ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1258 - (show annotations)
Sat Feb 23 14:14:38 2013 UTC (7 years ago) by ph10
File size: 124299 byte(s)
Document PCRE's \p{Lu} behaviour as different from Perl when /i is specified (I 
think Perl has changed).
1 .TH PCREPATTERN 3 "23 February 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 The original operation of PCRE was on strings of one-byte characters. However,
24 there is now also support for UTF-8 strings in the original library, an
25 extra library that supports 16-bit and UTF-16 character strings, and a
26 third library that supports 32-bit and UTF-32 character strings. To use these
27 features, PCRE must be built to include appropriate support. When using UTF
28 strings you must either call the compiling function with the PCRE_UTF8,
29 PCRE_UTF16, or PCRE_UTF32 option, or the pattern must start with one of
30 these special sequences:
31 .sp
32 (*UTF8)
33 (*UTF16)
34 (*UTF32)
35 (*UTF)
36 .sp
37 (*UTF) is a generic sequence that can be used with any of the libraries.
38 Starting a pattern with such a sequence is equivalent to setting the relevant
39 option. This feature is not Perl-compatible. How setting a UTF mode affects
40 pattern matching is mentioned in several places below. There is also a summary
41 of features in the
42 .\" HREF
43 \fBpcreunicode\fP
44 .\"
45 page.
46 .P
47 Another special sequence that may appear at the start of a pattern or in
48 combination with (*UTF8), (*UTF16), (*UTF32) or (*UTF) is:
49 .sp
50 (*UCP)
51 .sp
52 This has the same effect as setting the PCRE_UCP option: it causes sequences
53 such as \ed and \ew to use Unicode properties to determine character types,
54 instead of recognizing only characters with codes less than 128 via a lookup
55 table.
56 .P
57 If a pattern starts with (*NO_START_OPT), it has the same effect as setting the
58 PCRE_NO_START_OPTIMIZE option either at compile or matching time. There are
59 also some more of these special sequences that are concerned with the handling
60 of newlines; they are described below.
61 .P
62 The remainder of this document discusses the patterns that are supported by
63 PCRE when one its main matching functions, \fBpcre_exec()\fP (8-bit) or
64 \fBpcre[16|32]_exec()\fP (16- or 32-bit), is used. PCRE also has alternative
65 matching functions, \fBpcre_dfa_exec()\fP and \fBpcre[16|32_dfa_exec()\fP,
66 which match using a different algorithm that is not Perl-compatible. Some of
67 the features discussed below are not available when DFA matching is used. The
68 advantages and disadvantages of the alternative functions, and how they differ
69 from the normal functions, are discussed in the
70 .\" HREF
71 \fBpcrematching\fP
72 .\"
73 page.
74 .
75 .
77 .rs
78 .sp
79 PCRE can be compiled to run in an environment that uses EBCDIC as its character
80 code rather than ASCII or Unicode (typically a mainframe system). In the
81 sections below, character code values are ASCII or Unicode; in an EBCDIC
82 environment these characters may have different code values, and there are no
83 code points greater than 255.
84 .
85 .
86 .\" HTML <a name="newlines"></a>
88 .rs
89 .sp
90 PCRE supports five different conventions for indicating line breaks in
91 strings: a single CR (carriage return) character, a single LF (linefeed)
92 character, the two-character sequence CRLF, any of the three preceding, or any
93 Unicode newline sequence. The
94 .\" HREF
95 \fBpcreapi\fP
96 .\"
97 page has
98 .\" HTML <a href="pcreapi.html#newlines">
99 .\" </a>
100 further discussion
101 .\"
102 about newlines, and shows how to set the newline convention in the
103 \fIoptions\fP arguments for the compiling and matching functions.
104 .P
105 It is also possible to specify a newline convention by starting a pattern
106 string with one of the following five sequences:
107 .sp
108 (*CR) carriage return
109 (*LF) linefeed
110 (*CRLF) carriage return, followed by linefeed
111 (*ANYCRLF) any of the three above
112 (*ANY) all Unicode newline sequences
113 .sp
114 These override the default and the options given to the compiling function. For
115 example, on a Unix system where LF is the default newline sequence, the pattern
116 .sp
117 (*CR)a.b
118 .sp
119 changes the convention to CR. That pattern matches "a\enb" because LF is no
120 longer a newline. Note that these special settings, which are not
121 Perl-compatible, are recognized only at the very start of a pattern, and that
122 they must be in upper case. If more than one of them is present, the last one
123 is used.
124 .P
125 The newline convention affects where the circumflex and dollar assertions are
126 true. It also affects the interpretation of the dot metacharacter when
127 PCRE_DOTALL is not set, and the behaviour of \eN. However, it does not affect
128 what the \eR escape sequence matches. By default, this is any Unicode newline
129 sequence, for Perl compatibility. However, this can be changed; see the
130 description of \eR in the section entitled
131 .\" HTML <a href="#newlineseq">
132 .\" </a>
133 "Newline sequences"
134 .\"
135 below. A change of \eR setting can be combined with a change of newline
136 convention.
137 .
138 .
140 .rs
141 .sp
142 A regular expression is a pattern that is matched against a subject string from
143 left to right. Most characters stand for themselves in a pattern, and match the
144 corresponding characters in the subject. As a trivial example, the pattern
145 .sp
146 The quick brown fox
147 .sp
148 matches a portion of a subject string that is identical to itself. When
149 caseless matching is specified (the PCRE_CASELESS option), letters are matched
150 independently of case. In a UTF mode, PCRE always understands the concept of
151 case for characters whose values are less than 128, so caseless matching is
152 always possible. For characters with higher values, the concept of case is
153 supported if PCRE is compiled with Unicode property support, but not otherwise.
154 If you want to use caseless matching for characters 128 and above, you must
155 ensure that PCRE is compiled with Unicode property support as well as with
156 UTF support.
157 .P
158 The power of regular expressions comes from the ability to include alternatives
159 and repetitions in the pattern. These are encoded in the pattern by the use of
160 \fImetacharacters\fP, which do not stand for themselves but instead are
161 interpreted in some special way.
162 .P
163 There are two different sets of metacharacters: those that are recognized
164 anywhere in the pattern except within square brackets, and those that are
165 recognized within square brackets. Outside square brackets, the metacharacters
166 are as follows:
167 .sp
168 \e general escape character with several uses
169 ^ assert start of string (or line, in multiline mode)
170 $ assert end of string (or line, in multiline mode)
171 . match any character except newline (by default)
172 [ start character class definition
173 | start of alternative branch
174 ( start subpattern
175 ) end subpattern
176 ? extends the meaning of (
177 also 0 or 1 quantifier
178 also quantifier minimizer
179 * 0 or more quantifier
180 + 1 or more quantifier
181 also "possessive quantifier"
182 { start min/max quantifier
183 .sp
184 Part of a pattern that is in square brackets is called a "character class". In
185 a character class the only metacharacters are:
186 .sp
187 \e general escape character
188 ^ negate the class, but only if the first character
189 - indicates character range
190 .\" JOIN
191 [ POSIX character class (only if followed by POSIX
192 syntax)
193 ] terminates the character class
194 .sp
195 The following sections describe the use of each of the metacharacters.
196 .
197 .
199 .rs
200 .sp
201 The backslash character has several uses. Firstly, if it is followed by a
202 character that is not a number or a letter, it takes away any special meaning
203 that character may have. This use of backslash as an escape character applies
204 both inside and outside character classes.
205 .P
206 For example, if you want to match a * character, you write \e* in the pattern.
207 This escaping action applies whether or not the following character would
208 otherwise be interpreted as a metacharacter, so it is always safe to precede a
209 non-alphanumeric with backslash to specify that it stands for itself. In
210 particular, if you want to match a backslash, you write \e\e.
211 .P
212 In a UTF mode, only ASCII numbers and letters have any special meaning after a
213 backslash. All other characters (in particular, those whose codepoints are
214 greater than 127) are treated as literals.
215 .P
216 If a pattern is compiled with the PCRE_EXTENDED option, white space in the
217 pattern (other than in a character class) and characters between a # outside
218 a character class and the next newline are ignored. An escaping backslash can
219 be used to include a white space or # character as part of the pattern.
220 .P
221 If you want to remove the special meaning from a sequence of characters, you
222 can do so by putting them between \eQ and \eE. This is different from Perl in
223 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
224 Perl, $ and @ cause variable interpolation. Note the following examples:
225 .sp
226 Pattern PCRE matches Perl matches
227 .sp
228 .\" JOIN
229 \eQabc$xyz\eE abc$xyz abc followed by the
230 contents of $xyz
231 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
232 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
233 .sp
234 The \eQ...\eE sequence is recognized both inside and outside character classes.
235 An isolated \eE that is not preceded by \eQ is ignored. If \eQ is not followed
236 by \eE later in the pattern, the literal interpretation continues to the end of
237 the pattern (that is, \eE is assumed at the end). If the isolated \eQ is inside
238 a character class, this causes an error, because the character class is not
239 terminated.
240 .
241 .
242 .\" HTML <a name="digitsafterbackslash"></a>
243 .SS "Non-printing characters"
244 .rs
245 .sp
246 A second use of backslash provides a way of encoding non-printing characters
247 in patterns in a visible manner. There is no restriction on the appearance of
248 non-printing characters, apart from the binary zero that terminates a pattern,
249 but when a pattern is being prepared by text editing, it is often easier to use
250 one of the following escape sequences than the binary character it represents:
251 .sp
252 \ea alarm, that is, the BEL character (hex 07)
253 \ecx "control-x", where x is any ASCII character
254 \ee escape (hex 1B)
255 \ef form feed (hex 0C)
256 \en linefeed (hex 0A)
257 \er carriage return (hex 0D)
258 \et tab (hex 09)
259 \eddd character with octal code ddd, or back reference
260 \exhh character with hex code hh
261 \ex{hhh..} character with hex code hhh.. (non-JavaScript mode)
262 \euhhhh character with hex code hhhh (JavaScript mode only)
263 .sp
264 The precise effect of \ecx on ASCII characters is as follows: if x is a lower
265 case letter, it is converted to upper case. Then bit 6 of the character (hex
266 40) is inverted. Thus \ecA to \ecZ become hex 01 to hex 1A (A is 41, Z is 5A),
267 but \ec{ becomes hex 3B ({ is 7B), and \ec; becomes hex 7B (; is 3B). If the
268 data item (byte or 16-bit value) following \ec has a value greater than 127, a
269 compile-time error occurs. This locks out non-ASCII characters in all modes.
270 .P
271 The \ec facility was designed for use with ASCII characters, but with the
272 extension to Unicode it is even less useful than it once was. It is, however,
273 recognized when PCRE is compiled in EBCDIC mode, where data items are always
274 bytes. In this mode, all values are valid after \ec. If the next character is a
275 lower case letter, it is converted to upper case. Then the 0xc0 bits of the
276 byte are inverted. Thus \ecA becomes hex 01, as in ASCII (A is C1), but because
277 the EBCDIC letters are disjoint, \ecZ becomes hex 29 (Z is E9), and other
278 characters also generate different values.
279 .P
280 By default, after \ex, from zero to two hexadecimal digits are read (letters
281 can be in upper or lower case). Any number of hexadecimal digits may appear
282 between \ex{ and }, but the character code is constrained as follows:
283 .sp
284 8-bit non-UTF mode less than 0x100
285 8-bit UTF-8 mode less than 0x10ffff and a valid codepoint
286 16-bit non-UTF mode less than 0x10000
287 16-bit UTF-16 mode less than 0x10ffff and a valid codepoint
288 32-bit non-UTF mode less than 0x80000000
289 32-bit UTF-32 mode less than 0x10ffff and a valid codepoint
290 .sp
291 Invalid Unicode codepoints are the range 0xd800 to 0xdfff (the so-called
292 "surrogate" codepoints), and 0xffef.
293 .P
294 If characters other than hexadecimal digits appear between \ex{ and }, or if
295 there is no terminating }, this form of escape is not recognized. Instead, the
296 initial \ex will be interpreted as a basic hexadecimal escape, with no
297 following digits, giving a character whose value is zero.
298 .P
299 If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \ex is
300 as just described only when it is followed by two hexadecimal digits.
301 Otherwise, it matches a literal "x" character. In JavaScript mode, support for
302 code points greater than 256 is provided by \eu, which must be followed by
303 four hexadecimal digits; otherwise it matches a literal "u" character.
304 Character codes specified by \eu in JavaScript mode are constrained in the same
305 was as those specified by \ex in non-JavaScript mode.
306 .P
307 Characters whose value is less than 256 can be defined by either of the two
308 syntaxes for \ex (or by \eu in JavaScript mode). There is no difference in the
309 way they are handled. For example, \exdc is exactly the same as \ex{dc} (or
310 \eu00dc in JavaScript mode).
311 .P
312 After \e0 up to two further octal digits are read. If there are fewer than two
313 digits, just those that are present are used. Thus the sequence \e0\ex\e07
314 specifies two binary zeros followed by a BEL character (code value 7). Make
315 sure you supply two digits after the initial zero if the pattern character that
316 follows is itself an octal digit.
317 .P
318 The handling of a backslash followed by a digit other than 0 is complicated.
319 Outside a character class, PCRE reads it and any following digits as a decimal
320 number. If the number is less than 10, or if there have been at least that many
321 previous capturing left parentheses in the expression, the entire sequence is
322 taken as a \fIback reference\fP. A description of how this works is given
323 .\" HTML <a href="#backreferences">
324 .\" </a>
325 later,
326 .\"
327 following the discussion of
328 .\" HTML <a href="#subpattern">
329 .\" </a>
330 parenthesized subpatterns.
331 .\"
332 .P
333 Inside a character class, or if the decimal number is greater than 9 and there
334 have not been that many capturing subpatterns, PCRE re-reads up to three octal
335 digits following the backslash, and uses them to generate a data character. Any
336 subsequent digits stand for themselves. The value of the character is
337 constrained in the same way as characters specified in hexadecimal.
338 For example:
339 .sp
340 \e040 is another way of writing an ASCII space
341 .\" JOIN
342 \e40 is the same, provided there are fewer than 40
343 previous capturing subpatterns
344 \e7 is always a back reference
345 .\" JOIN
346 \e11 might be a back reference, or another way of
347 writing a tab
348 \e011 is always a tab
349 \e0113 is a tab followed by the character "3"
350 .\" JOIN
351 \e113 might be a back reference, otherwise the
352 character with octal code 113
353 .\" JOIN
354 \e377 might be a back reference, otherwise
355 the value 255 (decimal)
356 .\" JOIN
357 \e81 is either a back reference, or a binary zero
358 followed by the two characters "8" and "1"
359 .sp
360 Note that octal values of 100 or greater must not be introduced by a leading
361 zero, because no more than three octal digits are ever read.
362 .P
363 All the sequences that define a single character value can be used both inside
364 and outside character classes. In addition, inside a character class, \eb is
365 interpreted as the backspace character (hex 08).
