ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1213 - (show annotations)
Wed Nov 7 17:29:40 2012 UTC (7 years, 4 months ago) by ph10
File size: 124061 byte(s)
Add more words about $.
1 .TH PCREPATTERN 3 "07 November 2012" "PCRE 8.32"
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 an
26 extra 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 .sp
36 Starting a pattern with such a sequence is equivalent to setting the relevant
37 option. This feature is not Perl-compatible. How setting a UTF mode affects
38 pattern matching is mentioned in several places below. There is also a summary
39 of features in the
40 .\" HREF
41 \fBpcreunicode\fP
42 .\"
43 page.
44 .P
45 Another special sequence that may appear at the start of a pattern or in
46 combination with (*UTF8) or (*UTF16) or (*UTF32) is:
47 .sp
48 (*UCP)
49 .sp
50 This has the same effect as setting the PCRE_UCP option: it causes sequences
51 such as \ed and \ew to use Unicode properties to determine character types,
52 instead of recognizing only characters with codes less than 128 via a lookup
53 table.
54 .P
55 If a pattern starts with (*NO_START_OPT), it has the same effect as setting the
56 PCRE_NO_START_OPTIMIZE option either at compile or matching time. There are
57 also some more of these special sequences that are concerned with the handling
58 of newlines; they are described below.
59 .P
60 The remainder of this document discusses the patterns that are supported by
61 PCRE when one its main matching functions, \fBpcre_exec()\fP (8-bit) or
62 \fBpcre[16|32]_exec()\fP (16- or 32-bit), is used. PCRE also has alternative
63 matching functions, \fBpcre_dfa_exec()\fP and \fBpcre[16|32_dfa_exec()\fP,
64 which match using a different algorithm that is not Perl-compatible. Some of
65 the features discussed below are not available when DFA matching is used. The
66 advantages and disadvantages of the alternative functions, and how they differ
67 from the normal functions, are discussed in the
68 .\" HREF
69 \fBpcrematching\fP
70 .\"
71 page.
72 .
73 .
75 .rs
76 .sp
77 PCRE can be compiled to run in an environment that uses EBCDIC as its character
78 code rather than ASCII or Unicode (typically a mainframe system). In the
79 sections below, character code values are ASCII or Unicode; in an EBCDIC
80 environment these characters may have different code values, and there are no
81 code points greater than 255.
82 .
83 .
84 .\" HTML <a name="newlines"></a>
86 .rs
87 .sp
88 PCRE supports five different conventions for indicating line breaks in
89 strings: a single CR (carriage return) character, a single LF (linefeed)
90 character, the two-character sequence CRLF, any of the three preceding, or any
91 Unicode newline sequence. The
92 .\" HREF
93 \fBpcreapi\fP
94 .\"
95 page has
96 .\" HTML <a href="pcreapi.html#newlines">
97 .\" </a>
98 further discussion
99 .\"
100 about newlines, and shows how to set the newline convention in the
101 \fIoptions\fP arguments for the compiling and matching functions.
102 .P
103 It is also possible to specify a newline convention by starting a pattern
104 string with one of the following five sequences:
105 .sp
106 (*CR) carriage return
107 (*LF) linefeed
108 (*CRLF) carriage return, followed by linefeed
109 (*ANYCRLF) any of the three above
110 (*ANY) all Unicode newline sequences
111 .sp
112 These override the default and the options given to the compiling function. For
113 example, on a Unix system where LF is the default newline sequence, the pattern
114 .sp
115 (*CR)a.b
116 .sp
117 changes the convention to CR. That pattern matches "a\enb" because LF is no
118 longer a newline. Note that these special settings, which are not
119 Perl-compatible, are recognized only at the very start of a pattern, and that
120 they must be in upper case. If more than one of them is present, the last one
121 is used.
122 .P
123 The newline convention affects where the circumflex and dollar assertions are
124 true. It also affects the interpretation of the dot metacharacter when
125 PCRE_DOTALL is not set, and the behaviour of \eN. However, it does not affect
126 what the \eR escape sequence matches. By default, this is any Unicode newline
127 sequence, for Perl compatibility. However, this can be changed; see the
128 description of \eR in the section entitled
129 .\" HTML <a href="#newlineseq">
130 .\" </a>
131 "Newline sequences"
132 .\"
133 below. A change of \eR setting can be combined with a change of newline
134 convention.
135 .
136 .
138 .rs
139 .sp
140 A regular expression is a pattern that is matched against a subject string from
141 left to right. Most characters stand for themselves in a pattern, and match the
142 corresponding characters in the subject. As a trivial example, the pattern
143 .sp
144 The quick brown fox
145 .sp
146 matches a portion of a subject string that is identical to itself. When
147 caseless matching is specified (the PCRE_CASELESS option), letters are matched
148 independently of case. In a UTF mode, PCRE always understands the concept of
149 case for characters whose values are less than 128, so caseless matching is
150 always possible. For characters with higher values, the concept of case is
151 supported if PCRE is compiled with Unicode property support, but not otherwise.
152 If you want to use caseless matching for characters 128 and above, you must
153 ensure that PCRE is compiled with Unicode property support as well as with
154 UTF support.
155 .P
156 The power of regular expressions comes from the ability to include alternatives
157 and repetitions in the pattern. These are encoded in the pattern by the use of
158 \fImetacharacters\fP, which do not stand for themselves but instead are
159 interpreted in some special way.
160 .P
161 There are two different sets of metacharacters: those that are recognized
162 anywhere in the pattern except within square brackets, and those that are
163 recognized within square brackets. Outside square brackets, the metacharacters
164 are as follows:
165 .sp
166 \e general escape character with several uses
167 ^ assert start of string (or line, in multiline mode)
168 $ assert end of string (or line, in multiline mode)
169 . match any character except newline (by default)
170 [ start character class definition
171 | start of alternative branch
172 ( start subpattern
173 ) end subpattern
174 ? extends the meaning of (
175 also 0 or 1 quantifier
176 also quantifier minimizer
177 * 0 or more quantifier
178 + 1 or more quantifier
179 also "possessive quantifier"
180 { start min/max quantifier
181 .sp
182 Part of a pattern that is in square brackets is called a "character class". In
183 a character class the only metacharacters are:
184 .sp
185 \e general escape character
186 ^ negate the class, but only if the first character
187 - indicates character range
188 .\" JOIN
189 [ POSIX character class (only if followed by POSIX
190 syntax)
191 ] terminates the character class
192 .sp
193 The following sections describe the use of each of the metacharacters.
194 .
195 .
197 .rs
198 .sp
199 The backslash character has several uses. Firstly, if it is followed by a
200 character that is not a number or a letter, it takes away any special meaning
201 that character may have. This use of backslash as an escape character applies
202 both inside and outside character classes.
203 .P
204 For example, if you want to match a * character, you write \e* in the pattern.
205 This escaping action applies whether or not the following character would
206 otherwise be interpreted as a metacharacter, so it is always safe to precede a
207 non-alphanumeric with backslash to specify that it stands for itself. In
208 particular, if you want to match a backslash, you write \e\e.
209 .P
210 In a UTF mode, only ASCII numbers and letters have any special meaning after a
211 backslash. All other characters (in particular, those whose codepoints are
212 greater than 127) are treated as literals.
213 .P
214 If a pattern is compiled with the PCRE_EXTENDED option, white space in the
215 pattern (other than in a character class) and characters between a # outside
216 a character class and the next newline are ignored. An escaping backslash can
217 be used to include a white space or # character as part of the pattern.
218 .P
219 If you want to remove the special meaning from a sequence of characters, you
220 can do so by putting them between \eQ and \eE. This is different from Perl in
221 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
222 Perl, $ and @ cause variable interpolation. Note the following examples:
223 .sp
224 Pattern PCRE matches Perl matches
225 .sp
226 .\" JOIN
227 \eQabc$xyz\eE abc$xyz abc followed by the
228 contents of $xyz
229 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
230 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
231 .sp
232 The \eQ...\eE sequence is recognized both inside and outside character classes.
233 An isolated \eE that is not preceded by \eQ is ignored. If \eQ is not followed
234 by \eE later in the pattern, the literal interpretation continues to the end of
235 the pattern (that is, \eE is assumed at the end). If the isolated \eQ is inside
236 a character class, this causes an error, because the character class is not
237 terminated.
238 .
239 .
240 .\" HTML <a name="digitsafterbackslash"></a>
241 .SS "Non-printing characters"
242 .rs
243 .sp
244 A second use of backslash provides a way of encoding non-printing characters
245 in patterns in a visible manner. There is no restriction on the appearance of
246 non-printing characters, apart from the binary zero that terminates a pattern,
247 but when a pattern is being prepared by text editing, it is often easier to use
248 one of the following escape sequences than the binary character it represents:
249 .sp
250 \ea alarm, that is, the BEL character (hex 07)
251 \ecx "control-x", where x is any ASCII character
252 \ee escape (hex 1B)
253 \ef form feed (hex 0C)
254 \en linefeed (hex 0A)
255 \er carriage return (hex 0D)
256 \et tab (hex 09)
257 \eddd character with octal code ddd, or back reference
258 \exhh character with hex code hh
259 \ex{hhh..} character with hex code hhh.. (non-JavaScript mode)
260 \euhhhh character with hex code hhhh (JavaScript mode only)
261 .sp
262 The precise effect of \ecx on ASCII characters is as follows: if x is a lower
263 case letter, it is converted to upper case. Then bit 6 of the character (hex
264 40) is inverted. Thus \ecA to \ecZ become hex 01 to hex 1A (A is 41, Z is 5A),
265 but \ec{ becomes hex 3B ({ is 7B), and \ec; becomes hex 7B (; is 3B). If the
266 data item (byte or 16-bit value) following \ec has a value greater than 127, a
267 compile-time error occurs. This locks out non-ASCII characters in all modes.
268 .P
269 The \ec facility was designed for use with ASCII characters, but with the
270 extension to Unicode it is even less useful than it once was. It is, however,
271 recognized when PCRE is compiled in EBCDIC mode, where data items are always
272 bytes. In this mode, all values are valid after \ec. If the next character is a
273 lower case letter, it is converted to upper case. Then the 0xc0 bits of the
274 byte are inverted. Thus \ecA becomes hex 01, as in ASCII (A is C1), but because
275 the EBCDIC letters are disjoint, \ecZ becomes hex 29 (Z is E9), and other
276 characters also generate different values.
277 .P
278 By default, after \ex, from zero to two hexadecimal digits are read (letters
279 can be in upper or lower case). Any number of hexadecimal digits may appear
280 between \ex{ and }, but the character code is constrained as follows:
281 .sp
282 8-bit non-UTF mode less than 0x100
283 8-bit UTF-8 mode less than 0x10ffff and a valid codepoint
284 16-bit non-UTF mode less than 0x10000
285 16-bit UTF-16 mode less than 0x10ffff and a valid codepoint
286 32-bit non-UTF mode less than 0x80000000
287 32-bit UTF-32 mode less than 0x10ffff and a valid codepoint
288 .sp
289 Invalid Unicode codepoints are the range 0xd800 to 0xdfff (the so-called
290 "surrogate" codepoints), and 0xffef.
291 .P
292 If characters other than hexadecimal digits appear between \ex{ and }, or if
293 there is no terminating }, this form of escape is not recognized. Instead, the
294 initial \ex will be interpreted as a basic hexadecimal escape, with no
295 following digits, giving a character whose value is zero.
296 .P
297 If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \ex is
298 as just described only when it is followed by two hexadecimal digits.
299 Otherwise, it matches a literal "x" character. In JavaScript mode, support for
300 code points greater than 256 is provided by \eu, which must be followed by
301 four hexadecimal digits; otherwise it matches a literal "u" character.
302 Character codes specified by \eu in JavaScript mode are constrained in the same
303 was as those specified by \ex in non-JavaScript mode.
304 .P
305 Characters whose value is less than 256 can be defined by either of the two
306 syntaxes for \ex (or by \eu in JavaScript mode). There is no difference in the
307 way they are handled. For example, \exdc is exactly the same as \ex{dc} (or
308 \eu00dc in JavaScript mode).
309 .P
310 After \e0 up to two further octal digits are read. If there are fewer than two
311 digits, just those that are present are used. Thus the sequence \e0\ex\e07
312 specifies two binary zeros followed by a BEL character (code value 7). Make
313 sure you supply two digits after the initial zero if the pattern character that
314 follows is itself an octal digit.
315 .P
316 The handling of a backslash followed by a digit other than 0 is complicated.
317 Outside a character class, PCRE reads it and any following digits as a decimal
318 number. If the number is less than 10, or if there have been at least that many
319 previous capturing left parentheses in the expression, the entire sequence is
320 taken as a \fIback reference\fP. A description of how this works is given
321 .\" HTML <a href="#backreferences">
322 .\" </a>
323 later,
324 .\"
325 following the discussion of
326 .\" HTML <a href="#subpattern">
327 .\" </a>
328 parenthesized subpatterns.
329 .\"
330 .P
331 Inside a character class, or if the decimal number is greater than 9 and there
332 have not been that many capturing subpatterns, PCRE re-reads up to three octal
333 digits following the backslash, and uses them to generate a data character. Any
334 subsequent digits stand for themselves. The value of the character is
335 constrained in the same way as characters specified in hexadecimal.
336 For example:
337 .sp
338 \e040 is another way of writing an ASCII space
339 .\" JOIN
340 \e40 is the same, provided there are fewer than 40
341 previous capturing subpatterns
342 \e7 is always a back reference
343 .\" JOIN
344 \e11 might be a back reference, or another way of
345 writing a tab
346 \e011 is always a tab
347 \e0113 is a tab followed by the character "3"
348 .\" JOIN
349 \e113 might be a back reference, otherwise the
350 character with octal code 113
351 .\" JOIN
352 \e377 might be a back reference, otherwise
353 the value 255 (decimal)
354 .\" JOIN
355 \e81 is either a back reference, or a binary zero
356 followed by the two characters "8" and "1"
357 .sp
358 Note that octal values of 100 or greater must not be introduced by a leading
359 zero, because no more than three octal digits are ever read.
360 .P
361 All the sequences that define a single character value can be used both inside
362 and outside character classes. In addition, inside a character class, \eb is
363 interpreted as the backspace character (hex 08).
364 .P
365 \eN is not allowed in a character class. \eB, \eR, and \eX are not special
366 inside a character class. Like other unrecognized escape sequences, they are
367 treated as the literal characters "B", "R", and "X" by default, but cause an
368 error if the PCRE_EXTRA option is set. Outside a character class, these
369 sequences have different meanings.
370 .
371 .
372 .SS "Unsupported escape sequences"
373 .rs
374 .sp
375 In Perl, the sequences \el, \eL, \eu, and \eU are recognized by its string
376 handler and used to modify the case of following characters. By default, PCRE
377 does not support these escape sequences. However, if the PCRE_JAVASCRIPT_COMPAT
378 option is set, \eU matches a "U" character, and \eu can be used to define a
379 character by code point, as described in the previous section.
