2340 .rs
2341 .sp
2342 If the condition is not in any of the above formats, it must be an assertion.
2343 This may be a positive or negative lookahead or lookbehind assertion. Consider
2344 this pattern, again containing non-significant white space, and with the two
2345 alternatives on the second line:
2346 .sp
2347 (?(?=[^a-z]*[a-z])
2348 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
2349 .sp
2350 The condition is a positive lookahead assertion that matches an optional
2351 sequence of non-letters followed by a letter. In other words, it tests for the
2352 presence of at least one letter in the subject. If a letter is found, the
2353 subject is matched against the first alternative; otherwise it is matched
2354 against the second. This pattern matches strings in one of the two forms
2355 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2356 .
2357 .
2358 .\" HTML <a name="comments"></a>
2360 .rs
2361 .sp
2362 There are two ways of including comments in patterns that are processed by
2363 PCRE. In both cases, the start of the comment must not be in a character class,
2364 nor in the middle of any other sequence of related characters such as (?: or a
2365 subpattern name or number. The characters that make up a comment play no part
2366 in the pattern matching.
2367 .P
2368 The sequence (?# marks the start of a comment that continues up to the next
2369 closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2370 option is set, an unescaped # character also introduces a comment, which in
2371 this case continues to immediately after the next newline character or
2372 character sequence in the pattern. Which characters are interpreted as newlines
2373 is controlled by the options passed to a compiling function or by a special
2374 sequence at the start of the pattern, as described in the section entitled
2375 .\" HTML <a href="#newlines">
2376 .\" </a>
2377 "Newline conventions"
2378 .\"
2379 above. Note that the end of this type of comment is a literal newline sequence
2380 in the pattern; escape sequences that happen to represent a newline do not
2381 count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2382 default newline convention is in force:
2383 .sp
2384 abc #comment \en still comment
2385 .sp
2386 On encountering the # character, \fBpcre_compile()\fP skips along, looking for
2387 a newline in the pattern. The sequence \en is still literal at this stage, so
2388 it does not terminate the comment. Only an actual character with the code value
2389 0x0a (the default newline) does so.
2390 .
2391 .
2392 .\" HTML <a name="recursion"></a>
2394 .rs
2395 .sp
2396 Consider the problem of matching a string in parentheses, allowing for
2397 unlimited nested parentheses. Without the use of recursion, the best that can
2398 be done is to use a pattern that matches up to some fixed depth of nesting. It
2399 is not possible to handle an arbitrary nesting depth.
2400 .P
2401 For some time, Perl has provided a facility that allows regular expressions to
2402 recurse (amongst other things). It does this by interpolating Perl code in the
2403 expression at run time, and the code can refer to the expression itself. A Perl
2404 pattern using code interpolation to solve the parentheses problem can be
2405 created like this:
2406 .sp
2407 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
2408 .sp
2409 The (?p{...}) item interpolates Perl code at run time, and in this case refers
2410 recursively to the pattern in which it appears.
2411 .P
2412 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2413 supports special syntax for recursion of the entire pattern, and also for
2414 individual subpattern recursion. After its introduction in PCRE and Python,
2415 this kind of recursion was subsequently introduced into Perl at release 5.10.
2416 .P
2417 A special item that consists of (? followed by a number greater than zero and a
2418 closing parenthesis is a recursive subroutine call of the subpattern of the
2419 given number, provided that it occurs inside that subpattern. (If not, it is a
2420 .\" HTML <a href="#subpatternsassubroutines">
2421 .\" </a>
2422 non-recursive subroutine
2423 .\"
2424 call, which is described in the next section.) The special item (?R) or (?0) is
2425 a recursive call of the entire regular expression.
2426 .P
2427 This PCRE pattern solves the nested parentheses problem (assume the
2428 PCRE_EXTENDED option is set so that white space is ignored):
2429 .sp
2430 \e( ( [^()]++ | (?R) )* \e)
2431 .sp
2432 First it matches an opening parenthesis. Then it matches any number of
2433 substrings which can either be a sequence of non-parentheses, or a recursive
2434 match of the pattern itself (that is, a correctly parenthesized substring).
2435 Finally there is a closing parenthesis. Note the use of a possessive quantifier
2436 to avoid backtracking into sequences of non-parentheses.
2437 .P
2438 If this were part of a larger pattern, you would not want to recurse the entire
2439 pattern, so instead you could use this:
2440 .sp
2441 ( \e( ( [^()]++ | (?1) )* \e) )
2442 .sp
2443 We have put the pattern into parentheses, and caused the recursion to refer to
2444 them instead of the whole pattern.
