1 This file contains the PCRE man page that describes the regular expressions
2 supported by PCRE version 7.0. Note that not all of the features are relevant
3 in the context of Exim. In particular, the version of PCRE that is compiled
4 with Exim does not include UTF-8 support, there is no mechanism for changing
5 the options with which the PCRE functions are called, and features such as
6 callout are not accessible.
7 -----------------------------------------------------------------------------
9 PCREPATTERN(3) PCREPATTERN(3)
13 PCRE - Perl-compatible regular expressions
16 PCRE REGULAR EXPRESSION DETAILS
18 The syntax and semantics of the regular expressions supported by PCRE
19 are described below. Regular expressions are also described in the Perl
20 documentation and in a number of books, some of which have copious
21 examples. Jeffrey Friedl's "Mastering Regular Expressions", published
22 by O'Reilly, covers regular expressions in great detail. This descrip-
23 tion of PCRE's regular expressions is intended as reference material.
25 The original operation of PCRE was on strings of one-byte characters.
26 However, there is now also support for UTF-8 character strings. To use
27 this, you must build PCRE to include UTF-8 support, and then call
28 pcre_compile() with the PCRE_UTF8 option. How this affects pattern
29 matching is mentioned in several places below. There is also a summary
30 of UTF-8 features in the section on UTF-8 support in the main pcre
33 The remainder of this document discusses the patterns that are sup-
34 ported by PCRE when its main matching function, pcre_exec(), is used.
35 From release 6.0, PCRE offers a second matching function,
36 pcre_dfa_exec(), which matches using a different algorithm that is not
37 Perl-compatible. The advantages and disadvantages of the alternative
38 function, and how it differs from the normal function, are discussed in
39 the pcrematching page.
42 CHARACTERS AND METACHARACTERS
44 A regular expression is a pattern that is matched against a subject
45 string from left to right. Most characters stand for themselves in a
46 pattern, and match the corresponding characters in the subject. As a
47 trivial example, the pattern
51 matches a portion of a subject string that is identical to itself. When
52 caseless matching is specified (the PCRE_CASELESS option), letters are
53 matched independently of case. In UTF-8 mode, PCRE always understands
54 the concept of case for characters whose values are less than 128, so
55 caseless matching is always possible. For characters with higher val-
56 ues, the concept of case is supported if PCRE is compiled with Unicode
57 property support, but not otherwise. If you want to use caseless
58 matching for characters 128 and above, you must ensure that PCRE is
59 compiled with Unicode property support as well as with UTF-8 support.
61 The power of regular expressions comes from the ability to include
62 alternatives and repetitions in the pattern. These are encoded in the
63 pattern by the use of metacharacters, which do not stand for themselves
64 but instead are interpreted in some special way.
66 There are two different sets of metacharacters: those that are recog-
67 nized anywhere in the pattern except within square brackets, and those
68 that are recognized within square brackets. Outside square brackets,
69 the metacharacters are as follows:
71 \ general escape character with several uses
72 ^ assert start of string (or line, in multiline mode)
73 $ assert end of string (or line, in multiline mode)
74 . match any character except newline (by default)
75 [ start character class definition
76 | start of alternative branch
79 ? extends the meaning of (
80 also 0 or 1 quantifier
81 also quantifier minimizer
82 * 0 or more quantifier
83 + 1 or more quantifier
84 also "possessive quantifier"
85 { start min/max quantifier
87 Part of a pattern that is in square brackets is called a "character
88 class". In a character class the only metacharacters are:
90 \ general escape character
91 ^ negate the class, but only if the first character
92 - indicates character range
93 [ POSIX character class (only if followed by POSIX
95 ] terminates the character class
97 The following sections describe the use of each of the metacharacters.
102 The backslash character has several uses. Firstly, if it is followed by
103 a non-alphanumeric character, it takes away any special meaning that
104 character may have. This use of backslash as an escape character
105 applies both inside and outside character classes.
