ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 91 - (show annotations)
Sat Feb 24 21:41:34 2007 UTC (8 years, 7 months ago) by nigel
File size: 68437 byte(s)
Error occurred while calculating annotation data.
Load pcre-6.7 into code/trunk.
3 PCRE - Perl-compatible regular expressions
5 .rs
6 .sp
7 The syntax and semantics of the regular expressions supported by PCRE are
8 described below. Regular expressions are also described in the Perl
9 documentation and in a number of books, some of which have copious examples.
10 Jeffrey Friedl's "Mastering Regular Expressions", published by O'Reilly, covers
11 regular expressions in great detail. This description of PCRE's regular
12 expressions is intended as reference material.
13 .P
14 The original operation of PCRE was on strings of one-byte characters. However,
15 there is now also support for UTF-8 character strings. To use this, you must
16 build PCRE to include UTF-8 support, and then call \fBpcre_compile()\fP with
17 the PCRE_UTF8 option. How this affects pattern matching is mentioned in several
18 places below. There is also a summary of UTF-8 features in the
19 .\" HTML <a href="pcre.html#utf8support">
20 .\" </a>
21 section on UTF-8 support
22 .\"
23 in the main
24 .\" HREF
25 \fBpcre\fP
26 .\"
27 page.
28 .P
29 The remainder of this document discusses the patterns that are supported by
30 PCRE when its main matching function, \fBpcre_exec()\fP, is used.
31 From release 6.0, PCRE offers a second matching function,
32 \fBpcre_dfa_exec()\fP, which matches using a different algorithm that is not
33 Perl-compatible. The advantages and disadvantages of the alternative function,
34 and how it differs from the normal function, are discussed in the
35 .\" HREF
36 \fBpcrematching\fP
37 .\"
38 page.
39 .P
40 A regular expression is a pattern that is matched against a subject string from
41 left to right. Most characters stand for themselves in a pattern, and match the
42 corresponding characters in the subject. As a trivial example, the pattern
43 .sp
44 The quick brown fox
45 .sp
46 matches a portion of a subject string that is identical to itself. When
47 caseless matching is specified (the PCRE_CASELESS option), letters are matched
48 independently of case. In UTF-8 mode, PCRE always understands the concept of
49 case for characters whose values are less than 128, so caseless matching is
50 always possible. For characters with higher values, the concept of case is
51 supported if PCRE is compiled with Unicode property support, but not otherwise.
52 If you want to use caseless matching for characters 128 and above, you must
53 ensure that PCRE is compiled with Unicode property support as well as with
54 UTF-8 support.
55 .P
56 The power of regular expressions comes from the ability to include alternatives
57 and repetitions in the pattern. These are encoded in the pattern by the use of
58 \fImetacharacters\fP, which do not stand for themselves but instead are
59 interpreted in some special way.
60 .P
61 There are two different sets of metacharacters: those that are recognized
62 anywhere in the pattern except within square brackets, and those that are
63 recognized in square brackets. Outside square brackets, the metacharacters are
64 as follows:
65 .sp
66 \e general escape character with several uses
67 ^ assert start of string (or line, in multiline mode)
68 $ assert end of string (or line, in multiline mode)
69 . match any character except newline (by default)
70 [ start character class definition
71 | start of alternative branch
72 ( start subpattern
73 ) end subpattern
74 ? extends the meaning of (
75 also 0 or 1 quantifier
76 also quantifier minimizer
77 * 0 or more quantifier
78 + 1 or more quantifier
79 also "possessive quantifier"
80 { start min/max quantifier
81 .sp
82 Part of a pattern that is in square brackets is called a "character class". In
83 a character class the only metacharacters are:
84 .sp
85 \e general escape character
86 ^ negate the class, but only if the first character
87 - indicates character range
88 .\" JOIN
89 [ POSIX character class (only if followed by POSIX
90 syntax)
91 ] terminates the character class
92 .sp
93 The following sections describe the use of each of the metacharacters.
94 .
96 .rs
97 .sp
98 The backslash character has several uses. Firstly, if it is followed by a
99 non-alphanumeric character, it takes away any special meaning that character
100 may have. This use of backslash as an escape character applies both inside and
101 outside character classes.
102 .P
103 For example, if you want to match a * character, you write \e* in the pattern.
104 This escaping action applies whether or not the following character would
105 otherwise be interpreted as a metacharacter, so it is always safe to precede a
106 non-alphanumeric with backslash to specify that it stands for itself. In
107 particular, if you want to match a backslash, you write \e\e.
108 .P
109 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
110 pattern (other than in a character class) and characters between a # outside
111 a character class and the next newline are ignored. An escaping backslash can
112 be used to include a whitespace or # character as part of the pattern.
113 .P
114 If you want to remove the special meaning from a sequence of characters, you
115 can do so by putting them between \eQ and \eE. This is different from Perl in
116 that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
117 Perl, $ and @ cause variable interpolation. Note the following examples:
118 .sp
119 Pattern PCRE matches Perl matches
120 .sp
121 .\" JOIN
122 \eQabc$xyz\eE abc$xyz abc followed by the
123 contents of $xyz
124 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
125 \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
126 .sp
127 The \eQ...\eE sequence is recognized both inside and outside character classes.
128 .
129 .
130 .\" HTML <a name="digitsafterbackslash"></a>
131 .SS "Non-printing characters"
132 .rs
133 .sp
134 A second use of backslash provides a way of encoding non-printing characters
135 in patterns in a visible manner. There is no restriction on the appearance of
136 non-printing characters, apart from the binary zero that terminates a pattern,
137 but when a pattern is being prepared by text editing, it is usually easier to
138 use one of the following escape sequences than the binary character it
139 represents:
140 .sp
141 \ea alarm, that is, the BEL character (hex 07)
142 \ecx "control-x", where x is any character
143 \ee escape (hex 1B)
144 \ef formfeed (hex 0C)
145 \en newline (hex 0A)
146 \er carriage return (hex 0D)
147 \et tab (hex 09)
148 \eddd character with octal code ddd, or backreference
149 \exhh character with hex code hh
150 \ex{hhh..} character with hex code hhh..
151 .sp
152 The precise effect of \ecx is as follows: if x is a lower case letter, it
153 is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
154 Thus \ecz becomes hex 1A, but \ec{ becomes hex 3B, while \ec; becomes hex
155 7B.
156 .P
157 After \ex, from zero to two hexadecimal digits are read (letters can be in
158 upper or lower case). Any number of hexadecimal digits may appear between \ex{
159 and }, but the value of the character code must be less than 256 in non-UTF-8
160 mode, and less than 2**31 in UTF-8 mode (that is, the maximum hexadecimal value
161 is 7FFFFFFF). If characters other than hexadecimal digits appear between \ex{
162 and }, or if there is no terminating }, this form of escape is not recognized.
163 Instead, the initial \ex will be interpreted as a basic hexadecimal escape,
164 with no following digits, giving a character whose value is zero.
165 .P
166 Characters whose value is less than 256 can be defined by either of the two
167 syntaxes for \ex. There is no difference in the way they are handled. For
168 example, \exdc is exactly the same as \ex{dc}.
169 .P
170 After \e0 up to two further octal digits are read. If there are fewer than two
171 digits, just those that are present are used. Thus the sequence \e0\ex\e07
172 specifies two binary zeros followed by a BEL character (code value 7). Make
173 sure you supply two digits after the initial zero if the pattern character that
174 follows is itself an octal digit.
175 .P
176 The handling of a backslash followed by a digit other than 0 is complicated.
177 Outside a character class, PCRE reads it and any following digits as a decimal
178 number. If the number is less than 10, or if there have been at least that many
179 previous capturing left parentheses in the expression, the entire sequence is
180 taken as a \fIback reference\fP. A description of how this works is given
181 .\" HTML <a href="#backreferences">
182 .\" </a>
183 later,
184 .\"
185 following the discussion of
186 .\" HTML <a href="#subpattern">
187 .\" </a>
188 parenthesized subpatterns.