2445 .P
2446 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2447 is made easier by the use of relative references. Instead of (?1) in the
2448 pattern above you can write (?-2) to refer to the second most recently opened
2449 parentheses preceding the recursion. In other words, a negative number counts
2450 capturing parentheses leftwards from the point at which it is encountered.
2451 .P
2452 It is also possible to refer to subsequently opened parentheses, by writing
2453 references such as (?+2). However, these cannot be recursive because the
2454 reference is not inside the parentheses that are referenced. They are always
2455 .\" HTML <a href="#subpatternsassubroutines">
2456 .\" </a>
2457 non-recursive subroutine
2458 .\"
2459 calls, as described in the next section.
2460 .P
2461 An alternative approach is to use named parentheses instead. The Perl syntax
2462 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
2463 could rewrite the above example as follows:
2464 .sp
2465 (?<pn> \e( ( [^()]++ | (?&pn) )* \e) )
2466 .sp
2467 If there is more than one subpattern with the same name, the earliest one is
2468 used.
2469 .P
2470 This particular example pattern that we have been looking at contains nested
2471 unlimited repeats, and so the use of a possessive quantifier for matching
2472 strings of non-parentheses is important when applying the pattern to strings
2473 that do not match. For example, when this pattern is applied to
2474 .sp
2475 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2476 .sp
2477 it yields "no match" quickly. However, if a possessive quantifier is not used,
2478 the match runs for a very long time indeed because there are so many different
2479 ways the + and * repeats can carve up the subject, and all have to be tested
2480 before failure can be reported.
2481 .P
2482 At the end of a match, the values of capturing parentheses are those from
2483 the outermost level. If you want to obtain intermediate values, a callout
2484 function can be used (see below and the
2485 .\" HREF
2486 \fBpcrecallout\fP
2487 .\"
2488 documentation). If the pattern above is matched against
2489 .sp
2490 (ab(cd)ef)
2491 .sp
2492 the value for the inner capturing parentheses (numbered 2) is "ef", which is
2493 the last value taken on at the top level. If a capturing subpattern is not
2494 matched at the top level, its final captured value is unset, even if it was
2495 (temporarily) set at a deeper level during the matching process.
2496 .P
2497 If there are more than 15 capturing parentheses in a pattern, PCRE has to
2498 obtain extra memory to store data during a recursion, which it does by using
2499 \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no memory can
2500 be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2501 .P
2502 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2503 Consider this pattern, which matches text in angle brackets, allowing for
2504 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2505 recursing), whereas any characters are permitted at the outer level.
2506 .sp
2507 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
2508 .sp
2509 In this pattern, (?(R) is the start of a conditional subpattern, with two
2510 different alternatives for the recursive and non-recursive cases. The (?R) item
2511 is the actual recursive call.
2512 .
2513 .
2514 .\" HTML <a name="recursiondifference"></a>
2515 .SS "Differences in recursion processing between PCRE and Perl"
2516 .rs
2517 .sp
2518 Recursion processing in PCRE differs from Perl in two important ways. In PCRE
2519 (like Python, but unlike Perl), a recursive subpattern call is always treated
2520 as an atomic group. That is, once it has matched some of the subject string, it
2521 is never re-entered, even if it contains untried alternatives and there is a
2522 subsequent matching failure. This can be illustrated by the following pattern,
2523 which purports to match a palindromic string that contains an odd number of
2524 characters (for example, "a", "aba", "abcba", "abcdcba"):
2525 .sp
2526 ^(.|(.)(?1)\e2)$
2527 .sp
2528 The idea is that it either matches a single character, or two identical
2529 characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2530 it does not if the pattern is longer than three characters. Consider the
2531 subject string "abcba":
2532 .P
2533 At the top level, the first character is matched, but as it is not at the end
2534 of the string, the first alternative fails; the second alternative is taken
2535 and the recursion kicks in. The recursive call to subpattern 1 successfully
2536 matches the next character ("b"). (Note that the beginning and end of line
2537 tests are not part of the recursion).