366 .P
367 \eN is not allowed in a character class. \eB, \eR, and \eX are not special
368 inside a character class. Like other unrecognized escape sequences, they are
369 treated as the literal characters "B", "R", and "X" by default, but cause an
370 error if the PCRE_EXTRA option is set. Outside a character class, these
371 sequences have different meanings.
372 .
373 .
374 .SS "Unsupported escape sequences"
375 .rs
376 .sp
377 In Perl, the sequences \el, \eL, \eu, and \eU are recognized by its string
378 handler and used to modify the case of following characters. By default, PCRE
379 does not support these escape sequences. However, if the PCRE_JAVASCRIPT_COMPAT
380 option is set, \eU matches a "U" character, and \eu can be used to define a
381 character by code point, as described in the previous section.
382 .
383 .
384 .SS "Absolute and relative back references"
385 .rs
386 .sp
387 The sequence \eg followed by an unsigned or a negative number, optionally
388 enclosed in braces, is an absolute or relative back reference. A named back
389 reference can be coded as \eg{name}. Back references are discussed
390 .\" HTML <a href="#backreferences">
391 .\" </a>
392 later,
393 .\"
394 following the discussion of
395 .\" HTML <a href="#subpattern">
396 .\" </a>
397 parenthesized subpatterns.
398 .\"
399 .
400 .
401 .SS "Absolute and relative subroutine calls"
402 .rs
403 .sp
404 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
405 a number enclosed either in angle brackets or single quotes, is an alternative
406 syntax for referencing a subpattern as a "subroutine". Details are discussed
407 .\" HTML <a href="#onigurumasubroutines">
408 .\" </a>
409 later.
410 .\"
411 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
412 synonymous. The former is a back reference; the latter is a
413 .\" HTML <a href="#subpatternsassubroutines">
414 .\" </a>
415 subroutine
416 .\"
417 call.
418 .
419 .
420 .\" HTML <a name="genericchartypes"></a>
421 .SS "Generic character types"
422 .rs
423 .sp
424 Another use of backslash is for specifying generic character types:
425 .sp
426 \ed any decimal digit
427 \eD any character that is not a decimal digit
428 \eh any horizontal white space character
429 \eH any character that is not a horizontal white space character
430 \es any white space character
431 \eS any character that is not a white space character
432 \ev any vertical white space character
433 \eV any character that is not a vertical white space character
434 \ew any "word" character
435 \eW any "non-word" character
436 .sp
437 There is also the single sequence \eN, which matches a non-newline character.
438 This is the same as
439 .\" HTML <a href="#fullstopdot">
440 .\" </a>
441 the "." metacharacter
442 .\"
443 when PCRE_DOTALL is not set. Perl also uses \eN to match characters by name;
444 PCRE does not support this.
445 .P
446 Each pair of lower and upper case escape sequences partitions the complete set
447 of characters into two disjoint sets. Any given character matches one, and only
448 one, of each pair. The sequences can appear both inside and outside character
449 classes. They each match one character of the appropriate type. If the current
450 matching point is at the end of the subject string, all of them fail, because
451 there is no character to match.
452 .P
453 For compatibility with Perl, \es does not match the VT character (code 11).
454 This makes it different from the the POSIX "space" class. The \es characters
455 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
456 included in a Perl script, \es may match the VT character. In PCRE, it never
457 does.
458 .P
459 A "word" character is an underscore or any character that is a letter or digit.
460 By default, the definition of letters and digits is controlled by PCRE's
461 low-valued character tables, and may vary if locale-specific matching is taking
462 place (see
463 .\" HTML <a href="pcreapi.html#localesupport">
464 .\" </a>
465 "Locale support"
466 .\"
467 in the
468 .\" HREF
469 \fBpcreapi\fP
470 .\"
471 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
472 or "french" in Windows, some character codes greater than 128 are used for
473 accented letters, and these are then matched by \ew. The use of locales with
474 Unicode is discouraged.
475 .P
476 By default, in a UTF mode, characters with values greater than 128 never match
477 \ed, \es, or \ew, and always match \eD, \eS, and \eW. These sequences retain
478 their original meanings from before UTF support was available, mainly for
479 efficiency reasons. However, if PCRE is compiled with Unicode property support,
480 and the PCRE_UCP option is set, the behaviour is changed so that Unicode
481 properties are used to determine character types, as follows:
482 .sp
483 \ed any character that \ep{Nd} matches (decimal digit)
484 \es any character that \ep{Z} matches, plus HT, LF, FF, CR
485 \ew any character that \ep{L} or \ep{N} matches, plus underscore
486 .sp
487 The upper case escapes match the inverse sets of characters. Note that \ed
488 matches only decimal digits, whereas \ew matches any Unicode digit, as well as
489 any Unicode letter, and underscore. Note also that PCRE_UCP affects \eb, and
490 \eB because they are defined in terms of \ew and \eW. Matching these sequences
491 is noticeably slower when PCRE_UCP is set.
492 .P
493 The sequences \eh, \eH, \ev, and \eV are features that were added to Perl at
494 release 5.10. In contrast to the other sequences, which match only ASCII
495 characters by default, these always match certain high-valued codepoints,
496 whether or not PCRE_UCP is set. The horizontal space characters are:
497 .sp
498 U+0009 Horizontal tab (HT)
499 U+0020 Space
500 U+00A0 Non-break space
501 U+1680 Ogham space mark
502 U+180E Mongolian vowel separator
503 U+2000 En quad
504 U+2001 Em quad
505 U+2002 En space
506 U+2003 Em space
507 U+2004 Three-per-em space
508 U+2005 Four-per-em space
509 U+2006 Six-per-em space
510 U+2007 Figure space
511 U+2008 Punctuation space
512 U+2009 Thin space
513 U+200A Hair space
514 U+202F Narrow no-break space
515 U+205F Medium mathematical space
516 U+3000 Ideographic space
517 .sp
518 The vertical space characters are:
519 .sp
520 U+000A Linefeed (LF)
521 U+000B Vertical tab (VT)
522 U+000C Form feed (FF)
523 U+000D Carriage return (CR)
524 U+0085 Next line (NEL)
525 U+2028 Line separator
526 U+2029 Paragraph separator
527 .sp
528 In 8-bit, non-UTF-8 mode, only the characters with codepoints less than 256 are
529 relevant.
530 .
531 .
532 .\" HTML <a name="newlineseq"></a>
533 .SS "Newline sequences"
534 .rs
535 .sp
536 Outside a character class, by default, the escape sequence \eR matches any
537 Unicode newline sequence. In 8-bit non-UTF-8 mode \eR is equivalent to the
538 following:
539 .sp
540 (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
541 .sp
542 This is an example of an "atomic group", details of which are given
543 .\" HTML <a href="#atomicgroup">
544 .\" </a>
545 below.
546 .\"
547 This particular group matches either the two-character sequence CR followed by
548 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
549 U+000B), FF (form feed, U+000C), CR (carriage return, U+000D), or NEL (next
550 line, U+0085). The two-character sequence is treated as a single unit that
551 cannot be split.
552 .P
553 In other modes, two additional characters whose codepoints are greater than 255
554 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
555 Unicode character property support is not needed for these characters to be
556 recognized.
557 .P
558 It is possible to restrict \eR to match only CR, LF, or CRLF (instead of the
559 complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
560 either at compile time or when the pattern is matched. (BSR is an abbrevation
561 for "backslash R".) This can be made the default when PCRE is built; if this is
562 the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
563 It is also possible to specify these settings by starting a pattern string with
564 one of the following sequences:
565 .sp
566 (*BSR_ANYCRLF) CR, LF, or CRLF only
567 (*BSR_UNICODE) any Unicode newline sequence
568 .sp
569 These override the default and the options given to the compiling function, but
570 they can themselves be overridden by options given to a matching function. Note
571 that these special settings, which are not Perl-compatible, are recognized only
572 at the very start of a pattern, and that they must be in upper case. If more
573 than one of them is present, the last one is used. They can be combined with a
574 change of newline convention; for example, a pattern can start with:
575 .sp
577 .sp
578 They can also be combined with the (*UTF8), (*UTF16), (*UTF32), (*UTF) or
579 (*UCP) special sequences. Inside a character class, \eR is treated as an
580 unrecognized escape sequence, and so matches the letter "R" by default, but
581 causes an error if PCRE_EXTRA is set.
582 .
583 .
584 .\" HTML <a name="uniextseq"></a>
585 .SS Unicode character properties
586 .rs
587 .sp
588 When PCRE is built with Unicode character property support, three additional
589 escape sequences that match characters with specific properties are available.
590 When in 8-bit non-UTF-8 mode, these sequences are of course limited to testing
591 characters whose codepoints are less than 256, but they do work in this mode.
592 The extra escape sequences are:
593 .sp
594 \ep{\fIxx\fP} a character with the \fIxx\fP property
595 \eP{\fIxx\fP} a character without the \fIxx\fP property
596 \eX a Unicode extended grapheme cluster
597 .sp
598 The property names represented by \fIxx\fP above are limited to the Unicode
599 script names, the general category properties, "Any", which matches any
600 character (including newline), and some special PCRE properties (described
601 in the
602 .\" HTML <a href="#extraprops">
603 .\" </a>
604 next section).
605 .\"
606 Other Perl properties such as "InMusicalSymbols" are not currently supported by
607 PCRE. Note that \eP{Any} does not match any characters, so always causes a
608 match failure.
609 .P
610 Sets of Unicode characters are defined as belonging to certain scripts. A
611 character from one of these sets can be matched using a script name. For
612 example:
613 .sp
614 \ep{Greek}
615 \eP{Han}
616 .sp
617 Those that are not part of an identified script are lumped together as
618 "Common". The current list of scripts is:
619 .P
620 Arabic,
621 Armenian,
622 Avestan,
623 Balinese,
624 Bamum,
625 Batak,
626 Bengali,
627 Bopomofo,
628 Brahmi,
629 Braille,
630 Buginese,
631 Buhid,
632 Canadian_Aboriginal,
633 Carian,
634 Chakma,
635 Cham,
636 Cherokee,
637 Common,
638 Coptic,
639 Cuneiform,
640 Cypriot,
641 Cyrillic,
642 Deseret,
643 Devanagari,
644 Egyptian_Hieroglyphs,
645 Ethiopic,
646 Georgian,
647 Glagolitic,
648 Gothic,
649 Greek,
650 Gujarati,
651 Gurmukhi,
652 Han,
653 Hangul,
654 Hanunoo,
655 Hebrew,
656 Hiragana,
657 Imperial_Aramaic,
658 Inherited,
659 Inscriptional_Pahlavi,
660 Inscriptional_Parthian,
661 Javanese,
662 Kaithi,
663 Kannada,
664 Katakana,
665 Kayah_Li,
666 Kharoshthi,
667 Khmer,
668 Lao,
669 Latin,
670 Lepcha,
671 Limbu,
672 Linear_B,
673 Lisu,
674 Lycian,
675 Lydian,
676 Malayalam,
677 Mandaic,
678 Meetei_Mayek,
679 Meroitic_Cursive,
680 Meroitic_Hieroglyphs,
681 Miao,
682 Mongolian,
683 Myanmar,
684 New_Tai_Lue,
685 Nko,
686 Ogham,
687 Old_Italic,
688 Old_Persian,
689 Old_South_Arabian,
690 Old_Turkic,
691 Ol_Chiki,
692 Oriya,
693 Osmanya,
694 Phags_Pa,
695 Phoenician,
696 Rejang,
697 Runic,
698 Samaritan,
699 Saurashtra,
700 Sharada,
701 Shavian,
702 Sinhala,
703 Sora_Sompeng,
704 Sundanese,
705 Syloti_Nagri,
706 Syriac,
707 Tagalog,
708 Tagbanwa,
709 Tai_Le,
710 Tai_Tham,
711 Tai_Viet,
712 Takri,
713 Tamil,
714 Telugu,
715 Thaana,
716 Thai,
717 Tibetan,
718 Tifinagh,
719 Ugaritic,
720 Vai,
721 Yi.
722 .P
723 Each character has exactly one Unicode general category property, specified by
724 a two-letter abbreviation. For compatibility with Perl, negation can be
725 specified by including a circumflex between the opening brace and the property
726 name. For example, \ep{^Lu} is the same as \eP{Lu}.
727 .P
728 If only one letter is specified with \ep or \eP, it includes all the general
729 category properties that start with that letter. In this case, in the absence
730 of negation, the curly brackets in the escape sequence are optional; these two
731 examples have the same effect:
732 .sp
733 \ep{L}
734 \epL
735 .sp
736 The following general category property codes are supported:
737 .sp
738 C Other
739 Cc Control
740 Cf Format
741 Cn Unassigned
742 Co Private use
743 Cs Surrogate
744 .sp
745 L Letter
746 Ll Lower case letter
747 Lm Modifier letter
748 Lo Other letter
749 Lt Title case letter
750 Lu Upper case letter
751 .sp
752 M Mark
753 Mc Spacing mark
754 Me Enclosing mark
755 Mn Non-spacing mark
756 .sp
757 N Number
758 Nd Decimal number
759 Nl Letter number
760 No Other number
761 .sp
762 P Punctuation
763 Pc Connector punctuation
764 Pd Dash punctuation
765 Pe Close punctuation
766 Pf Final punctuation
767 Pi Initial punctuation
768 Po Other punctuation
769 Ps Open punctuation
770 .sp
771 S Symbol
772 Sc Currency symbol
773 Sk Modifier symbol
774 Sm Mathematical symbol
775 So Other symbol
776 .sp
777 Z Separator
778 Zl Line separator
779 Zp Paragraph separator
780 Zs Space separator
781 .sp
782 The special property L& is also supported: it matches a character that has
783 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
784 a modifier or "other".
785 .P
786 The Cs (Surrogate) property applies only to characters in the range U+D800 to
787 U+DFFF. Such characters are not valid in Unicode strings and so
788 cannot be tested by PCRE, unless UTF validity checking has been turned off
789 (see the discussion of PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK and
790 PCRE_NO_UTF32_CHECK in the
791 .\" HREF
792 \fBpcreapi\fP
793 .\"
794 page). Perl does not support the Cs property.
795 .P
796 The long synonyms for property names that Perl supports (such as \ep{Letter})
797 are not supported by PCRE, nor is it permitted to prefix any of these
798 properties with "Is".
799 .P
800 No character that is in the Unicode table has the Cn (unassigned) property.
801 Instead, this property is assumed for any code point that is not in the
802 Unicode table.
803 .P
804 Specifying caseless matching does not affect these escape sequences. For
805 example, \ep{Lu} always matches only upper case letters. This is different from
806 the behaviour of current versions of Perl.
807 .P
808 Matching characters by Unicode property is not fast, because PCRE has to do a
809 multistage table lookup in order to find a character's property. That is why
810 the traditional escape sequences such as \ed and \ew do not use Unicode
811 properties in PCRE by default, though you can make them do so by setting the
812 PCRE_UCP option or by starting the pattern with (*UCP).
813 .
814 .