189 .\"
190 .P
191 Inside a character class, or if the decimal number is greater than 9 and there
192 have not been that many capturing subpatterns, PCRE re-reads up to three octal
193 digits following the backslash, ane uses them to generate a data character. Any
194 subsequent digits stand for themselves. In non-UTF-8 mode, the value of a
195 character specified in octal must be less than \e400. In UTF-8 mode, values up
196 to \e777 are permitted. For example:
197 .sp
198 \e040 is another way of writing a space
199 .\" JOIN
200 \e40 is the same, provided there are fewer than 40
201 previous capturing subpatterns
202 \e7 is always a back reference
203 .\" JOIN
204 \e11 might be a back reference, or another way of
205 writing a tab
206 \e011 is always a tab
207 \e0113 is a tab followed by the character "3"
208 .\" JOIN
209 \e113 might be a back reference, otherwise the
210 character with octal code 113
211 .\" JOIN
212 \e377 might be a back reference, otherwise
213 the byte consisting entirely of 1 bits
214 .\" JOIN
215 \e81 is either a back reference, or a binary zero
216 followed by the two characters "8" and "1"
217 .sp
218 Note that octal values of 100 or greater must not be introduced by a leading
219 zero, because no more than three octal digits are ever read.
220 .P
221 All the sequences that define a single character value can be used both inside
222 and outside character classes. In addition, inside a character class, the
223 sequence \eb is interpreted as the backspace character (hex 08), and the
224 sequence \eX is interpreted as the character "X". Outside a character class,
225 these sequences have different meanings
226 .\" HTML <a href="#uniextseq">
227 .\" </a>
228 (see below).
229 .\"
230 .
231 .
232 .SS "Generic character types"
233 .rs
234 .sp
235 The third use of backslash is for specifying generic character types. The
236 following are always recognized:
237 .sp
238 \ed any decimal digit
239 \eD any character that is not a decimal digit
240 \es any whitespace character
241 \eS any character that is not a whitespace character
242 \ew any "word" character
243 \eW any "non-word" character
244 .sp
245 Each pair of escape sequences partitions the complete set of characters into
246 two disjoint sets. Any given character matches one, and only one, of each pair.
247 .P
248 These character type sequences can appear both inside and outside character
249 classes. They each match one character of the appropriate type. If the current
250 matching point is at the end of the subject string, all of them fail, since
251 there is no character to match.
252 .P
253 For compatibility with Perl, \es does not match the VT character (code 11).
254 This makes it different from the the POSIX "space" class. The \es characters
255 are HT (9), LF (10), FF (12), CR (13), and space (32). (If "use locale;" is
256 included in a Perl script, \es may match the VT character. In PCRE, it never
257 does.)
258 .P
259 A "word" character is an underscore or any character less than 256 that is a
260 letter or digit. The definition of letters and digits is controlled by PCRE's
261 low-valued character tables, and may vary if locale-specific matching is taking
262 place (see
263 .\" HTML <a href="pcreapi.html#localesupport">
264 .\" </a>
265 "Locale support"
266 .\"
267 in the
268 .\" HREF
269 \fBpcreapi\fP
270 .\"
271 page). For example, in the "fr_FR" (French) locale, some character codes
272 greater than 128 are used for accented letters, and these are matched by \ew.
273 .P
274 In UTF-8 mode, characters with values greater than 128 never match \ed, \es, or
275 \ew, and always match \eD, \eS, and \eW. This is true even when Unicode
276 character property support is available. The use of locales with Unicode is
277 discouraged.
278 .
279 .
280 .\" HTML <a name="uniextseq"></a>
281 .SS Unicode character properties
282 .rs
283 .sp
284 When PCRE is built with Unicode character property support, three additional
285 escape sequences to match character properties are available when UTF-8 mode
286 is selected. They are:
287 .sp
288 \ep{\fIxx\fP} a character with the \fIxx\fP property
289 \eP{\fIxx\fP} a character without the \fIxx\fP property
290 \eX an extended Unicode sequence
291 .sp
292 The property names represented by \fIxx\fP above are limited to the Unicode
293 script names, the general category properties, and "Any", which matches any
294 character (including newline). Other properties such as "InMusicalSymbols" are
295 not currently supported by PCRE. Note that \eP{Any} does not match any
296 characters, so always causes a match failure.
297 .P
298 Sets of Unicode characters are defined as belonging to certain scripts. A
299 character from one of these sets can be matched using a script name. For
300 example:
301 .sp
302 \ep{Greek}
303 \eP{Han}
304 .sp
305 Those that are not part of an identified script are lumped together as
306 "Common". The current list of scripts is:
307 .P
308 Arabic,
309 Armenian,
310 Bengali,
311 Bopomofo,
312 Braille,
313 Buginese,
314 Buhid,
315 Canadian_Aboriginal,
316 Cherokee,
317 Common,
318 Coptic,
319 Cypriot,
320 Cyrillic,
321 Deseret,
322 Devanagari,
323 Ethiopic,
324 Georgian,
325 Glagolitic,
326 Gothic,
327 Greek,
328 Gujarati,
329 Gurmukhi,
330 Han,
331 Hangul,
332 Hanunoo,
333 Hebrew,
334 Hiragana,
335 Inherited,
336 Kannada,
337 Katakana,
338 Kharoshthi,
339 Khmer,
340 Lao,
341 Latin,
342 Limbu,
343 Linear_B,
344 Malayalam,
345 Mongolian,
346 Myanmar,
347 New_Tai_Lue,
348 Ogham,
349 Old_Italic,
350 Old_Persian,
351 Oriya,
352 Osmanya,
353 Runic,
354 Shavian,
355 Sinhala,
356 Syloti_Nagri,
357 Syriac,
358 Tagalog,
359 Tagbanwa,
360 Tai_Le,
361 Tamil,
362 Telugu,
363 Thaana,
364 Thai,
365 Tibetan,
366 Tifinagh,
367 Ugaritic,
368 Yi.
369 .P
370 Each character has exactly one general category property, specified by a
371 two-letter abbreviation. For compatibility with Perl, negation can be specified
372 by including a circumflex between the opening brace and the property name. For
373 example, \ep{^Lu} is the same as \eP{Lu}.
374 .P
375 If only one letter is specified with \ep or \eP, it includes all the general
376 category properties that start with that letter. In this case, in the absence
377 of negation, the curly brackets in the escape sequence are optional; these two
378 examples have the same effect:
379 .sp
380 \ep{L}
381 \epL
382 .sp
383 The following general category property codes are supported:
384 .sp
385 C Other
386 Cc Control
387 Cf Format
388 Cn Unassigned
389 Co Private use
390 Cs Surrogate
391 .sp
392 L Letter
393 Ll Lower case letter
394 Lm Modifier letter
395 Lo Other letter
396 Lt Title case letter
397 Lu Upper case letter
398 .sp
399 M Mark
400 Mc Spacing mark
401 Me Enclosing mark
402 Mn Non-spacing mark
403 .sp
404 N Number
405 Nd Decimal number
406 Nl Letter number
407 No Other number
408 .sp
409 P Punctuation
410 Pc Connector punctuation
411 Pd Dash punctuation
412 Pe Close punctuation
413 Pf Final punctuation
414 Pi Initial punctuation
415 Po Other punctuation
416 Ps Open punctuation
417 .sp
418 S Symbol
419 Sc Currency symbol
420 Sk Modifier symbol
421 Sm Mathematical symbol
422 So Other symbol
423 .sp
424 Z Separator
425 Zl Line separator
426 Zp Paragraph separator
427 Zs Space separator
428 .sp
429 The special property L& is also supported: it matches a character that has
430 the Lu, Ll, or Lt property, in other words, a letter that is not classified as
431 a modifier or "other".
432 .P
433 The long synonyms for these properties that Perl supports (such as \ep{Letter})
434 are not supported by PCRE, nor is it permitted to prefix any of these
435 properties with "Is".
436 .P
437 No character that is in the Unicode table has the Cn (unassigned) property.
438 Instead, this property is assumed for any code point that is not in the
439 Unicode table.
440 .P
441 Specifying caseless matching does not affect these escape sequences. For
442 example, \ep{Lu} always matches only upper case letters.