380 .
381 .
382 .SS "Absolute and relative back references"
383 .rs
384 .sp
385 The sequence \eg followed by an unsigned or a negative number, optionally
386 enclosed in braces, is an absolute or relative back reference. A named back
387 reference can be coded as \eg{name}. Back references are discussed
388 .\" HTML <a href="#backreferences">
389 .\" </a>
390 later,
391 .\"
392 following the discussion of
393 .\" HTML <a href="#subpattern">
394 .\" </a>
395 parenthesized subpatterns.
396 .\"
397 .
398 .
399 .SS "Absolute and relative subroutine calls"
400 .rs
401 .sp
402 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
403 a number enclosed either in angle brackets or single quotes, is an alternative
404 syntax for referencing a subpattern as a "subroutine". Details are discussed
405 .\" HTML <a href="#onigurumasubroutines">
406 .\" </a>
407 later.
408 .\"
409 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
410 synonymous. The former is a back reference; the latter is a
411 .\" HTML <a href="#subpatternsassubroutines">
412 .\" </a>
413 subroutine
414 .\"
415 call.
416 .
417 .
418 .\" HTML <a name="genericchartypes"></a>
419 .SS "Generic character types"
420 .rs
421 .sp
422 Another use of backslash is for specifying generic character types:
423 .sp
424 \ed any decimal digit
425 \eD any character that is not a decimal digit
426 \eh any horizontal white space character
427 \eH any character that is not a horizontal white space character
428 \es any white space character
429 \eS any character that is not a white space character
430 \ev any vertical white space character
431 \eV any character that is not a vertical white space character
432 \ew any "word" character
433 \eW any "non-word" character
434 .sp
435 There is also the single sequence \eN, which matches a non-newline character.
436 This is the same as
437 .\" HTML <a href="#fullstopdot">
438 .\" </a>
439 the "." metacharacter
440 .\"
441 when PCRE_DOTALL is not set. Perl also uses \eN to match characters by name;
442 PCRE does not support this.
443 .P
444 Each pair of lower and upper case escape sequences partitions the complete set
445 of characters into two disjoint sets. Any given character matches one, and only
446 one, of each pair. The sequences can appear both inside and outside character
447 classes. They each match one character of the appropriate type. If the current
448 matching point is at the end of the subject string, all of them fail, because
449 there is no character to match.
450 .P
451 For compatibility with Perl, \es does not match the VT character (code 11).
452 This makes it different from the the POSIX "space" class. The \es characters
453 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
454 included in a Perl script, \es may match the VT character. In PCRE, it never
455 does.
456 .P
457 A "word" character is an underscore or any character that is a letter or digit.
458 By default, the definition of letters and digits is controlled by PCRE's
459 low-valued character tables, and may vary if locale-specific matching is taking
460 place (see
461 .\" HTML <a href="pcreapi.html#localesupport">
462 .\" </a>
463 "Locale support"
464 .\"
465 in the
466 .\" HREF
467 \fBpcreapi\fP
468 .\"
469 page). For example, in a French locale such as "fr_FR" in Unix-like systems,
470 or "french" in Windows, some character codes greater than 128 are used for
471 accented letters, and these are then matched by \ew. The use of locales with
472 Unicode is discouraged.
473 .P
474 By default, in a UTF mode, characters with values greater than 128 never match
475 \ed, \es, or \ew, and always match \eD, \eS, and \eW. These sequences retain
476 their original meanings from before UTF support was available, mainly for
477 efficiency reasons. However, if PCRE is compiled with Unicode property support,
478 and the PCRE_UCP option is set, the behaviour is changed so that Unicode
479 properties are used to determine character types, as follows:
480 .sp
481 \ed any character that \ep{Nd} matches (decimal digit)
482 \es any character that \ep{Z} matches, plus HT, LF, FF, CR
483 \ew any character that \ep{L} or \ep{N} matches, plus underscore
484 .sp
485 The upper case escapes match the inverse sets of characters. Note that \ed
486 matches only decimal digits, whereas \ew matches any Unicode digit, as well as
487 any Unicode letter, and underscore. Note also that PCRE_UCP affects \eb, and
488 \eB because they are defined in terms of \ew and \eW. Matching these sequences
489 is noticeably slower when PCRE_UCP is set.
490 .P
491 The sequences \eh, \eH, \ev, and \eV are features that were added to Perl at
492 release 5.10. In contrast to the other sequences, which match only ASCII
493 characters by default, these always match certain high-valued codepoints,
494 whether or not PCRE_UCP is set. The horizontal space characters are:
495 .sp
496 U+0009 Horizontal tab (HT)
497 U+0020 Space
498 U+00A0 Non-break space
499 U+1680 Ogham space mark
500 U+180E Mongolian vowel separator
501 U+2000 En quad
502 U+2001 Em quad
503 U+2002 En space
504 U+2003 Em space
505 U+2004 Three-per-em space
506 U+2005 Four-per-em space
507 U+2006 Six-per-em space
508 U+2007 Figure space
509 U+2008 Punctuation space
510 U+2009 Thin space
511 U+200A Hair space
512 U+202F Narrow no-break space
513 U+205F Medium mathematical space
514 U+3000 Ideographic space
515 .sp
516 The vertical space characters are:
517 .sp
518 U+000A Linefeed (LF)
519 U+000B Vertical tab (VT)
520 U+000C Form feed (FF)
521 U+000D Carriage return (CR)
522 U+0085 Next line (NEL)
523 U+2028 Line separator
524 U+2029 Paragraph separator
525 .sp
526 In 8-bit, non-UTF-8 mode, only the characters with codepoints less than 256 are
527 relevant.
528 .
529 .
530 .\" HTML <a name="newlineseq"></a>
531 .SS "Newline sequences"
532 .rs
533 .sp
534 Outside a character class, by default, the escape sequence \eR matches any
535 Unicode newline sequence. In 8-bit non-UTF-8 mode \eR is equivalent to the
536 following:
537 .sp
538 (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
539 .sp
540 This is an example of an "atomic group", details of which are given
541 .\" HTML <a href="#atomicgroup">
542 .\" </a>
543 below.
544 .\"
545 This particular group matches either the two-character sequence CR followed by
546 LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
547 U+000B), FF (form feed, U+000C), CR (carriage return, U+000D), or NEL (next
548 line, U+0085). The two-character sequence is treated as a single unit that
549 cannot be split.
550 .P
551 In other modes, two additional characters whose codepoints are greater than 255
552 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
553 Unicode character property support is not needed for these characters to be
554 recognized.
555 .P
556 It is possible to restrict \eR to match only CR, LF, or CRLF (instead of the
557 complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
558 either at compile time or when the pattern is matched. (BSR is an abbrevation
559 for "backslash R".) This can be made the default when PCRE is built; if this is
560 the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
561 It is also possible to specify these settings by starting a pattern string with
562 one of the following sequences:
563 .sp
564 (*BSR_ANYCRLF) CR, LF, or CRLF only
565 (*BSR_UNICODE) any Unicode newline sequence
566 .sp
567 These override the default and the options given to the compiling function, but
568 they can themselves be overridden by options given to a matching function. Note
569 that these special settings, which are not Perl-compatible, are recognized only
570 at the very start of a pattern, and that they must be in upper case. If more
571 than one of them is present, the last one is used. They can be combined with a
572 change of newline convention; for example, a pattern can start with:
573 .sp
575 .sp
576 They can also be combined with the (*UTF8), (*UTF16), (*UTF32) or (*UCP) special
577 sequences. Inside a character class, \eR is treated as an unrecognized escape
578 sequence, and so matches the letter "R" by default, but causes an error if
579 PCRE_EXTRA is set.
580 .
581 .
582 .\" HTML <a name="uniextseq"></a>
583 .SS Unicode character properties
584 .rs
585 .sp
586 When PCRE is built with Unicode character property support, three additional
587 escape sequences that match characters with specific properties are available.
588 When in 8-bit non-UTF-8 mode, these sequences are of course limited to testing
589 characters whose codepoints are less than 256, but they do work in this mode.
590 The extra escape sequences are:
591 .sp
592 \ep{\fIxx\fP} a character with the \fIxx\fP property
593 \eP{\fIxx\fP} a character without the \fIxx\fP property
594 \eX a Unicode extended grapheme cluster
595 .sp
596 The property names represented by \fIxx\fP above are limited to the Unicode
597 script names, the general category properties, "Any", which matches any
598 character (including newline), and some special PCRE properties (described
599 in the
600 .\" HTML <a href="#extraprops">
601 .\" </a>
602 next section).
603 .\"
604 Other Perl properties such as "InMusicalSymbols" are not currently supported by
605 PCRE. Note that \eP{Any} does not match any characters, so always causes a
606 match failure.
607 .P
608 Sets of Unicode characters are defined as belonging to certain scripts. A
609 character from one of these sets can be matched using a script name. For
610 example:
611 .sp
612 \ep{Greek}
613 \eP{Han}
614 .sp
615 Those that are not part of an identified script are lumped together as
616 "Common". The current list of scripts is:
617 .P
618 Arabic,
619 Armenian,
620 Avestan,
621 Balinese,
622 Bamum,
623 Batak,
624 Bengali,
625 Bopomofo,
626 Brahmi,
627 Braille,
628 Buginese,
629 Buhid,
630 Canadian_Aboriginal,
631 Carian,
632 Chakma,
633 Cham,
634 Cherokee,
635 Common,
636 Coptic,
637 Cuneiform,
638 Cypriot,
639 Cyrillic,
640 Deseret,
641 Devanagari,
642 Egyptian_Hieroglyphs,
643 Ethiopic,
644 Georgian,
645 Glagolitic,
646 Gothic,
647 Greek,
648 Gujarati,
649 Gurmukhi,
650 Han,
651 Hangul,
652 Hanunoo,
653 Hebrew,
654 Hiragana,
655 Imperial_Aramaic,
656 Inherited,
657 Inscriptional_Pahlavi,
658 Inscriptional_Parthian,
659 Javanese,
660 Kaithi,
661 Kannada,
662 Katakana,
663 Kayah_Li,
664 Kharoshthi,
665 Khmer,
666 Lao,
667 Latin,
668 Lepcha,
669 Limbu,
670 Linear_B,
671 Lisu,
672 Lycian,
673 Lydian,
674 Malayalam,
675 Mandaic,
676 Meetei_Mayek,
677 Meroitic_Cursive,
678 Meroitic_Hieroglyphs,
679 Miao,
680 Mongolian,
681 Myanmar,
682 New_Tai_Lue,
683 Nko,
684 Ogham,
685 Old_Italic,
686 Old_Persian,
687 Old_South_Arabian,
688 Old_Turkic,
689 Ol_Chiki,
690 Oriya,
691 Osmanya,
692 Phags_Pa,
693 Phoenician,
694 Rejang,
695 Runic,
696 Samaritan,
697 Saurashtra,
698 Sharada,
699 Shavian,
700 Sinhala,
701 Sora_Sompeng,
702 Sundanese,
703 Syloti_Nagri,
704 Syriac,
705 Tagalog,
706 Tagbanwa,
707 Tai_Le,
708 Tai_Tham,
709 Tai_Viet,
710 Takri,
711 Tamil,
712 Telugu,
713 Thaana,
714 Thai,
715 Tibetan,
716 Tifinagh,
717 Ugaritic,
718 Vai,
719 Yi.
720 .P
721 Each character has exactly one Unicode general category property, specified by
722 a two-letter abbreviation. For compatibility with Perl, negation can be
723 specified by including a circumflex between the opening brace and the property
724 name. For example, \ep{^Lu} is the same as \eP{Lu}.
725 .P
726 If only one letter is specified with \ep or \eP, it includes all the general
727 category properties that start with that letter. In this case, in the absence
728 of negation, the curly brackets in the escape sequence are optional; these two
729 examples have the same effect:
730 .sp
731 \ep{L}
732 \epL
733 .sp
734 The following general category property codes are supported:
735 .sp
736 C Other
737 Cc Control
738 Cf Format
739 Cn Unassigned
740 Co Private use
741 Cs Surrogate
742 .sp
743 L Letter
744 Ll Lower case letter
745 Lm Modifier letter
746 Lo Other letter
747 Lt Title case letter
748 Lu Upper case letter
749 .sp
750 M Mark
751 Mc Spacing mark
752 Me Enclosing mark
753 Mn Non-spacing mark
754 .sp
755 N Number
756 Nd Decimal number
757 Nl Letter number
758 No Other number
759 .sp
760 P Punctuation
761 Pc Connector punctuation
762 Pd Dash punctuation
763 Pe Close punctuation
764 Pf Final punctuation
765 Pi Initial punctuation
766 Po Other punctuation
767 Ps Open punctuation
768 .sp
769 S Symbol
770 Sc Currency symbol
771 Sk Modifier symbol
772 Sm Mathematical symbol
773 So Other symbol
774 .sp
775 Z Separator
776 Zl Line separator
777 Zp Paragraph separator
778 Zs Space separator
779 .sp
780 The special property L& is also supported: it matches a character that has
781 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
782 a modifier or "other".
783 .P
784 The Cs (Surrogate) property applies only to characters in the range U+D800 to
785 U+DFFF. Such characters are not valid in Unicode strings and so
786 cannot be tested by PCRE, unless UTF validity checking has been turned off
787 (see the discussion of PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK and
788 PCRE_NO_UTF32_CHECK in the
789 .\" HREF
790 \fBpcreapi\fP
791 .\"
792 page). Perl does not support the Cs property.
793 .P
794 The long synonyms for property names that Perl supports (such as \ep{Letter})
795 are not supported by PCRE, nor is it permitted to prefix any of these
796 properties with "Is".
797 .P
798 No character that is in the Unicode table has the Cn (unassigned) property.
799 Instead, this property is assumed for any code point that is not in the
800 Unicode table.
801 .P
802 Specifying caseless matching does not affect these escape sequences. For
803 example, \ep{Lu} always matches only upper case letters.
804 .P
805 Matching characters by Unicode property is not fast, because PCRE has to do a
806 multistage table lookup in order to find a character's property. That is why
807 the traditional escape sequences such as \ed and \ew do not use Unicode
808 properties in PCRE by default, though you can make them do so by setting the
809 PCRE_UCP option or by starting the pattern with (*UCP).
810 .
811 .
812 .SS Extended grapheme clusters
813 .rs
814 .sp
815 The \eX escape matches any number of Unicode characters that form an "extended
816 grapheme cluster", and treats the sequence as an atomic group
817 .\" HTML <a href="#atomicgroup">
818 .\" </a>
819 (see below).
820 .\"
821 Up to and including release 8.31, PCRE matched an earlier, simpler definition
822 that was equivalent to
823 .sp
824 (?>\ePM\epM*)
825 .sp
826 That is, it matched a character without the "mark" property, followed by zero
827 or more characters with the "mark" property. Characters with the "mark"
828 property are typically non-spacing accents that affect the preceding character.