2538 .P
2539 Back at the top level, the next character ("c") is compared with what
2540 subpattern 2 matched, which was "a". This fails. Because the recursion is
2541 treated as an atomic group, there are now no backtracking points, and so the
2542 entire match fails. (Perl is able, at this point, to re-enter the recursion and
2543 try the second alternative.) However, if the pattern is written with the
2544 alternatives in the other order, things are different:
2545 .sp
2546 ^((.)(?1)\e2|.)$
2547 .sp
2548 This time, the recursing alternative is tried first, and continues to recurse
2549 until it runs out of characters, at which point the recursion fails. But this
2550 time we do have another alternative to try at the higher level. That is the big
2551 difference: in the previous case the remaining alternative is at a deeper
2552 recursion level, which PCRE cannot use.
2553 .P
2554 To change the pattern so that it matches all palindromic strings, not just
2555 those with an odd number of characters, it is tempting to change the pattern to
2556 this:
2557 .sp
2558 ^((.)(?1)\e2|.?)$
2559 .sp
2560 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2561 deeper recursion has matched a single character, it cannot be entered again in
2562 order to match an empty string. The solution is to separate the two cases, and
2563 write out the odd and even cases as alternatives at the higher level:
2564 .sp
2565 ^(?:((.)(?1)\e2|)|((.)(?3)\e4|.))
2566 .sp
2567 If you want to match typical palindromic phrases, the pattern has to ignore all
2568 non-word characters, which can be done like this:
2569 .sp
2570 ^\eW*+(?:((.)\eW*+(?1)\eW*+\e2|)|((.)\eW*+(?3)\eW*+\e4|\eW*+.\eW*+))\eW*+$
2571 .sp
2572 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2573 man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2574 the use of the possessive quantifier *+ to avoid backtracking into sequences of
2575 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2576 more) to match typical phrases, and Perl takes so long that you think it has
2577 gone into a loop.
2578 .P
2579 \fBWARNING\fP: The palindrome-matching patterns above work only if the subject
2580 string does not start with a palindrome that is shorter than the entire string.
2581 For example, although "abcba" is correctly matched, if the subject is "ababa",
2582 PCRE finds the palindrome "aba" at the start, then fails at top level because
2583 the end of the string does not follow. Once again, it cannot jump back into the
2584 recursion to try other alternatives, so the entire match fails.
2585 .P
2586 The second way in which PCRE and Perl differ in their recursion processing is
2587 in the handling of captured values. In Perl, when a subpattern is called
2588 recursively or as a subpattern (see the next section), it has no access to any
2589 values that were captured outside the recursion, whereas in PCRE these values
2590 can be referenced. Consider this pattern:
2591 .sp
2592 ^(.)(\e1|a(?2))
2593 .sp
2594 In PCRE, this pattern matches "bab". The first capturing parentheses match "b",
2595 then in the second group, when the back reference \e1 fails to match "b", the
2596 second alternative matches "a" and then recurses. In the recursion, \e1 does
2597 now match "b" and so the whole match succeeds. In Perl, the pattern fails to
2598 match because inside the recursive call \e1 cannot access the externally set
2599 value.
2600 .
2601 .
2602 .\" HTML <a name="subpatternsassubroutines"></a>
2604 .rs
2605 .sp
2606 If the syntax for a recursive subpattern call (either by number or by
2607 name) is used outside the parentheses to which it refers, it operates like a
2608 subroutine in a programming language. The called subpattern may be defined
2609 before or after the reference. A numbered reference can be absolute or
2610 relative, as in these examples:
2611 .sp
2612 (...(absolute)...)...(?2)...
2613 (...(relative)...)...(?-1)...
2614 (...(?+1)...(relative)...
2615 .sp
2616 An earlier example pointed out that the pattern
2617 .sp
2618 (sens|respons)e and \e1ibility
2619 .sp
2620 matches "sense and sensibility" and "response and responsibility", but not
2621 "sense and responsibility". If instead the pattern
2622 .sp
2623 (sens|respons)e and (?1)ibility
2624 .sp
2625 is used, it does match "sense and responsibility" as well as the other two
2626 strings. Another example is given in the discussion of DEFINE above.
2627 .P
2628 All subroutine calls, whether recursive or not, are always treated as atomic
2629 groups. That is, once a subroutine has matched some of the subject string, it
2630 is never re-entered, even if it contains untried alternatives and there is a
2631 subsequent matching failure. Any capturing parentheses that are set during the
2632 subroutine call revert to their previous values afterwards.
2633 .P
2634 Processing options such as case-independence are fixed when a subpattern is
2635 defined, so if it is used as a subroutine, such options cannot be changed for
2636 different calls. For example, consider this pattern:
2637 .sp
2638 (abc)(?i:(?-1))
2639 .sp
2640 It matches "abcabc". It does not match "abcABC" because the change of
2641 processing option does not affect the called subpattern.