443 .P
444 The \eX escape matches any number of Unicode characters that form an extended
445 Unicode sequence. \eX is equivalent to
446 .sp
447 (?>\ePM\epM*)
448 .sp
449 That is, it matches a character without the "mark" property, followed by zero
450 or more characters with the "mark" property, and treats the sequence as an
451 atomic group
452 .\" HTML <a href="#atomicgroup">
453 .\" </a>
454 (see below).
455 .\"
456 Characters with the "mark" property are typically accents that affect the
457 preceding character.
458 .P
459 Matching characters by Unicode property is not fast, because PCRE has to search
460 a structure that contains data for over fifteen thousand characters. That is
461 why the traditional escape sequences such as \ed and \ew do not use Unicode
462 properties in PCRE.
463 .
464 .
465 .\" HTML <a name="smallassertions"></a>
466 .SS "Simple assertions"
467 .rs
468 .sp
469 The fourth use of backslash is for certain simple assertions. An assertion
470 specifies a condition that has to be met at a particular point in a match,
471 without consuming any characters from the subject string. The use of
472 subpatterns for more complicated assertions is described
473 .\" HTML <a href="#bigassertions">
474 .\" </a>
475 below.
476 .\"
477 The backslashed assertions are:
478 .sp
479 \eb matches at a word boundary
480 \eB matches when not at a word boundary
481 \eA matches at start of subject
482 \eZ matches at end of subject or before newline at end
483 \ez matches at end of subject
484 \eG matches at first matching position in subject
485 .sp
486 These assertions may not appear in character classes (but note that \eb has a
487 different meaning, namely the backspace character, inside a character class).
488 .P
489 A word boundary is a position in the subject string where the current character
490 and the previous character do not both match \ew or \eW (i.e. one matches
491 \ew and the other matches \eW), or the start or end of the string if the
492 first or last character matches \ew, respectively.
493 .P
494 The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
495 dollar (described in the next section) in that they only ever match at the very
496 start and end of the subject string, whatever options are set. Thus, they are
497 independent of multiline mode. These three assertions are not affected by the
498 PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
499 circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
500 argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
501 at a point other than the beginning of the subject, \eA can never match. The
502 difference between \eZ and \ez is that \eZ matches before a newline at the end
503 of the string as well as at the very end, whereas \ez matches only at the end.
504 .P
505 The \eG assertion is true only when the current matching position is at the
506 start point of the match, as specified by the \fIstartoffset\fP argument of
507 \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
508 non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
509 arguments, you can mimic Perl's /g option, and it is in this kind of
510 implementation where \eG can be useful.
511 .P
512 Note, however, that PCRE's interpretation of \eG, as the start of the current
513 match, is subtly different from Perl's, which defines it as the end of the
514 previous match. In Perl, these can be different when the previously matched
515 string was empty. Because PCRE does just one match at a time, it cannot
516 reproduce this behaviour.
517 .P
518 If all the alternatives of a pattern begin with \eG, the expression is anchored
519 to the starting match position, and the "anchored" flag is set in the compiled
520 regular expression.
521 .
522 .
524 .rs
525 .sp
526 Outside a character class, in the default matching mode, the circumflex
527 character is an assertion that is true only if the current matching point is
528 at the start of the subject string. If the \fIstartoffset\fP argument of
529 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
530 option is unset. Inside a character class, circumflex has an entirely different
531 meaning
532 .\" HTML <a href="#characterclass">
533 .\" </a>
534 (see below).
535 .\"
536 .P
537 Circumflex need not be the first character of the pattern if a number of
538 alternatives are involved, but it should be the first thing in each alternative
539 in which it appears if the pattern is ever to match that branch. If all
540 possible alternatives start with a circumflex, that is, if the pattern is
541 constrained to match only at the start of the subject, it is said to be an
542 "anchored" pattern. (There are also other constructs that can cause a pattern
543 to be anchored.)
544 .P
545 A dollar character is an assertion that is true only if the current matching
546 point is at the end of the subject string, or immediately before a newline
547 at the end of the string (by default). Dollar need not be the last character of
548 the pattern if a number of alternatives are involved, but it should be the last
549 item in any branch in which it appears. Dollar has no special meaning in a
550 character class.
551 .P
552 The meaning of dollar can be changed so that it matches only at the very end of
553 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
554 does not affect the \eZ assertion.
555 .P
556 The meanings of the circumflex and dollar characters are changed if the
557 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
558 immediately after internal newlines as well as at the start of the subject
559 string. It does not match after a newline that ends the string. A dollar
560 matches before any newlines in the string, as well as at the very end, when
561 PCRE_MULTILINE is set. When newline is specified as the two-character
562 sequence CRLF, isolated CR and LF characters do not indicate newlines.
563 .P
564 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
565 \en represents a newline) in multiline mode, but not otherwise. Consequently,
566 patterns that are anchored in single line mode because all branches start with
567 ^ are not anchored in multiline mode, and a match for circumflex is possible
568 when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
569 PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
570 .P
571 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
572 end of the subject in both modes, and if all branches of a pattern start with
573 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
574 .
575 .
577 .rs
578 .sp
579 Outside a character class, a dot in the pattern matches any one character in
580 the subject string except (by default) a character that signifies the end of a
581 line. In UTF-8 mode, the matched character may be more than one byte long. When
582 a line ending is defined as a single character (CR or LF), dot never matches
583 that character; when the two-character sequence CRLF is used, dot does not
584 match CR if it is immediately followed by LF, but otherwise it matches all
585 characters (including isolated CRs and LFs).
586 .P
587 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
588 option is set, a dot matches any one character, without exception. If newline
589 is defined as the two-character sequence CRLF, it takes two dots to match it.
590 .P
591 The handling of dot is entirely independent of the handling of circumflex and
592 dollar, the only relationship being that they both involve newlines. Dot has no
593 special meaning in a character class.
594 .
595 .
597 .rs
598 .sp
599 Outside a character class, the escape sequence \eC matches any one byte, both
600 in and out of UTF-8 mode. Unlike a dot, it always matches CR and LF. The
601 feature is provided in Perl in order to match individual bytes in UTF-8 mode.
602 Because it breaks up UTF-8 characters into individual bytes, what remains in
603 the string may be a malformed UTF-8 string. For this reason, the \eC escape
604 sequence is best avoided.
605 .P
606 PCRE does not allow \eC to appear in lookbehind assertions
607 .\" HTML <a href="#lookbehind">
608 .\" </a>
609 (described below),
610 .\"
611 because in UTF-8 mode this would make it impossible to calculate the length of
612 the lookbehind.
613 .
614 .
615 .\" HTML <a name="characterclass"></a>
617 .rs
618 .sp
619 An opening square bracket introduces a character class, terminated by a closing
620 square bracket. A closing square bracket on its own is not special. If a
621 closing square bracket is required as a member of the class, it should be the
622 first data character in the class (after an initial circumflex, if present) or
623 escaped with a backslash.
624 .P
625 A character class matches a single character in the subject. In UTF-8 mode, the
626 character may occupy more than one byte. A matched character must be in the set
627 of characters defined by the class, unless the first character in the class
628 definition is a circumflex, in which case the subject character must not be in
629 the set defined by the class. If a circumflex is actually required as a member
630 of the class, ensure it is not the first character, or escape it with a
631 backslash.
632 .P
633 For example, the character class [aeiou] matches any lower case vowel, while
634 [^aeiou] matches any character that is not a lower case vowel. Note that a
635 circumflex is just a convenient notation for specifying the characters that
636 are in the class by enumerating those that are not. A class that starts with a
637 circumflex is not an assertion: it still consumes a character from the subject
638 string, and therefore it fails if the current pointer is at the end of the
639 string.
640 .P
641 In UTF-8 mode, characters with values greater than 255 can be included in a
642 class as a literal string of bytes, or by using the \ex{ escaping mechanism.
643 .P
644 When caseless matching is set, any letters in a class represent both their
645 upper case and lower case versions, so for example, a caseless [aeiou] matches
646 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
647 caseful version would. In UTF-8 mode, PCRE always understands the concept of
648 case for characters whose values are less than 128, so caseless matching is
649 always possible. For characters with higher values, the concept of case is
650 supported if PCRE is compiled with Unicode property support, but not otherwise.