829 .P
830 This simple definition was extended in Unicode to include more complicated
831 kinds of composite character by giving each character a grapheme breaking
832 property, and creating rules that use these properties to define the boundaries
833 of extended grapheme clusters. In releases of PCRE later than 8.31, \eX matches
834 one of these clusters.
835 .P
836 \eX always matches at least one character. Then it decides whether to add
837 additional characters according to the following rules for ending a cluster:
838 .P
839 1. End at the end of the subject string.
840 .P
841 2. Do not end between CR and LF; otherwise end after any control character.
842 .P
843 3. Do not break Hangul (a Korean script) syllable sequences. Hangul characters
844 are of five types: L, V, T, LV, and LVT. An L character may be followed by an
845 L, V, LV, or LVT character; an LV or V character may be followed by a V or T
846 character; an LVT or T character may be follwed only by a T character.
847 .P
848 4. Do not end before extending characters or spacing marks. Characters with
849 the "mark" property always have the "extend" grapheme breaking property.
850 .P
851 5. Do not end after prepend characters.
852 .P
853 6. Otherwise, end the cluster.
854 .
855 .
856 .\" HTML <a name="extraprops"></a>
857 .SS PCRE's additional properties
858 .rs
859 .sp
860 As well as the standard Unicode properties described above, PCRE supports four
861 more that make it possible to convert traditional escape sequences such as \ew
862 and \es and POSIX character classes to use Unicode properties. PCRE uses these
863 non-standard, non-Perl properties internally when PCRE_UCP is set. They are:
864 .sp
865 Xan Any alphanumeric character
866 Xps Any POSIX space character
867 Xsp Any Perl space character
868 Xwd Any Perl "word" character
869 .sp
870 Xan matches characters that have either the L (letter) or the N (number)
871 property. Xps matches the characters tab, linefeed, vertical tab, form feed, or
872 carriage return, and any other character that has the Z (separator) property.
873 Xsp is the same as Xps, except that vertical tab is excluded. Xwd matches the
874 same characters as Xan, plus underscore.
875 .
876 .
877 .\" HTML <a name="resetmatchstart"></a>
878 .SS "Resetting the match start"
879 .rs
880 .sp
881 The escape sequence \eK causes any previously matched characters not to be
882 included in the final matched sequence. For example, the pattern:
883 .sp
884 foo\eKbar
885 .sp
886 matches "foobar", but reports that it has matched "bar". This feature is
887 similar to a lookbehind assertion
888 .\" HTML <a href="#lookbehind">
889 .\" </a>
890 (described below).
891 .\"
892 However, in this case, the part of the subject before the real match does not
893 have to be of fixed length, as lookbehind assertions do. The use of \eK does
894 not interfere with the setting of
895 .\" HTML <a href="#subpattern">
896 .\" </a>
897 captured substrings.
898 .\"
899 For example, when the pattern
900 .sp
901 (foo)\eKbar
902 .sp
903 matches "foobar", the first substring is still set to "foo".
904 .P
905 Perl documents that the use of \eK within assertions is "not well defined". In
906 PCRE, \eK is acted upon when it occurs inside positive assertions, but is
907 ignored in negative assertions.
908 .
909 .
910 .\" HTML <a name="smallassertions"></a>
911 .SS "Simple assertions"
912 .rs
913 .sp
914 The final use of backslash is for certain simple assertions. An assertion
915 specifies a condition that has to be met at a particular point in a match,
916 without consuming any characters from the subject string. The use of
917 subpatterns for more complicated assertions is described
918 .\" HTML <a href="#bigassertions">
919 .\" </a>
920 below.
921 .\"
922 The backslashed assertions are:
923 .sp
924 \eb matches at a word boundary
925 \eB matches when not at a word boundary
926 \eA matches at the start of the subject
927 \eZ matches at the end of the subject
928 also matches before a newline at the end of the subject
929 \ez matches only at the end of the subject
930 \eG matches at the first matching position in the subject
931 .sp
932 Inside a character class, \eb has a different meaning; it matches the backspace
933 character. If any other of these assertions appears in a character class, by
934 default it matches the corresponding literal character (for example, \eB
935 matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
936 escape sequence" error is generated instead.
937 .P
938 A word boundary is a position in the subject string where the current character
939 and the previous character do not both match \ew or \eW (i.e. one matches
940 \ew and the other matches \eW), or the start or end of the string if the
941 first or last character matches \ew, respectively. In a UTF mode, the meanings
942 of \ew and \eW can be changed by setting the PCRE_UCP option. When this is
943 done, it also affects \eb and \eB. Neither PCRE nor Perl has a separate "start
944 of word" or "end of word" metasequence. However, whatever follows \eb normally
945 determines which it is. For example, the fragment \eba matches "a" at the start
946 of a word.
947 .P
948 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
949 dollar (described in the next section) in that they only ever match at the very
950 start and end of the subject string, whatever options are set. Thus, they are
951 independent of multiline mode. These three assertions are not affected by the
952 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
953 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
954 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
955 at a point other than the beginning of the subject, \eA can never match. The
956 difference between \eZ and \ez is that \eZ matches before a newline at the end
957 of the string as well as at the very end, whereas \ez matches only at the end.
958 .P
959 The \eG assertion is true only when the current matching position is at the
960 start point of the match, as specified by the \fIstartoffset\fP argument of
961 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
962 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
963 arguments, you can mimic Perl's /g option, and it is in this kind of
964 implementation where \eG can be useful.
965 .P
966 Note, however, that PCRE's interpretation of \eG, as the start of the current
967 match, is subtly different from Perl's, which defines it as the end of the
968 previous match. In Perl, these can be different when the previously matched
969 string was empty. Because PCRE does just one match at a time, it cannot
970 reproduce this behaviour.
971 .P
972 If all the alternatives of a pattern begin with \eG, the expression is anchored
973 to the starting match position, and the "anchored" flag is set in the compiled
974 regular expression.
975 .
976 .
978 .rs
979 .sp
980 The circumflex and dollar metacharacters are zero-width assertions. That is,
981 they test for a particular condition being true without consuming any
982 characters from the subject string.
983 .P
984 Outside a character class, in the default matching mode, the circumflex
985 character is an assertion that is true only if the current matching point is at
986 the start of the subject string. If the \fIstartoffset\fP argument of
987 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
988 option is unset. Inside a character class, circumflex has an entirely different
989 meaning
990 .\" HTML <a href="#characterclass">
991 .\" </a>
992 (see below).
993 .\"
994 .P
995 Circumflex need not be the first character of the pattern if a number of
996 alternatives are involved, but it should be the first thing in each alternative
997 in which it appears if the pattern is ever to match that branch. If all
998 possible alternatives start with a circumflex, that is, if the pattern is
999 constrained to match only at the start of the subject, it is said to be an
1000 "anchored" pattern. (There are also other constructs that can cause a pattern
1001 to be anchored.)
1002 .P
1003 The dollar character is an assertion that is true only if the current matching
1004 point is at the end of the subject string, or immediately before a newline at
1005 the end of the string (by default). Note, however, that it does not actually
1006 match the newline. Dollar need not be the last character of the pattern if a
1007 number of alternatives are involved, but it should be the last item in any
1008 branch in which it appears. Dollar has no special meaning in a character class.
1009 .P
1010 The meaning of dollar can be changed so that it matches only at the very end of
1011 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
1012 does not affect the \eZ assertion.
1013 .P
1014 The meanings of the circumflex and dollar characters are changed if the
1015 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
1016 immediately after internal newlines as well as at the start of the subject
1017 string. It does not match after a newline that ends the string. A dollar
1018 matches before any newlines in the string, as well as at the very end, when
1019 PCRE_MULTILINE is set. When newline is specified as the two-character
1020 sequence CRLF, isolated CR and LF characters do not indicate newlines.
1021 .P
1022 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
1023 \en represents a newline) in multiline mode, but not otherwise. Consequently,
1024 patterns that are anchored in single line mode because all branches start with
1025 ^ are not anchored in multiline mode, and a match for circumflex is possible
1026 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
1027 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
1028 .P
1029 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
1030 end of the subject in both modes, and if all branches of a pattern start with
1031 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
1032 .
1033 .
1034 .\" HTML <a name="fullstopdot"></a>
1036 .rs
1037 .sp
1038 Outside a character class, a dot in the pattern matches any one character in
1039 the subject string except (by default) a character that signifies the end of a
1040 line.
1041 .P
1042 When a line ending is defined as a single character, dot never matches that
1043 character; when the two-character sequence CRLF is used, dot does not match CR
1044 if it is immediately followed by LF, but otherwise it matches all characters
1045 (including isolated CRs and LFs). When any Unicode line endings are being
1046 recognized, dot does not match CR or LF or any of the other line ending
1047 characters.
1048 .P
1049 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
1050 option is set, a dot matches any one character, without exception. If the
1051 two-character sequence CRLF is present in the subject string, it takes two dots
1052 to match it.
1053 .P
1054 The handling of dot is entirely independent of the handling of circumflex and
1055 dollar, the only relationship being that they both involve newlines. Dot has no
1056 special meaning in a character class.
1057 .P
1058 The escape sequence \eN behaves like a dot, except that it is not affected by
1059 the PCRE_DOTALL option. In other words, it matches any character except one
1060 that signifies the end of a line. Perl also uses \eN to match characters by
1061 name; PCRE does not support this.
1062 .
1063 .
1065 .rs
1066 .sp
1067 Outside a character class, the escape sequence \eC matches any one data unit,
1068 whether or not a UTF mode is set. In the 8-bit library, one data unit is one
1069 byte; in the 16-bit library it is a 16-bit unit; in the 32-bit library it is
1070 a 32-bit unit. Unlike a dot, \eC always
1071 matches line-ending characters. The feature is provided in Perl in order to
1072 match individual bytes in UTF-8 mode, but it is unclear how it can usefully be
1073 used. Because \eC breaks up characters into individual data units, matching one
1074 unit with \eC in a UTF mode means that the rest of the string may start with a
1075 malformed UTF character. This has undefined results, because PCRE assumes that
1076 it is dealing with valid UTF strings (and by default it checks this at the
1077 start of processing unless the PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK or
1078 PCRE_NO_UTF32_CHECK option is used).
1079 .P
1080 PCRE does not allow \eC to appear in lookbehind assertions
1081 .\" HTML <a href="#lookbehind">
1082 .\" </a>
1083 (described below)
1084 .\"
1085 in a UTF mode, because this would make it impossible to calculate the length of
1086 the lookbehind.
1087 .P
1088 In general, the \eC escape sequence is best avoided. However, one
1089 way of using it that avoids the problem of malformed UTF characters is to use a
1090 lookahead to check the length of the next character, as in this pattern, which
1091 could be used with a UTF-8 string (ignore white space and line breaks):
1092 .sp
1093 (?| (?=[\ex00-\ex7f])(\eC) |
1094 (?=[\ex80-\ex{7ff}])(\eC)(\eC) |
1095 (?=[\ex{800}-\ex{ffff}])(\eC)(\eC)(\eC) |
1096 (?=[\ex{10000}-\ex{1fffff}])(\eC)(\eC)(\eC)(\eC))
1097 .sp
1098 A group that starts with (?| resets the capturing parentheses numbers in each
1099 alternative (see
1100 .\" HTML <a href="#dupsubpatternnumber">
1101 .\" </a>
1102 "Duplicate Subpattern Numbers"
1103 .\"
1104 below). The assertions at the start of each branch check the next UTF-8
1105 character for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The
1106 character's individual bytes are then captured by the appropriate number of
1107 groups.
1108 .
1109 .
1110 .\" HTML <a name="characterclass"></a>
1112 .rs
1113 .sp
1114 An opening square bracket introduces a character class, terminated by a closing
1115 square bracket. A closing square bracket on its own is not special by default.
1116 However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
1117 bracket causes a compile-time error. If a closing square bracket is required as
1118 a member of the class, it should be the first data character in the class
1119 (after an initial circumflex, if present) or escaped with a backslash.
1120 .P
1121 A character class matches a single character in the subject. In a UTF mode, the
1122 character may be more than one data unit long. A matched character must be in
1123 the set of characters defined by the class, unless the first character in the
1124 class definition is a circumflex, in which case the subject character must not
1125 be in the set defined by the class. If a circumflex is actually required as a
1126 member of the class, ensure it is not the first character, or escape it with a
1127 backslash.
1128 .P
1129 For example, the character class [aeiou] matches any lower case vowel, while
1130 [^aeiou] matches any character that is not a lower case vowel. Note that a
1131 circumflex is just a convenient notation for specifying the characters that
1132 are in the class by enumerating those that are not. A class that starts with a
1133 circumflex is not an assertion; it still consumes a character from the subject
1134 string, and therefore it fails if the current pointer is at the end of the
1135 string.
1136 .P
1137 In UTF-8 (UTF-16, UTF-32) mode, characters with values greater than 255 (0xffff)
1138 can be included in a class as a literal string of data units, or by using the
1139 \ex{ escaping mechanism.
1140 .P
1141 When caseless matching is set, any letters in a class represent both their
1142 upper case and lower case versions, so for example, a caseless [aeiou] matches
1143 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1144 caseful version would. In a UTF mode, PCRE always understands the concept of
1145 case for characters whose values are less than 128, so caseless matching is
1146 always possible. For characters with higher values, the concept of case is
1147 supported if PCRE is compiled with Unicode property support, but not otherwise.
1148 If you want to use caseless matching in a UTF mode for characters 128 and
1149 above, you must ensure that PCRE is compiled with Unicode property support as
1150 well as with UTF support.
1151 .P
1152 Characters that might indicate line breaks are never treated in any special way
1153 when matching character classes, whatever line-ending sequence is in use, and
1154 whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1155 such as [^a] always matches one of these characters.
1156 .P
1157 The minus (hyphen) character can be used to specify a range of characters in a
1158 character class. For example, [d-m] matches any letter between d and m,
1159 inclusive. If a minus character is required in a class, it must be escaped with
1160 a backslash or appear in a position where it cannot be interpreted as
1161 indicating a range, typically as the first or last character in the class.
1162 .P
1163 It is not possible to have the literal character "]" as the end character of a
1164 range. A pattern such as [W-]46] is interpreted as a class of two characters
1165 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1166 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1167 the end of range, so [W-\e]46] is interpreted as a class containing a range
1168 followed by two other characters. The octal or hexadecimal representation of
1169 "]" can also be used to end a range.
1170 .P
1171 Ranges operate in the collating sequence of character values. They can also be
1172 used for characters specified numerically, for example [\e000-\e037]. Ranges
1173 can include any characters that are valid for the current mode.
1174 .P
1175 If a range that includes letters is used when caseless matching is set, it
1176 matches the letters in either case. For example, [W-c] is equivalent to
1177 [][\e\e^_`wxyzabc], matched caselessly, and in a non-UTF mode, if character
1178 tables for a French locale are in use, [\exc8-\excb] matches accented E
1179 characters in both cases. In UTF modes, PCRE supports the concept of case for
1180 characters with values greater than 128 only when it is compiled with Unicode
1181 property support.