2642 .
2643 .
2644 .\" HTML <a name="onigurumasubroutines"></a>
2646 .rs
2647 .sp
2648 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
2649 a number enclosed either in angle brackets or single quotes, is an alternative
2650 syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2651 are two of the examples used above, rewritten using this syntax:
2652 .sp
2653 (?<pn> \e( ( (?>[^()]+) | \eg<pn> )* \e) )
2654 (sens|respons)e and \eg'1'ibility
2655 .sp
2656 PCRE supports an extension to Oniguruma: if a number is preceded by a
2657 plus or a minus sign it is taken as a relative reference. For example:
2658 .sp
2659 (abc)(?i:\eg<-1>)
2660 .sp
2661 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
2662 synonymous. The former is a back reference; the latter is a subroutine call.
2663 .
2664 .
2666 .rs
2667 .sp
2668 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2669 code to be obeyed in the middle of matching a regular expression. This makes it
2670 possible, amongst other things, to extract different substrings that match the
2671 same pair of parentheses when there is a repetition.
2672 .P
2673 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2674 code. The feature is called "callout". The caller of PCRE provides an external
2675 function by putting its entry point in the global variable \fIpcre_callout\fP
2676 (8-bit library) or \fIpcre[16|32]_callout\fP (16-bit or 32-bit library).
2677 By default, this variable contains NULL, which disables all calling out.
2678 .P
2679 Within a regular expression, (?C) indicates the points at which the external
2680 function is to be called. If you want to identify different callout points, you
2681 can put a number less than 256 after the letter C. The default value is zero.
2682 For example, this pattern has two callout points:
2683 .sp
2684 (?C1)abc(?C2)def
2685 .sp
2686 If the PCRE_AUTO_CALLOUT flag is passed to a compiling function, callouts are
2687 automatically installed before each item in the pattern. They are all numbered
2688 255. If there is a conditional group in the pattern whose condition is an
2689 assertion, an additional callout is inserted just before the condition. An
2690 explicit callout may also be set at this position, as in this example:
2691 .sp
2692 (?(?C9)(?=a)abc|def)
2693 .sp
2694 Note that this applies only to assertion conditions, not to other types of
2695 condition.
2696 .P
2697 During matching, when PCRE reaches a callout point, the external function is
2698 called. It is provided with the number of the callout, the position in the
2699 pattern, and, optionally, one item of data originally supplied by the caller of
2700 the matching function. The callout function may cause matching to proceed, to
2701 backtrack, or to fail altogether. A complete description of the interface to
2702 the callout function is given in the
2703 .\" HREF
2704 \fBpcrecallout\fP
2705 .\"
2706 documentation.
2707 .
2708 .
2709 .\" HTML <a name="backtrackcontrol"></a>
2711 .rs
2712 .sp
2713 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2714 are still described in the Perl documentation as "experimental and subject to
2715 change or removal in a future version of Perl". It goes on to say: "Their usage
2716 in production code should be noted to avoid problems during upgrades." The same
2717 remarks apply to the PCRE features described in this section.
2718 .P
2719 The new verbs make use of what was previously invalid syntax: an opening
2720 parenthesis followed by an asterisk. They are generally of the form
2721 (*VERB) or (*VERB:NAME). Some may take either form, possibly behaving
2722 differently depending on whether or not a name is present. A name is any
2723 sequence of characters that does not include a closing parenthesis. The maximum
2724 length of name is 255 in the 8-bit library and 65535 in the 16-bit and 32-bit
2725 libraries. If the name is empty, that is, if the closing parenthesis
2726 immediately follows the colon, the effect is as if the colon were not there.
2727 Any number of these verbs may occur in a pattern.
2728 .P
2729 Since these verbs are specifically related to backtracking, most of them can be
2730 used only when the pattern is to be matched using one of the traditional
2731 matching functions, because these use a backtracking algorithm. With the
2732 exception of (*FAIL), which behaves like a failing negative assertion, the
2733 backtracking control verbs cause an error if encountered by a DFA matching
2734 function.
2735 .P
2736 The behaviour of these verbs in
2737 .\" HTML <a href="#btrepeat">
2738 .\" </a>
2739 repeated groups,
2740 .\"
2741 .\" HTML <a href="#btassert">
2742 .\" </a>
2743 assertions,
2744 .\"
2745 and in
2746 .\" HTML <a href="#btsub">
2747 .\" </a>
2748 subpatterns called as subroutines
2749 .\"
2750 (whether or not recursively) is documented below.