815 .SS Extended grapheme clusters
816 .rs
817 .sp
818 The \eX escape matches any number of Unicode characters that form an "extended
819 grapheme cluster", and treats the sequence as an atomic group
820 .\" HTML <a href="#atomicgroup">
821 .\" </a>
822 (see below).
823 .\"
824 Up to and including release 8.31, PCRE matched an earlier, simpler definition
825 that was equivalent to
826 .sp
827 (?>\ePM\epM*)
828 .sp
829 That is, it matched a character without the "mark" property, followed by zero
830 or more characters with the "mark" property. Characters with the "mark"
831 property are typically non-spacing accents that affect the preceding character.
832 .P
833 This simple definition was extended in Unicode to include more complicated
834 kinds of composite character by giving each character a grapheme breaking
835 property, and creating rules that use these properties to define the boundaries
836 of extended grapheme clusters. In releases of PCRE later than 8.31, \eX matches
837 one of these clusters.
838 .P
839 \eX always matches at least one character. Then it decides whether to add
840 additional characters according to the following rules for ending a cluster:
841 .P
842 1. End at the end of the subject string.
843 .P
844 2. Do not end between CR and LF; otherwise end after any control character.
845 .P
846 3. Do not break Hangul (a Korean script) syllable sequences. Hangul characters
847 are of five types: L, V, T, LV, and LVT. An L character may be followed by an
848 L, V, LV, or LVT character; an LV or V character may be followed by a V or T
849 character; an LVT or T character may be follwed only by a T character.
850 .P
851 4. Do not end before extending characters or spacing marks. Characters with
852 the "mark" property always have the "extend" grapheme breaking property.
853 .P
854 5. Do not end after prepend characters.
855 .P
856 6. Otherwise, end the cluster.
857 .
858 .
859 .\" HTML <a name="extraprops"></a>
860 .SS PCRE's additional properties
861 .rs
862 .sp
863 As well as the standard Unicode properties described above, PCRE supports four
864 more that make it possible to convert traditional escape sequences such as \ew
865 and \es and POSIX character classes to use Unicode properties. PCRE uses these
866 non-standard, non-Perl properties internally when PCRE_UCP is set. They are:
867 .sp
868 Xan Any alphanumeric character
869 Xps Any POSIX space character
870 Xsp Any Perl space character
871 Xwd Any Perl "word" character
872 .sp
873 Xan matches characters that have either the L (letter) or the N (number)
874 property. Xps matches the characters tab, linefeed, vertical tab, form feed, or
875 carriage return, and any other character that has the Z (separator) property.
876 Xsp is the same as Xps, except that vertical tab is excluded. Xwd matches the
877 same characters as Xan, plus underscore.
878 .
879 .
880 .\" HTML <a name="resetmatchstart"></a>
881 .SS "Resetting the match start"
882 .rs
883 .sp
884 The escape sequence \eK causes any previously matched characters not to be
885 included in the final matched sequence. For example, the pattern:
886 .sp
887 foo\eKbar
888 .sp
889 matches "foobar", but reports that it has matched "bar". This feature is
890 similar to a lookbehind assertion
891 .\" HTML <a href="#lookbehind">
892 .\" </a>
893 (described below).
894 .\"
895 However, in this case, the part of the subject before the real match does not
896 have to be of fixed length, as lookbehind assertions do. The use of \eK does
897 not interfere with the setting of
898 .\" HTML <a href="#subpattern">
899 .\" </a>
900 captured substrings.
901 .\"
902 For example, when the pattern
903 .sp
904 (foo)\eKbar
905 .sp
906 matches "foobar", the first substring is still set to "foo".
907 .P
908 Perl documents that the use of \eK within assertions is "not well defined". In
909 PCRE, \eK is acted upon when it occurs inside positive assertions, but is
910 ignored in negative assertions.
911 .
912 .
913 .\" HTML <a name="smallassertions"></a>
914 .SS "Simple assertions"
915 .rs
916 .sp
917 The final use of backslash is for certain simple assertions. An assertion
918 specifies a condition that has to be met at a particular point in a match,
919 without consuming any characters from the subject string. The use of
920 subpatterns for more complicated assertions is described
921 .\" HTML <a href="#bigassertions">
922 .\" </a>
923 below.
924 .\"
925 The backslashed assertions are:
926 .sp
927 \eb matches at a word boundary
928 \eB matches when not at a word boundary
929 \eA matches at the start of the subject
930 \eZ matches at the end of the subject
931 also matches before a newline at the end of the subject
932 \ez matches only at the end of the subject
933 \eG matches at the first matching position in the subject
934 .sp
935 Inside a character class, \eb has a different meaning; it matches the backspace
936 character. If any other of these assertions appears in a character class, by
937 default it matches the corresponding literal character (for example, \eB
938 matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
939 escape sequence" error is generated instead.
940 .P
941 A word boundary is a position in the subject string where the current character
942 and the previous character do not both match \ew or \eW (i.e. one matches
943 \ew and the other matches \eW), or the start or end of the string if the
944 first or last character matches \ew, respectively. In a UTF mode, the meanings
945 of \ew and \eW can be changed by setting the PCRE_UCP option. When this is
946 done, it also affects \eb and \eB. Neither PCRE nor Perl has a separate "start
947 of word" or "end of word" metasequence. However, whatever follows \eb normally
948 determines which it is. For example, the fragment \eba matches "a" at the start
949 of a word.
950 .P
951 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
952 dollar (described in the next section) in that they only ever match at the very
953 start and end of the subject string, whatever options are set. Thus, they are
954 independent of multiline mode. These three assertions are not affected by the
955 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
956 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
957 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
958 at a point other than the beginning of the subject, \eA can never match. The
959 difference between \eZ and \ez is that \eZ matches before a newline at the end
960 of the string as well as at the very end, whereas \ez matches only at the end.
961 .P
962 The \eG assertion is true only when the current matching position is at the
963 start point of the match, as specified by the \fIstartoffset\fP argument of
964 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
965 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
966 arguments, you can mimic Perl's /g option, and it is in this kind of
967 implementation where \eG can be useful.
968 .P
969 Note, however, that PCRE's interpretation of \eG, as the start of the current
970 match, is subtly different from Perl's, which defines it as the end of the
971 previous match. In Perl, these can be different when the previously matched
972 string was empty. Because PCRE does just one match at a time, it cannot
973 reproduce this behaviour.
974 .P
975 If all the alternatives of a pattern begin with \eG, the expression is anchored
976 to the starting match position, and the "anchored" flag is set in the compiled
977 regular expression.
978 .
979 .
981 .rs
982 .sp
983 The circumflex and dollar metacharacters are zero-width assertions. That is,
984 they test for a particular condition being true without consuming any
985 characters from the subject string.
986 .P
987 Outside a character class, in the default matching mode, the circumflex
988 character is an assertion that is true only if the current matching point is at
989 the start of the subject string. If the \fIstartoffset\fP argument of
990 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
991 option is unset. Inside a character class, circumflex has an entirely different
992 meaning
993 .\" HTML <a href="#characterclass">
994 .\" </a>
995 (see below).
996 .\"
997 .P
998 Circumflex need not be the first character of the pattern if a number of
999 alternatives are involved, but it should be the first thing in each alternative
1000 in which it appears if the pattern is ever to match that branch. If all
1001 possible alternatives start with a circumflex, that is, if the pattern is
1002 constrained to match only at the start of the subject, it is said to be an
1003 "anchored" pattern. (There are also other constructs that can cause a pattern
1004 to be anchored.)
1005 .P
1006 The dollar character is an assertion that is true only if the current matching
1007 point is at the end of the subject string, or immediately before a newline at
1008 the end of the string (by default). Note, however, that it does not actually
1009 match the newline. Dollar need not be the last character of the pattern if a
1010 number of alternatives are involved, but it should be the last item in any
1011 branch in which it appears. Dollar has no special meaning in a character class.
1012 .P
1013 The meaning of dollar can be changed so that it matches only at the very end of
1014 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
1015 does not affect the \eZ assertion.
1016 .P
1017 The meanings of the circumflex and dollar characters are changed if the
1018 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
1019 immediately after internal newlines as well as at the start of the subject
1020 string. It does not match after a newline that ends the string. A dollar
1021 matches before any newlines in the string, as well as at the very end, when
1022 PCRE_MULTILINE is set. When newline is specified as the two-character
1023 sequence CRLF, isolated CR and LF characters do not indicate newlines.
1024 .P
1025 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
1026 \en represents a newline) in multiline mode, but not otherwise. Consequently,
1027 patterns that are anchored in single line mode because all branches start with
1028 ^ are not anchored in multiline mode, and a match for circumflex is possible
1029 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
1030 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
1031 .P
1032 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
1033 end of the subject in both modes, and if all branches of a pattern start with
1034 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
1035 .
1036 .
1037 .\" HTML <a name="fullstopdot"></a>
1039 .rs
1040 .sp
1041 Outside a character class, a dot in the pattern matches any one character in
1042 the subject string except (by default) a character that signifies the end of a
1043 line.
1044 .P
1045 When a line ending is defined as a single character, dot never matches that
1046 character; when the two-character sequence CRLF is used, dot does not match CR
1047 if it is immediately followed by LF, but otherwise it matches all characters
1048 (including isolated CRs and LFs). When any Unicode line endings are being
1049 recognized, dot does not match CR or LF or any of the other line ending
1050 characters.
1051 .P
1052 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
1053 option is set, a dot matches any one character, without exception. If the
1054 two-character sequence CRLF is present in the subject string, it takes two dots
1055 to match it.
1056 .P
1057 The handling of dot is entirely independent of the handling of circumflex and
1058 dollar, the only relationship being that they both involve newlines. Dot has no
1059 special meaning in a character class.
1060 .P
1061 The escape sequence \eN behaves like a dot, except that it is not affected by
1062 the PCRE_DOTALL option. In other words, it matches any character except one
1063 that signifies the end of a line. Perl also uses \eN to match characters by
1064 name; PCRE does not support this.
1065 .
1066 .
1068 .rs
1069 .sp
1070 Outside a character class, the escape sequence \eC matches any one data unit,
1071 whether or not a UTF mode is set. In the 8-bit library, one data unit is one
1072 byte; in the 16-bit library it is a 16-bit unit; in the 32-bit library it is
1073 a 32-bit unit. Unlike a dot, \eC always
1074 matches line-ending characters. The feature is provided in Perl in order to
1075 match individual bytes in UTF-8 mode, but it is unclear how it can usefully be
1076 used. Because \eC breaks up characters into individual data units, matching one
1077 unit with \eC in a UTF mode means that the rest of the string may start with a
1078 malformed UTF character. This has undefined results, because PCRE assumes that
1079 it is dealing with valid UTF strings (and by default it checks this at the
1080 start of processing unless the PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK or
1081 PCRE_NO_UTF32_CHECK option is used).
1082 .P
1083 PCRE does not allow \eC to appear in lookbehind assertions
1084 .\" HTML <a href="#lookbehind">
1085 .\" </a>
1086 (described below)
1087 .\"
1088 in a UTF mode, because this would make it impossible to calculate the length of
1089 the lookbehind.
1090 .P
1091 In general, the \eC escape sequence is best avoided. However, one
1092 way of using it that avoids the problem of malformed UTF characters is to use a
1093 lookahead to check the length of the next character, as in this pattern, which
1094 could be used with a UTF-8 string (ignore white space and line breaks):
1095 .sp
1096 (?| (?=[\ex00-\ex7f])(\eC) |
1097 (?=[\ex80-\ex{7ff}])(\eC)(\eC) |
1098 (?=[\ex{800}-\ex{ffff}])(\eC)(\eC)(\eC) |
1099 (?=[\ex{10000}-\ex{1fffff}])(\eC)(\eC)(\eC)(\eC))
1100 .sp
1101 A group that starts with (?| resets the capturing parentheses numbers in each
1102 alternative (see
1103 .\" HTML <a href="#dupsubpatternnumber">
1104 .\" </a>
1105 "Duplicate Subpattern Numbers"
1106 .\"
1107 below). The assertions at the start of each branch check the next UTF-8
1108 character for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The
1109 character's individual bytes are then captured by the appropriate number of
1110 groups.
1111 .
1112 .
1113 .\" HTML <a name="characterclass"></a>
1115 .rs
1116 .sp
1117 An opening square bracket introduces a character class, terminated by a closing
1118 square bracket. A closing square bracket on its own is not special by default.
1119 However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
1120 bracket causes a compile-time error. If a closing square bracket is required as
1121 a member of the class, it should be the first data character in the class
1122 (after an initial circumflex, if present) or escaped with a backslash.
1123 .P
1124 A character class matches a single character in the subject. In a UTF mode, the
1125 character may be more than one data unit long. A matched character must be in
1126 the set of characters defined by the class, unless the first character in the
1127 class definition is a circumflex, in which case the subject character must not
1128 be in the set defined by the class. If a circumflex is actually required as a
1129 member of the class, ensure it is not the first character, or escape it with a
1130 backslash.
1131 .P
1132 For example, the character class [aeiou] matches any lower case vowel, while
1133 [^aeiou] matches any character that is not a lower case vowel. Note that a
1134 circumflex is just a convenient notation for specifying the characters that
1135 are in the class by enumerating those that are not. A class that starts with a
1136 circumflex is not an assertion; it still consumes a character from the subject
1137 string, and therefore it fails if the current pointer is at the end of the
1138 string.
1139 .P
1140 In UTF-8 (UTF-16, UTF-32) mode, characters with values greater than 255 (0xffff)
1141 can be included in a class as a literal string of data units, or by using the
1142 \ex{ escaping mechanism.
1143 .P
1144 When caseless matching is set, any letters in a class represent both their
1145 upper case and lower case versions, so for example, a caseless [aeiou] matches
1146 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1147 caseful version would. In a UTF mode, PCRE always understands the concept of
1148 case for characters whose values are less than 128, so caseless matching is
1149 always possible. For characters with higher values, the concept of case is
1150 supported if PCRE is compiled with Unicode property support, but not otherwise.
1151 If you want to use caseless matching in a UTF mode for characters 128 and
1152 above, you must ensure that PCRE is compiled with Unicode property support as
1153 well as with UTF support.
1154 .P
1155 Characters that might indicate line breaks are never treated in any special way
1156 when matching character classes, whatever line-ending sequence is in use, and
1157 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1158 such as [^a] always matches one of these characters.
1159 .P
1160 The minus (hyphen) character can be used to specify a range of characters in a
1161 character class. For example, [d-m] matches any letter between d and m,
1162 inclusive. If a minus character is required in a class, it must be escaped with
1163 a backslash or appear in a position where it cannot be interpreted as
1164 indicating a range, typically as the first or last character in the class.
1165 .P
1166 It is not possible to have the literal character "]" as the end character of a
1167 range. A pattern such as [W-]46] is interpreted as a class of two characters
1168 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1169 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1170 the end of range, so [W-\e]46] is interpreted as a class containing a range
1171 followed by two other characters. The octal or hexadecimal representation of
1172 "]" can also be used to end a range.