1182 .P
1183 The character escape sequences \ed, \eD, \eh, \eH, \ep, \eP, \es, \eS, \ev,
1184 \eV, \ew, and \eW may appear in a character class, and add the characters that
1185 they match to the class. For example, [\edABCDEF] matches any hexadecimal
1186 digit. In UTF modes, the PCRE_UCP option affects the meanings of \ed, \es, \ew
1187 and their upper case partners, just as it does when they appear outside a
1188 character class, as described in the section entitled
1189 .\" HTML <a href="#genericchartypes">
1190 .\" </a>
1191 "Generic character types"
1192 .\"
1193 above. The escape sequence \eb has a different meaning inside a character
1194 class; it matches the backspace character. The sequences \eB, \eN, \eR, and \eX
1195 are not special inside a character class. Like any other unrecognized escape
1196 sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1197 default, but cause an error if the PCRE_EXTRA option is set.
1198 .P
1199 A circumflex can conveniently be used with the upper case character types to
1200 specify a more restricted set of characters than the matching lower case type.
1201 For example, the class [^\eW_] matches any letter or digit, but not underscore,
1202 whereas [\ew] includes underscore. A positive character class should be read as
1203 "something OR something OR ..." and a negative class as "NOT something AND NOT
1204 something AND NOT ...".
1205 .P
1206 The only metacharacters that are recognized in character classes are backslash,
1207 hyphen (only where it can be interpreted as specifying a range), circumflex
1208 (only at the start), opening square bracket (only when it can be interpreted as
1209 introducing a POSIX class name - see the next section), and the terminating
1210 closing square bracket. However, escaping other non-alphanumeric characters
1211 does no harm.
1212 .
1213 .
1215 .rs
1216 .sp
1217 Perl supports the POSIX notation for character classes. This uses names
1218 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1219 this notation. For example,
1220 .sp
1221 [01[:alpha:]%]
1222 .sp
1223 matches "0", "1", any alphabetic character, or "%". The supported class names
1224 are:
1225 .sp
1226 alnum letters and digits
1227 alpha letters
1228 ascii character codes 0 - 127
1229 blank space or tab only
1230 cntrl control characters
1231 digit decimal digits (same as \ed)
1232 graph printing characters, excluding space
1233 lower lower case letters
1234 print printing characters, including space
1235 punct printing characters, excluding letters and digits and space
1236 space white space (not quite the same as \es)
1237 upper upper case letters
1238 word "word" characters (same as \ew)
1239 xdigit hexadecimal digits
1240 .sp
1241 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
1242 space (32). Notice that this list includes the VT character (code 11). This
1243 makes "space" different to \es, which does not include VT (for Perl
1244 compatibility).
1245 .P
1246 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1247 5.8. Another Perl extension is negation, which is indicated by a ^ character
1248 after the colon. For example,
1249 .sp
1250 [12[:^digit:]]
1251 .sp
1252 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1253 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1254 supported, and an error is given if they are encountered.
1255 .P
1256 By default, in UTF modes, characters with values greater than 128 do not match
1257 any of the POSIX character classes. However, if the PCRE_UCP option is passed
1258 to \fBpcre_compile()\fP, some of the classes are changed so that Unicode
1259 character properties are used. This is achieved by replacing the POSIX classes
1260 by other sequences, as follows:
1261 .sp
1262 [:alnum:] becomes \ep{Xan}
1263 [:alpha:] becomes \ep{L}
1264 [:blank:] becomes \eh
1265 [:digit:] becomes \ep{Nd}
1266 [:lower:] becomes \ep{Ll}
1267 [:space:] becomes \ep{Xps}
1268 [:upper:] becomes \ep{Lu}
1269 [:word:] becomes \ep{Xwd}
1270 .sp
1271 Negated versions, such as [:^alpha:] use \eP instead of \ep. The other POSIX
1272 classes are unchanged, and match only characters with code points less than
1273 128.
1274 .
1275 .
1277 .rs
1278 .sp
1279 Vertical bar characters are used to separate alternative patterns. For example,
1280 the pattern
1281 .sp
1282 gilbert|sullivan
1283 .sp
1284 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1285 and an empty alternative is permitted (matching the empty string). The matching
1286 process tries each alternative in turn, from left to right, and the first one
1287 that succeeds is used. If the alternatives are within a subpattern
1288 .\" HTML <a href="#subpattern">
1289 .\" </a>
1290 (defined below),
1291 .\"
1292 "succeeds" means matching the rest of the main pattern as well as the
1293 alternative in the subpattern.
1294 .
1295 .
1297 .rs
1298 .sp
1299 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1300 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1301 the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1302 The option letters are
1303 .sp
1304 i for PCRE_CASELESS
1306 s for PCRE_DOTALL
1307 x for PCRE_EXTENDED
1308 .sp
1309 For example, (?im) sets caseless, multiline matching. It is also possible to
1310 unset these options by preceding the letter with a hyphen, and a combined
1311 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1312 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1313 permitted. If a letter appears both before and after the hyphen, the option is
1314 unset.
1315 .P
1316 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1317 changed in the same way as the Perl-compatible options by using the characters
1318 J, U and X respectively.
1319 .P
1320 When one of these option changes occurs at top level (that is, not inside
1321 subpattern parentheses), the change applies to the remainder of the pattern
1322 that follows. If the change is placed right at the start of a pattern, PCRE
1323 extracts it into the global options (and it will therefore show up in data
1324 extracted by the \fBpcre_fullinfo()\fP function).
1325 .P
1326 An option change within a subpattern (see below for a description of
1327 subpatterns) affects only that part of the subpattern that follows it, so
1328 .sp
1329 (a(?i)b)c
1330 .sp
1331 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1332 By this means, options can be made to have different settings in different
1333 parts of the pattern. Any changes made in one alternative do carry on
1334 into subsequent branches within the same subpattern. For example,
1335 .sp
1336 (a(?i)b|c)
1337 .sp
1338 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1339 branch is abandoned before the option setting. This is because the effects of
1340 option settings happen at compile time. There would be some very weird
1341 behaviour otherwise.
1342 .P
1343 \fBNote:\fP There are other PCRE-specific options that can be set by the
1344 application when the compiling or matching functions are called. In some cases
1345 the pattern can contain special leading sequences such as (*CRLF) to override
1346 what the application has set or what has been defaulted. Details are given in
1347 the section entitled
1348 .\" HTML <a href="#newlineseq">
1349 .\" </a>
1350 "Newline sequences"
1351 .\"
1352 above. There are also the (*UTF8), (*UTF16),(*UTF32) and (*UCP) leading
1353 sequences that can be used to set UTF and Unicode property modes; they are
1354 equivalent to setting the PCRE_UTF8, PCRE_UTF16, PCRE_UTF32 and the PCRE_UCP
1355 options, respectively.
1356 .
1357 .
1358 .\" HTML <a name="subpattern"></a>
1360 .rs
1361 .sp
1362 Subpatterns are delimited by parentheses (round brackets), which can be nested.
1363 Turning part of a pattern into a subpattern does two things:
1364 .sp
1365 1. It localizes a set of alternatives. For example, the pattern
1366 .sp
1367 cat(aract|erpillar|)
1368 .sp
1369 matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1370 match "cataract", "erpillar" or an empty string.
1371 .sp
1372 2. It sets up the subpattern as a capturing subpattern. This means that, when
1373 the whole pattern matches, that portion of the subject string that matched the
1374 subpattern is passed back to the caller via the \fIovector\fP argument of the
1375 matching function. (This applies only to the traditional matching functions;
1376 the DFA matching functions do not support capturing.)
1377 .P
1378 Opening parentheses are counted from left to right (starting from 1) to obtain
1379 numbers for the capturing subpatterns. For example, if the string "the red
1380 king" is matched against the pattern
1381 .sp
1382 the ((red|white) (king|queen))
1383 .sp
1384 the captured substrings are "red king", "red", and "king", and are numbered 1,
1385 2, and 3, respectively.
1386 .P
1387 The fact that plain parentheses fulfil two functions is not always helpful.
1388 There are often times when a grouping subpattern is required without a
1389 capturing requirement. If an opening parenthesis is followed by a question mark
1390 and a colon, the subpattern does not do any capturing, and is not counted when
1391 computing the number of any subsequent capturing subpatterns. For example, if
1392 the string "the white queen" is matched against the pattern
1393 .sp
1394 the ((?:red|white) (king|queen))
1395 .sp
1396 the captured substrings are "white queen" and "queen", and are numbered 1 and
1397 2. The maximum number of capturing subpatterns is 65535.
1398 .P
1399 As a convenient shorthand, if any option settings are required at the start of
1400 a non-capturing subpattern, the option letters may appear between the "?" and
1401 the ":". Thus the two patterns
1402 .sp
1403 (?i:saturday|sunday)
1404 (?:(?i)saturday|sunday)
1405 .sp
1406 match exactly the same set of strings. Because alternative branches are tried
1407 from left to right, and options are not reset until the end of the subpattern
1408 is reached, an option setting in one branch does affect subsequent branches, so
1409 the above patterns match "SUNDAY" as well as "Saturday".
1410 .
1411 .
1412 .\" HTML <a name="dupsubpatternnumber"></a>
1414 .rs
1415 .sp
1416 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1417 the same numbers for its capturing parentheses. Such a subpattern starts with
1418 (?| and is itself a non-capturing subpattern. For example, consider this
1419 pattern:
1420 .sp
1421 (?|(Sat)ur|(Sun))day
1422 .sp
1423 Because the two alternatives are inside a (?| group, both sets of capturing
1424 parentheses are numbered one. Thus, when the pattern matches, you can look
1425 at captured substring number one, whichever alternative matched. This construct
1426 is useful when you want to capture part, but not all, of one of a number of
1427 alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1428 number is reset at the start of each branch. The numbers of any capturing
1429 parentheses that follow the subpattern start after the highest number used in
1430 any branch. The following example is taken from the Perl documentation. The
1431 numbers underneath show in which buffer the captured content will be stored.
1432 .sp
1433 # before ---------------branch-reset----------- after
1434 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1435 # 1 2 2 3 2 3 4
1436 .sp
1437 A back reference to a numbered subpattern uses the most recent value that is
1438 set for that number by any subpattern. The following pattern matches "abcabc"
1439 or "defdef":
1440 .sp
1441 /(?|(abc)|(def))\e1/
1442 .sp
1443 In contrast, a subroutine call to a numbered subpattern always refers to the
1444 first one in the pattern with the given number. The following pattern matches
1445 "abcabc" or "defabc":
1446 .sp
1447 /(?|(abc)|(def))(?1)/
1448 .sp
1449 If a
1450 .\" HTML <a href="#conditions">
1451 .\" </a>
1452 condition test
1453 .\"
1454 for a subpattern's having matched refers to a non-unique number, the test is
1455 true if any of the subpatterns of that number have matched.
1456 .P
1457 An alternative approach to using this "branch reset" feature is to use
1458 duplicate named subpatterns, as described in the next section.
1459 .
1460 .
1462 .rs
1463 .sp
1464 Identifying capturing parentheses by number is simple, but it can be very hard
1465 to keep track of the numbers in complicated regular expressions. Furthermore,
1466 if an expression is modified, the numbers may change. To help with this
1467 difficulty, PCRE supports the naming of subpatterns. This feature was not
1468 added to Perl until release 5.10. Python had the feature earlier, and PCRE
1469 introduced it at release 4.0, using the Python syntax. PCRE now supports both
1470 the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1471 have different names, but PCRE does not.
1472 .P
1473 In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1474 (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1475 parentheses from other parts of the pattern, such as
1476 .\" HTML <a href="#backreferences">
1477 .\" </a>
1478 back references,
1479 .\"
1480 .\" HTML <a href="#recursion">
1481 .\" </a>
1482 recursion,
1483 .\"
1484 and
1485 .\" HTML <a href="#conditions">
1486 .\" </a>
1487 conditions,
1488 .\"
1489 can be made by name as well as by number.
1490 .P
1491 Names consist of up to 32 alphanumeric characters and underscores. Named
1492 capturing parentheses are still allocated numbers as well as names, exactly as
1493 if the names were not present. The PCRE API provides function calls for
1494 extracting the name-to-number translation table from a compiled pattern. There
1495 is also a convenience function for extracting a captured substring by name.
1496 .P
1497 By default, a name must be unique within a pattern, but it is possible to relax
1498 this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1499 names are also always permitted for subpatterns with the same number, set up as
1500 described in the previous section.) Duplicate names can be useful for patterns
1501 where only one instance of the named parentheses can match. Suppose you want to
1502 match the name of a weekday, either as a 3-letter abbreviation or as the full
1503 name, and in both cases you want to extract the abbreviation. This pattern
1504 (ignoring the line breaks) does the job:
1505 .sp
1506 (?<DN>Mon|Fri|Sun)(?:day)?|
1507 (?<DN>Tue)(?:sday)?|
1508 (?<DN>Wed)(?:nesday)?|
1509 (?<DN>Thu)(?:rsday)?|
1510 (?<DN>Sat)(?:urday)?
1511 .sp
1512 There are five capturing substrings, but only one is ever set after a match.
1513 (An alternative way of solving this problem is to use a "branch reset"
1514 subpattern, as described in the previous section.)
1515 .P
1516 The convenience function for extracting the data by name returns the substring
1517 for the first (and in this example, the only) subpattern of that name that
1518 matched. This saves searching to find which numbered subpattern it was.
1519 .P
1520 If you make a back reference to a non-unique named subpattern from elsewhere in
1521 the pattern, the one that corresponds to the first occurrence of the name is
1522 used. In the absence of duplicate numbers (see the previous section) this is
1523 the one with the lowest number. If you use a named reference in a condition
1524 test (see the
1525 .\"
1526 .\" HTML <a href="#conditions">
1527 .\" </a>
1528 section about conditions
1529 .\"
1530 below), either to check whether a subpattern has matched, or to check for
1531 recursion, all subpatterns with the same name are tested. If the condition is
1532 true for any one of them, the overall condition is true. This is the same
1533 behaviour as testing by number. For further details of the interfaces for
1534 handling named subpatterns, see the
1535 .\" HREF
1536 \fBpcreapi\fP
1537 .\"
1538 documentation.
1539 .P
1540 \fBWarning:\fP You cannot use different names to distinguish between two
1541 subpatterns with the same number because PCRE uses only the numbers when
1542 matching. For this reason, an error is given at compile time if different names
1543 are given to subpatterns with the same number. However, you can give the same
1544 name to subpatterns with the same number, even when PCRE_DUPNAMES is not set.
1545 .
1546 .