2751 .
2752 .
2753 .\" HTML <a name="nooptimize"></a>
2754 .SS "Optimizations that affect backtracking verbs"
2755 .rs
2756 .sp
2757 PCRE contains some optimizations that are used to speed up matching by running
2758 some checks at the start of each match attempt. For example, it may know the
2759 minimum length of matching subject, or that a particular character must be
2760 present. When one of these optimizations bypasses the running of a match, any
2761 included backtracking verbs will not, of course, be processed. You can suppress
2762 the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2763 when calling \fBpcre_compile()\fP or \fBpcre_exec()\fP, or by starting the
2764 pattern with (*NO_START_OPT). There is more discussion of this option in the
2765 section entitled
2766 .\" HTML <a href="pcreapi.html#execoptions">
2767 .\" </a>
2768 "Option bits for \fBpcre_exec()\fP"
2769 .\"
2770 in the
2771 .\" HREF
2772 \fBpcreapi\fP
2773 .\"
2774 documentation.
2775 .P
2776 Experiments with Perl suggest that it too has similar optimizations, sometimes
2777 leading to anomalous results.
2778 .
2779 .
2780 .SS "Verbs that act immediately"
2781 .rs
2782 .sp
2783 The following verbs act as soon as they are encountered. They may not be
2784 followed by a name.
2785 .sp
2786 (*ACCEPT)
2787 .sp
2788 This verb causes the match to end successfully, skipping the remainder of the
2789 pattern. However, when it is inside a subpattern that is called as a
2790 subroutine, only that subpattern is ended successfully. Matching then continues
2791 at the outer level. If (*ACCEPT) in triggered in a positive assertion, the
2792 assertion succeeds; in a negative assertion, the assertion fails.
2793 .P
2794 If (*ACCEPT) is inside capturing parentheses, the data so far is captured. For
2795 example:
2796 .sp
2797 A((?:A|B(*ACCEPT)|C)D)
2798 .sp
2799 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2800 the outer parentheses.
2801 .sp
2802 (*FAIL) or (*F)
2803 .sp
2804 This verb causes a matching failure, forcing backtracking to occur. It is
2805 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2806 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2807 Perl features that are not present in PCRE. The nearest equivalent is the
2808 callout feature, as for example in this pattern:
2809 .sp
2810 a+(?C)(*FAIL)
2811 .sp
2812 A match with the string "aaaa" always fails, but the callout is taken before
2813 each backtrack happens (in this example, 10 times).
2814 .
2815 .
2816 .SS "Recording which path was taken"
2817 .rs
2818 .sp
2819 There is one verb whose main purpose is to track how a match was arrived at,
2820 though it also has a secondary use in conjunction with advancing the match
2821 starting point (see (*SKIP) below).
2822 .sp
2823 (*MARK:NAME) or (*:NAME)
2824 .sp
2825 A name is always required with this verb. There may be as many instances of
2826 (*MARK) as you like in a pattern, and their names do not have to be unique.
2827 .P
2828 When a match succeeds, the name of the last-encountered (*MARK:NAME),
2829 (*PRUNE:NAME), or (*THEN:NAME) on the matching path is passed back to the
2830 caller as described in the section entitled
2831 .\" HTML <a href="pcreapi.html#extradata">
2832 .\" </a>
2833 "Extra data for \fBpcre_exec()\fP"
2834 .\"
2835 in the
2836 .\" HREF
2837 \fBpcreapi\fP
2838 .\"
2839 documentation. Here is an example of \fBpcretest\fP output, where the /K
2840 modifier requests the retrieval and outputting of (*MARK) data:
2841 .sp
2842 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2843 data> XY
2844 0: XY
2845 MK: A
2846 XZ
2847 0: XZ
2848 MK: B
2849 .sp
2850 The (*MARK) name is tagged with "MK:" in this output, and in this example it
2851 indicates which of the two alternatives matched. This is a more efficient way
2852 of obtaining this information than putting each alternative in its own
2853 capturing parentheses.
2854 .P
2855 If a verb with a name is encountered in a positive assertion that is true, the
2856 name is recorded and passed back if it is the last-encountered. This does not
2857 happen for negative assertions or failing positive assertions.