651 If you want to use caseless matching for characters 128 and above, you must
652 ensure that PCRE is compiled with Unicode property support as well as with
653 UTF-8 support.
654 .P
655 Characters that might indicate line breaks (CR and LF) are never treated in any
656 special way when matching character classes, whatever line-ending sequence is
657 in use, and whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is
658 used. A class such as [^a] always matches one of these characters.
659 .P
660 The minus (hyphen) character can be used to specify a range of characters in a
661 character class. For example, [d-m] matches any letter between d and m,
662 inclusive. If a minus character is required in a class, it must be escaped with
663 a backslash or appear in a position where it cannot be interpreted as
664 indicating a range, typically as the first or last character in the class.
665 .P
666 It is not possible to have the literal character "]" as the end character of a
667 range. A pattern such as [W-]46] is interpreted as a class of two characters
668 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
669 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
670 the end of range, so [W-\e]46] is interpreted as a class containing a range
671 followed by two other characters. The octal or hexadecimal representation of
672 "]" can also be used to end a range.
673 .P
674 Ranges operate in the collating sequence of character values. They can also be
675 used for characters specified numerically, for example [\e000-\e037]. In UTF-8
676 mode, ranges can include characters whose values are greater than 255, for
677 example [\ex{100}-\ex{2ff}].
678 .P
679 If a range that includes letters is used when caseless matching is set, it
680 matches the letters in either case. For example, [W-c] is equivalent to
681 [][\e\e^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
682 tables for the "fr_FR" locale are in use, [\exc8-\excb] matches accented E
683 characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
684 characters with values greater than 128 only when it is compiled with Unicode
685 property support.
686 .P
687 The character types \ed, \eD, \ep, \eP, \es, \eS, \ew, and \eW may also appear
688 in a character class, and add the characters that they match to the class. For
689 example, [\edABCDEF] matches any hexadecimal digit. A circumflex can
690 conveniently be used with the upper case character types to specify a more
691 restricted set of characters than the matching lower case type. For example,
692 the class [^\eW_] matches any letter or digit, but not underscore.
693 .P
694 The only metacharacters that are recognized in character classes are backslash,
695 hyphen (only where it can be interpreted as specifying a range), circumflex
696 (only at the start), opening square bracket (only when it can be interpreted as
697 introducing a POSIX class name - see the next section), and the terminating
698 closing square bracket. However, escaping other non-alphanumeric characters
699 does no harm.
700 .
701 .
703 .rs
704 .sp
705 Perl supports the POSIX notation for character classes. This uses names
706 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
707 this notation. For example,
708 .sp
709 [01[:alpha:]%]
710 .sp
711 matches "0", "1", any alphabetic character, or "%". The supported class names
712 are
713 .sp
714 alnum letters and digits
715 alpha letters
716 ascii character codes 0 - 127
717 blank space or tab only
718 cntrl control characters
719 digit decimal digits (same as \ed)
720 graph printing characters, excluding space
721 lower lower case letters
722 print printing characters, including space
723 punct printing characters, excluding letters and digits
724 space white space (not quite the same as \es)
725 upper upper case letters
726 word "word" characters (same as \ew)
727 xdigit hexadecimal digits
728 .sp
729 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
730 space (32). Notice that this list includes the VT character (code 11). This
731 makes "space" different to \es, which does not include VT (for Perl
732 compatibility).
733 .P
734 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
735 5.8. Another Perl extension is negation, which is indicated by a ^ character
736 after the colon. For example,
737 .sp
738 [12[:^digit:]]
739 .sp
740 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
741 syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
742 supported, and an error is given if they are encountered.
743 .P
744 In UTF-8 mode, characters with values greater than 128 do not match any of
745 the POSIX character classes.
746 .
747 .
749 .rs
750 .sp
751 Vertical bar characters are used to separate alternative patterns. For example,
752 the pattern
753 .sp
754 gilbert|sullivan
755 .sp
756 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
757 and an empty alternative is permitted (matching the empty string). The matching
758 process tries each alternative in turn, from left to right, and the first one
759 that succeeds is used. If the alternatives are within a subpattern
760 .\" HTML <a href="#subpattern">
761 .\" </a>
762 (defined below),
763 .\"
764 "succeeds" means matching the rest of the main pattern as well as the
765 alternative in the subpattern.
766 .
767 .
769 .rs
770 .sp
772 PCRE_EXTENDED options can be changed from within the pattern by a sequence of
773 Perl option letters enclosed between "(?" and ")". The option letters are
774 .sp
777 s for PCRE_DOTALL
779 .sp
780 For example, (?im) sets caseless, multiline matching. It is also possible to
781 unset these options by preceding the letter with a hyphen, and a combined
782 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
783 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
784 permitted. If a letter appears both before and after the hyphen, the option is
785 unset.
786 .P
787 When an option change occurs at top level (that is, not inside subpattern
788 parentheses), the change applies to the remainder of the pattern that follows.
789 If the change is placed right at the start of a pattern, PCRE extracts it into
790 the global options (and it will therefore show up in data extracted by the
791 \fBpcre_fullinfo()\fP function).
792 .P
793 An option change within a subpattern affects only that part of the current
794 pattern that follows it, so
795 .sp
796 (a(?i)b)c
797 .sp
798 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
799 By this means, options can be made to have different settings in different
800 parts of the pattern. Any changes made in one alternative do carry on
801 into subsequent branches within the same subpattern. For example,
802 .sp
803 (a(?i)b|c)
804 .sp
805 matches "ab", "aB", "c", and "C", even though when matching "C" the first
806 branch is abandoned before the option setting. This is because the effects of
807 option settings happen at compile time. There would be some very weird
808 behaviour otherwise.
809 .P
810 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
811 changed in the same way as the Perl-compatible options by using the characters
812 J, U and X respectively.
813 .
814 .
815 .\" HTML <a name="subpattern"></a>
817 .rs
818 .sp
819 Subpatterns are delimited by parentheses (round brackets), which can be nested.
820 Turning part of a pattern into a subpattern does two things:
821 .sp
822 1. It localizes a set of alternatives. For example, the pattern
823 .sp
824 cat(aract|erpillar|)
825 .sp
826 matches one of the words "cat", "cataract", or "caterpillar". Without the
827 parentheses, it would match "cataract", "erpillar" or the empty string.
828 .sp
829 2. It sets up the subpattern as a capturing subpattern. This means that, when
830 the whole pattern matches, that portion of the subject string that matched the
831 subpattern is passed back to the caller via the \fIovector\fP argument of
832 \fBpcre_exec()\fP. Opening parentheses are counted from left to right (starting
833 from 1) to obtain numbers for the capturing subpatterns.
834 .P
835 For example, if the string "the red king" is matched against the pattern
836 .sp
837 the ((red|white) (king|queen))
838 .sp
839 the captured substrings are "red king", "red", and "king", and are numbered 1,
840 2, and 3, respectively.
841 .P
842 The fact that plain parentheses fulfil two functions is not always helpful.
843 There are often times when a grouping subpattern is required without a
844 capturing requirement. If an opening parenthesis is followed by a question mark
845 and a colon, the subpattern does not do any capturing, and is not counted when
846 computing the number of any subsequent capturing subpatterns. For example, if
847 the string "the white queen" is matched against the pattern
848 .sp
849 the ((?:red|white) (king|queen))
850 .sp
851 the captured substrings are "white queen" and "queen", and are numbered 1 and
852 2. The maximum number of capturing subpatterns is 65535, and the maximum depth
853 of nesting of all subpatterns, both capturing and non-capturing, is 200.
854 .P
855 As a convenient shorthand, if any option settings are required at the start of
856 a non-capturing subpattern, the option letters may appear between the "?" and
857 the ":". Thus the two patterns
858 .sp
859 (?i:saturday|sunday)
860 (?:(?i)saturday|sunday)
861 .sp
862 match exactly the same set of strings. Because alternative branches are tried
863 from left to right, and options are not reset until the end of the subpattern
864 is reached, an option setting in one branch does affect subsequent branches, so
865 the above patterns match "SUNDAY" as well as "Saturday".
866 .
867 .