1548 .rs
1549 .sp
1550 Repetition is specified by quantifiers, which can follow any of the following
1551 items:
1552 .sp
1553 a literal data character
1554 the dot metacharacter
1555 the \eC escape sequence
1556 the \eX escape sequence
1557 the \eR escape sequence
1558 an escape such as \ed or \epL that matches a single character
1559 a character class
1560 a back reference (see next section)
1561 a parenthesized subpattern (including assertions)
1562 a subroutine call to a subpattern (recursive or otherwise)
1563 .sp
1564 The general repetition quantifier specifies a minimum and maximum number of
1565 permitted matches, by giving the two numbers in curly brackets (braces),
1566 separated by a comma. The numbers must be less than 65536, and the first must
1567 be less than or equal to the second. For example:
1568 .sp
1569 z{2,4}
1570 .sp
1571 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1572 character. If the second number is omitted, but the comma is present, there is
1573 no upper limit; if the second number and the comma are both omitted, the
1574 quantifier specifies an exact number of required matches. Thus
1575 .sp
1576 [aeiou]{3,}
1577 .sp
1578 matches at least 3 successive vowels, but may match many more, while
1579 .sp
1580 \ed{8}
1581 .sp
1582 matches exactly 8 digits. An opening curly bracket that appears in a position
1583 where a quantifier is not allowed, or one that does not match the syntax of a
1584 quantifier, is taken as a literal character. For example, {,6} is not a
1585 quantifier, but a literal string of four characters.
1586 .P
1587 In UTF modes, quantifiers apply to characters rather than to individual data
1588 units. Thus, for example, \ex{100}{2} matches two characters, each of
1589 which is represented by a two-byte sequence in a UTF-8 string. Similarly,
1590 \eX{3} matches three Unicode extended grapheme clusters, each of which may be
1591 several data units long (and they may be of different lengths).
1592 .P
1593 The quantifier {0} is permitted, causing the expression to behave as if the
1594 previous item and the quantifier were not present. This may be useful for
1595 subpatterns that are referenced as
1596 .\" HTML <a href="#subpatternsassubroutines">
1597 .\" </a>
1598 subroutines
1599 .\"
1600 from elsewhere in the pattern (but see also the section entitled
1601 .\" HTML <a href="#subdefine">
1602 .\" </a>
1603 "Defining subpatterns for use by reference only"
1604 .\"
1605 below). Items other than subpatterns that have a {0} quantifier are omitted
1606 from the compiled pattern.
1607 .P
1608 For convenience, the three most common quantifiers have single-character
1609 abbreviations:
1610 .sp
1611 * is equivalent to {0,}
1612 + is equivalent to {1,}
1613 ? is equivalent to {0,1}
1614 .sp
1615 It is possible to construct infinite loops by following a subpattern that can
1616 match no characters with a quantifier that has no upper limit, for example:
1617 .sp
1618 (a?)*
1619 .sp
1620 Earlier versions of Perl and PCRE used to give an error at compile time for
1621 such patterns. However, because there are cases where this can be useful, such
1622 patterns are now accepted, but if any repetition of the subpattern does in fact
1623 match no characters, the loop is forcibly broken.
1624 .P
1625 By default, the quantifiers are "greedy", that is, they match as much as
1626 possible (up to the maximum number of permitted times), without causing the
1627 rest of the pattern to fail. The classic example of where this gives problems
1628 is in trying to match comments in C programs. These appear between /* and */
1629 and within the comment, individual * and / characters may appear. An attempt to
1630 match C comments by applying the pattern
1631 .sp
1632 /\e*.*\e*/
1633 .sp
1634 to the string
1635 .sp
1636 /* first comment */ not comment /* second comment */
1637 .sp
1638 fails, because it matches the entire string owing to the greediness of the .*
1639 item.
1640 .P
1641 However, if a quantifier is followed by a question mark, it ceases to be
1642 greedy, and instead matches the minimum number of times possible, so the
1643 pattern
1644 .sp
1645 /\e*.*?\e*/
1646 .sp
1647 does the right thing with the C comments. The meaning of the various
1648 quantifiers is not otherwise changed, just the preferred number of matches.
1649 Do not confuse this use of question mark with its use as a quantifier in its
1650 own right. Because it has two uses, it can sometimes appear doubled, as in
1651 .sp
1652 \ed??\ed
1653 .sp
1654 which matches one digit by preference, but can match two if that is the only
1655 way the rest of the pattern matches.
1656 .P
1657 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1658 the quantifiers are not greedy by default, but individual ones can be made
1659 greedy by following them with a question mark. In other words, it inverts the
1660 default behaviour.
1661 .P
1662 When a parenthesized subpattern is quantified with a minimum repeat count that
1663 is greater than 1 or with a limited maximum, more memory is required for the
1664 compiled pattern, in proportion to the size of the minimum or maximum.
1665 .P
1666 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1667 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1668 implicitly anchored, because whatever follows will be tried against every
1669 character position in the subject string, so there is no point in retrying the
1670 overall match at any position after the first. PCRE normally treats such a
1671 pattern as though it were preceded by \eA.
1672 .P
1673 In cases where it is known that the subject string contains no newlines, it is
1674 worth setting PCRE_DOTALL in order to obtain this optimization, or
1675 alternatively using ^ to indicate anchoring explicitly.
1676 .P
1677 However, there are some cases where the optimization cannot be used. When .*
1678 is inside capturing parentheses that are the subject of a back reference
1679 elsewhere in the pattern, a match at the start may fail where a later one
1680 succeeds. Consider, for example:
1681 .sp
1682 (.*)abc\e1
1683 .sp
1684 If the subject is "xyz123abc123" the match point is the fourth character. For
1685 this reason, such a pattern is not implicitly anchored.
1686 .P
1687 Another case where implicit anchoring is not applied is when the leading .* is
1688 inside an atomic group. Once again, a match at the start may fail where a later
1689 one succeeds. Consider this pattern:
1690 .sp
1691 (?>.*?a)b
1692 .sp
1693 It matches "ab" in the subject "aab". The use of the backtracking control verbs
1694 (*PRUNE) and (*SKIP) also disable this optimization.
1695 .P
1696 When a capturing subpattern is repeated, the value captured is the substring
1697 that matched the final iteration. For example, after
1698 .sp
1699 (tweedle[dume]{3}\es*)+
1700 .sp
1701 has matched "tweedledum tweedledee" the value of the captured substring is
1702 "tweedledee". However, if there are nested capturing subpatterns, the
1703 corresponding captured values may have been set in previous iterations. For
1704 example, after
1705 .sp
1706 /(a|(b))+/
1707 .sp
1708 matches "aba" the value of the second captured substring is "b".
1709 .
1710 .
1711 .\" HTML <a name="atomicgroup"></a>
1713 .rs
1714 .sp
1715 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1716 repetition, failure of what follows normally causes the repeated item to be
1717 re-evaluated to see if a different number of repeats allows the rest of the
1718 pattern to match. Sometimes it is useful to prevent this, either to change the
1719 nature of the match, or to cause it fail earlier than it otherwise might, when
1720 the author of the pattern knows there is no point in carrying on.
1721 .P
1722 Consider, for example, the pattern \ed+foo when applied to the subject line
1723 .sp
1724 123456bar
1725 .sp
1726 After matching all 6 digits and then failing to match "foo", the normal
1727 action of the matcher is to try again with only 5 digits matching the \ed+
1728 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1729 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1730 that once a subpattern has matched, it is not to be re-evaluated in this way.
1731 .P
1732 If we use atomic grouping for the previous example, the matcher gives up
1733 immediately on failing to match "foo" the first time. The notation is a kind of
1734 special parenthesis, starting with (?> as in this example:
1735 .sp
1736 (?>\ed+)foo
1737 .sp
1738 This kind of parenthesis "locks up" the part of the pattern it contains once
1739 it has matched, and a failure further into the pattern is prevented from
1740 backtracking into it. Backtracking past it to previous items, however, works as
1741 normal.
1742 .P
1743 An alternative description is that a subpattern of this type matches the string
1744 of characters that an identical standalone pattern would match, if anchored at
1745 the current point in the subject string.
1746 .P
1747 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1748 the above example can be thought of as a maximizing repeat that must swallow
1749 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1750 number of digits they match in order to make the rest of the pattern match,
1751 (?>\ed+) can only match an entire sequence of digits.
1752 .P
1753 Atomic groups in general can of course contain arbitrarily complicated
1754 subpatterns, and can be nested. However, when the subpattern for an atomic
1755 group is just a single repeated item, as in the example above, a simpler
1756 notation, called a "possessive quantifier" can be used. This consists of an
1757 additional + character following a quantifier. Using this notation, the
1758 previous example can be rewritten as
1759 .sp
1760 \ed++foo
1761 .sp
1762 Note that a possessive quantifier can be used with an entire group, for
1763 example:
1764 .sp
1765 (abc|xyz){2,3}+
1766 .sp
1767 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1768 option is ignored. They are a convenient notation for the simpler forms of
1769 atomic group. However, there is no difference in the meaning of a possessive
1770 quantifier and the equivalent atomic group, though there may be a performance
1771 difference; possessive quantifiers should be slightly faster.
1772 .P
1773 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1774 Jeffrey Friedl originated the idea (and the name) in the first edition of his
1775 book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1776 package, and PCRE copied it from there. It ultimately found its way into Perl
1777 at release 5.10.
1778 .P
1779 PCRE has an optimization that automatically "possessifies" certain simple
1780 pattern constructs. For example, the sequence A+B is treated as A++B because
1781 there is no point in backtracking into a sequence of A's when B must follow.
1782 .P
1783 When a pattern contains an unlimited repeat inside a subpattern that can itself
1784 be repeated an unlimited number of times, the use of an atomic group is the
1785 only way to avoid some failing matches taking a very long time indeed. The
1786 pattern
1787 .sp
1788 (\eD+|<\ed+>)*[!?]
1789 .sp
1790 matches an unlimited number of substrings that either consist of non-digits, or
1791 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1792 quickly. However, if it is applied to
1793 .sp
1794 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1795 .sp
1796 it takes a long time before reporting failure. This is because the string can
1797 be divided between the internal \eD+ repeat and the external * repeat in a
1798 large number of ways, and all have to be tried. (The example uses [!?] rather
1799 than a single character at the end, because both PCRE and Perl have an
1800 optimization that allows for fast failure when a single character is used. They
1801 remember the last single character that is required for a match, and fail early
1802 if it is not present in the string.) If the pattern is changed so that it uses
1803 an atomic group, like this:
1804 .sp
1805 ((?>\eD+)|<\ed+>)*[!?]
1806 .sp
1807 sequences of non-digits cannot be broken, and failure happens quickly.
1808 .
1809 .
1810 .\" HTML <a name="backreferences"></a>
1812 .rs
1813 .sp
1814 Outside a character class, a backslash followed by a digit greater than 0 (and
1815 possibly further digits) is a back reference to a capturing subpattern earlier
1816 (that is, to its left) in the pattern, provided there have been that many
1817 previous capturing left parentheses.
1818 .P
1819 However, if the decimal number following the backslash is less than 10, it is
1820 always taken as a back reference, and causes an error only if there are not
1821 that many capturing left parentheses in the entire pattern. In other words, the
1822 parentheses that are referenced need not be to the left of the reference for
1823 numbers less than 10. A "forward back reference" of this type can make sense
1824 when a repetition is involved and the subpattern to the right has participated
1825 in an earlier iteration.
1826 .P
1827 It is not possible to have a numerical "forward back reference" to a subpattern
1828 whose number is 10 or more using this syntax because a sequence such as \e50 is
1829 interpreted as a character defined in octal. See the subsection entitled
1830 "Non-printing characters"
1831 .\" HTML <a href="#digitsafterbackslash">
1832 .\" </a>
1833 above
1834 .\"
1835 for further details of the handling of digits following a backslash. There is
1836 no such problem when named parentheses are used. A back reference to any
1837 subpattern is possible using named parentheses (see below).
1838 .P
1839 Another way of avoiding the ambiguity inherent in the use of digits following a
1840 backslash is to use the \eg escape sequence. This escape must be followed by an
1841 unsigned number or a negative number, optionally enclosed in braces. These
1842 examples are all identical:
1843 .sp
1844 (ring), \e1
1845 (ring), \eg1
1846 (ring), \eg{1}
1847 .sp
1848 An unsigned number specifies an absolute reference without the ambiguity that
1849 is present in the older syntax. It is also useful when literal digits follow
1850 the reference. A negative number is a relative reference. Consider this
1851 example:
1852 .sp
1853 (abc(def)ghi)\eg{-1}
1854 .sp
1855 The sequence \eg{-1} is a reference to the most recently started capturing
1856 subpattern before \eg, that is, is it equivalent to \e2 in this example.
1857 Similarly, \eg{-2} would be equivalent to \e1. The use of relative references
1858 can be helpful in long patterns, and also in patterns that are created by
1859 joining together fragments that contain references within themselves.
1860 .P
1861 A back reference matches whatever actually matched the capturing subpattern in
1862 the current subject string, rather than anything matching the subpattern
1863 itself (see
1864 .\" HTML <a href="#subpatternsassubroutines">
1865 .\" </a>
1866 "Subpatterns as subroutines"
1867 .\"
1868 below for a way of doing that). So the pattern
1869 .sp
1870 (sens|respons)e and \e1ibility
1871 .sp
1872 matches "sense and sensibility" and "response and responsibility", but not
1873 "sense and responsibility". If caseful matching is in force at the time of the
1874 back reference, the case of letters is relevant. For example,
1875 .sp
1876 ((?i)rah)\es+\e1
1877 .sp
1878 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1879 capturing subpattern is matched caselessly.
1880 .P
1881 There are several different ways of writing back references to named
1882 subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
1883 \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1884 back reference syntax, in which \eg can be used for both numeric and named
1885 references, is also supported. We could rewrite the above example in any of
1886 the following ways:
1887 .sp
1888 (?<p1>(?i)rah)\es+\ek<p1>
1889 (?'p1'(?i)rah)\es+\ek{p1}
1890 (?P<p1>(?i)rah)\es+(?P=p1)
1891 (?<p1>(?i)rah)\es+\eg{p1}
1892 .sp
1893 A subpattern that is referenced by name may appear in the pattern before or
1894 after the reference.
1895 .P
1896 There may be more than one back reference to the same subpattern. If a
1897 subpattern has not actually been used in a particular match, any back
1898 references to it always fail by default. For example, the pattern
1899 .sp
1900 (a|(bc))\e2
1901 .sp
1902 always fails if it starts to match "a" rather than "bc". However, if the
1903 PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
1904 unset value matches an empty string.
1905 .P
1906 Because there may be many capturing parentheses in a pattern, all digits
1907 following a backslash are taken as part of a potential back reference number.
1908 If the pattern continues with a digit character, some delimiter must be used to
1909 terminate the back reference. If the PCRE_EXTENDED option is set, this can be
1910 white space. Otherwise, the \eg{ syntax or an empty comment (see
1911 .\" HTML <a href="#comments">
1912 .\" </a>
1913 "Comments"
1914 .\"
1915 below) can be used.
1916 .