2858 .P
2859 After a partial match or a failed match, the last encountered name in the
2860 entire match process is returned. For example:
2861 .sp
2862 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2863 data> XP
2864 No match, mark = B
2865 .sp
2866 Note that in this unanchored example the mark is retained from the match
2867 attempt that started at the letter "X" in the subject. Subsequent match
2868 attempts starting at "P" and then with an empty string do not get as far as the
2869 (*MARK) item, but nevertheless do not reset it.
2870 .P
2871 If you are interested in (*MARK) values after failed matches, you should
2872 probably set the PCRE_NO_START_OPTIMIZE option
2873 .\" HTML <a href="#nooptimize">
2874 .\" </a>
2875 (see above)
2876 .\"
2877 to ensure that the match is always attempted.
2878 .
2879 .
2880 .SS "Verbs that act after backtracking"
2881 .rs
2882 .sp
2883 The following verbs do nothing when they are encountered. Matching continues
2884 with what follows, but if there is no subsequent match, causing a backtrack to
2885 the verb, a failure is forced. That is, backtracking cannot pass to the left of
2886 the verb. However, when one of these verbs appears inside an atomic group or an
2887 assertion that is true, its effect is confined to that group, because once the
2888 group has been matched, there is never any backtracking into it. In this
2889 situation, backtracking can "jump back" to the left of the entire atomic group
2890 or assertion. (Remember also, as stated above, that this localization also
2891 applies in subroutine calls.)
2892 .P
2893 These verbs differ in exactly what kind of failure occurs when backtracking
2894 reaches them. The behaviour described below is what happens when the verb is
2895 not in a subroutine or an assertion. Subsequent sections cover these special
2896 cases.
2897 .sp
2898 (*COMMIT)
2899 .sp
2900 This verb, which may not be followed by a name, causes the whole match to fail
2901 outright if there is a later matching failure that causes backtracking to reach
2902 it. Even if the pattern is unanchored, no further attempts to find a match by
2903 advancing the starting point take place. If (*COMMIT) is the only backtracking
2904 verb that is encountered, once it has been passed \fBpcre_exec()\fP is
2905 committed to finding a match at the current starting point, or not at all. For
2906 example:
2907 .sp
2908 a+(*COMMIT)b
2909 .sp
2910 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2911 dynamic anchor, or "I've started, so I must finish." The name of the most
2912 recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
2913 match failure.
2914 .P
2915 If there is more than one backtracking verb in a pattern, a different one that
2916 follows (*COMMIT) may be triggered first, so merely passing (*COMMIT) during a
2917 match does not always guarantee that a match must be at this starting point.
2918 .P
2919 Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
2920 unless PCRE's start-of-match optimizations are turned off, as shown in this
2921 \fBpcretest\fP example:
2922 .sp
2923 re> /(*COMMIT)abc/
2924 data> xyzabc
2925 0: abc
2926 xyzabc\eY
2927 No match
2928 .sp
2929 PCRE knows that any match must start with "a", so the optimization skips along
2930 the subject to "a" before running the first match attempt, which succeeds. When
2931 the optimization is disabled by the \eY escape in the second subject, the match
2932 starts at "x" and so the (*COMMIT) causes it to fail without trying any other
2933 starting points.
2934 .sp
2935 (*PRUNE) or (*PRUNE:NAME)
2936 .sp
2937 This verb causes the match to fail at the current starting position in the
2938 subject if there is a later matching failure that causes backtracking to reach
2939 it. If the pattern is unanchored, the normal "bumpalong" advance to the next
2940 starting character then happens. Backtracking can occur as usual to the left of
2941 (*PRUNE), before it is reached, or when matching to the right of (*PRUNE), but
2942 if there is no match to the right, backtracking cannot cross (*PRUNE). In
2943 simple cases, the use of (*PRUNE) is just an alternative to an atomic group or
2944 possessive quantifier, but there are some uses of (*PRUNE) that cannot be
2945 expressed in any other way. In an anchored pattern (*PRUNE) has the same effect
2946 as (*COMMIT).
2947 .P
2948 The behaviour of (*PRUNE:NAME) is the not the same as (*MARK:NAME)(*PRUNE).
2949 It is like (*MARK:NAME) in that the name is remembered for passing back to the
2950 caller. However, (*SKIP:NAME) searches only for names set with (*MARK).