869 .rs
870 .sp
871 Identifying capturing parentheses by number is simple, but it can be very hard
872 to keep track of the numbers in complicated regular expressions. Furthermore,
873 if an expression is modified, the numbers may change. To help with this
874 difficulty, PCRE supports the naming of subpatterns, something that Perl does
875 not provide. The Python syntax (?P<name>...) is used. References to capturing
876 parentheses from other parts of the pattern, such as
877 .\" HTML <a href="#backreferences">
878 .\" </a>
879 backreferences,
880 .\"
881 .\" HTML <a href="#recursion">
882 .\" </a>
883 recursion,
884 .\"
885 and
886 .\" HTML <a href="#conditions">
887 .\" </a>
888 conditions,
889 .\"
890 can be made by name as well as by number.
891 .P
892 Names consist of up to 32 alphanumeric characters and underscores. Named
893 capturing parentheses are still allocated numbers as well as names. The PCRE
894 API provides function calls for extracting the name-to-number translation table
895 from a compiled pattern. There is also a convenience function for extracting a
896 captured substring by name.
897 .P
898 By default, a name must be unique within a pattern, but it is possible to relax
899 this constraint by setting the PCRE_DUPNAMES option at compile time. This can
900 be useful for patterns where only one instance of the named parentheses can
901 match. Suppose you want to match the name of a weekday, either as a 3-letter
902 abbreviation or as the full name, and in both cases you want to extract the
903 abbreviation. This pattern (ignoring the line breaks) does the job:
904 .sp
905 (?P<DN>Mon|Fri|Sun)(?:day)?|
906 (?P<DN>Tue)(?:sday)?|
907 (?P<DN>Wed)(?:nesday)?|
908 (?P<DN>Thu)(?:rsday)?|
909 (?P<DN>Sat)(?:urday)?
910 .sp
911 There are five capturing substrings, but only one is ever set after a match.
912 The convenience function for extracting the data by name returns the substring
913 for the first, and in this example, the only, subpattern of that name that
914 matched. This saves searching to find which numbered subpattern it was. If you
915 make a reference to a non-unique named subpattern from elsewhere in the
916 pattern, the one that corresponds to the lowest number is used. For further
917 details of the interfaces for handling named subpatterns, see the
918 .\" HREF
919 \fBpcreapi\fP
920 .\"
921 documentation.
922 .
923 .
925 .rs
926 .sp
927 Repetition is specified by quantifiers, which can follow any of the following
928 items:
929 .sp
930 a literal data character
931 the . metacharacter
932 the \eC escape sequence
933 the \eX escape sequence (in UTF-8 mode with Unicode properties)
934 an escape such as \ed that matches a single character
935 a character class
936 a back reference (see next section)
937 a parenthesized subpattern (unless it is an assertion)
938 .sp
939 The general repetition quantifier specifies a minimum and maximum number of
940 permitted matches, by giving the two numbers in curly brackets (braces),
941 separated by a comma. The numbers must be less than 65536, and the first must
942 be less than or equal to the second. For example:
943 .sp
944 z{2,4}
945 .sp
946 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
947 character. If the second number is omitted, but the comma is present, there is
948 no upper limit; if the second number and the comma are both omitted, the
949 quantifier specifies an exact number of required matches. Thus
950 .sp
951 [aeiou]{3,}
952 .sp
953 matches at least 3 successive vowels, but may match many more, while
954 .sp
955 \ed{8}
956 .sp
957 matches exactly 8 digits. An opening curly bracket that appears in a position
958 where a quantifier is not allowed, or one that does not match the syntax of a
959 quantifier, is taken as a literal character. For example, {,6} is not a
960 quantifier, but a literal string of four characters.
961 .P
962 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
963 bytes. Thus, for example, \ex{100}{2} matches two UTF-8 characters, each of
964 which is represented by a two-byte sequence. Similarly, when Unicode property
965 support is available, \eX{3} matches three Unicode extended sequences, each of
966 which may be several bytes long (and they may be of different lengths).
967 .P
968 The quantifier {0} is permitted, causing the expression to behave as if the
969 previous item and the quantifier were not present.
970 .P
971 For convenience (and historical compatibility) the three most common
972 quantifiers have single-character abbreviations:
973 .sp
974 * is equivalent to {0,}
975 + is equivalent to {1,}
976 ? is equivalent to {0,1}
977 .sp
978 It is possible to construct infinite loops by following a subpattern that can
979 match no characters with a quantifier that has no upper limit, for example:
980 .sp
981 (a?)*
982 .sp
983 Earlier versions of Perl and PCRE used to give an error at compile time for
984 such patterns. However, because there are cases where this can be useful, such
985 patterns are now accepted, but if any repetition of the subpattern does in fact
986 match no characters, the loop is forcibly broken.
987 .P
988 By default, the quantifiers are "greedy", that is, they match as much as
989 possible (up to the maximum number of permitted times), without causing the
990 rest of the pattern to fail. The classic example of where this gives problems
991 is in trying to match comments in C programs. These appear between /* and */
992 and within the comment, individual * and / characters may appear. An attempt to
993 match C comments by applying the pattern
994 .sp
995 /\e*.*\e*/
996 .sp
997 to the string
998 .sp
999 /* first comment */ not comment /* second comment */
1000 .sp
1001 fails, because it matches the entire string owing to the greediness of the .*
1002 item.
1003 .P
1004 However, if a quantifier is followed by a question mark, it ceases to be
1005 greedy, and instead matches the minimum number of times possible, so the
1006 pattern
1007 .sp
1008 /\e*.*?\e*/
1009 .sp
1010 does the right thing with the C comments. The meaning of the various
1011 quantifiers is not otherwise changed, just the preferred number of matches.
1012 Do not confuse this use of question mark with its use as a quantifier in its
1013 own right. Because it has two uses, it can sometimes appear doubled, as in
1014 .sp
1015 \ed??\ed
1016 .sp
1017 which matches one digit by preference, but can match two if that is the only
1018 way the rest of the pattern matches.
1019 .P
1020 If the PCRE_UNGREEDY option is set (an option which is not available in Perl),
1021 the quantifiers are not greedy by default, but individual ones can be made
1022 greedy by following them with a question mark. In other words, it inverts the
1023 default behaviour.
1024 .P
1025 When a parenthesized subpattern is quantified with a minimum repeat count that
1026 is greater than 1 or with a limited maximum, more memory is required for the
1027 compiled pattern, in proportion to the size of the minimum or maximum.
1028 .P
1029 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1030 to Perl's /s) is set, thus allowing the . to match newlines, the pattern is
1031 implicitly anchored, because whatever follows will be tried against every
1032 character position in the subject string, so there is no point in retrying the
1033 overall match at any position after the first. PCRE normally treats such a
1034 pattern as though it were preceded by \eA.
1035 .P
1036 In cases where it is known that the subject string contains no newlines, it is
1037 worth setting PCRE_DOTALL in order to obtain this optimization, or
1038 alternatively using ^ to indicate anchoring explicitly.
1039 .P
1040 However, there is one situation where the optimization cannot be used. When .*
1041 is inside capturing parentheses that are the subject of a backreference
1042 elsewhere in the pattern, a match at the start may fail, and a later one
1043 succeed. Consider, for example:
1044 .sp
1045 (.*)abc\e1
1046 .sp
1047 If the subject is "xyz123abc123" the match point is the fourth character. For
1048 this reason, such a pattern is not implicitly anchored.
1049 .P
1050 When a capturing subpattern is repeated, the value captured is the substring
1051 that matched the final iteration. For example, after
1052 .sp
1053 (tweedle[dume]{3}\es*)+
1054 .sp
1055 has matched "tweedledum tweedledee" the value of the captured substring is
1056 "tweedledee". However, if there are nested capturing subpatterns, the
1057 corresponding captured values may have been set in previous iterations. For
1058 example, after
1059 .sp
1060 /(a|(b))+/
1061 .sp
1062 matches "aba" the value of the second captured substring is "b".
1063 .
1064 .
1065 .\" HTML <a name="atomicgroup"></a>
1067 .rs
1068 .sp
1069 With both maximizing and minimizing repetition, failure of what follows
1070 normally causes the repeated item to be re-evaluated to see if a different
1071 number of repeats allows the rest of the pattern to match. Sometimes it is
1072 useful to prevent this, either to change the nature of the match, or to cause
1073 it fail earlier than it otherwise might, when the author of the pattern knows
1074 there is no point in carrying on.