107 For example, if you want to match a * character, you write \* in the
108 pattern. This escaping action applies whether or not the following
109 character would otherwise be interpreted as a metacharacter, so it is
110 always safe to precede a non-alphanumeric with backslash to specify
111 that it stands for itself. In particular, if you want to match a back-
114 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in
115 the pattern (other than in a character class) and characters between a
116 # outside a character class and the next newline are ignored. An escap-
117 ing backslash can be used to include a whitespace or # character as
120 If you want to remove the special meaning from a sequence of charac-
121 ters, you can do so by putting them between \Q and \E. This is differ-
122 ent from Perl in that $ and @ are handled as literals in \Q...\E
123 sequences in PCRE, whereas in Perl, $ and @ cause variable interpola-
124 tion. Note the following examples:
126 Pattern PCRE matches Perl matches
128 \Qabc$xyz\E abc$xyz abc followed by the
130 \Qabc\$xyz\E abc\$xyz abc\$xyz
131 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
133 The \Q...\E sequence is recognized both inside and outside character
136 Non-printing characters
138 A second use of backslash provides a way of encoding non-printing char-
139 acters in patterns in a visible manner. There is no restriction on the
140 appearance of non-printing characters, apart from the binary zero that
141 terminates a pattern, but when a pattern is being prepared by text
142 editing, it is usually easier to use one of the following escape
143 sequences than the binary character it represents:
145 \a alarm, that is, the BEL character (hex 07)
146 \cx "control-x", where x is any character
150 \r carriage return (hex 0D)
152 \ddd character with octal code ddd, or backreference
153 \xhh character with hex code hh
154 \x{hhh..} character with hex code hhh..
156 The precise effect of \cx is as follows: if x is a lower case letter,
157 it is converted to upper case. Then bit 6 of the character (hex 40) is
158 inverted. Thus \cz becomes hex 1A, but \c{ becomes hex 3B, while \c;
161 After \x, from zero to two hexadecimal digits are read (letters can be
162 in upper or lower case). Any number of hexadecimal digits may appear
163 between \x{ and }, but the value of the character code must be less
164 than 256 in non-UTF-8 mode, and less than 2**31 in UTF-8 mode (that is,
165 the maximum hexadecimal value is 7FFFFFFF). If characters other than
166 hexadecimal digits appear between \x{ and }, or if there is no termi-
167 nating }, this form of escape is not recognized. Instead, the initial
168 \x will be interpreted as a basic hexadecimal escape, with no following
169 digits, giving a character whose value is zero.
171 Characters whose value is less than 256 can be defined by either of the
172 two syntaxes for \x. There is no difference in the way they are han-
173 dled. For example, \xdc is exactly the same as \x{dc}.
175 After \0 up to two further octal digits are read. If there are fewer
176 than two digits, just those that are present are used. Thus the
177 sequence \0\x\07 specifies two binary zeros followed by a BEL character
178 (code value 7). Make sure you supply two digits after the initial zero
179 if the pattern character that follows is itself an octal digit.
181 The handling of a backslash followed by a digit other than 0 is compli-
182 cated. Outside a character class, PCRE reads it and any following dig-
183 its as a decimal number. If the number is less than 10, or if there
184 have been at least that many previous capturing left parentheses in the
185 expression, the entire sequence is taken as a back reference. A
186 description of how this works is given later, following the discussion
187 of parenthesized subpatterns.
189 Inside a character class, or if the decimal number is greater than 9
190 and there have not been that many capturing subpatterns, PCRE re-reads
191 up to three octal digits following the backslash, and uses them to gen-
192 erate a data character. Any subsequent digits stand for themselves. In
193 non-UTF-8 mode, the value of a character specified in octal must be
194 less than \400. In UTF-8 mode, values up to \777 are permitted. For
197 \040 is another way of writing a space
198 \40 is the same, provided there are fewer than 40
199 previous capturing subpatterns
200 \7 is always a back reference
201 \11 might be a back reference, or another way of
204 \0113 is a tab followed by the character "3"
205 \113 might be a back reference, otherwise the
206 character with octal code 113
207 \377 might be a back reference, otherwise
208 the byte consisting entirely of 1 bits
209 \81 is either a back reference, or a binary zero
210 followed by the two characters "8" and "1"
212 Note that octal values of 100 or greater must not be introduced by a
213 leading zero, because no more than three octal digits are ever read.
215 All the sequences that define a single character value can be used both
216 inside and outside character classes. In addition, inside a character
217 class, the sequence \b is interpreted as the backspace character (hex
218 08), and the sequences \R and \X are interpreted as the characters "R"
219 and "X", respectively. Outside a character class, these sequences have
220 different meanings (see below).
222 Absolute and relative back references
224 The sequence \g followed by a positive or negative number, optionally
225 enclosed in braces, is an absolute or relative back reference. Back
226 references are discussed later, following the discussion of parenthe-
229 Generic character types
231 Another use of backslash is for specifying generic character types. The
232 following are always recognized:
235 \D any character that is not a decimal digit
236 \s any whitespace character
237 \S any character that is not a whitespace character
238 \w any "word" character
239 \W any "non-word" character
241 Each pair of escape sequences partitions the complete set of characters
242 into two disjoint sets. Any given character matches one, and only one,
245 These character type sequences can appear both inside and outside char-
246 acter classes. They each match one character of the appropriate type.
247 If the current matching point is at the end of the subject string, all
248 of them fail, since there is no character to match.