2951 .sp
2952 (*SKIP)
2953 .sp
2954 This verb, when given without a name, is like (*PRUNE), except that if the
2955 pattern is unanchored, the "bumpalong" advance is not to the next character,
2956 but to the position in the subject where (*SKIP) was encountered. (*SKIP)
2957 signifies that whatever text was matched leading up to it cannot be part of a
2958 successful match. Consider:
2959 .sp
2960 a+(*SKIP)b
2961 .sp
2962 If the subject is "aaaac...", after the first match attempt fails (starting at
2963 the first character in the string), the starting point skips on to start the
2964 next attempt at "c". Note that a possessive quantifer does not have the same
2965 effect as this example; although it would suppress backtracking during the
2966 first match attempt, the second attempt would start at the second character
2967 instead of skipping on to "c".
2968 .sp
2969 (*SKIP:NAME)
2970 .sp
2971 When (*SKIP) has an associated name, its behaviour is modified. When it is
2972 triggered, the previous path through the pattern is searched for the most
2973 recent (*MARK) that has the same name. If one is found, the "bumpalong" advance
2974 is to the subject position that corresponds to that (*MARK) instead of to where
2975 (*SKIP) was encountered. If no (*MARK) with a matching name is found, the
2976 (*SKIP) is ignored.
2977 .P
2978 Note that (*SKIP:NAME) searches only for names set by (*MARK:NAME). It ignores
2979 names that are set by (*PRUNE:NAME) or (*THEN:NAME).
2980 .sp
2981 (*THEN) or (*THEN:NAME)
2982 .sp
2983 This verb causes a skip to the next innermost alternative when backtracking
2984 reaches it. That is, it cancels any further backtracking within the current
2985 alternative. Its name comes from the observation that it can be used for a
2986 pattern-based if-then-else block:
2987 .sp
2988 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2989 .sp
2990 If the COND1 pattern matches, FOO is tried (and possibly further items after
2991 the end of the group if FOO succeeds); on failure, the matcher skips to the
2992 second alternative and tries COND2, without backtracking into COND1. If that
2993 succeeds and BAR fails, COND3 is tried. If subsequently BAZ fails, there are no
2994 more alternatives, so there is a backtrack to whatever came before the entire
2995 group. If (*THEN) is not inside an alternation, it acts like (*PRUNE).
2996 .P
2997 The behaviour of (*THEN:NAME) is the not the same as (*MARK:NAME)(*THEN).
2998 It is like (*MARK:NAME) in that the name is remembered for passing back to the
2999 caller. However, (*SKIP:NAME) searches only for names set with (*MARK).
3000 .P
3001 A subpattern that does not contain a | character is just a part of the
3002 enclosing alternative; it is not a nested alternation with only one
3003 alternative. The effect of (*THEN) extends beyond such a subpattern to the
3004 enclosing alternative. Consider this pattern, where A, B, etc. are complex
3005 pattern fragments that do not contain any | characters at this level:
3006 .sp
3007 A (B(*THEN)C) | D
3008 .sp
3009 If A and B are matched, but there is a failure in C, matching does not
3010 backtrack into A; instead it moves to the next alternative, that is, D.
3011 However, if the subpattern containing (*THEN) is given an alternative, it
3012 behaves differently:
3013 .sp
3014 A (B(*THEN)C | (*FAIL)) | D
3015 .sp
3016 The effect of (*THEN) is now confined to the inner subpattern. After a failure
3017 in C, matching moves to (*FAIL), which causes the whole subpattern to fail
3018 because there are no more alternatives to try. In this case, matching does now
3019 backtrack into A.
3020 .P
3021 Note that a conditional subpattern is not considered as having two
3022 alternatives, because only one is ever used. In other words, the | character in
3023 a conditional subpattern has a different meaning. Ignoring white space,
3024 consider:
3025 .sp
3026 ^.*? (?(?=a) a | b(*THEN)c )
3027 .sp
3028 If the subject is "ba", this pattern does not match. Because .*? is ungreedy,
3029 it initially matches zero characters. The condition (?=a) then fails, the
3030 character "b" is matched, but "c" is not. At this point, matching does not
3031 backtrack to .*? as might perhaps be expected from the presence of the |
3032 character. The conditional subpattern is part of the single alternative that
3033 comprises the whole pattern, and so the match fails. (If there was a backtrack
3034 into .*?, allowing it to match "b", the match would succeed.)
3035 .P
3036 The verbs just described provide four different "strengths" of control when
3037 subsequent matching fails. (*THEN) is the weakest, carrying on the match at the
3038 next alternative. (*PRUNE) comes next, failing the match at the current
3039 starting position, but allowing an advance to the next character (for an
3040 unanchored pattern). (*SKIP) is similar, except that the advance may be more
3041 than one character. (*COMMIT) is the strongest, causing the entire match to
3042 fail.