1075 .P
1076 Consider, for example, the pattern \ed+foo when applied to the subject line
1077 .sp
1078 123456bar
1079 .sp
1080 After matching all 6 digits and then failing to match "foo", the normal
1081 action of the matcher is to try again with only 5 digits matching the \ed+
1082 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1083 (a term taken from Jeffrey Friedl's book) provides the means for specifying
1084 that once a subpattern has matched, it is not to be re-evaluated in this way.
1085 .P
1086 If we use atomic grouping for the previous example, the matcher would give up
1087 immediately on failing to match "foo" the first time. The notation is a kind of
1088 special parenthesis, starting with (?> as in this example:
1089 .sp
1090 (?>\ed+)foo
1091 .sp
1092 This kind of parenthesis "locks up" the part of the pattern it contains once
1093 it has matched, and a failure further into the pattern is prevented from
1094 backtracking into it. Backtracking past it to previous items, however, works as
1095 normal.
1096 .P
1097 An alternative description is that a subpattern of this type matches the string
1098 of characters that an identical standalone pattern would match, if anchored at
1099 the current point in the subject string.
1100 .P
1101 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1102 the above example can be thought of as a maximizing repeat that must swallow
1103 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1104 number of digits they match in order to make the rest of the pattern match,
1105 (?>\ed+) can only match an entire sequence of digits.
1106 .P
1107 Atomic groups in general can of course contain arbitrarily complicated
1108 subpatterns, and can be nested. However, when the subpattern for an atomic
1109 group is just a single repeated item, as in the example above, a simpler
1110 notation, called a "possessive quantifier" can be used. This consists of an
1111 additional + character following a quantifier. Using this notation, the
1112 previous example can be rewritten as
1113 .sp
1114 \ed++foo
1115 .sp
1116 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1117 option is ignored. They are a convenient notation for the simpler forms of
1118 atomic group. However, there is no difference in the meaning or processing of a
1119 possessive quantifier and the equivalent atomic group.
1120 .P
1121 The possessive quantifier syntax is an extension to the Perl syntax. Jeffrey
1122 Friedl originated the idea (and the name) in the first edition of his book.
1123 Mike McCloskey liked it, so implemented it when he built Sun's Java package,
1124 and PCRE copied it from there.
1125 .P
1126 When a pattern contains an unlimited repeat inside a subpattern that can itself
1127 be repeated an unlimited number of times, the use of an atomic group is the
1128 only way to avoid some failing matches taking a very long time indeed. The
1129 pattern
1130 .sp
1131 (\eD+|<\ed+>)*[!?]
1132 .sp
1133 matches an unlimited number of substrings that either consist of non-digits, or
1134 digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1135 quickly. However, if it is applied to
1136 .sp
1137 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1138 .sp
1139 it takes a long time before reporting failure. This is because the string can
1140 be divided between the internal \eD+ repeat and the external * repeat in a
1141 large number of ways, and all have to be tried. (The example uses [!?] rather
1142 than a single character at the end, because both PCRE and Perl have an
1143 optimization that allows for fast failure when a single character is used. They
1144 remember the last single character that is required for a match, and fail early
1145 if it is not present in the string.) If the pattern is changed so that it uses
1146 an atomic group, like this:
1147 .sp
1148 ((?>\eD+)|<\ed+>)*[!?]
1149 .sp
1150 sequences of non-digits cannot be broken, and failure happens quickly.
1151 .
1152 .
1153 .\" HTML <a name="backreferences"></a>
1155 .rs
1156 .sp
1157 Outside a character class, a backslash followed by a digit greater than 0 (and
1158 possibly further digits) is a back reference to a capturing subpattern earlier
1159 (that is, to its left) in the pattern, provided there have been that many
1160 previous capturing left parentheses.
1161 .P
1162 However, if the decimal number following the backslash is less than 10, it is
1163 always taken as a back reference, and causes an error only if there are not
1164 that many capturing left parentheses in the entire pattern. In other words, the
1165 parentheses that are referenced need not be to the left of the reference for
1166 numbers less than 10. A "forward back reference" of this type can make sense
1167 when a repetition is involved and the subpattern to the right has participated
1168 in an earlier iteration.
1169 .P
1170 It is not possible to have a numerical "forward back reference" to subpattern
1171 whose number is 10 or more. However, a back reference to any subpattern is
1172 possible using named parentheses (see below). See also the subsection entitled
1173 "Non-printing characters"
1174 .\" HTML <a href="#digitsafterbackslash">
1175 .\" </a>
1176 above
1177 .\"
1178 for further details of the handling of digits following a backslash.
1179 .P
1180 A back reference matches whatever actually matched the capturing subpattern in
1181 the current subject string, rather than anything matching the subpattern
1182 itself (see
1183 .\" HTML <a href="#subpatternsassubroutines">
1184 .\" </a>
1185 "Subpatterns as subroutines"
1186 .\"
1187 below for a way of doing that). So the pattern
1188 .sp
1189 (sens|respons)e and \e1ibility
1190 .sp
1191 matches "sense and sensibility" and "response and responsibility", but not
1192 "sense and responsibility". If caseful matching is in force at the time of the
1193 back reference, the case of letters is relevant. For example,
1194 .sp
1195 ((?i)rah)\es+\e1
1196 .sp
1197 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1198 capturing subpattern is matched caselessly.
1199 .P
1200 Back references to named subpatterns use the Python syntax (?P=name). We could
1201 rewrite the above example as follows:
1202 .sp
1203 (?P<p1>(?i)rah)\es+(?P=p1)
1204 .sp
1205 A subpattern that is referenced by name may appear in the pattern before or
1206 after the reference.
1207 .P
1208 There may be more than one back reference to the same subpattern. If a
1209 subpattern has not actually been used in a particular match, any back
1210 references to it always fail. For example, the pattern
1211 .sp
1212 (a|(bc))\e2
1213 .sp
1214 always fails if it starts to match "a" rather than "bc". Because there may be
1215 many capturing parentheses in a pattern, all digits following the backslash are
1216 taken as part of a potential back reference number. If the pattern continues
1217 with a digit character, some delimiter must be used to terminate the back
1218 reference. If the PCRE_EXTENDED option is set, this can be whitespace.
1219 Otherwise an empty comment (see
1220 .\" HTML <a href="#comments">
1221 .\" </a>
1222 "Comments"
1223 .\"
1224 below) can be used.
1225 .P
1226 A back reference that occurs inside the parentheses to which it refers fails
1227 when the subpattern is first used, so, for example, (a\e1) never matches.
1228 However, such references can be useful inside repeated subpatterns. For
1229 example, the pattern
1230 .sp
1231 (a|b\e1)+
1232 .sp
1233 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1234 the subpattern, the back reference matches the character string corresponding
1235 to the previous iteration. In order for this to work, the pattern must be such
1236 that the first iteration does not need to match the back reference. This can be
1237 done using alternation, as in the example above, or by a quantifier with a
1238 minimum of zero.
1239 .
1240 .
1241 .\" HTML <a name="bigassertions"></a>
1243 .rs
1244 .sp
1245 An assertion is a test on the characters following or preceding the current
1246 matching point that does not actually consume any characters. The simple
1247 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
1248 .\" HTML <a href="#smallassertions">
1249 .\" </a>
1250 above.
1251 .\"
1252 .P
1253 More complicated assertions are coded as subpatterns. There are two kinds:
1254 those that look ahead of the current position in the subject string, and those
1255 that look behind it. An assertion subpattern is matched in the normal way,
1256 except that it does not cause the current matching position to be changed.
1257 .P
1258 Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1259 because it makes no sense to assert the same thing several times. If any kind
1260 of assertion contains capturing subpatterns within it, these are counted for
1261 the purposes of numbering the capturing subpatterns in the whole pattern.
1262 However, substring capturing is carried out only for positive assertions,
1263 because it does not make sense for negative assertions.
1264 .
1265 .