1917 .SS "Recursive back references"
1918 .rs
1919 .sp
1920 A back reference that occurs inside the parentheses to which it refers fails
1921 when the subpattern is first used, so, for example, (a\e1) never matches.
1922 However, such references can be useful inside repeated subpatterns. For
1923 example, the pattern
1924 .sp
1925 (a|b\e1)+
1926 .sp
1927 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1928 the subpattern, the back reference matches the character string corresponding
1929 to the previous iteration. In order for this to work, the pattern must be such
1930 that the first iteration does not need to match the back reference. This can be
1931 done using alternation, as in the example above, or by a quantifier with a
1932 minimum of zero.
1933 .P
1934 Back references of this type cause the group that they reference to be treated
1935 as an
1936 .\" HTML <a href="#atomicgroup">
1937 .\" </a>
1938 atomic group.
1939 .\"
1940 Once the whole group has been matched, a subsequent matching failure cannot
1941 cause backtracking into the middle of the group.
1942 .
1943 .
1944 .\" HTML <a name="bigassertions"></a>
1946 .rs
1947 .sp
1948 An assertion is a test on the characters following or preceding the current
1949 matching point that does not actually consume any characters. The simple
1950 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
1951 .\" HTML <a href="#smallassertions">
1952 .\" </a>
1953 above.
1954 .\"
1955 .P
1956 More complicated assertions are coded as subpatterns. There are two kinds:
1957 those that look ahead of the current position in the subject string, and those
1958 that look behind it. An assertion subpattern is matched in the normal way,
1959 except that it does not cause the current matching position to be changed.
1960 .P
1961 Assertion subpatterns are not capturing subpatterns. If such an assertion
1962 contains capturing subpatterns within it, these are counted for the purposes of
1963 numbering the capturing subpatterns in the whole pattern. However, substring
1964 capturing is carried out only for positive assertions, because it does not make
1965 sense for negative assertions.
1966 .P
1967 For compatibility with Perl, assertion subpatterns may be repeated; though
1968 it makes no sense to assert the same thing several times, the side effect of
1969 capturing parentheses may occasionally be useful. In practice, there only three
1970 cases:
1971 .sp
1972 (1) If the quantifier is {0}, the assertion is never obeyed during matching.
1973 However, it may contain internal capturing parenthesized groups that are called
1974 from elsewhere via the
1975 .\" HTML <a href="#subpatternsassubroutines">
1976 .\" </a>
1977 subroutine mechanism.
1978 .\"
1979 .sp
1980 (2) If quantifier is {0,n} where n is greater than zero, it is treated as if it
1981 were {0,1}. At run time, the rest of the pattern match is tried with and
1982 without the assertion, the order depending on the greediness of the quantifier.
1983 .sp
1984 (3) If the minimum repetition is greater than zero, the quantifier is ignored.
1985 The assertion is obeyed just once when encountered during matching.
1986 .
1987 .
1988 .SS "Lookahead assertions"
1989 .rs
1990 .sp
1991 Lookahead assertions start with (?= for positive assertions and (?! for
1992 negative assertions. For example,
1993 .sp
1994 \ew+(?=;)
1995 .sp
1996 matches a word followed by a semicolon, but does not include the semicolon in
1997 the match, and
1998 .sp
1999 foo(?!bar)
2000 .sp
2001 matches any occurrence of "foo" that is not followed by "bar". Note that the
2002 apparently similar pattern
2003 .sp
2004 (?!foo)bar
2005 .sp
2006 does not find an occurrence of "bar" that is preceded by something other than
2007 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
2008 (?!foo) is always true when the next three characters are "bar". A
2009 lookbehind assertion is needed to achieve the other effect.
2010 .P
2011 If you want to force a matching failure at some point in a pattern, the most
2012 convenient way to do it is with (?!) because an empty string always matches, so
2013 an assertion that requires there not to be an empty string must always fail.
2014 The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
2015 .
2016 .
2017 .\" HTML <a name="lookbehind"></a>
2018 .SS "Lookbehind assertions"
2019 .rs
2020 .sp
2021 Lookbehind assertions start with (?<= for positive assertions and (?<! for
2022 negative assertions. For example,
2023 .sp
2024 (?<!foo)bar
2025 .sp
2026 does find an occurrence of "bar" that is not preceded by "foo". The contents of
2027 a lookbehind assertion are restricted such that all the strings it matches must
2028 have a fixed length. However, if there are several top-level alternatives, they
2029 do not all have to have the same fixed length. Thus
2030 .sp
2031 (?<=bullock|donkey)
2032 .sp
2033 is permitted, but
2034 .sp
2035 (?<!dogs?|cats?)
2036 .sp
2037 causes an error at compile time. Branches that match different length strings
2038 are permitted only at the top level of a lookbehind assertion. This is an
2039 extension compared with Perl, which requires all branches to match the same
2040 length of string. An assertion such as
2041 .sp
2042 (?<=ab(c|de))
2043 .sp
2044 is not permitted, because its single top-level branch can match two different
2045 lengths, but it is acceptable to PCRE if rewritten to use two top-level
2046 branches:
2047 .sp
2048 (?<=abc|abde)
2049 .sp
2050 In some cases, the escape sequence \eK
2051 .\" HTML <a href="#resetmatchstart">
2052 .\" </a>
2053 (see above)
2054 .\"
2055 can be used instead of a lookbehind assertion to get round the fixed-length
2056 restriction.
2057 .P
2058 The implementation of lookbehind assertions is, for each alternative, to
2059 temporarily move the current position back by the fixed length and then try to
2060 match. If there are insufficient characters before the current position, the
2061 assertion fails.
2062 .P
2063 In a UTF mode, PCRE does not allow the \eC escape (which matches a single data
2064 unit even in a UTF mode) to appear in lookbehind assertions, because it makes
2065 it impossible to calculate the length of the lookbehind. The \eX and \eR
2066 escapes, which can match different numbers of data units, are also not
2067 permitted.
2068 .P
2069 .\" HTML <a href="#subpatternsassubroutines">
2070 .\" </a>
2071 "Subroutine"
2072 .\"
2073 calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
2074 as the subpattern matches a fixed-length string.
2075 .\" HTML <a href="#recursion">
2076 .\" </a>
2077 Recursion,
2078 .\"
2079 however, is not supported.
2080 .P
2081 Possessive quantifiers can be used in conjunction with lookbehind assertions to
2082 specify efficient matching of fixed-length strings at the end of subject
2083 strings. Consider a simple pattern such as
2084 .sp
2085 abcd$
2086 .sp
2087 when applied to a long string that does not match. Because matching proceeds
2088 from left to right, PCRE will look for each "a" in the subject and then see if
2089 what follows matches the rest of the pattern. If the pattern is specified as
2090 .sp
2091 ^.*abcd$
2092 .sp
2093 the initial .* matches the entire string at first, but when this fails (because
2094 there is no following "a"), it backtracks to match all but the last character,
2095 then all but the last two characters, and so on. Once again the search for "a"
2096 covers the entire string, from right to left, so we are no better off. However,
2097 if the pattern is written as
2098 .sp
2099 ^.*+(?<=abcd)
2100 .sp
2101 there can be no backtracking for the .*+ item; it can match only the entire
2102 string. The subsequent lookbehind assertion does a single test on the last four
2103 characters. If it fails, the match fails immediately. For long strings, this
2104 approach makes a significant difference to the processing time.
2105 .
2106 .
2107 .SS "Using multiple assertions"
2108 .rs
2109 .sp
2110 Several assertions (of any sort) may occur in succession. For example,
2111 .sp
2112 (?<=\ed{3})(?<!999)foo
2113 .sp
2114 matches "foo" preceded by three digits that are not "999". Notice that each of
2115 the assertions is applied independently at the same point in the subject
2116 string. First there is a check that the previous three characters are all
2117 digits, and then there is a check that the same three characters are not "999".
2118 This pattern does \fInot\fP match "foo" preceded by six characters, the first
2119 of which are digits and the last three of which are not "999". For example, it
2120 doesn't match "123abcfoo". A pattern to do that is
2121 .sp
2122 (?<=\ed{3}...)(?<!999)foo
2123 .sp
2124 This time the first assertion looks at the preceding six characters, checking
2125 that the first three are digits, and then the second assertion checks that the
2126 preceding three characters are not "999".
2127 .P
2128 Assertions can be nested in any combination. For example,
2129 .sp
2130 (?<=(?<!foo)bar)baz
2131 .sp
2132 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
2133 preceded by "foo", while
2134 .sp
2135 (?<=\ed{3}(?!999)...)foo
2136 .sp
2137 is another pattern that matches "foo" preceded by three digits and any three
2138 characters that are not "999".
2139 .
2140 .
2141 .\" HTML <a name="conditions"></a>
2143 .rs
2144 .sp
2145 It is possible to cause the matching process to obey a subpattern
2146 conditionally or to choose between two alternative subpatterns, depending on
2147 the result of an assertion, or whether a specific capturing subpattern has
2148 already been matched. The two possible forms of conditional subpattern are:
2149 .sp
2150 (?(condition)yes-pattern)
2151 (?(condition)yes-pattern|no-pattern)
2152 .sp
2153 If the condition is satisfied, the yes-pattern is used; otherwise the
2154 no-pattern (if present) is used. If there are more than two alternatives in the
2155 subpattern, a compile-time error occurs. Each of the two alternatives may
2156 itself contain nested subpatterns of any form, including conditional
2157 subpatterns; the restriction to two alternatives applies only at the level of
2158 the condition. This pattern fragment is an example where the alternatives are
2159 complex:
2160 .sp
2161 (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
2162 .sp
2163 .P
2164 There are four kinds of condition: references to subpatterns, references to
2165 recursion, a pseudo-condition called DEFINE, and assertions.
2166 .
2167 .SS "Checking for a used subpattern by number"
2168 .rs
2169 .sp
2170 If the text between the parentheses consists of a sequence of digits, the
2171 condition is true if a capturing subpattern of that number has previously
2172 matched. If there is more than one capturing subpattern with the same number
2173 (see the earlier
2174 .\"
2175 .\" HTML <a href="#recursion">
2176 .\" </a>
2177 section about duplicate subpattern numbers),
2178 .\"
2179 the condition is true if any of them have matched. An alternative notation is
2180 to precede the digits with a plus or minus sign. In this case, the subpattern
2181 number is relative rather than absolute. The most recently opened parentheses
2182 can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
2183 loops it can also make sense to refer to subsequent groups. The next
2184 parentheses to be opened can be referenced as (?(+1), and so on. (The value
2185 zero in any of these forms is not used; it provokes a compile-time error.)
2186 .P
2187 Consider the following pattern, which contains non-significant white space to
2188 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2189 three parts for ease of discussion:
2190 .sp
2191 ( \e( )? [^()]+ (?(1) \e) )
2192 .sp
2193 The first part matches an optional opening parenthesis, and if that
2194 character is present, sets it as the first captured substring. The second part
2195 matches one or more characters that are not parentheses. The third part is a
2196 conditional subpattern that tests whether or not the first set of parentheses
2197 matched. If they did, that is, if subject started with an opening parenthesis,
2198 the condition is true, and so the yes-pattern is executed and a closing
2199 parenthesis is required. Otherwise, since no-pattern is not present, the
2200 subpattern matches nothing. In other words, this pattern matches a sequence of
2201 non-parentheses, optionally enclosed in parentheses.
2202 .P
2203 If you were embedding this pattern in a larger one, you could use a relative
2204 reference:
2205 .sp
2206 ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
2207 .sp
2208 This makes the fragment independent of the parentheses in the larger pattern.
2209 .
2210 .SS "Checking for a used subpattern by name"
2211 .rs
2212 .sp
2213 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
2214 subpattern by name. For compatibility with earlier versions of PCRE, which had
2215 this facility before Perl, the syntax (?(name)...) is also recognized. However,
2216 there is a possible ambiguity with this syntax, because subpattern names may
2217 consist entirely of digits. PCRE looks first for a named subpattern; if it
2218 cannot find one and the name consists entirely of digits, PCRE looks for a
2219 subpattern of that number, which must be greater than zero. Using subpattern
2220 names that consist entirely of digits is not recommended.
2221 .P
2222 Rewriting the above example to use a named subpattern gives this:
2223 .sp
2224 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
2225 .sp
2226 If the name used in a condition of this kind is a duplicate, the test is
2227 applied to all subpatterns of the same name, and is true if any one of them has
2228 matched.
2229 .
2230 .SS "Checking for pattern recursion"
2231 .rs
2232 .sp
2233 If the condition is the string (R), and there is no subpattern with the name R,
2234 the condition is true if a recursive call to the whole pattern or any
2235 subpattern has been made. If digits or a name preceded by ampersand follow the
2236 letter R, for example:
2237 .sp
2238 (?(R3)...) or (?(R&name)...)
2239 .sp
2240 the condition is true if the most recent recursion is into a subpattern whose
2241 number or name is given. This condition does not check the entire recursion
2242 stack. If the name used in a condition of this kind is a duplicate, the test is
2243 applied to all subpatterns of the same name, and is true if any one of them is
2244 the most recent recursion.
2245 .P
2246 At "top level", all these recursion test conditions are false.
2247 .\" HTML <a href="#recursion">
2248 .\" </a>
2249 The syntax for recursive patterns
2250 .\"
2251 is described below.
2252 .
2253 .\" HTML <a name="subdefine"></a>
2254 .SS "Defining subpatterns for use by reference only"
2255 .rs
2256 .sp
2257 If the condition is the string (DEFINE), and there is no subpattern with the
2258 name DEFINE, the condition is always false. In this case, there may be only one
2259 alternative in the subpattern. It is always skipped if control reaches this
2260 point in the pattern; the idea of DEFINE is that it can be used to define
2261 subroutines that can be referenced from elsewhere. (The use of
2262 .\" HTML <a href="#subpatternsassubroutines">
2263 .\" </a>
2264 subroutines
2265 .\"
2266 is described below.) For example, a pattern to match an IPv4 address such as
2267 "" could be written like this (ignore white space and line
2268 breaks):
2269 .sp
2270 (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
2271 \eb (?&byte) (\e.(?&byte)){3} \eb
2272 .sp
2273 The first part of the pattern is a DEFINE group inside which a another group
2274 named "byte" is defined. This matches an individual component of an IPv4
2275 address (a number less than 256). When matching takes place, this part of the
2276 pattern is skipped because DEFINE acts like a false condition. The rest of the
2277 pattern uses references to the named group to match the four dot-separated
2278 components of an IPv4 address, insisting on a word boundary at each end.
2279 .
2280 .SS "Assertion conditions"
2281 .rs
2282 .sp
2283 If the condition is not in any of the above formats, it must be an assertion.
2284 This may be a positive or negative lookahead or lookbehind assertion. Consider
2285 this pattern, again containing non-significant white space, and with the two
2286 alternatives on the second line:
2287 .sp
2288 (?(?=[^a-z]*[a-z])
2289 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
2290 .sp
2291 The condition is a positive lookahead assertion that matches an optional
2292 sequence of non-letters followed by a letter. In other words, it tests for the
2293 presence of at least one letter in the subject. If a letter is found, the
2294 subject is matched against the first alternative; otherwise it is matched
2295 against the second. This pattern matches strings in one of the two forms
2296 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2297 .