1173 .P
1174 Ranges operate in the collating sequence of character values. They can also be
1175 used for characters specified numerically, for example [\e000-\e037]. Ranges
1176 can include any characters that are valid for the current mode.
1177 .P
1178 If a range that includes letters is used when caseless matching is set, it
1179 matches the letters in either case. For example, [W-c] is equivalent to
1180 [][\e\e^_`wxyzabc], matched caselessly, and in a non-UTF mode, if character
1181 tables for a French locale are in use, [\exc8-\excb] matches accented E
1182 characters in both cases. In UTF modes, PCRE supports the concept of case for
1183 characters with values greater than 128 only when it is compiled with Unicode
1184 property support.
1185 .P
1186 The character escape sequences \ed, \eD, \eh, \eH, \ep, \eP, \es, \eS, \ev,
1187 \eV, \ew, and \eW may appear in a character class, and add the characters that
1188 they match to the class. For example, [\edABCDEF] matches any hexadecimal
1189 digit. In UTF modes, the PCRE_UCP option affects the meanings of \ed, \es, \ew
1190 and their upper case partners, just as it does when they appear outside a
1191 character class, as described in the section entitled
1192 .\" HTML <a href="#genericchartypes">
1193 .\" </a>
1194 "Generic character types"
1195 .\"
1196 above. The escape sequence \eb has a different meaning inside a character
1197 class; it matches the backspace character. The sequences \eB, \eN, \eR, and \eX
1198 are not special inside a character class. Like any other unrecognized escape
1199 sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1200 default, but cause an error if the PCRE_EXTRA option is set.
1201 .P
1202 A circumflex can conveniently be used with the upper case character types to
1203 specify a more restricted set of characters than the matching lower case type.
1204 For example, the class [^\eW_] matches any letter or digit, but not underscore,
1205 whereas [\ew] includes underscore. A positive character class should be read as
1206 "something OR something OR ..." and a negative class as "NOT something AND NOT
1207 something AND NOT ...".
1208 .P
1209 The only metacharacters that are recognized in character classes are backslash,
1210 hyphen (only where it can be interpreted as specifying a range), circumflex
1211 (only at the start), opening square bracket (only when it can be interpreted as
1212 introducing a POSIX class name - see the next section), and the terminating
1213 closing square bracket. However, escaping other non-alphanumeric characters
1214 does no harm.
1215 .
1216 .
1218 .rs
1219 .sp
1220 Perl supports the POSIX notation for character classes. This uses names
1221 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1222 this notation. For example,
1223 .sp
1224 [01[:alpha:]%]
1225 .sp
1226 matches "0", "1", any alphabetic character, or "%". The supported class names
1227 are:
1228 .sp
1229 alnum letters and digits
1230 alpha letters
1231 ascii character codes 0 - 127
1232 blank space or tab only
1233 cntrl control characters
1234 digit decimal digits (same as \ed)
1235 graph printing characters, excluding space
1236 lower lower case letters
1237 print printing characters, including space
1238 punct printing characters, excluding letters and digits and space
1239 space white space (not quite the same as \es)
1240 upper upper case letters
1241 word "word" characters (same as \ew)
1242 xdigit hexadecimal digits
1243 .sp
1244 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
1245 space (32). Notice that this list includes the VT character (code 11). This
1246 makes "space" different to \es, which does not include VT (for Perl
1247 compatibility).
1248 .P
1249 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1250 5.8. Another Perl extension is negation, which is indicated by a ^ character
1251 after the colon. For example,
1252 .sp
1253 [12[:^digit:]]
1254 .sp
1255 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1256 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1257 supported, and an error is given if they are encountered.
1258 .P
1259 By default, in UTF modes, characters with values greater than 128 do not match
1260 any of the POSIX character classes. However, if the PCRE_UCP option is passed
1261 to \fBpcre_compile()\fP, some of the classes are changed so that Unicode
1262 character properties are used. This is achieved by replacing the POSIX classes
1263 by other sequences, as follows:
1264 .sp
1265 [:alnum:] becomes \ep{Xan}
1266 [:alpha:] becomes \ep{L}
1267 [:blank:] becomes \eh
1268 [:digit:] becomes \ep{Nd}
1269 [:lower:] becomes \ep{Ll}
1270 [:space:] becomes \ep{Xps}
1271 [:upper:] becomes \ep{Lu}
1272 [:word:] becomes \ep{Xwd}
1273 .sp
1274 Negated versions, such as [:^alpha:] use \eP instead of \ep. The other POSIX
1275 classes are unchanged, and match only characters with code points less than
1276 128.
1277 .
1278 .
1280 .rs
1281 .sp
1282 Vertical bar characters are used to separate alternative patterns. For example,
1283 the pattern
1284 .sp
1285 gilbert|sullivan
1286 .sp
1287 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1288 and an empty alternative is permitted (matching the empty string). The matching
1289 process tries each alternative in turn, from left to right, and the first one
1290 that succeeds is used. If the alternatives are within a subpattern
1291 .\" HTML <a href="#subpattern">
1292 .\" </a>
1293 (defined below),
1294 .\"
1295 "succeeds" means matching the rest of the main pattern as well as the
1296 alternative in the subpattern.
1297 .
1298 .
1300 .rs
1301 .sp
1302 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1303 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1304 the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1305 The option letters are
1306 .sp
1307 i for PCRE_CASELESS
1309 s for PCRE_DOTALL
1310 x for PCRE_EXTENDED
1311 .sp
1312 For example, (?im) sets caseless, multiline matching. It is also possible to
1313 unset these options by preceding the letter with a hyphen, and a combined
1314 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1315 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1316 permitted. If a letter appears both before and after the hyphen, the option is
1317 unset.
1318 .P
1319 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1320 changed in the same way as the Perl-compatible options by using the characters
1321 J, U and X respectively.
1322 .P
1323 When one of these option changes occurs at top level (that is, not inside
1324 subpattern parentheses), the change applies to the remainder of the pattern
1325 that follows. If the change is placed right at the start of a pattern, PCRE
1326 extracts it into the global options (and it will therefore show up in data
1327 extracted by the \fBpcre_fullinfo()\fP function).
1328 .P
1329 An option change within a subpattern (see below for a description of
1330 subpatterns) affects only that part of the subpattern that follows it, so
1331 .sp
1332 (a(?i)b)c
1333 .sp
1334 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1335 By this means, options can be made to have different settings in different
1336 parts of the pattern. Any changes made in one alternative do carry on
1337 into subsequent branches within the same subpattern. For example,
1338 .sp
1339 (a(?i)b|c)
1340 .sp
1341 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1342 branch is abandoned before the option setting. This is because the effects of
1343 option settings happen at compile time. There would be some very weird
1344 behaviour otherwise.
1345 .P
1346 \fBNote:\fP There are other PCRE-specific options that can be set by the
1347 application when the compiling or matching functions are called. In some cases
1348 the pattern can contain special leading sequences such as (*CRLF) to override
1349 what the application has set or what has been defaulted. Details are given in
1350 the section entitled
1351 .\" HTML <a href="#newlineseq">
1352 .\" </a>
1353 "Newline sequences"
1354 .\"
1355 above. There are also the (*UTF8), (*UTF16),(*UTF32), and (*UCP) leading
1356 sequences that can be used to set UTF and Unicode property modes; they are
1357 equivalent to setting the PCRE_UTF8, PCRE_UTF16, PCRE_UTF32 and the PCRE_UCP
1358 options, respectively. The (*UTF) sequence is a generic version that can be
1359 used with any of the libraries.
1360 .
1361 .
1362 .\" HTML <a name="subpattern"></a>
1364 .rs
1365 .sp
1366 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1367 Turning part of a pattern into a subpattern does two things:
1368 .sp
1369 1. It localizes a set of alternatives. For example, the pattern
1370 .sp
1371 cat(aract|erpillar|)
1372 .sp
1373 matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1374 match "cataract", "erpillar" or an empty string.
1375 .sp
1376 2. It sets up the subpattern as a capturing subpattern. This means that, when
1377 the whole pattern matches, that portion of the subject string that matched the
1378 subpattern is passed back to the caller via the \fIovector\fP argument of the
1379 matching function. (This applies only to the traditional matching functions;
1380 the DFA matching functions do not support capturing.)
1381 .P
1382 Opening parentheses are counted from left to right (starting from 1) to obtain
1383 numbers for the capturing subpatterns. For example, if the string "the red
1384 king" is matched against the pattern
1385 .sp
1386 the ((red|white) (king|queen))
1387 .sp
1388 the captured substrings are "red king", "red", and "king", and are numbered 1,
1389 2, and 3, respectively.
1390 .P
1391 The fact that plain parentheses fulfil two functions is not always helpful.
1392 There are often times when a grouping subpattern is required without a
1393 capturing requirement. If an opening parenthesis is followed by a question mark
1394 and a colon, the subpattern does not do any capturing, and is not counted when
1395 computing the number of any subsequent capturing subpatterns. For example, if
1396 the string "the white queen" is matched against the pattern
1397 .sp
1398 the ((?:red|white) (king|queen))
1399 .sp
1400 the captured substrings are "white queen" and "queen", and are numbered 1 and
1401 2. The maximum number of capturing subpatterns is 65535.
1402 .P
1403 As a convenient shorthand, if any option settings are required at the start of
1404 a non-capturing subpattern, the option letters may appear between the "?" and
1405 the ":". Thus the two patterns
1406 .sp
1407 (?i:saturday|sunday)
1408 (?:(?i)saturday|sunday)
1409 .sp
1410 match exactly the same set of strings. Because alternative branches are tried
1411 from left to right, and options are not reset until the end of the subpattern
1412 is reached, an option setting in one branch does affect subsequent branches, so
1413 the above patterns match "SUNDAY" as well as "Saturday".
1414 .
1415 .
1416 .\" HTML <a name="dupsubpatternnumber"></a>
1418 .rs
1419 .sp
1420 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1421 the same numbers for its capturing parentheses. Such a subpattern starts with
1422 (?| and is itself a non-capturing subpattern. For example, consider this
1423 pattern:
1424 .sp
1425 (?|(Sat)ur|(Sun))day
1426 .sp
1427 Because the two alternatives are inside a (?| group, both sets of capturing
1428 parentheses are numbered one. Thus, when the pattern matches, you can look
1429 at captured substring number one, whichever alternative matched. This construct
1430 is useful when you want to capture part, but not all, of one of a number of
1431 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1432 number is reset at the start of each branch. The numbers of any capturing
1433 parentheses that follow the subpattern start after the highest number used in
1434 any branch. The following example is taken from the Perl documentation. The
1435 numbers underneath show in which buffer the captured content will be stored.
1436 .sp
1437 # before ---------------branch-reset----------- after
1438 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1439 # 1 2 2 3 2 3 4
1440 .sp
1441 A back reference to a numbered subpattern uses the most recent value that is
1442 set for that number by any subpattern. The following pattern matches "abcabc"
1443 or "defdef":
1444 .sp
1445 /(?|(abc)|(def))\e1/
1446 .sp
1447 In contrast, a subroutine call to a numbered subpattern always refers to the
1448 first one in the pattern with the given number. The following pattern matches
1449 "abcabc" or "defabc":
1450 .sp
1451 /(?|(abc)|(def))(?1)/
1452 .sp
1453 If a
1454 .\" HTML <a href="#conditions">
1455 .\" </a>
1456 condition test
1457 .\"
1458 for a subpattern's having matched refers to a non-unique number, the test is
1459 true if any of the subpatterns of that number have matched.
1460 .P
1461 An alternative approach to using this "branch reset" feature is to use
1462 duplicate named subpatterns, as described in the next section.
1463 .
1464 .
1466 .rs
1467 .sp
1468 Identifying capturing parentheses by number is simple, but it can be very hard
1469 to keep track of the numbers in complicated regular expressions. Furthermore,
1470 if an expression is modified, the numbers may change. To help with this
1471 difficulty, PCRE supports the naming of subpatterns. This feature was not
1472 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1473 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1474 the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1475 have different names, but PCRE does not.
1476 .P
1477 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1478 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1479 parentheses from other parts of the pattern, such as
1480 .\" HTML <a href="#backreferences">
1481 .\" </a>
1482 back references,
1483 .\"
1484 .\" HTML <a href="#recursion">
1485 .\" </a>
1486 recursion,
1487 .\"
1488 and
1489 .\" HTML <a href="#conditions">
1490 .\" </a>
1491 conditions,
1492 .\"
1493 can be made by name as well as by number.
1494 .P
1495 Names consist of up to 32 alphanumeric characters and underscores. Named
1496 capturing parentheses are still allocated numbers as well as names, exactly as
1497 if the names were not present. The PCRE API provides function calls for
1498 extracting the name-to-number translation table from a compiled pattern. There
1499 is also a convenience function for extracting a captured substring by name.
1500 .P
1501 By default, a name must be unique within a pattern, but it is possible to relax
1502 this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1503 names are also always permitted for subpatterns with the same number, set up as
1504 described in the previous section.) Duplicate names can be useful for patterns
1505 where only one instance of the named parentheses can match. Suppose you want to
1506 match the name of a weekday, either as a 3-letter abbreviation or as the full
1507 name, and in both cases you want to extract the abbreviation. This pattern
1508 (ignoring the line breaks) does the job:
1509 .sp
1510 (?<DN>Mon|Fri|Sun)(?:day)?|
1511 (?<DN>Tue)(?:sday)?|
1512 (?<DN>Wed)(?:nesday)?|
1513 (?<DN>Thu)(?:rsday)?|
1514 (?<DN>Sat)(?:urday)?
1515 .sp
1516 There are five capturing substrings, but only one is ever set after a match.
1517 (An alternative way of solving this problem is to use a "branch reset"
1518 subpattern, as described in the previous section.)
1519 .P
1520 The convenience function for extracting the data by name returns the substring
1521 for the first (and in this example, the only) subpattern of that name that
1522 matched. This saves searching to find which numbered subpattern it was.
1523 .P
1524 If you make a back reference to a non-unique named subpattern from elsewhere in
1525 the pattern, the one that corresponds to the first occurrence of the name is
1526 used. In the absence of duplicate numbers (see the previous section) this is
1527 the one with the lowest number. If you use a named reference in a condition
1528 test (see the
1529 .\"
1530 .\" HTML <a href="#conditions">
1531 .\" </a>
1532 section about conditions
1533 .\"
1534 below), either to check whether a subpattern has matched, or to check for
1535 recursion, all subpatterns with the same name are tested. If the condition is
1536 true for any one of them, the overall condition is true. This is the same
1537 behaviour as testing by number. For further details of the interfaces for
1538 handling named subpatterns, see the
1539 .\" HREF
1540 \fBpcreapi\fP
1541 .\"
1542 documentation.
1543 .P
1544 \fBWarning:\fP You cannot use different names to distinguish between two
1545 subpatterns with the same number because PCRE uses only the numbers when
1546 matching. For this reason, an error is given at compile time if different names
1547 are given to subpatterns with the same number. However, you can give the same
1548 name to subpatterns with the same number, even when PCRE_DUPNAMES is not set.