3043 .
3044 .
3045 .SS "More than one backtracking verb"
3046 .rs
3047 .sp
3048 If more than one backtracking verb is present in a pattern, the one that is
3049 backtracked onto first acts. For example, consider this pattern, where A, B,
3050 etc. are complex pattern fragments:
3051 .sp
3053 .sp
3054 If A matches but B fails, the backtrack to (*COMMIT) causes the entire match to
3055 fail. However, if A and B match, but C fails, the backtrack to (*THEN) causes
3056 the next alternative (ABD) to be tried. This behaviour is consistent, but is
3057 not always the same as Perl's. It means that if two or more backtracking verbs
3058 appear in succession, all the the last of them has no effect. Consider this
3059 example:
3060 .sp
3061 ...(*COMMIT)(*PRUNE)...
3062 .sp
3063 If there is a matching failure to the right, backtracking onto (*PRUNE) causes
3064 it to be triggered, and its action is taken. There can never be a backtrack
3065 onto (*COMMIT).
3066 .
3067 .
3068 .\" HTML <a name="btrepeat"></a>
3069 .SS "Backtracking verbs in repeated groups"
3070 .rs
3071 .sp
3072 PCRE differs from Perl in its handling of backtracking verbs in repeated
3073 groups. For example, consider:
3074 .sp
3075 /(a(*COMMIT)b)+ac/
3076 .sp
3077 If the subject is "abac", Perl matches, but PCRE fails because the (*COMMIT) in
3078 the second repeat of the group acts.
3079 .
3080 .
3081 .\" HTML <a name="btassert"></a>
3082 .SS "Backtracking verbs in assertions"
3083 .rs
3084 .sp
3085 (*FAIL) in an assertion has its normal effect: it forces an immediate backtrack.
3086 .P
3087 (*ACCEPT) in a positive assertion causes the assertion to succeed without any
3088 further processing. In a negative assertion, (*ACCEPT) causes the assertion to
3089 fail without any further processing.
3090 .P
3091 The other backtracking verbs are not treated specially if they appear in a
3092 positive assertion. In particular, (*THEN) skips to the next alternative in the
3093 innermost enclosing group that has alternations, whether or not this is within
3094 the assertion.
3095 .P
3096 Negative assertions are, however, different, in order to ensure that changing a
3097 positive assertion into a negative assertion changes its result. Backtracking
3098 into (*COMMIT), (*SKIP), or (*PRUNE) causes a negative assertion to be true,
3099 without considering any further alternative branches in the assertion.
3100 Backtracking into (*THEN) causes it to skip to the next enclosing alternative
3101 within the assertion (the normal behaviour), but if the assertion does not have
3102 such an alternative, (*THEN) behaves like (*PRUNE).
3103 .
3104 .
3105 .\" HTML <a name="btsub"></a>
3106 .SS "Backtracking verbs in subroutines"
3107 .rs
3108 .sp
3109 These behaviours occur whether or not the subpattern is called recursively.
3110 Perl's treatment of subroutines is different in some cases.
3111 .P
3112 (*FAIL) in a subpattern called as a subroutine has its normal effect: it forces
3113 an immediate backtrack.
3114 .P
3115 (*ACCEPT) in a subpattern called as a subroutine causes the subroutine match to
3116 succeed without any further processing. Matching then continues after the
3117 subroutine call.
3118 .P
3119 (*COMMIT), (*SKIP), and (*PRUNE) in a subpattern called as a subroutine cause
3120 the subroutine match to fail.
3121 .P
3122 (*THEN) skips to the next alternative in the innermost enclosing group within
3123 the subpattern that has alternatives. If there is no such group within the
3124 subpattern, (*THEN) causes the subroutine match to fail.
3125 .
3126 .
3127 .SH "SEE ALSO"
3128 .rs
3129 .sp
3130 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3),
3131 \fBpcresyntax\fP(3), \fBpcre\fP(3), \fBpcre16(3)\fP, \fBpcre32(3)\fP.
3132 .
3133 .
3135 .rs
3136 .sp
3137 .nf
3138 Philip Hazel
3139 University Computing Service
3140 Cambridge CB2 3QH, England.
3141 .fi
3142 .
3143 .
3145 .rs
3146 .sp
3147 .nf
3148 Last updated: 26 April 2013
3149 Copyright (c) 1997-2013 University of Cambridge.
3150 .fi


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

  ViewVC Help
Powered by ViewVC 1.1.5