1266 .SS "Lookahead assertions"
1267 .rs
1268 .sp
1269 Lookahead assertions start with (?= for positive assertions and (?! for
1270 negative assertions. For example,
1271 .sp
1272 \ew+(?=;)
1273 .sp
1274 matches a word followed by a semicolon, but does not include the semicolon in
1275 the match, and
1276 .sp
1277 foo(?!bar)
1278 .sp
1279 matches any occurrence of "foo" that is not followed by "bar". Note that the
1280 apparently similar pattern
1281 .sp
1282 (?!foo)bar
1283 .sp
1284 does not find an occurrence of "bar" that is preceded by something other than
1285 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1286 (?!foo) is always true when the next three characters are "bar". A
1287 lookbehind assertion is needed to achieve the other effect.
1288 .P
1289 If you want to force a matching failure at some point in a pattern, the most
1290 convenient way to do it is with (?!) because an empty string always matches, so
1291 an assertion that requires there not to be an empty string must always fail.
1292 .
1293 .
1294 .\" HTML <a name="lookbehind"></a>
1295 .SS "Lookbehind assertions"
1296 .rs
1297 .sp
1298 Lookbehind assertions start with (?<= for positive assertions and (?<! for
1299 negative assertions. For example,
1300 .sp
1301 (?<!foo)bar
1302 .sp
1303 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1304 a lookbehind assertion are restricted such that all the strings it matches must
1305 have a fixed length. However, if there are several top-level alternatives, they
1306 do not all have to have the same fixed length. Thus
1307 .sp
1308 (?<=bullock|donkey)
1309 .sp
1310 is permitted, but
1311 .sp
1312 (?<!dogs?|cats?)
1313 .sp
1314 causes an error at compile time. Branches that match different length strings
1315 are permitted only at the top level of a lookbehind assertion. This is an
1316 extension compared with Perl (at least for 5.8), which requires all branches to
1317 match the same length of string. An assertion such as
1318 .sp
1319 (?<=ab(c|de))
1320 .sp
1321 is not permitted, because its single top-level branch can match two different
1322 lengths, but it is acceptable if rewritten to use two top-level branches:
1323 .sp
1324 (?<=abc|abde)
1325 .sp
1326 The implementation of lookbehind assertions is, for each alternative, to
1327 temporarily move the current position back by the fixed width and then try to
1328 match. If there are insufficient characters before the current position, the
1329 match is deemed to fail.
1330 .P
1331 PCRE does not allow the \eC escape (which matches a single byte in UTF-8 mode)
1332 to appear in lookbehind assertions, because it makes it impossible to calculate
1333 the length of the lookbehind. The \eX escape, which can match different numbers
1334 of bytes, is also not permitted.
1335 .P
1336 Atomic groups can be used in conjunction with lookbehind assertions to specify
1337 efficient matching at the end of the subject string. Consider a simple pattern
1338 such as
1339 .sp
1340 abcd$
1341 .sp
1342 when applied to a long string that does not match. Because matching proceeds
1343 from left to right, PCRE will look for each "a" in the subject and then see if
1344 what follows matches the rest of the pattern. If the pattern is specified as
1345 .sp
1346 ^.*abcd$
1347 .sp
1348 the initial .* matches the entire string at first, but when this fails (because
1349 there is no following "a"), it backtracks to match all but the last character,
1350 then all but the last two characters, and so on. Once again the search for "a"
1351 covers the entire string, from right to left, so we are no better off. However,
1352 if the pattern is written as
1353 .sp
1354 ^(?>.*)(?<=abcd)
1355 .sp
1356 or, equivalently, using the possessive quantifier syntax,
1357 .sp
1358 ^.*+(?<=abcd)
1359 .sp
1360 there can be no backtracking for the .* item; it can match only the entire
1361 string. The subsequent lookbehind assertion does a single test on the last four
1362 characters. If it fails, the match fails immediately. For long strings, this
1363 approach makes a significant difference to the processing time.
1364 .
1365 .
1366 .SS "Using multiple assertions"
1367 .rs
1368 .sp
1369 Several assertions (of any sort) may occur in succession. For example,
1370 .sp
1371 (?<=\ed{3})(?<!999)foo
1372 .sp
1373 matches "foo" preceded by three digits that are not "999". Notice that each of
1374 the assertions is applied independently at the same point in the subject
1375 string. First there is a check that the previous three characters are all
1376 digits, and then there is a check that the same three characters are not "999".
1377 This pattern does \fInot\fP match "foo" preceded by six characters, the first
1378 of which are digits and the last three of which are not "999". For example, it
1379 doesn't match "123abcfoo". A pattern to do that is
1380 .sp
1381 (?<=\ed{3}...)(?<!999)foo
1382 .sp
1383 This time the first assertion looks at the preceding six characters, checking
1384 that the first three are digits, and then the second assertion checks that the
1385 preceding three characters are not "999".
1386 .P
1387 Assertions can be nested in any combination. For example,
1388 .sp
1389 (?<=(?<!foo)bar)baz
1390 .sp
1391 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1392 preceded by "foo", while
1393 .sp
1394 (?<=\ed{3}(?!999)...)foo
1395 .sp
1396 is another pattern that matches "foo" preceded by three digits and any three
1397 characters that are not "999".
1398 .
1399 .
1400 .\" HTML <a name="conditions"></a>
1402 .rs
1403 .sp
1404 It is possible to cause the matching process to obey a subpattern
1405 conditionally or to choose between two alternative subpatterns, depending on
1406 the result of an assertion, or whether a previous capturing subpattern matched
1407 or not. The two possible forms of conditional subpattern are
1408 .sp
1409 (?(condition)yes-pattern)
1410 (?(condition)yes-pattern|no-pattern)
1411 .sp
1412 If the condition is satisfied, the yes-pattern is used; otherwise the
1413 no-pattern (if present) is used. If there are more than two alternatives in the
1414 subpattern, a compile-time error occurs.
1415 .P
1416 There are three kinds of condition. If the text between the parentheses
1417 consists of a sequence of digits, or a sequence of alphanumeric characters and
1418 underscores, the condition is satisfied if the capturing subpattern of that
1419 number or name has previously matched. There is a possible ambiguity here,
1420 because subpattern names may consist entirely of digits. PCRE looks first for a
1421 named subpattern; if it cannot find one and the text consists entirely of
1422 digits, it looks for a subpattern of that number, which must be greater than
1423 zero. Using subpattern names that consist entirely of digits is not
1424 recommended.
1425 .P
1426 Consider the following pattern, which contains non-significant white space to
1427 make it more readable (assume the PCRE_EXTENDED option) and to divide it into
1428 three parts for ease of discussion:
1429 .sp
1430 ( \e( )? [^()]+ (?(1) \e) )
1431 .sp
1432 The first part matches an optional opening parenthesis, and if that
1433 character is present, sets it as the first captured substring. The second part
1434 matches one or more characters that are not parentheses. The third part is a
1435 conditional subpattern that tests whether the first set of parentheses matched
1436 or not. If they did, that is, if subject started with an opening parenthesis,
1437 the condition is true, and so the yes-pattern is executed and a closing
1438 parenthesis is required. Otherwise, since no-pattern is not present, the
1439 subpattern matches nothing. In other words, this pattern matches a sequence of
1440 non-parentheses, optionally enclosed in parentheses. Rewriting it to use a
1441 named subpattern gives this:
1442 .sp
1443 (?P<OPEN> \e( )? [^()]+ (?(OPEN) \e) )
1444 .sp
1445 If the condition is the string (R), and there is no subpattern with the name R,
1446 the condition is satisfied if a recursive call to the pattern or subpattern has
1447 been made. At "top level", the condition is false. This is a PCRE extension.
1448 Recursive patterns are described in the next section.
1449 .P
1450 If the condition is not a sequence of digits or (R), it must be an assertion.
1451 This may be a positive or negative lookahead or lookbehind assertion. Consider
1452 this pattern, again containing non-significant white space, and with the two
1453 alternatives on the second line:
1454 .sp
1455 (?(?=[^a-z]*[a-z])
1456 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
1457 .sp
1458 The condition is a positive lookahead assertion that matches an optional
1459 sequence of non-letters followed by a letter. In other words, it tests for the
1460 presence of at least one letter in the subject. If a letter is found, the
1461 subject is matched against the first alternative; otherwise it is matched
1462 against the second. This pattern matches strings in one of the two forms
1463 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
1464 .