1549 .
1550 .
1552 .rs
1553 .sp
1554 Repetition is specified by quantifiers, which can follow any of the following
1555 items:
1556 .sp
1557 a literal data character
1558 the dot metacharacter
1559 the \eC escape sequence
1560 the \eX escape sequence
1561 the \eR escape sequence
1562 an escape such as \ed or \epL that matches a single character
1563 a character class
1564 a back reference (see next section)
1565 a parenthesized subpattern (including assertions)
1566 a subroutine call to a subpattern (recursive or otherwise)
1567 .sp
1568 The general repetition quantifier specifies a minimum and maximum number of
1569 permitted matches, by giving the two numbers in curly brackets (braces),
1570 separated by a comma. The numbers must be less than 65536, and the first must
1571 be less than or equal to the second. For example:
1572 .sp
1573 z{2,4}
1574 .sp
1575 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1576 character. If the second number is omitted, but the comma is present, there is
1577 no upper limit; if the second number and the comma are both omitted, the
1578 quantifier specifies an exact number of required matches. Thus
1579 .sp
1580 [aeiou]{3,}
1581 .sp
1582 matches at least 3 successive vowels, but may match many more, while
1583 .sp
1584 \ed{8}
1585 .sp
1586 matches exactly 8 digits. An opening curly bracket that appears in a position
1587 where a quantifier is not allowed, or one that does not match the syntax of a
1588 quantifier, is taken as a literal character. For example, {,6} is not a
1589 quantifier, but a literal string of four characters.
1590 .P
1591 In UTF modes, quantifiers apply to characters rather than to individual data
1592 units. Thus, for example, \ex{100}{2} matches two characters, each of
1593 which is represented by a two-byte sequence in a UTF-8 string. Similarly,
1594 \eX{3} matches three Unicode extended grapheme clusters, each of which may be
1595 several data units long (and they may be of different lengths).
1596 .P
1597 The quantifier {0} is permitted, causing the expression to behave as if the
1598 previous item and the quantifier were not present. This may be useful for
1599 subpatterns that are referenced as
1600 .\" HTML <a href="#subpatternsassubroutines">
1601 .\" </a>
1602 subroutines
1603 .\"
1604 from elsewhere in the pattern (but see also the section entitled
1605 .\" HTML <a href="#subdefine">
1606 .\" </a>
1607 "Defining subpatterns for use by reference only"
1608 .\"
1609 below). Items other than subpatterns that have a {0} quantifier are omitted
1610 from the compiled pattern.
1611 .P
1612 For convenience, the three most common quantifiers have single-character
1613 abbreviations:
1614 .sp
1615 * is equivalent to {0,}
1616 + is equivalent to {1,}
1617 ? is equivalent to {0,1}
1618 .sp
1619 It is possible to construct infinite loops by following a subpattern that can
1620 match no characters with a quantifier that has no upper limit, for example:
1621 .sp
1622 (a?)*
1623 .sp
1624 Earlier versions of Perl and PCRE used to give an error at compile time for
1625 such patterns. However, because there are cases where this can be useful, such
1626 patterns are now accepted, but if any repetition of the subpattern does in fact
1627 match no characters, the loop is forcibly broken.
1628 .P
1629 By default, the quantifiers are "greedy", that is, they match as much as
1630 possible (up to the maximum number of permitted times), without causing the
1631 rest of the pattern to fail. The classic example of where this gives problems
1632 is in trying to match comments in C programs. These appear between /* and */
1633 and within the comment, individual * and / characters may appear. An attempt to
1634 match C comments by applying the pattern
1635 .sp
1636 /\e*.*\e*/
1637 .sp
1638 to the string
1639 .sp
1640 /* first comment */ not comment /* second comment */
1641 .sp
1642 fails, because it matches the entire string owing to the greediness of the .*
1643 item.
1644 .P
1645 However, if a quantifier is followed by a question mark, it ceases to be
1646 greedy, and instead matches the minimum number of times possible, so the
1647 pattern
1648 .sp
1649 /\e*.*?\e*/
1650 .sp
1651 does the right thing with the C comments. The meaning of the various
1652 quantifiers is not otherwise changed, just the preferred number of matches.
1653 Do not confuse this use of question mark with its use as a quantifier in its
1654 own right. Because it has two uses, it can sometimes appear doubled, as in
1655 .sp
1656 \ed??\ed
1657 .sp
1658 which matches one digit by preference, but can match two if that is the only
1659 way the rest of the pattern matches.
1660 .P
1661 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1662 the quantifiers are not greedy by default, but individual ones can be made
1663 greedy by following them with a question mark. In other words, it inverts the
1664 default behaviour.
1665 .P
1666 When a parenthesized subpattern is quantified with a minimum repeat count that
1667 is greater than 1 or with a limited maximum, more memory is required for the
1668 compiled pattern, in proportion to the size of the minimum or maximum.
1669 .P
1670 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1671 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1672 implicitly anchored, because whatever follows will be tried against every
1673 character position in the subject string, so there is no point in retrying the
1674 overall match at any position after the first. PCRE normally treats such a
1675 pattern as though it were preceded by \eA.
1676 .P
1677 In cases where it is known that the subject string contains no newlines, it is
1678 worth setting PCRE_DOTALL in order to obtain this optimization, or
1679 alternatively using ^ to indicate anchoring explicitly.
1680 .P
1681 However, there are some cases where the optimization cannot be used. When .*
1682 is inside capturing parentheses that are the subject of a back reference
1683 elsewhere in the pattern, a match at the start may fail where a later one
1684 succeeds. Consider, for example:
1685 .sp
1686 (.*)abc\e1
1687 .sp
1688 If the subject is "xyz123abc123" the match point is the fourth character. For
1689 this reason, such a pattern is not implicitly anchored.
1690 .P
1691 Another case where implicit anchoring is not applied is when the leading .* is
1692 inside an atomic group. Once again, a match at the start may fail where a later
1693 one succeeds. Consider this pattern:
1694 .sp
1695 (?>.*?a)b
1696 .sp
1697 It matches "ab" in the subject "aab". The use of the backtracking control verbs
1698 (*PRUNE) and (*SKIP) also disable this optimization.
1699 .P
1700 When a capturing subpattern is repeated, the value captured is the substring
1701 that matched the final iteration. For example, after
1702 .sp
1703 (tweedle[dume]{3}\es*)+
1704 .sp
1705 has matched "tweedledum tweedledee" the value of the captured substring is
1706 "tweedledee". However, if there are nested capturing subpatterns, the
1707 corresponding captured values may have been set in previous iterations. For
1708 example, after
1709 .sp
1710 /(a|(b))+/
1711 .sp
1712 matches "aba" the value of the second captured substring is "b".
1713 .
1714 .
1715 .\" HTML <a name="atomicgroup"></a>
1717 .rs
1718 .sp
1719 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1720 repetition, failure of what follows normally causes the repeated item to be
1721 re-evaluated to see if a different number of repeats allows the rest of the
1722 pattern to match. Sometimes it is useful to prevent this, either to change the
1723 nature of the match, or to cause it fail earlier than it otherwise might, when
1724 the author of the pattern knows there is no point in carrying on.
1725 .P
1726 Consider, for example, the pattern \ed+foo when applied to the subject line
1727 .sp
1728 123456bar
1729 .sp
1730 After matching all 6 digits and then failing to match "foo", the normal
1731 action of the matcher is to try again with only 5 digits matching the \ed+
1732 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1733 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1734 that once a subpattern has matched, it is not to be re-evaluated in this way.
1735 .P
1736 If we use atomic grouping for the previous example, the matcher gives up
1737 immediately on failing to match "foo" the first time. The notation is a kind of
1738 special parenthesis, starting with (?> as in this example:
1739 .sp
1740 (?>\ed+)foo
1741 .sp
1742 This kind of parenthesis "locks up" the part of the pattern it contains once
1743 it has matched, and a failure further into the pattern is prevented from
1744 backtracking into it. Backtracking past it to previous items, however, works as
1745 normal.
1746 .P
1747 An alternative description is that a subpattern of this type matches the string
1748 of characters that an identical standalone pattern would match, if anchored at
1749 the current point in the subject string.
1750 .P
1751 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1752 the above example can be thought of as a maximizing repeat that must swallow
1753 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1754 number of digits they match in order to make the rest of the pattern match,
1755 (?>\ed+) can only match an entire sequence of digits.
1756 .P
1757 Atomic groups in general can of course contain arbitrarily complicated
1758 subpatterns, and can be nested. However, when the subpattern for an atomic
1759 group is just a single repeated item, as in the example above, a simpler
1760 notation, called a "possessive quantifier" can be used. This consists of an
1761 additional + character following a quantifier. Using this notation, the
1762 previous example can be rewritten as
1763 .sp
1764 \ed++foo
1765 .sp
1766 Note that a possessive quantifier can be used with an entire group, for
1767 example:
1768 .sp
1769 (abc|xyz){2,3}+
1770 .sp
1771 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1772 option is ignored. They are a convenient notation for the simpler forms of
1773 atomic group. However, there is no difference in the meaning of a possessive
1774 quantifier and the equivalent atomic group, though there may be a performance
1775 difference; possessive quantifiers should be slightly faster.
1776 .P
1777 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1778 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1779 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1780 package, and PCRE copied it from there. It ultimately found its way into Perl
1781 at release 5.10.
1782 .P
1783 PCRE has an optimization that automatically "possessifies" certain simple
1784 pattern constructs. For example, the sequence A+B is treated as A++B because
1785 there is no point in backtracking into a sequence of A's when B must follow.
1786 .P
1787 When a pattern contains an unlimited repeat inside a subpattern that can itself
1788 be repeated an unlimited number of times, the use of an atomic group is the
1789 only way to avoid some failing matches taking a very long time indeed. The
1790 pattern
1791 .sp
1792 (\eD+|<\ed+>)*[!?]
1793 .sp
1794 matches an unlimited number of substrings that either consist of non-digits, or
1795 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1796 quickly. However, if it is applied to
1797 .sp
1798 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1799 .sp
1800 it takes a long time before reporting failure. This is because the string can
1801 be divided between the internal \eD+ repeat and the external * repeat in a
1802 large number of ways, and all have to be tried. (The example uses [!?] rather
1803 than a single character at the end, because both PCRE and Perl have an
1804 optimization that allows for fast failure when a single character is used. They
1805 remember the last single character that is required for a match, and fail early
1806 if it is not present in the string.) If the pattern is changed so that it uses
1807 an atomic group, like this:
1808 .sp
1809 ((?>\eD+)|<\ed+>)*[!?]
1810 .sp
1811 sequences of non-digits cannot be broken, and failure happens quickly.
1812 .
1813 .
1814 .\" HTML <a name="backreferences"></a>
1816 .rs
1817 .sp
1818 Outside a character class, a backslash followed by a digit greater than 0 (and
1819 possibly further digits) is a back reference to a capturing subpattern earlier
1820 (that is, to its left) in the pattern, provided there have been that many
1821 previous capturing left parentheses.
1822 .P
1823 However, if the decimal number following the backslash is less than 10, it is
1824 always taken as a back reference, and causes an error only if there are not
1825 that many capturing left parentheses in the entire pattern. In other words, the
1826 parentheses that are referenced need not be to the left of the reference for
1827 numbers less than 10. A "forward back reference" of this type can make sense
1828 when a repetition is involved and the subpattern to the right has participated
1829 in an earlier iteration.
1830 .P
1831 It is not possible to have a numerical "forward back reference" to a subpattern
1832 whose number is 10 or more using this syntax because a sequence such as \e50 is
1833 interpreted as a character defined in octal. See the subsection entitled
1834 "Non-printing characters"
1835 .\" HTML <a href="#digitsafterbackslash">
1836 .\" </a>
1837 above
1838 .\"
1839 for further details of the handling of digits following a backslash. There is
1840 no such problem when named parentheses are used. A back reference to any
1841 subpattern is possible using named parentheses (see below).
1842 .P
1843 Another way of avoiding the ambiguity inherent in the use of digits following a
1844 backslash is to use the \eg escape sequence. This escape must be followed by an
1845 unsigned number or a negative number, optionally enclosed in braces. These
1846 examples are all identical:
1847 .sp
1848 (ring), \e1
1849 (ring), \eg1
1850 (ring), \eg{1}
1851 .sp
1852 An unsigned number specifies an absolute reference without the ambiguity that
1853 is present in the older syntax. It is also useful when literal digits follow
1854 the reference. A negative number is a relative reference. Consider this
1855 example:
1856 .sp
1857 (abc(def)ghi)\eg{-1}
1858 .sp
1859 The sequence \eg{-1} is a reference to the most recently started capturing
1860 subpattern before \eg, that is, is it equivalent to \e2 in this example.
1861 Similarly, \eg{-2} would be equivalent to \e1. The use of relative references
1862 can be helpful in long patterns, and also in patterns that are created by
1863 joining together fragments that contain references within themselves.
1864 .P
1865 A back reference matches whatever actually matched the capturing subpattern in
1866 the current subject string, rather than anything matching the subpattern
1867 itself (see
1868 .\" HTML <a href="#subpatternsassubroutines">
1869 .\" </a>
1870 "Subpatterns as subroutines"
1871 .\"
1872 below for a way of doing that). So the pattern
1873 .sp
1874 (sens|respons)e and \e1ibility
1875 .sp
1876 matches "sense and sensibility" and "response and responsibility", but not
1877 "sense and responsibility". If caseful matching is in force at the time of the
1878 back reference, the case of letters is relevant. For example,
1879 .sp
1880 ((?i)rah)\es+\e1
1881 .sp
1882 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1883 capturing subpattern is matched caselessly.
1884 .P
1885 There are several different ways of writing back references to named
1886 subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
1887 \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1888 back reference syntax, in which \eg can be used for both numeric and named
1889 references, is also supported. We could rewrite the above example in any of
1890 the following ways:
1891 .sp
1892 (?<p1>(?i)rah)\es+\ek<p1>
1893 (?'p1'(?i)rah)\es+\ek{p1}
1894 (?P<p1>(?i)rah)\es+(?P=p1)
1895 (?<p1>(?i)rah)\es+\eg{p1}
1896 .sp
1897 A subpattern that is referenced by name may appear in the pattern before or
1898 after the reference.
1899 .P
1900 There may be more than one back reference to the same subpattern. If a
1901 subpattern has not actually been used in a particular match, any back
1902 references to it always fail by default. For example, the pattern
1903 .sp
1904 (a|(bc))\e2
1905 .sp
1906 always fails if it starts to match "a" rather than "bc". However, if the
1907 PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
1908 unset value matches an empty string.
1909 .P
1910 Because there may be many capturing parentheses in a pattern, all digits
1911 following a backslash are taken as part of a potential back reference number.