2298 .
2299 .\" HTML <a name="comments"></a>
2301 .rs
2302 .sp
2303 There are two ways of including comments in patterns that are processed by
2304 PCRE. In both cases, the start of the comment must not be in a character class,
2305 nor in the middle of any other sequence of related characters such as (?: or a
2306 subpattern name or number. The characters that make up a comment play no part
2307 in the pattern matching.
2308 .P
2309 The sequence (?# marks the start of a comment that continues up to the next
2310 closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2311 option is set, an unescaped # character also introduces a comment, which in
2312 this case continues to immediately after the next newline character or
2313 character sequence in the pattern. Which characters are interpreted as newlines
2314 is controlled by the options passed to a compiling function or by a special
2315 sequence at the start of the pattern, as described in the section entitled
2316 .\" HTML <a href="#newlines">
2317 .\" </a>
2318 "Newline conventions"
2319 .\"
2320 above. Note that the end of this type of comment is a literal newline sequence
2321 in the pattern; escape sequences that happen to represent a newline do not
2322 count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2323 default newline convention is in force:
2324 .sp
2325 abc #comment \en still comment
2326 .sp
2327 On encountering the # character, \fBpcre_compile()\fP skips along, looking for
2328 a newline in the pattern. The sequence \en is still literal at this stage, so
2329 it does not terminate the comment. Only an actual character with the code value
2330 0x0a (the default newline) does so.
2331 .
2332 .
2333 .\" HTML <a name="recursion"></a>
2335 .rs
2336 .sp
2337 Consider the problem of matching a string in parentheses, allowing for
2338 unlimited nested parentheses. Without the use of recursion, the best that can
2339 be done is to use a pattern that matches up to some fixed depth of nesting. It
2340 is not possible to handle an arbitrary nesting depth.
2341 .P
2342 For some time, Perl has provided a facility that allows regular expressions to
2343 recurse (amongst other things). It does this by interpolating Perl code in the
2344 expression at run time, and the code can refer to the expression itself. A Perl
2345 pattern using code interpolation to solve the parentheses problem can be
2346 created like this:
2347 .sp
2348 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
2349 .sp
2350 The (?p{...}) item interpolates Perl code at run time, and in this case refers
2351 recursively to the pattern in which it appears.
2352 .P
2353 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2354 supports special syntax for recursion of the entire pattern, and also for
2355 individual subpattern recursion. After its introduction in PCRE and Python,
2356 this kind of recursion was subsequently introduced into Perl at release 5.10.
2357 .P
2358 A special item that consists of (? followed by a number greater than zero and a
2359 closing parenthesis is a recursive subroutine call of the subpattern of the
2360 given number, provided that it occurs inside that subpattern. (If not, it is a
2361 .\" HTML <a href="#subpatternsassubroutines">
2362 .\" </a>
2363 non-recursive subroutine
2364 .\"
2365 call, which is described in the next section.) The special item (?R) or (?0) is
2366 a recursive call of the entire regular expression.
2367 .P
2368 This PCRE pattern solves the nested parentheses problem (assume the
2369 PCRE_EXTENDED option is set so that white space is ignored):
2370 .sp
2371 \e( ( [^()]++ | (?R) )* \e)
2372 .sp
2373 First it matches an opening parenthesis. Then it matches any number of
2374 substrings which can either be a sequence of non-parentheses, or a recursive
2375 match of the pattern itself (that is, a correctly parenthesized substring).
2376 Finally there is a closing parenthesis. Note the use of a possessive quantifier
2377 to avoid backtracking into sequences of non-parentheses.
2378 .P
2379 If this were part of a larger pattern, you would not want to recurse the entire
2380 pattern, so instead you could use this:
2381 .sp
2382 ( \e( ( [^()]++ | (?1) )* \e) )
2383 .sp
2384 We have put the pattern into parentheses, and caused the recursion to refer to
2385 them instead of the whole pattern.
2386 .P
2387 In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2388 is made easier by the use of relative references. Instead of (?1) in the
2389 pattern above you can write (?-2) to refer to the second most recently opened
2390 parentheses preceding the recursion. In other words, a negative number counts
2391 capturing parentheses leftwards from the point at which it is encountered.
2392 .P
2393 It is also possible to refer to subsequently opened parentheses, by writing
2394 references such as (?+2). However, these cannot be recursive because the
2395 reference is not inside the parentheses that are referenced. They are always
2396 .\" HTML <a href="#subpatternsassubroutines">
2397 .\" </a>
2398 non-recursive subroutine
2399 .\"
2400 calls, as described in the next section.
2401 .P
2402 An alternative approach is to use named parentheses instead. The Perl syntax
2403 for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
2404 could rewrite the above example as follows:
2405 .sp
2406 (?<pn> \e( ( [^()]++ | (?&pn) )* \e) )
2407 .sp
2408 If there is more than one subpattern with the same name, the earliest one is
2409 used.
2410 .P
2411 This particular example pattern that we have been looking at contains nested
2412 unlimited repeats, and so the use of a possessive quantifier for matching
2413 strings of non-parentheses is important when applying the pattern to strings
2414 that do not match. For example, when this pattern is applied to
2415 .sp
2416 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2417 .sp
2418 it yields "no match" quickly. However, if a possessive quantifier is not used,
2419 the match runs for a very long time indeed because there are so many different
2420 ways the + and * repeats can carve up the subject, and all have to be tested
2421 before failure can be reported.
2422 .P
2423 At the end of a match, the values of capturing parentheses are those from
2424 the outermost level. If you want to obtain intermediate values, a callout
2425 function can be used (see below and the
2426 .\" HREF
2427 \fBpcrecallout\fP
2428 .\"
2429 documentation). If the pattern above is matched against
2430 .sp
2431 (ab(cd)ef)
2432 .sp
2433 the value for the inner capturing parentheses (numbered 2) is "ef", which is
2434 the last value taken on at the top level. If a capturing subpattern is not
2435 matched at the top level, its final captured value is unset, even if it was
2436 (temporarily) set at a deeper level during the matching process.
2437 .P
2438 If there are more than 15 capturing parentheses in a pattern, PCRE has to
2439 obtain extra memory to store data during a recursion, which it does by using
2440 \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no memory can
2441 be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2442 .P
2443 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2444 Consider this pattern, which matches text in angle brackets, allowing for
2445 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2446 recursing), whereas any characters are permitted at the outer level.
2447 .sp
2448 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
2449 .sp
2450 In this pattern, (?(R) is the start of a conditional subpattern, with two
2451 different alternatives for the recursive and non-recursive cases. The (?R) item
2452 is the actual recursive call.
2453 .
2454 .
2455 .\" HTML <a name="recursiondifference"></a>
2456 .SS "Differences in recursion processing between PCRE and Perl"
2457 .rs
2458 .sp
2459 Recursion processing in PCRE differs from Perl in two important ways. In PCRE
2460 (like Python, but unlike Perl), a recursive subpattern call is always treated
2461 as an atomic group. That is, once it has matched some of the subject string, it
2462 is never re-entered, even if it contains untried alternatives and there is a
2463 subsequent matching failure. This can be illustrated by the following pattern,
2464 which purports to match a palindromic string that contains an odd number of
2465 characters (for example, "a", "aba", "abcba", "abcdcba"):
2466 .sp
2467 ^(.|(.)(?1)\e2)$
2468 .sp
2469 The idea is that it either matches a single character, or two identical
2470 characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2471 it does not if the pattern is longer than three characters. Consider the
2472 subject string "abcba":
2473 .P
2474 At the top level, the first character is matched, but as it is not at the end
2475 of the string, the first alternative fails; the second alternative is taken
2476 and the recursion kicks in. The recursive call to subpattern 1 successfully
2477 matches the next character ("b"). (Note that the beginning and end of line
2478 tests are not part of the recursion).
2479 .P
2480 Back at the top level, the next character ("c") is compared with what
2481 subpattern 2 matched, which was "a". This fails. Because the recursion is
2482 treated as an atomic group, there are now no backtracking points, and so the
2483 entire match fails. (Perl is able, at this point, to re-enter the recursion and
2484 try the second alternative.) However, if the pattern is written with the
2485 alternatives in the other order, things are different:
2486 .sp
2487 ^((.)(?1)\e2|.)$
2488 .sp
2489 This time, the recursing alternative is tried first, and continues to recurse
2490 until it runs out of characters, at which point the recursion fails. But this
2491 time we do have another alternative to try at the higher level. That is the big
2492 difference: in the previous case the remaining alternative is at a deeper
2493 recursion level, which PCRE cannot use.
2494 .P
2495 To change the pattern so that it matches all palindromic strings, not just
2496 those with an odd number of characters, it is tempting to change the pattern to
2497 this:
2498 .sp
2499 ^((.)(?1)\e2|.?)$
2500 .sp
2501 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2502 deeper recursion has matched a single character, it cannot be entered again in
2503 order to match an empty string. The solution is to separate the two cases, and
2504 write out the odd and even cases as alternatives at the higher level:
2505 .sp
2506 ^(?:((.)(?1)\e2|)|((.)(?3)\e4|.))
2507 .sp
2508 If you want to match typical palindromic phrases, the pattern has to ignore all
2509 non-word characters, which can be done like this:
2510 .sp
2511 ^\eW*+(?:((.)\eW*+(?1)\eW*+\e2|)|((.)\eW*+(?3)\eW*+\e4|\eW*+.\eW*+))\eW*+$
2512 .sp
2513 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2514 man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2515 the use of the possessive quantifier *+ to avoid backtracking into sequences of
2516 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2517 more) to match typical phrases, and Perl takes so long that you think it has
2518 gone into a loop.
2519 .P
2520 \fBWARNING\fP: The palindrome-matching patterns above work only if the subject
2521 string does not start with a palindrome that is shorter than the entire string.
2522 For example, although "abcba" is correctly matched, if the subject is "ababa",
2523 PCRE finds the palindrome "aba" at the start, then fails at top level because
2524 the end of the string does not follow. Once again, it cannot jump back into the
2525 recursion to try other alternatives, so the entire match fails.
2526 .P
2527 The second way in which PCRE and Perl differ in their recursion processing is
2528 in the handling of captured values. In Perl, when a subpattern is called
2529 recursively or as a subpattern (see the next section), it has no access to any
2530 values that were captured outside the recursion, whereas in PCRE these values
2531 can be referenced. Consider this pattern:
2532 .sp
2533 ^(.)(\e1|a(?2))
2534 .sp
2535 In PCRE, this pattern matches "bab". The first capturing parentheses match "b",
2536 then in the second group, when the back reference \e1 fails to match "b", the
2537 second alternative matches "a" and then recurses. In the recursion, \e1 does
2538 now match "b" and so the whole match succeeds. In Perl, the pattern fails to
2539 match because inside the recursive call \e1 cannot access the externally set
2540 value.
2541 .
2542 .
2543 .\" HTML <a name="subpatternsassubroutines"></a>
2545 .rs
2546 .sp
2547 If the syntax for a recursive subpattern call (either by number or by
2548 name) is used outside the parentheses to which it refers, it operates like a
2549 subroutine in a programming language. The called subpattern may be defined
2550 before or after the reference. A numbered reference can be absolute or
2551 relative, as in these examples:
2552 .sp
2553 (...(absolute)...)...(?2)...
2554 (...(relative)...)...(?-1)...
2555 (...(?+1)...(relative)...
2556 .sp
2557 An earlier example pointed out that the pattern
2558 .sp
2559 (sens|respons)e and \e1ibility
2560 .sp
2561 matches "sense and sensibility" and "response and responsibility", but not
2562 "sense and responsibility". If instead the pattern
2563 .sp
2564 (sens|respons)e and (?1)ibility
2565 .sp
2566 is used, it does match "sense and responsibility" as well as the other two
2567 strings. Another example is given in the discussion of DEFINE above.
2568 .P
2569 All subroutine calls, whether recursive or not, are always treated as atomic
2570 groups. That is, once a subroutine has matched some of the subject string, it
2571 is never re-entered, even if it contains untried alternatives and there is a
2572 subsequent matching failure. Any capturing parentheses that are set during the
2573 subroutine call revert to their previous values afterwards.
2574 .P
2575 Processing options such as case-independence are fixed when a subpattern is
2576 defined, so if it is used as a subroutine, such options cannot be changed for
2577 different calls. For example, consider this pattern:
2578 .sp
2579 (abc)(?i:(?-1))
2580 .sp
2581 It matches "abcabc". It does not match "abcABC" because the change of
2582 processing option does not affect the called subpattern.
2583 .
2584 .
2585 .\" HTML <a name="onigurumasubroutines"></a>
2587 .rs
2588 .sp
2589 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
2590 a number enclosed either in angle brackets or single quotes, is an alternative
2591 syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2592 are two of the examples used above, rewritten using this syntax:
2593 .sp
2594 (?<pn> \e( ( (?>[^()]+) | \eg<pn> )* \e) )
2595 (sens|respons)e and \eg'1'ibility
2596 .sp
2597 PCRE supports an extension to Oniguruma: if a number is preceded by a
2598 plus or a minus sign it is taken as a relative reference. For example:
2599 .sp
2600 (abc)(?i:\eg<-1>)
2601 .sp
2602 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
2603 synonymous. The former is a back reference; the latter is a subroutine call.
2604 .
2605 .
2607 .rs
2608 .sp
2609 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2610 code to be obeyed in the middle of matching a regular expression. This makes it
2611 possible, amongst other things, to extract different substrings that match the
2612 same pair of parentheses when there is a repetition.
2613 .P
2614 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2615 code. The feature is called "callout". The caller of PCRE provides an external
2616 function by putting its entry point in the global variable \fIpcre_callout\fP
2617 (8-bit library) or \fIpcre[16|32]_callout\fP (16-bit or 32-bit library).
2618 By default, this variable contains NULL, which disables all calling out.
2619 .P
2620 Within a regular expression, (?C) indicates the points at which the external
2621 function is to be called. If you want to identify different callout points, you
2622 can put a number less than 256 after the letter C. The default value is zero.
2623 For example, this pattern has two callout points:
2624 .sp
2625 (?C1)abc(?C2)def
2626 .sp
2627 If the PCRE_AUTO_CALLOUT flag is passed to a compiling function, callouts are
2628 automatically installed before each item in the pattern. They are all numbered
2629 255.
2630 .P
2631 During matching, when PCRE reaches a callout point, the external function is
2632 called. It is provided with the number of the callout, the position in the
2633 pattern, and, optionally, one item of data originally supplied by the caller of
2634 the matching function. The callout function may cause matching to proceed, to
2635 backtrack, or to fail altogether. A complete description of the interface to
2636 the callout function is given in the
2637 .\" HREF
2638 \fBpcrecallout\fP
2639 .\"
2640 documentation.