1465 .
1466 .\" HTML <a name="comments"></a>
1468 .rs
1469 .sp
1470 The sequence (?# marks the start of a comment that continues up to the next
1471 closing parenthesis. Nested parentheses are not permitted. The characters
1472 that make up a comment play no part in the pattern matching at all.
1473 .P
1474 If the PCRE_EXTENDED option is set, an unescaped # character outside a
1475 character class introduces a comment that continues to immediately after the
1476 next newline in the pattern.
1477 .
1478 .
1479 .\" HTML <a name="recursion"></a>
1481 .rs
1482 .sp
1483 Consider the problem of matching a string in parentheses, allowing for
1484 unlimited nested parentheses. Without the use of recursion, the best that can
1485 be done is to use a pattern that matches up to some fixed depth of nesting. It
1486 is not possible to handle an arbitrary nesting depth. Perl provides a facility
1487 that allows regular expressions to recurse (amongst other things). It does this
1488 by interpolating Perl code in the expression at run time, and the code can
1489 refer to the expression itself. A Perl pattern to solve the parentheses problem
1490 can be created like this:
1491 .sp
1492 $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
1493 .sp
1494 The (?p{...}) item interpolates Perl code at run time, and in this case refers
1495 recursively to the pattern in which it appears. Obviously, PCRE cannot support
1496 the interpolation of Perl code. Instead, it supports some special syntax for
1497 recursion of the entire pattern, and also for individual subpattern recursion.
1498 .P
1499 The special item that consists of (? followed by a number greater than zero and
1500 a closing parenthesis is a recursive call of the subpattern of the given
1501 number, provided that it occurs inside that subpattern. (If not, it is a
1502 "subroutine" call, which is described in the next section.) The special item
1503 (?R) is a recursive call of the entire regular expression.
1504 .P
1505 A recursive subpattern call is always treated as an atomic group. That is, once
1506 it has matched some of the subject string, it is never re-entered, even if
1507 it contains untried alternatives and there is a subsequent matching failure.
1508 .P
1509 This PCRE pattern solves the nested parentheses problem (assume the
1510 PCRE_EXTENDED option is set so that white space is ignored):
1511 .sp
1512 \e( ( (?>[^()]+) | (?R) )* \e)
1513 .sp
1514 First it matches an opening parenthesis. Then it matches any number of
1515 substrings which can either be a sequence of non-parentheses, or a recursive
1516 match of the pattern itself (that is, a correctly parenthesized substring).
1517 Finally there is a closing parenthesis.
1518 .P
1519 If this were part of a larger pattern, you would not want to recurse the entire
1520 pattern, so instead you could use this:
1521 .sp
1522 ( \e( ( (?>[^()]+) | (?1) )* \e) )
1523 .sp
1524 We have put the pattern into parentheses, and caused the recursion to refer to
1525 them instead of the whole pattern. In a larger pattern, keeping track of
1526 parenthesis numbers can be tricky. It may be more convenient to use named
1527 parentheses instead. For this, PCRE uses (?P>name), which is an extension to
1528 the Python syntax that PCRE uses for named parentheses (Perl does not provide
1529 named parentheses). We could rewrite the above example as follows:
1530 .sp
1531 (?P<pn> \e( ( (?>[^()]+) | (?P>pn) )* \e) )
1532 .sp
1533 This particular example pattern contains nested unlimited repeats, and so the
1534 use of atomic grouping for matching strings of non-parentheses is important
1535 when applying the pattern to strings that do not match. For example, when this
1536 pattern is applied to
1537 .sp
1538 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
1539 .sp
1540 it yields "no match" quickly. However, if atomic grouping is not used,
1541 the match runs for a very long time indeed because there are so many different
1542 ways the + and * repeats can carve up the subject, and all have to be tested
1543 before failure can be reported.
1544 .P
1545 At the end of a match, the values set for any capturing subpatterns are those
1546 from the outermost level of the recursion at which the subpattern value is set.
1547 If you want to obtain intermediate values, a callout function can be used (see
1548 the next section and the
1549 .\" HREF
1550 \fBpcrecallout\fP
1551 .\"
1552 documentation). If the pattern above is matched against
1553 .sp
1554 (ab(cd)ef)
1555 .sp
1556 the value for the capturing parentheses is "ef", which is the last value taken
1557 on at the top level. If additional parentheses are added, giving
1558 .sp
1559 \e( ( ( (?>[^()]+) | (?R) )* ) \e)
1560 ^ ^
1561 ^ ^
1562 .sp
1563 the string they capture is "ab(cd)ef", the contents of the top level
1564 parentheses. If there are more than 15 capturing parentheses in a pattern, PCRE
1565 has to obtain extra memory to store data during a recursion, which it does by
1566 using \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no
1567 memory can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
1568 .P
1569 Do not confuse the (?R) item with the condition (R), which tests for recursion.
1570 Consider this pattern, which matches text in angle brackets, allowing for
1571 arbitrary nesting. Only digits are allowed in nested brackets (that is, when
1572 recursing), whereas any characters are permitted at the outer level.
1573 .sp
1574 < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
1575 .sp
1576 In this pattern, (?(R) is the start of a conditional subpattern, with two
1577 different alternatives for the recursive and non-recursive cases. The (?R) item
1578 is the actual recursive call.
1579 .
1580 .
1581 .\" HTML <a name="subpatternsassubroutines"></a>
1583 .rs
1584 .sp
1585 If the syntax for a recursive subpattern reference (either by number or by
1586 name) is used outside the parentheses to which it refers, it operates like a
1587 subroutine in a programming language. An earlier example pointed out that the
1588 pattern
1589 .sp
1590 (sens|respons)e and \e1ibility
1591 .sp
1592 matches "sense and sensibility" and "response and responsibility", but not
1593 "sense and responsibility". If instead the pattern
1594 .sp
1595 (sens|respons)e and (?1)ibility
1596 .sp
1597 is used, it does match "sense and responsibility" as well as the other two
1598 strings. Such references, if given numerically, must follow the subpattern to
1599 which they refer. However, named references can refer to later subpatterns.
1600 .P
1601 Like recursive subpatterns, a "subroutine" call is always treated as an atomic
1602 group. That is, once it has matched some of the subject string, it is never
1603 re-entered, even if it contains untried alternatives and there is a subsequent
1604 matching failure.
1605 .
1606 .
1608 .rs
1609 .sp
1610 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
1611 code to be obeyed in the middle of matching a regular expression. This makes it
1612 possible, amongst other things, to extract different substrings that match the
1613 same pair of parentheses when there is a repetition.
1614 .P
1615 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
1616 code. The feature is called "callout". The caller of PCRE provides an external
1617 function by putting its entry point in the global variable \fIpcre_callout\fP.
1618 By default, this variable contains NULL, which disables all calling out.
1619 .P
1620 Within a regular expression, (?C) indicates the points at which the external
1621 function is to be called. If you want to identify different callout points, you
1622 can put a number less than 256 after the letter C. The default value is zero.
1623 For example, this pattern has two callout points:
1624 .sp
1625 (?C1)\dabc(?C2)def
1626 .sp
1627 If the PCRE_AUTO_CALLOUT flag is passed to \fBpcre_compile()\fP, callouts are
1628 automatically installed before each item in the pattern. They are all numbered
1629 255.
1630 .P
1631 During matching, when PCRE reaches a callout point (and \fIpcre_callout\fP is
1632 set), the external function is called. It is provided with the number of the
1633 callout, the position in the pattern, and, optionally, one item of data
1634 originally supplied by the caller of \fBpcre_exec()\fP. The callout function
1635 may cause matching to proceed, to backtrack, or to fail altogether. A complete
1636 description of the interface to the callout function is given in the
1637 .\" HREF
1638 \fBpcrecallout\fP
1639 .\"
1640 documentation.
1641 .P
1642 .in 0
1643 Last updated: 06 June 2006
1644 .br
1645 Copyright (c) 1997-2006 University of Cambridge.

  ViewVC Help
Powered by ViewVC 1.1.5