1912 If the pattern continues with a digit character, some delimiter must be used to
1913 terminate the back reference. If the PCRE_EXTENDED option is set, this can be
1914 white space. Otherwise, the \eg{ syntax or an empty comment (see
1915 .\" HTML <a href="#comments">
1916 .\" </a>
1917 "Comments"
1918 .\"
1919 below) can be used.
1920 .
1921 .SS "Recursive back references"
1922 .rs
1923 .sp
1924 A back reference that occurs inside the parentheses to which it refers fails
1925 when the subpattern is first used, so, for example, (a\e1) never matches.
1926 However, such references can be useful inside repeated subpatterns. For
1927 example, the pattern
1928 .sp
1929 (a|b\e1)+
1930 .sp
1931 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1932 the subpattern, the back reference matches the character string corresponding
1933 to the previous iteration. In order for this to work, the pattern must be such
1934 that the first iteration does not need to match the back reference. This can be
1935 done using alternation, as in the example above, or by a quantifier with a
1936 minimum of zero.
1937 .P
1938 Back references of this type cause the group that they reference to be treated
1939 as an
1940 .\" HTML <a href="#atomicgroup">
1941 .\" </a>
1942 atomic group.
1943 .\"
1944 Once the whole group has been matched, a subsequent matching failure cannot
1945 cause backtracking into the middle of the group.
1946 .
1947 .
1948 .\" HTML <a name="bigassertions"></a>
1950 .rs
1951 .sp
1952 An assertion is a test on the characters following or preceding the current
1953 matching point that does not actually consume any characters. The simple
1954 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
1955 .\" HTML <a href="#smallassertions">
1956 .\" </a>
1957 above.
1958 .\"
1959 .P
1960 More complicated assertions are coded as subpatterns. There are two kinds:
1961 those that look ahead of the current position in the subject string, and those
1962 that look behind it. An assertion subpattern is matched in the normal way,
1963 except that it does not cause the current matching position to be changed.
1964 .P
1965 Assertion subpatterns are not capturing subpatterns. If such an assertion
1966 contains capturing subpatterns within it, these are counted for the purposes of
1967 numbering the capturing subpatterns in the whole pattern. However, substring
1968 capturing is carried out only for positive assertions, because it does not make
1969 sense for negative assertions.
1970 .P
1971 For compatibility with Perl, assertion subpatterns may be repeated; though
1972 it makes no sense to assert the same thing several times, the side effect of
1973 capturing parentheses may occasionally be useful. In practice, there only three
1974 cases:
1975 .sp
1976 (1) If the quantifier is {0}, the assertion is never obeyed during matching.
1977 However, it may contain internal capturing parenthesized groups that are called
1978 from elsewhere via the
1979 .\" HTML <a href="#subpatternsassubroutines">
1980 .\" </a>
1981 subroutine mechanism.
1982 .\"
1983 .sp
1984 (2) If quantifier is {0,n} where n is greater than zero, it is treated as if it
1985 were {0,1}. At run time, the rest of the pattern match is tried with and
1986 without the assertion, the order depending on the greediness of the quantifier.
1987 .sp
1988 (3) If the minimum repetition is greater than zero, the quantifier is ignored.
1989 The assertion is obeyed just once when encountered during matching.
1990 .
1991 .
1992 .SS "Lookahead assertions"
1993 .rs
1994 .sp
1995 Lookahead assertions start with (?= for positive assertions and (?! for
1996 negative assertions. For example,
1997 .sp
1998 \ew+(?=;)
1999 .sp
2000 matches a word followed by a semicolon, but does not include the semicolon in
2001 the match, and
2002 .sp
2003 foo(?!bar)
2004 .sp
2005 matches any occurrence of "foo" that is not followed by "bar". Note that the
2006 apparently similar pattern
2007 .sp
2008 (?!foo)bar
2009 .sp
2010 does not find an occurrence of "bar" that is preceded by something other than
2011 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
2012 (?!foo) is always true when the next three characters are "bar". A
2013 lookbehind assertion is needed to achieve the other effect.
2014 .P
2015 If you want to force a matching failure at some point in a pattern, the most
2016 convenient way to do it is with (?!) because an empty string always matches, so
2017 an assertion that requires there not to be an empty string must always fail.
2018 The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
2019 .
2020 .
2021 .\" HTML <a name="lookbehind"></a>
2022 .SS "Lookbehind assertions"
2023 .rs
2024 .sp
2025 Lookbehind assertions start with (?<= for positive assertions and (?<! for
2026 negative assertions. For example,
2027 .sp
2028 (?<!foo)bar
2029 .sp
2030 does find an occurrence of "bar" that is not preceded by "foo". The contents of
2031 a lookbehind assertion are restricted such that all the strings it matches must
2032 have a fixed length. However, if there are several top-level alternatives, they
2033 do not all have to have the same fixed length. Thus
2034 .sp
2035 (?<=bullock|donkey)
2036 .sp
2037 is permitted, but
2038 .sp
2039 (?<!dogs?|cats?)
2040 .sp
2041 causes an error at compile time. Branches that match different length strings
2042 are permitted only at the top level of a lookbehind assertion. This is an
2043 extension compared with Perl, which requires all branches to match the same
2044 length of string. An assertion such as
2045 .sp
2046 (?<=ab(c|de))
2047 .sp
2048 is not permitted, because its single top-level branch can match two different
2049 lengths, but it is acceptable to PCRE if rewritten to use two top-level
2050 branches:
2051 .sp
2052 (?<=abc|abde)
2053 .sp
2054 In some cases, the escape sequence \eK
2055 .\" HTML <a href="#resetmatchstart">
2056 .\" </a>
2057 (see above)
2058 .\"
2059 can be used instead of a lookbehind assertion to get round the fixed-length
2060 restriction.
2061 .P
2062 The implementation of lookbehind assertions is, for each alternative, to
2063 temporarily move the current position back by the fixed length and then try to
2064 match. If there are insufficient characters before the current position, the
2065 assertion fails.
2066 .P
2067 In a UTF mode, PCRE does not allow the \eC escape (which matches a single data
2068 unit even in a UTF mode) to appear in lookbehind assertions, because it makes
2069 it impossible to calculate the length of the lookbehind. The \eX and \eR
2070 escapes, which can match different numbers of data units, are also not
2071 permitted.
2072 .P
2073 .\" HTML <a href="#subpatternsassubroutines">
2074 .\" </a>
2075 "Subroutine"
2076 .\"
2077 calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
2078 as the subpattern matches a fixed-length string.
2079 .\" HTML <a href="#recursion">
2080 .\" </a>
2081 Recursion,
2082 .\"
2083 however, is not supported.
2084 .P
2085 Possessive quantifiers can be used in conjunction with lookbehind assertions to
2086 specify efficient matching of fixed-length strings at the end of subject
2087 strings. Consider a simple pattern such as
2088 .sp
2089 abcd$
2090 .sp
2091 when applied to a long string that does not match. Because matching proceeds
2092 from left to right, PCRE will look for each "a" in the subject and then see if
2093 what follows matches the rest of the pattern. If the pattern is specified as
2094 .sp
2095 ^.*abcd$
2096 .sp
2097 the initial .* matches the entire string at first, but when this fails (because
2098 there is no following "a"), it backtracks to match all but the last character,
2099 then all but the last two characters, and so on. Once again the search for "a"
2100 covers the entire string, from right to left, so we are no better off. However,
2101 if the pattern is written as
2102 .sp
2103 ^.*+(?<=abcd)
2104 .sp
2105 there can be no backtracking for the .*+ item; it can match only the entire
2106 string. The subsequent lookbehind assertion does a single test on the last four
2107 characters. If it fails, the match fails immediately. For long strings, this
2108 approach makes a significant difference to the processing time.
2109 .
2110 .
2111 .SS "Using multiple assertions"
2112 .rs
2113 .sp
2114 Several assertions (of any sort) may occur in succession. For example,
2115 .sp
2116 (?<=\ed{3})(?<!999)foo
2117 .sp
2118 matches "foo" preceded by three digits that are not "999". Notice that each of
2119 the assertions is applied independently at the same point in the subject
2120 string. First there is a check that the previous three characters are all
2121 digits, and then there is a check that the same three characters are not "999".
2122 This pattern does \fInot\fP match "foo" preceded by six characters, the first
2123 of which are digits and the last three of which are not "999". For example, it
2124 doesn't match "123abcfoo". A pattern to do that is
2125 .sp
2126 (?<=\ed{3}...)(?<!999)foo
2127 .sp
2128 This time the first assertion looks at the preceding six characters, checking
2129 that the first three are digits, and then the second assertion checks that the
2130 preceding three characters are not "999".
2131 .P
2132 Assertions can be nested in any combination. For example,
2133 .sp
2134 (?<=(?<!foo)bar)baz
2135 .sp
2136 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
2137 preceded by "foo", while
2138 .sp
2139 (?<=\ed{3}(?!999)...)foo
2140 .sp
2141 is another pattern that matches "foo" preceded by three digits and any three
2142 characters that are not "999".
2143 .
2144 .
2145 .\" HTML <a name="conditions"></a>
2147 .rs
2148 .sp
2149 It is possible to cause the matching process to obey a subpattern
2150 conditionally or to choose between two alternative subpatterns, depending on
2151 the result of an assertion, or whether a specific capturing subpattern has
2152 already been matched. The two possible forms of conditional subpattern are:
2153 .sp
2154 (?(condition)yes-pattern)
2155 (?(condition)yes-pattern|no-pattern)
2156 .sp
2157 If the condition is satisfied, the yes-pattern is used; otherwise the
2158 no-pattern (if present) is used. If there are more than two alternatives in the
2159 subpattern, a compile-time error occurs. Each of the two alternatives may
2160 itself contain nested subpatterns of any form, including conditional
2161 subpatterns; the restriction to two alternatives applies only at the level of
2162 the condition. This pattern fragment is an example where the alternatives are
2163 complex:
2164 .sp
2165 (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
2166 .sp
2167 .P
2168 There are four kinds of condition: references to subpatterns, references to
2169 recursion, a pseudo-condition called DEFINE, and assertions.
2170 .
2171 .SS "Checking for a used subpattern by number"
2172 .rs
2173 .sp
2174 If the text between the parentheses consists of a sequence of digits, the
2175 condition is true if a capturing subpattern of that number has previously
2176 matched. If there is more than one capturing subpattern with the same number
2177 (see the earlier
2178 .\"
2179 .\" HTML <a href="#recursion">
2180 .\" </a>
2181 section about duplicate subpattern numbers),
2182 .\"
2183 the condition is true if any of them have matched. An alternative notation is
2184 to precede the digits with a plus or minus sign. In this case, the subpattern
2185 number is relative rather than absolute. The most recently opened parentheses
2186 can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
2187 loops it can also make sense to refer to subsequent groups. The next
2188 parentheses to be opened can be referenced as (?(+1), and so on. (The value
2189 zero in any of these forms is not used; it provokes a compile-time error.)
2190 .P
2191 Consider the following pattern, which contains non-significant white space to
2192 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2193 three parts for ease of discussion:
2194 .sp
2195 ( \e( )? [^()]+ (?(1) \e) )
2196 .sp
2197 The first part matches an optional opening parenthesis, and if that
2198 character is present, sets it as the first captured substring. The second part
2199 matches one or more characters that are not parentheses. The third part is a
2200 conditional subpattern that tests whether or not the first set of parentheses
2201 matched. If they did, that is, if subject started with an opening parenthesis,
2202 the condition is true, and so the yes-pattern is executed and a closing
2203 parenthesis is required. Otherwise, since no-pattern is not present, the
2204 subpattern matches nothing. In other words, this pattern matches a sequence of
2205 non-parentheses, optionally enclosed in parentheses.
2206 .P
2207 If you were embedding this pattern in a larger one, you could use a relative
2208 reference:
2209 .sp
2210 ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
2211 .sp
2212 This makes the fragment independent of the parentheses in the larger pattern.
2213 .
2214 .SS "Checking for a used subpattern by name"
2215 .rs
2216 .sp
2217 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
2218 subpattern by name. For compatibility with earlier versions of PCRE, which had
2219 this facility before Perl, the syntax (?(name)...) is also recognized. However,
2220 there is a possible ambiguity with this syntax, because subpattern names may
2221 consist entirely of digits. PCRE looks first for a named subpattern; if it
2222 cannot find one and the name consists entirely of digits, PCRE looks for a
2223 subpattern of that number, which must be greater than zero. Using subpattern
2224 names that consist entirely of digits is not recommended.
2225 .P
2226 Rewriting the above example to use a named subpattern gives this:
2227 .sp
2228 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
2229 .sp
2230 If the name used in a condition of this kind is a duplicate, the test is
2231 applied to all subpatterns of the same name, and is true if any one of them has
2232 matched.
2233 .
2234 .SS "Checking for pattern recursion"
2235 .rs
2236 .sp
2237 If the condition is the string (R), and there is no subpattern with the name R,
2238 the condition is true if a recursive call to the whole pattern or any
2239 subpattern has been made. If digits or a name preceded by ampersand follow the
2240 letter R, for example:
2241 .sp
2242 (?(R3)...) or (?(R&name)...)
2243 .sp
2244 the condition is true if the most recent recursion is into a subpattern whose
2245 number or name is given. This condition does not check the entire recursion
2246 stack. If the name used in a condition of this kind is a duplicate, the test is
2247 applied to all subpatterns of the same name, and is true if any one of them is
2248 the most recent recursion.
2249 .P
2250 At "top level", all these recursion test conditions are false.
2251 .\" HTML <a href="#recursion">
2252 .\" </a>
2253 The syntax for recursive patterns
2254 .\"
2255 is described below.
2256 .
2257 .\" HTML <a name="subdefine"></a>
2258 .SS "Defining subpatterns for use by reference only"
2259 .rs
2260 .sp
2261 If the condition is the string (DEFINE), and there is no subpattern with the
2262 name DEFINE, the condition is always false. In this case, there may be only one
2263 alternative in the subpattern. It is always skipped if control reaches this
2264 point in the pattern; the idea of DEFINE is that it can be used to define
2265 subroutines that can be referenced from elsewhere. (The use of
2266 .\" HTML <a href="#subpatternsassubroutines">
2267 .\" </a>
2268 subroutines
2269 .\"
2270 is described below.) For example, a pattern to match an IPv4 address such as
2271 "" could be written like this (ignore white space and line
2272 breaks):
2273 .sp
2274 (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
2275 \eb (?&byte) (\e.(?&byte)){3} \eb
2276 .sp
2277 The first part of the pattern is a DEFINE group inside which a another group
2278 named "byte" is defined. This matches an individual component of an IPv4
2279 address (a number less than 256). When matching takes place, this part of the
2280 pattern is skipped because DEFINE acts like a false condition. The rest of the
2281 pattern uses references to the named group to match the four dot-separated
2282 components of an IPv4 address, insisting on a word boundary at each end.
2283 .