2641 .
2642 .
2643 .\" HTML <a name="backtrackcontrol"></a>
2645 .rs
2646 .sp
2647 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2648 are described in the Perl documentation as "experimental and subject to change
2649 or removal in a future version of Perl". It goes on to say: "Their usage in
2650 production code should be noted to avoid problems during upgrades." The same
2651 remarks apply to the PCRE features described in this section.
2652 .P
2653 Since these verbs are specifically related to backtracking, most of them can be
2654 used only when the pattern is to be matched using one of the traditional
2655 matching functions, which use a backtracking algorithm. With the exception of
2656 (*FAIL), which behaves like a failing negative assertion, they cause an error
2657 if encountered by a DFA matching function.
2658 .P
2659 If any of these verbs are used in an assertion or in a subpattern that is
2660 called as a subroutine (whether or not recursively), their effect is confined
2661 to that subpattern; it does not extend to the surrounding pattern, with one
2662 exception: the name from a *(MARK), (*PRUNE), or (*THEN) that is encountered in
2663 a successful positive assertion \fIis\fP passed back when a match succeeds
2664 (compare capturing parentheses in assertions). Note that such subpatterns are
2665 processed as anchored at the point where they are tested. Note also that Perl's
2666 treatment of subroutines and assertions is different in some cases.
2667 .P
2668 The new verbs make use of what was previously invalid syntax: an opening
2669 parenthesis followed by an asterisk. They are generally of the form
2670 (*VERB) or (*VERB:NAME). Some may take either form, with differing behaviour,
2671 depending on whether or not an argument is present. A name is any sequence of
2672 characters that does not include a closing parenthesis. The maximum length of
2673 name is 255 in the 8-bit library and 65535 in the 16-bit and 32-bit library.
2674 If the name is empty, that is, if the closing parenthesis immediately follows
2675 the colon, the effect is as if the colon were not there. Any number of these
2676 verbs may occur in a pattern.
2677 .
2678 .
2679 .\" HTML <a name="nooptimize"></a>
2680 .SS "Optimizations that affect backtracking verbs"
2681 .rs
2682 .sp
2683 PCRE contains some optimizations that are used to speed up matching by running
2684 some checks at the start of each match attempt. For example, it may know the
2685 minimum length of matching subject, or that a particular character must be
2686 present. When one of these optimizations suppresses the running of a match, any
2687 included backtracking verbs will not, of course, be processed. You can suppress
2688 the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2689 when calling \fBpcre_compile()\fP or \fBpcre_exec()\fP, or by starting the
2690 pattern with (*NO_START_OPT). There is more discussion of this option in the
2691 section entitled
2692 .\" HTML <a href="pcreapi.html#execoptions">
2693 .\" </a>
2694 "Option bits for \fBpcre_exec()\fP"
2695 .\"
2696 in the
2697 .\" HREF
2698 \fBpcreapi\fP
2699 .\"
2700 documentation.
2701 .P
2702 Experiments with Perl suggest that it too has similar optimizations, sometimes
2703 leading to anomalous results.
2704 .
2705 .
2706 .SS "Verbs that act immediately"
2707 .rs
2708 .sp
2709 The following verbs act as soon as they are encountered. They may not be
2710 followed by a name.
2711 .sp
2712 (*ACCEPT)
2713 .sp
2714 This verb causes the match to end successfully, skipping the remainder of the
2715 pattern. However, when it is inside a subpattern that is called as a
2716 subroutine, only that subpattern is ended successfully. Matching then continues
2717 at the outer level. If (*ACCEPT) is inside capturing parentheses, the data so
2718 far is captured. For example:
2719 .sp
2720 A((?:A|B(*ACCEPT)|C)D)
2721 .sp
2722 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2723 the outer parentheses.
2724 .sp
2725 (*FAIL) or (*F)
2726 .sp
2727 This verb causes a matching failure, forcing backtracking to occur. It is
2728 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2729 probably useful only when combined with (?{}) or (??{}). Those are, of course,
2730 Perl features that are not present in PCRE. The nearest equivalent is the
2731 callout feature, as for example in this pattern:
2732 .sp
2733 a+(?C)(*FAIL)
2734 .sp
2735 A match with the string "aaaa" always fails, but the callout is taken before
2736 each backtrack happens (in this example, 10 times).
2737 .
2738 .
2739 .SS "Recording which path was taken"
2740 .rs
2741 .sp
2742 There is one verb whose main purpose is to track how a match was arrived at,
2743 though it also has a secondary use in conjunction with advancing the match
2744 starting point (see (*SKIP) below).
2745 .sp
2746 (*MARK:NAME) or (*:NAME)
2747 .sp
2748 A name is always required with this verb. There may be as many instances of
2749 (*MARK) as you like in a pattern, and their names do not have to be unique.
2750 .P
2751 When a match succeeds, the name of the last-encountered (*MARK) on the matching
2752 path is passed back to the caller as described in the section entitled
2753 .\" HTML <a href="pcreapi.html#extradata">
2754 .\" </a>
2755 "Extra data for \fBpcre_exec()\fP"
2756 .\"
2757 in the
2758 .\" HREF
2759 \fBpcreapi\fP
2760 .\"
2761 documentation. Here is an example of \fBpcretest\fP output, where the /K
2762 modifier requests the retrieval and outputting of (*MARK) data:
2763 .sp
2764 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2765 data> XY
2766 0: XY
2767 MK: A
2768 XZ
2769 0: XZ
2770 MK: B
2771 .sp
2772 The (*MARK) name is tagged with "MK:" in this output, and in this example it
2773 indicates which of the two alternatives matched. This is a more efficient way
2774 of obtaining this information than putting each alternative in its own
2775 capturing parentheses.
2776 .P
2777 If (*MARK) is encountered in a positive assertion, its name is recorded and
2778 passed back if it is the last-encountered. This does not happen for negative
2779 assertions.
2780 .P
2781 After a partial match or a failed match, the name of the last encountered
2782 (*MARK) in the entire match process is returned. For example:
2783 .sp
2784 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2785 data> XP
2786 No match, mark = B
2787 .sp
2788 Note that in this unanchored example the mark is retained from the match
2789 attempt that started at the letter "X" in the subject. Subsequent match
2790 attempts starting at "P" and then with an empty string do not get as far as the
2791 (*MARK) item, but nevertheless do not reset it.
2792 .P
2793 If you are interested in (*MARK) values after failed matches, you should
2794 probably set the PCRE_NO_START_OPTIMIZE option
2795 .\" HTML <a href="#nooptimize">
2796 .\" </a>
2797 (see above)
2798 .\"
2799 to ensure that the match is always attempted.
2800 .
2801 .
2802 .SS "Verbs that act after backtracking"
2803 .rs
2804 .sp
2805 The following verbs do nothing when they are encountered. Matching continues
2806 with what follows, but if there is no subsequent match, causing a backtrack to
2807 the verb, a failure is forced. That is, backtracking cannot pass to the left of
2808 the verb. However, when one of these verbs appears inside an atomic group, its
2809 effect is confined to that group, because once the group has been matched,
2810 there is never any backtracking into it. In this situation, backtracking can
2811 "jump back" to the left of the entire atomic group. (Remember also, as stated
2812 above, that this localization also applies in subroutine calls and assertions.)
2813 .P
2814 These verbs differ in exactly what kind of failure occurs when backtracking
2815 reaches them.
2816 .sp
2817 (*COMMIT)
2818 .sp
2819 This verb, which may not be followed by a name, causes the whole match to fail
2820 outright if the rest of the pattern does not match. Even if the pattern is
2821 unanchored, no further attempts to find a match by advancing the starting point
2822 take place. Once (*COMMIT) has been passed, \fBpcre_exec()\fP is committed to
2823 finding a match at the current starting point, or not at all. For example:
2824 .sp
2825 a+(*COMMIT)b
2826 .sp
2827 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2828 dynamic anchor, or "I've started, so I must finish." The name of the most
2829 recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
2830 match failure.
2831 .P
2832 Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
2833 unless PCRE's start-of-match optimizations are turned off, as shown in this
2834 \fBpcretest\fP example:
2835 .sp
2836 re> /(*COMMIT)abc/
2837 data> xyzabc
2838 0: abc
2839 xyzabc\eY
2840 No match
2841 .sp
2842 PCRE knows that any match must start with "a", so the optimization skips along
2843 the subject to "a" before running the first match attempt, which succeeds. When
2844 the optimization is disabled by the \eY escape in the second subject, the match
2845 starts at "x" and so the (*COMMIT) causes it to fail without trying any other
2846 starting points.
2847 .sp
2848 (*PRUNE) or (*PRUNE:NAME)
2849 .sp
2850 This verb causes the match to fail at the current starting position in the
2851 subject if the rest of the pattern does not match. If the pattern is
2852 unanchored, the normal "bumpalong" advance to the next starting character then
2853 happens. Backtracking can occur as usual to the left of (*PRUNE), before it is
2854 reached, or when matching to the right of (*PRUNE), but if there is no match to
2855 the right, backtracking cannot cross (*PRUNE). In simple cases, the use of
2856 (*PRUNE) is just an alternative to an atomic group or possessive quantifier,
2857 but there are some uses of (*PRUNE) that cannot be expressed in any other way.
2858 The behaviour of (*PRUNE:NAME) is the same as (*MARK:NAME)(*PRUNE). In an
2859 anchored pattern (*PRUNE) has the same effect as (*COMMIT).
2860 .sp
2861 (*SKIP)
2862 .sp
2863 This verb, when given without a name, is like (*PRUNE), except that if the
2864 pattern is unanchored, the "bumpalong" advance is not to the next character,
2865 but to the position in the subject where (*SKIP) was encountered. (*SKIP)
2866 signifies that whatever text was matched leading up to it cannot be part of a
2867 successful match. Consider:
2868 .sp
2869 a+(*SKIP)b
2870 .sp
2871 If the subject is "aaaac...", after the first match attempt fails (starting at
2872 the first character in the string), the starting point skips on to start the
2873 next attempt at "c". Note that a possessive quantifer does not have the same
2874 effect as this example; although it would suppress backtracking during the
2875 first match attempt, the second attempt would start at the second character
2876 instead of skipping on to "c".
2877 .sp
2878 (*SKIP:NAME)
2879 .sp
2880 When (*SKIP) has an associated name, its behaviour is modified. If the
2881 following pattern fails to match, the previous path through the pattern is
2882 searched for the most recent (*MARK) that has the same name. If one is found,
2883 the "bumpalong" advance is to the subject position that corresponds to that
2884 (*MARK) instead of to where (*SKIP) was encountered. If no (*MARK) with a
2885 matching name is found, the (*SKIP) is ignored.
2886 .sp
2887 (*THEN) or (*THEN:NAME)
2888 .sp
2889 This verb causes a skip to the next innermost alternative if the rest of the
2890 pattern does not match. That is, it cancels pending backtracking, but only
2891 within the current alternative. Its name comes from the observation that it can
2892 be used for a pattern-based if-then-else block:
2893 .sp
2894 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2895 .sp
2896 If the COND1 pattern matches, FOO is tried (and possibly further items after
2897 the end of the group if FOO succeeds); on failure, the matcher skips to the
2898 second alternative and tries COND2, without backtracking into COND1. The
2899 behaviour of (*THEN:NAME) is exactly the same as (*MARK:NAME)(*THEN).
2900 If (*THEN) is not inside an alternation, it acts like (*PRUNE).
2901 .P
2902 Note that a subpattern that does not contain a | character is just a part of
2903 the enclosing alternative; it is not a nested alternation with only one
2904 alternative. The effect of (*THEN) extends beyond such a subpattern to the
2905 enclosing alternative. Consider this pattern, where A, B, etc. are complex
2906 pattern fragments that do not contain any | characters at this level:
2907 .sp
2908 A (B(*THEN)C) | D
2909 .sp
2910 If A and B are matched, but there is a failure in C, matching does not
2911 backtrack into A; instead it moves to the next alternative, that is, D.
2912 However, if the subpattern containing (*THEN) is given an alternative, it
2913 behaves differently:
2914 .sp
2915 A (B(*THEN)C | (*FAIL)) | D
2916 .sp
2917 The effect of (*THEN) is now confined to the inner subpattern. After a failure
2918 in C, matching moves to (*FAIL), which causes the whole subpattern to fail
2919 because there are no more alternatives to try. In this case, matching does now
2920 backtrack into A.
2921 .P
2922 Note also that a conditional subpattern is not considered as having two
2923 alternatives, because only one is ever used. In other words, the | character in
2924 a conditional subpattern has a different meaning. Ignoring white space,
2925 consider:
2926 .sp
2927 ^.*? (?(?=a) a | b(*THEN)c )
2928 .sp
2929 If the subject is "ba", this pattern does not match. Because .*? is ungreedy,
2930 it initially matches zero characters. The condition (?=a) then fails, the
2931 character "b" is matched, but "c" is not. At this point, matching does not
2932 backtrack to .*? as might perhaps be expected from the presence of the |
2933 character. The conditional subpattern is part of the single alternative that
2934 comprises the whole pattern, and so the match fails. (If there was a backtrack
2935 into .*?, allowing it to match "b", the match would succeed.)
2936 .P
2937 The verbs just described provide four different "strengths" of control when
2938 subsequent matching fails. (*THEN) is the weakest, carrying on the match at the
2939 next alternative. (*PRUNE) comes next, failing the match at the current
2940 starting position, but allowing an advance to the next character (for an
2941 unanchored pattern). (*SKIP) is similar, except that the advance may be more
2942 than one character. (*COMMIT) is the strongest, causing the entire match to
2943 fail.
2944 .P
2945 If more than one such verb is present in a pattern, the "strongest" one wins.
2946 For example, consider this pattern, where A, B, etc. are complex pattern
2947 fragments:
2948 .sp
2949 (A(*COMMIT)B(*THEN)C|D)
2950 .sp
2951 Once A has matched, PCRE is committed to this match, at the current starting
2952 position. If subsequently B matches, but C does not, the normal (*THEN) action
2953 of trying the next alternative (that is, D) does not happen because (*COMMIT)
2954 overrides.
2955 .
2956 .
2957 .SH "SEE ALSO"
2958 .rs
2959 .sp
2960 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3),
2961 \fBpcresyntax\fP(3), \fBpcre\fP(3), \fBpcre16(3)\fP, \fBpcre32(3)\fP.
2962 .
2963 .
2965 .rs
2966 .sp
2967 .nf
2968 Philip Hazel
2969 University Computing Service
2970 Cambridge CB2 3QH, England.
2971 .fi
2972 .
2973 .
2975 .rs
2976 .sp
2977 .nf
2978 Last updated: 07 November 2012
2979 Copyright (c) 1997-2012 University of Cambridge.
2980 .fi


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

  ViewVC Help
Powered by ViewVC 1.1.5