2284 .SS "Assertion conditions"
2285 .rs
2286 .sp
2287 If the condition is not in any of the above formats, it must be an assertion.
2288 This may be a positive or negative lookahead or lookbehind assertion. Consider
2289 this pattern, again containing non-significant white space, and with the two
2290 alternatives on the second line:
2291 .sp
2292 (?(?=[^a-z]*[a-z])
2293 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
2294 .sp
2295 The condition is a positive lookahead assertion that matches an optional
2296 sequence of non-letters followed by a letter. In other words, it tests for the
2297 presence of at least one letter in the subject. If a letter is found, the
2298 subject is matched against the first alternative; otherwise it is matched
2299 against the second. This pattern matches strings in one of the two forms
2300 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2301 .
2302 .
2303 .\" HTML <a name="comments"></a>
2305 .rs
2306 .sp
2307 There are two ways of including comments in patterns that are processed by
2308 PCRE. In both cases, the start of the comment must not be in a character class,
2309 nor in the middle of any other sequence of related characters such as (?: or a
2310 subpattern name or number. The characters that make up a comment play no part
2311 in the pattern matching.
2312 .P
2313 The sequence (?# marks the start of a comment that continues up to the next
2314 closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2315 option is set, an unescaped # character also introduces a comment, which in
2316 this case continues to immediately after the next newline character or
2317 character sequence in the pattern. Which characters are interpreted as newlines
2318 is controlled by the options passed to a compiling function or by a special
2319 sequence at the start of the pattern, as described in the section entitled
2320 .\" HTML <a href="#newlines">
2321 .\" </a>
2322 "Newline conventions"
2323 .\"
2324 above. Note that the end of this type of comment is a literal newline sequence
2325 in the pattern; escape sequences that happen to represent a newline do not
2326 count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2327 default newline convention is in force:
2328 .sp
2329 abc #comment \en still comment
2330 .sp
2331 On encountering the # character, \fBpcre_compile()\fP skips along, looking for
2332 a newline in the pattern. The sequence \en is still literal at this stage, so
2333 it does not terminate the comment. Only an actual character with the code value
2334 0x0a (the default newline) does so.
2335 .
2336 .
2337 .\" HTML <a name="recursion"></a>
2339 .rs
2340 .sp
2341 Consider the problem of matching a string in parentheses, allowing for
2342 unlimited nested parentheses. Without the use of recursion, the best that can
2343 be done is to use a pattern that matches up to some fixed depth of nesting. It
2344 is not possible to handle an arbitrary nesting depth.
2345 .P
2346 For some time, Perl has provided a facility that allows regular expressions to
2347 recurse (amongst other things). It does this by interpolating Perl code in the
2348 expression at run time, and the code can refer to the expression itself. A Perl
2349 pattern using code interpolation to solve the parentheses problem can be
2350 created like this:
2351 .sp
2352 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
2353 .sp
2354 The (?p{...}) item interpolates Perl code at run time, and in this case refers
2355 recursively to the pattern in which it appears.
2356 .P
2357 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2358 supports special syntax for recursion of the entire pattern, and also for
2359 individual subpattern recursion. After its introduction in PCRE and Python,
2360 this kind of recursion was subsequently introduced into Perl at release 5.10.
2361 .P
2362 A special item that consists of (? followed by a number greater than zero and a
2363 closing parenthesis is a recursive subroutine call of the subpattern of the
2364 given number, provided that it occurs inside that subpattern. (If not, it is a
2365 .\" HTML <a href="#subpatternsassubroutines">
2366 .\" </a>
2367 non-recursive subroutine
2368 .\"
2369 call, which is described in the next section.) The special item (?R) or (?0) is
2370 a recursive call of the entire regular expression.
2371 .P
2372 This PCRE pattern solves the nested parentheses problem (assume the
2373 PCRE_EXTENDED option is set so that white space is ignored):
2374 .sp
2375 \e( ( [^()]++ | (?R) )* \e)
2376 .sp
2377 First it matches an opening parenthesis. Then it matches any number of
2378 substrings which can either be a sequence of non-parentheses, or a recursive
2379 match of the pattern itself (that is, a correctly parenthesized substring).
2380 Finally there is a closing parenthesis. Note the use of a possessive quantifier
2381 to avoid backtracking into sequences of non-parentheses.
2382 .P
2383 If this were part of a larger pattern, you would not want to recurse the entire
2384 pattern, so instead you could use this:
2385 .sp
2386 ( \e( ( [^()]++ | (?1) )* \e) )
2387 .sp
2388 We have put the pattern into parentheses, and caused the recursion to refer to
2389 them instead of the whole pattern.
2390 .P
2391 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2392 is made easier by the use of relative references. Instead of (?1) in the
2393 pattern above you can write (?-2) to refer to the second most recently opened
2394 parentheses preceding the recursion. In other words, a negative number counts
2395 capturing parentheses leftwards from the point at which it is encountered.
2396 .P
2397 It is also possible to refer to subsequently opened parentheses, by writing
2398 references such as (?+2). However, these cannot be recursive because the
2399 reference is not inside the parentheses that are referenced. They are always
2400 .\" HTML <a href="#subpatternsassubroutines">
2401 .\" </a>
2402 non-recursive subroutine
2403 .\"
2404 calls, as described in the next section.
2405 .P
2406 An alternative approach is to use named parentheses instead. The Perl syntax
2407 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
2408 could rewrite the above example as follows:
2409 .sp
2410 (?<pn> \e( ( [^()]++ | (?&pn) )* \e) )
2411 .sp
2412 If there is more than one subpattern with the same name, the earliest one is
2413 used.
2414 .P
2415 This particular example pattern that we have been looking at contains nested
2416 unlimited repeats, and so the use of a possessive quantifier for matching
2417 strings of non-parentheses is important when applying the pattern to strings
2418 that do not match. For example, when this pattern is applied to
2419 .sp
2420 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2421 .sp
2422 it yields "no match" quickly. However, if a possessive quantifier is not used,
2423 the match runs for a very long time indeed because there are so many different
2424 ways the + and * repeats can carve up the subject, and all have to be tested
2425 before failure can be reported.
2426 .P
2427 At the end of a match, the values of capturing parentheses are those from
2428 the outermost level. If you want to obtain intermediate values, a callout
2429 function can be used (see below and the
2430 .\" HREF
2431 \fBpcrecallout\fP
2432 .\"
2433 documentation). If the pattern above is matched against
2434 .sp
2435 (ab(cd)ef)
2436 .sp
2437 the value for the inner capturing parentheses (numbered 2) is "ef", which is
2438 the last value taken on at the top level. If a capturing subpattern is not
2439 matched at the top level, its final captured value is unset, even if it was
2440 (temporarily) set at a deeper level during the matching process.
2441 .P
2442 If there are more than 15 capturing parentheses in a pattern, PCRE has to
2443 obtain extra memory to store data during a recursion, which it does by using
2444 \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no memory can
2445 be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2446 .P
2447 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2448 Consider this pattern, which matches text in angle brackets, allowing for
2449 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2450 recursing), whereas any characters are permitted at the outer level.
2451 .sp
2452 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
2453 .sp
2454 In this pattern, (?(R) is the start of a conditional subpattern, with two
2455 different alternatives for the recursive and non-recursive cases. The (?R) item
2456 is the actual recursive call.
2457 .
2458 .
2459 .\" HTML <a name="recursiondifference"></a>
2460 .SS "Differences in recursion processing between PCRE and Perl"
2461 .rs
2462 .sp
2463 Recursion processing in PCRE differs from Perl in two important ways. In PCRE
2464 (like Python, but unlike Perl), a recursive subpattern call is always treated
2465 as an atomic group. That is, once it has matched some of the subject string, it
2466 is never re-entered, even if it contains untried alternatives and there is a
2467 subsequent matching failure. This can be illustrated by the following pattern,
2468 which purports to match a palindromic string that contains an odd number of
2469 characters (for example, "a", "aba", "abcba", "abcdcba"):
2470 .sp
2471 ^(.|(.)(?1)\e2)$
2472 .sp
2473 The idea is that it either matches a single character, or two identical
2474 characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2475 it does not if the pattern is longer than three characters. Consider the
2476 subject string "abcba":
2477 .P
2478 At the top level, the first character is matched, but as it is not at the end
2479 of the string, the first alternative fails; the second alternative is taken
2480 and the recursion kicks in. The recursive call to subpattern 1 successfully
2481 matches the next character ("b"). (Note that the beginning and end of line
2482 tests are not part of the recursion).
2483 .P
2484 Back at the top level, the next character ("c") is compared with what
2485 subpattern 2 matched, which was "a". This fails. Because the recursion is
2486 treated as an atomic group, there are now no backtracking points, and so the
2487 entire match fails. (Perl is able, at this point, to re-enter the recursion and
2488 try the second alternative.) However, if the pattern is written with the
2489 alternatives in the other order, things are different:
2490 .sp
2491 ^((.)(?1)\e2|.)$
2492 .sp
2493 This time, the recursing alternative is tried first, and continues to recurse
2494 until it runs out of characters, at which point the recursion fails. But this
2495 time we do have another alternative to try at the higher level. That is the big
2496 difference: in the previous case the remaining alternative is at a deeper
2497 recursion level, which PCRE cannot use.
2498 .P
2499 To change the pattern so that it matches all palindromic strings, not just
2500 those with an odd number of characters, it is tempting to change the pattern to
2501 this:
2502 .sp
2503 ^((.)(?1)\e2|.?)$
2504 .sp
2505 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2506 deeper recursion has matched a single character, it cannot be entered again in
2507 order to match an empty string. The solution is to separate the two cases, and
2508 write out the odd and even cases as alternatives at the higher level:
2509 .sp
2510 ^(?:((.)(?1)\e2|)|((.)(?3)\e4|.))
2511 .sp
2512 If you want to match typical palindromic phrases, the pattern has to ignore all
2513 non-word characters, which can be done like this:
2514 .sp
2515 ^\eW*+(?:((.)\eW*+(?1)\eW*+\e2|)|((.)\eW*+(?3)\eW*+\e4|\eW*+.\eW*+))\eW*+$
2516 .sp
2517 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2518 man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2519 the use of the possessive quantifier *+ to avoid backtracking into sequences of
2520 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2521 more) to match typical phrases, and Perl takes so long that you think it has
2522 gone into a loop.
2523 .P
2524 \fBWARNING\fP: The palindrome-matching patterns above work only if the subject
2525 string does not start with a palindrome that is shorter than the entire string.
2526 For example, although "abcba" is correctly matched, if the subject is "ababa",
2527 PCRE finds the palindrome "aba" at the start, then fails at top level because
2528 the end of the string does not follow. Once again, it cannot jump back into the
2529 recursion to try other alternatives, so the entire match fails.
2530 .P
2531 The second way in which PCRE and Perl differ in their recursion processing is
2532 in the handling of captured values. In Perl, when a subpattern is called
2533 recursively or as a subpattern (see the next section), it has no access to any
2534 values that were captured outside the recursion, whereas in PCRE these values
2535 can be referenced. Consider this pattern:
2536 .sp
2537 ^(.)(\e1|a(?2))
2538 .sp
2539 In PCRE, this pattern matches "bab". The first capturing parentheses match "b",
2540 then in the second group, when the back reference \e1 fails to match "b", the
2541 second alternative matches "a" and then recurses. In the recursion, \e1 does
2542 now match "b" and so the whole match succeeds. In Perl, the pattern fails to
2543 match because inside the recursive call \e1 cannot access the externally set
2544 value.
2545 .
2546 .
2547 .\" HTML <a name="subpatternsassubroutines"></a>
2549 .rs
2550 .sp
2551 If the syntax for a recursive subpattern call (either by number or by
2552 name) is used outside the parentheses to which it refers, it operates like a
2553 subroutine in a programming language. The called subpattern may be defined
2554 before or after the reference. A numbered reference can be absolute or
2555 relative, as in these examples:
2556 .sp
2557 (...(absolute)...)...(?2)...
2558 (...(relative)...)...(?-1)...
2559 (...(?+1)...(relative)...
2560 .sp
2561 An earlier example pointed out that the pattern
2562 .sp
2563 (sens|respons)e and \e1ibility
2564 .sp
2565 matches "sense and sensibility" and "response and responsibility", but not
2566 "sense and responsibility". If instead the pattern
2567 .sp
2568 (sens|respons)e and (?1)ibility
2569 .sp
2570 is used, it does match "sense and responsibility" as well as the other two
2571 strings. Another example is given in the discussion of DEFINE above.
2572 .P
2573 All subroutine calls, whether recursive or not, are always treated as atomic
2574 groups. That is, once a subroutine has matched some of the subject string, it
2575 is never re-entered, even if it contains untried alternatives and there is a
2576 subsequent matching failure. Any capturing parentheses that are set during the
2577 subroutine call revert to their previous values afterwards.
2578 .P
2579 Processing options such as case-independence are fixed when a subpattern is
2580 defined, so if it is used as a subroutine, such options cannot be changed for
2581 different calls. For example, consider this pattern:
2582 .sp
2583 (abc)(?i:(?-1))
2584 .sp
2585 It matches "abcabc". It does not match "abcABC" because the change of
2586 processing option does not affect the called subpattern.
2587 .
2588 .
2589 .\" HTML <a name="onigurumasubroutines"></a>
2591 .rs
2592 .sp
2593 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
2594 a number enclosed either in angle brackets or single quotes, is an alternative
2595 syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2596 are two of the examples used above, rewritten using this syntax:
2597 .sp
2598 (?<pn> \e( ( (?>[^()]+) | \eg<pn> )* \e) )
2599 (sens|respons)e and \eg'1'ibility
2600 .sp
2601 PCRE supports an extension to Oniguruma: if a number is preceded by a
2602 plus or a minus sign it is taken as a relative reference. For example:
2603 .sp
2604 (abc)(?i:\eg<-1>)
2605 .sp
2606 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
2607 synonymous. The former is a back reference; the latter is a subroutine call.
2608 .
2609 .
2611 .rs
2612 .sp
2613 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2614 code to be obeyed in the middle of matching a regular expression. This makes it
2615 possible, amongst other things, to extract different substrings that match the
2616 same pair of parentheses when there is a repetition.
2617 .P
2618 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2619 code. The feature is called "callout". The caller of PCRE provides an external
2620 function by putting its entry point in the global variable \fIpcre_callout\fP
2621 (8-bit library) or \fIpcre[16|32]_callout\fP (16-bit or 32-bit library).
2622 By default, this variable contains NULL, which disables all calling out.
2623 .P
2624 Within a regular expression, (?C) indicates the points at which the external
2625 function is to be called. If you want to identify different callout points, you
2626 can put a number less than 256 after the letter C. The default value is zero.
2627 For example, this pattern has two callout points:
2628 .sp
2629 (?C1)abc(?C2)def
2630 .sp
2631 If the PCRE_AUTO_CALLOUT flag is passed to a compiling function, callouts are
2632 automatically installed before each item in the pattern. They are all numbered
2633 255.
2634 .P
2635 During matching, when PCRE reaches a callout point, the external function is
2636 called. It is provided with the number of the callout, the position in the
2637 pattern, and, optionally, one item of data originally supplied by the caller of
2638 the matching function. The callout function may cause matching to proceed, to
2639 backtrack, or to fail altogether. A complete description of the interface to
2640 the callout function is given in the
2641 .\" HREF
2642 \fBpcrecallout\fP
2643 .\"
2644 documentation.
2645 .
2646 .
2647 .\" HTML <a name="backtrackcontrol"></a>
2649 .rs
2650 .sp
2651 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2652 are described in the Perl documentation as "experimental and subject to change
2653 or removal in a future version of Perl". It goes on to say: "Their usage in
2654 production code should be noted to avoid problems during upgrades." The same
2655 remarks apply to the PCRE features described in this section.
2656 .P
2657 Since these verbs are specifically related to backtracking, most of them can be
2658 used only when the pattern is to be matched using one of the traditional
2659 matching functions, which use a backtracking algorithm. With the exception of
2660 (*FAIL), which behaves like a failing negative assertion, they cause an error
2661 if encountered by a DFA matching function.
2662 .P
2663 If any of these verbs are used in an assertion or in a subpattern that is
2664 called as a subroutine (whether or not recursively), their effect is confined
2665 to that subpattern; it does not extend to the surrounding pattern, with one
2666 exception: the name from a *(MARK), (*PRUNE), or (*THEN) that is encountered in
2667 a successful positive assertion \fIis\fP passed back when a match succeeds
2668 (compare capturing parentheses in assertions). Note that such subpatterns are
2669 processed as anchored at the point where they are tested. Note also that Perl's
2670 treatment of subroutines and assertions is different in some cases.
2671 .P
2672 The new verbs make use of what was previously invalid syntax: an opening
2673 parenthesis followed by an asterisk. They are generally of the form
2674 (*VERB) or (*VERB:NAME). Some may take either form, with differing behaviour,
2675 depending on whether or not an argument is present. A name is any sequence of
2676 characters that does not include a closing parenthesis. The maximum length of
2677 name is 255 in the 8-bit library and 65535 in the 16-bit and 32-bit library.
2678 If the name is empty, that is, if the closing parenthesis immediately follows
2679 the colon, the effect is as if the colon were not there. Any number of these
2680 verbs may occur in a pattern.
2681 .
2682 .
2683 .\" HTML <a name="nooptimize"></a>
2684 .SS "Optimizations that affect backtracking verbs"
2685 .rs
2686 .sp
2687 PCRE contains some optimizations that are used to speed up matching by running
2688 some checks at the start of each match attempt. For example, it may know the
2689 minimum length of matching subject, or that a particular character must be
2690 present. When one of these optimizations suppresses the running of a match, any
2691 included backtracking verbs will not, of course, be processed. You can suppress
2692 the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2693 when calling \fBpcre_compile()\fP or \fBpcre_exec()\fP, or by starting the
2694 pattern with (*NO_START_OPT). There is more discussion of this option in the
2695 section entitled
2696 .\" HTML <a href="pcreapi.html#execoptions">
2697 .\" </a>
2698 "Option bits for \fBpcre_exec()\fP"
2699 .\"
2700 in the
2701 .\" HREF
2702 \fBpcreapi\fP
2703 .\"
2704 documentation.
2705 .P
2706 Experiments with Perl suggest that it too has similar optimizations, sometimes
2707 leading to anomalous results.
2708 .
2709 .
2710 .SS "Verbs that act immediately"
2711 .rs
2712 .sp
2713 The following verbs act as soon as they are encountered. They may not be
2714 followed by a name.
2715 .sp
2716 (*ACCEPT)
2717 .sp
2718 This verb causes the match to end successfully, skipping the remainder of the
2719 pattern. However, when it is inside a subpattern that is called as a
2720 subroutine, only that subpattern is ended successfully. Matching then continues
2721 at the outer level. If (*ACCEPT) is inside capturing parentheses, the data so
2722 far is captured. For example:
2723 .sp
2724 A((?:A|B(*ACCEPT)|C)D)
2725 .sp
2726 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2727 the outer parentheses.
2728 .sp
2729 (*FAIL) or (*F)
2730 .sp
2731 This verb causes a matching failure, forcing backtracking to occur. It is
2732 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2733 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2734 Perl features that are not present in PCRE. The nearest equivalent is the
2735 callout feature, as for example in this pattern:
2736 .sp
2737 a+(?C)(*FAIL)
2738 .sp
2739 A match with the string "aaaa" always fails, but the callout is taken before
2740 each backtrack happens (in this example, 10 times).
2741 .
2742 .
2743 .SS "Recording which path was taken"
2744 .rs
2745 .sp
2746 There is one verb whose main purpose is to track how a match was arrived at,
2747 though it also has a secondary use in conjunction with advancing the match
2748 starting point (see (*SKIP) below).
2749 .sp
2750 (*MARK:NAME) or (*:NAME)
2751 .sp
2752 A name is always required with this verb. There may be as many instances of
2753 (*MARK) as you like in a pattern, and their names do not have to be unique.
2754 .P
2755 When a match succeeds, the name of the last-encountered (*MARK) on the matching
2756 path is passed back to the caller as described in the section entitled
2757 .\" HTML <a href="pcreapi.html#extradata">
2758 .\" </a>
2759 "Extra data for \fBpcre_exec()\fP"
2760 .\"
2761 in the
2762 .\" HREF
2763 \fBpcreapi\fP
2764 .\"
2765 documentation. Here is an example of \fBpcretest\fP output, where the /K
2766 modifier requests the retrieval and outputting of (*MARK) data:
2767 .sp
2768 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2769 data> XY
2770 0: XY
2771 MK: A
2772 XZ
2773 0: XZ
2774 MK: B
2775 .sp
2776 The (*MARK) name is tagged with "MK:" in this output, and in this example it
2777 indicates which of the two alternatives matched. This is a more efficient way
2778 of obtaining this information than putting each alternative in its own
2779 capturing parentheses.
2780 .P
2781 If (*MARK) is encountered in a positive assertion, its name is recorded and
2782 passed back if it is the last-encountered. This does not happen for negative
2783 assertions.
2784 .P
2785 After a partial match or a failed match, the name of the last encountered
2786 (*MARK) in the entire match process is returned. For example:
2787 .sp
2788 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2789 data> XP
2790 No match, mark = B
2791 .sp
2792 Note that in this unanchored example the mark is retained from the match
2793 attempt that started at the letter "X" in the subject. Subsequent match
2794 attempts starting at "P" and then with an empty string do not get as far as the
2795 (*MARK) item, but nevertheless do not reset it.
2796 .P
2797 If you are interested in (*MARK) values after failed matches, you should
2798 probably set the PCRE_NO_START_OPTIMIZE option
2799 .\" HTML <a href="#nooptimize">
2800 .\" </a>
2801 (see above)
2802 .\"
2803 to ensure that the match is always attempted.
2804 .
2805 .
2806 .SS "Verbs that act after backtracking"
2807 .rs
2808 .sp
2809 The following verbs do nothing when they are encountered. Matching continues
2810 with what follows, but if there is no subsequent match, causing a backtrack to
2811 the verb, a failure is forced. That is, backtracking cannot pass to the left of
2812 the verb. However, when one of these verbs appears inside an atomic group, its
2813 effect is confined to that group, because once the group has been matched,
2814 there is never any backtracking into it. In this situation, backtracking can
2815 "jump back" to the left of the entire atomic group. (Remember also, as stated
2816 above, that this localization also applies in subroutine calls and assertions.)
2817 .P
2818 These verbs differ in exactly what kind of failure occurs when backtracking
2819 reaches them.
2820 .sp
2821 (*COMMIT)
2822 .sp
2823 This verb, which may not be followed by a name, causes the whole match to fail
2824 outright if the rest of the pattern does not match. Even if the pattern is
2825 unanchored, no further attempts to find a match by advancing the starting point
2826 take place. Once (*COMMIT) has been passed, \fBpcre_exec()\fP is committed to
2827 finding a match at the current starting point, or not at all. For example:
2828 .sp
2829 a+(*COMMIT)b
2830 .sp
2831 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2832 dynamic anchor, or "I've started, so I must finish." The name of the most
2833 recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
2834 match failure.
2835 .P
2836 Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
2837 unless PCRE's start-of-match optimizations are turned off, as shown in this
2838 \fBpcretest\fP example:
2839 .sp
2840 re> /(*COMMIT)abc/
2841 data> xyzabc
2842 0: abc
2843 xyzabc\eY
2844 No match
2845 .sp
2846 PCRE knows that any match must start with "a", so the optimization skips along
2847 the subject to "a" before running the first match attempt, which succeeds. When
2848 the optimization is disabled by the \eY escape in the second subject, the match
2849 starts at "x" and so the (*COMMIT) causes it to fail without trying any other
2850 starting points.
2851 .sp
2852 (*PRUNE) or (*PRUNE:NAME)
2853 .sp
2854 This verb causes the match to fail at the current starting position in the
2855 subject if the rest of the pattern does not match. If the pattern is
2856 unanchored, the normal "bumpalong" advance to the next starting character then
2857 happens. Backtracking can occur as usual to the left of (*PRUNE), before it is
2858 reached, or when matching to the right of (*PRUNE), but if there is no match to
2859 the right, backtracking cannot cross (*PRUNE). In simple cases, the use of
2860 (*PRUNE) is just an alternative to an atomic group or possessive quantifier,
2861 but there are some uses of (*PRUNE) that cannot be expressed in any other way.
2862 The behaviour of (*PRUNE:NAME) is the same as (*MARK:NAME)(*PRUNE). In an
2863 anchored pattern (*PRUNE) has the same effect as (*COMMIT).
2864 .sp
2865 (*SKIP)
2866 .sp
2867 This verb, when given without a name, is like (*PRUNE), except that if the
2868 pattern is unanchored, the "bumpalong" advance is not to the next character,
2869 but to the position in the subject where (*SKIP) was encountered. (*SKIP)
2870 signifies that whatever text was matched leading up to it cannot be part of a
2871 successful match. Consider:
2872 .sp
2873 a+(*SKIP)b
2874 .sp
2875 If the subject is "aaaac...", after the first match attempt fails (starting at
2876 the first character in the string), the starting point skips on to start the
2877 next attempt at "c". Note that a possessive quantifer does not have the same
2878 effect as this example; although it would suppress backtracking during the
2879 first match attempt, the second attempt would start at the second character
2880 instead of skipping on to "c".
2881 .sp
2882 (*SKIP:NAME)
2883 .sp
2884 When (*SKIP) has an associated name, its behaviour is modified. If the
2885 following pattern fails to match, the previous path through the pattern is
2886 searched for the most recent (*MARK) that has the same name. If one is found,
2887 the "bumpalong" advance is to the subject position that corresponds to that
2888 (*MARK) instead of to where (*SKIP) was encountered. If no (*MARK) with a
2889 matching name is found, the (*SKIP) is ignored.
2890 .sp
2891 (*THEN) or (*THEN:NAME)
2892 .sp
2893 This verb causes a skip to the next innermost alternative if the rest of the
2894 pattern does not match. That is, it cancels pending backtracking, but only
2895 within the current alternative. Its name comes from the observation that it can
2896 be used for a pattern-based if-then-else block:
2897 .sp
2898 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2899 .sp
2900 If the COND1 pattern matches, FOO is tried (and possibly further items after
2901 the end of the group if FOO succeeds); on failure, the matcher skips to the
2902 second alternative and tries COND2, without backtracking into COND1. The
2903 behaviour of (*THEN:NAME) is exactly the same as (*MARK:NAME)(*THEN).
2904 If (*THEN) is not inside an alternation, it acts like (*PRUNE).
2905 .P
2906 Note that a subpattern that does not contain a | character is just a part of
2907 the enclosing alternative; it is not a nested alternation with only one
2908 alternative. The effect of (*THEN) extends beyond such a subpattern to the
2909 enclosing alternative. Consider this pattern, where A, B, etc. are complex
2910 pattern fragments that do not contain any | characters at this level:
2911 .sp
2912 A (B(*THEN)C) | D
2913 .sp
2914 If A and B are matched, but there is a failure in C, matching does not
2915 backtrack into A; instead it moves to the next alternative, that is, D.
2916 However, if the subpattern containing (*THEN) is given an alternative, it
2917 behaves differently:
2918 .sp
2919 A (B(*THEN)C | (*FAIL)) | D
2920 .sp
2921 The effect of (*THEN) is now confined to the inner subpattern. After a failure
2922 in C, matching moves to (*FAIL), which causes the whole subpattern to fail
2923 because there are no more alternatives to try. In this case, matching does now
2924 backtrack into A.
2925 .P
2926 Note also that a conditional subpattern is not considered as having two
2927 alternatives, because only one is ever used. In other words, the | character in
2928 a conditional subpattern has a different meaning. Ignoring white space,
2929 consider:
2930 .sp
2931 ^.*? (?(?=a) a | b(*THEN)c )
2932 .sp
2933 If the subject is "ba", this pattern does not match. Because .*? is ungreedy,
2934 it initially matches zero characters. The condition (?=a) then fails, the
2935 character "b" is matched, but "c" is not. At this point, matching does not
2936 backtrack to .*? as might perhaps be expected from the presence of the |
2937 character. The conditional subpattern is part of the single alternative that
2938 comprises the whole pattern, and so the match fails. (If there was a backtrack
2939 into .*?, allowing it to match "b", the match would succeed.)
2940 .P
2941 The verbs just described provide four different "strengths" of control when
2942 subsequent matching fails. (*THEN) is the weakest, carrying on the match at the
2943 next alternative. (*PRUNE) comes next, failing the match at the current
2944 starting position, but allowing an advance to the next character (for an
2945 unanchored pattern). (*SKIP) is similar, except that the advance may be more
2946 than one character. (*COMMIT) is the strongest, causing the entire match to
2947 fail.
2948 .P
2949 If more than one such verb is present in a pattern, the "strongest" one wins.
2950 For example, consider this pattern, where A, B, etc. are complex pattern
2951 fragments:
2952 .sp
2953 (A(*COMMIT)B(*THEN)C|D)
2954 .sp
2955 Once A has matched, PCRE is committed to this match, at the current starting
2956 position. If subsequently B matches, but C does not, the normal (*THEN) action
2957 of trying the next alternative (that is, D) does not happen because (*COMMIT)
2958 overrides.
2959 .
2960 .
2961 .SH "SEE ALSO"
2962 .rs
2963 .sp
2964 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3),
2965 \fBpcresyntax\fP(3), \fBpcre\fP(3), \fBpcre16(3)\fP, \fBpcre32(3)\fP.
2966 .
2967 .
2969 .rs
2970 .sp
2971 .nf
2972 Philip Hazel
2973 University Computing Service
2974 Cambridge CB2 3QH, England.
2975 .fi
2976 .
2977 .
2979 .rs
2980 .sp
2981 .nf
2982 Last updated: 23 February 2013
2983 Copyright (c) 1997-2013 University of Cambridge.
2984 .fi


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

  ViewVC Help
Powered by ViewVC 1.1.5