250 For compatibility with Perl, \s does not match the VT character (code
251 11). This makes it different from the the POSIX "space" class. The \s
252 characters are HT (9), LF (10), FF (12), CR (13), and space (32). (If
253 "use locale;" is included in a Perl script, \s may match the VT charac-
254 ter. In PCRE, it never does.)
256 A "word" character is an underscore or any character less than 256 that
257 is a letter or digit. The definition of letters and digits is con-
258 trolled by PCRE's low-valued character tables, and may vary if locale-
259 specific matching is taking place (see "Locale support" in the pcreapi
260 page). For example, in the "fr_FR" (French) locale, some character
261 codes greater than 128 are used for accented letters, and these are
264 In UTF-8 mode, characters with values greater than 128 never match \d,
265 \s, or \w, and always match \D, \S, and \W. This is true even when Uni-
266 code character property support is available. The use of locales with
267 Unicode is discouraged.
271 Outside a character class, the escape sequence \R matches any Unicode
272 newline sequence. This is an extension to Perl. In non-UTF-8 mode \R is
273 equivalent to the following:
275 (?>\r\n|\n|\x0b|\f|\r|\x85)
277 This is an example of an "atomic group", details of which are given
278 below. This particular group matches either the two-character sequence
279 CR followed by LF, or one of the single characters LF (linefeed,
280 U+000A), VT (vertical tab, U+000B), FF (formfeed, U+000C), CR (carriage
281 return, U+000D), or NEL (next line, U+0085). The two-character sequence
282 is treated as a single unit that cannot be split.
284 In UTF-8 mode, two additional characters whose codepoints are greater
285 than 255 are added: LS (line separator, U+2028) and PS (paragraph sepa-
286 rator, U+2029). Unicode character property support is not needed for
287 these characters to be recognized.
289 Inside a character class, \R matches the letter "R".
291 Unicode character properties
293 When PCRE is built with Unicode character property support, three addi-
294 tional escape sequences to match character properties are available
295 when UTF-8 mode is selected. They are:
297 \p{xx} a character with the xx property
298 \P{xx} a character without the xx property
299 \X an extended Unicode sequence
301 The property names represented by xx above are limited to the Unicode
302 script names, the general category properties, and "Any", which matches
303 any character (including newline). Other properties such as "InMusical-
304 Symbols" are not currently supported by PCRE. Note that \P{Any} does
305 not match any characters, so always causes a match failure.
307 Sets of Unicode characters are defined as belonging to certain scripts.
308 A character from one of these sets can be matched using a script name.
314 Those that are not part of an identified script are lumped together as
315 "Common". The current list of scripts is:
317 Arabic, Armenian, Balinese, Bengali, Bopomofo, Braille, Buginese,
318 Buhid, Canadian_Aboriginal, Cherokee, Common, Coptic, Cuneiform,
319 Cypriot, Cyrillic, Deseret, Devanagari, Ethiopic, Georgian, Glagolitic,
320 Gothic, Greek, Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hira-
321 gana, Inherited, Kannada, Katakana, Kharoshthi, Khmer, Lao, Latin,
322 Limbu, Linear_B, Malayalam, Mongolian, Myanmar, New_Tai_Lue, Nko,
323 Ogham, Old_Italic, Old_Persian, Oriya, Osmanya, Phags_Pa, Phoenician,
324 Runic, Shavian, Sinhala, Syloti_Nagri, Syriac, Tagalog, Tagbanwa,
325 Tai_Le, Tamil, Telugu, Thaana, Thai, Tibetan, Tifinagh, Ugaritic, Yi.
327 Each character has exactly one general category property, specified by
328 a two-letter abbreviation. For compatibility with Perl, negation can be
329 specified by including a circumflex between the opening brace and the
330 property name. For example, \p{^Lu} is the same as \P{Lu}.
332 If only one letter is specified with \p or \P, it includes all the gen-
333 eral category properties that start with that letter. In this case, in
334 the absence of negation, the curly brackets in the escape sequence are
335 optional; these two examples have the same effect:
340 The following general category property codes are supported:
367 Pc Connector punctuation
371 Pi Initial punctuation
378 Sm Mathematical symbol
383 Zp Paragraph separator
386 The special property L& is also supported: it matches a character that
387 has the Lu, Ll, or Lt property, in other words, a letter that is not
388 classified as a modifier or "other".
390 The long synonyms for these properties that Perl supports (such as
391 \p{Letter}) are not supported by PCRE, nor is it permitted to prefix
392 any of these properties with "Is".
394 No character that is in the Unicode table has the Cn (unassigned) prop-
395 erty. Instead, this property is assumed for any code point that is not
396 in the Unicode table.
398 Specifying caseless matching does not affect these escape sequences.
399 For example, \p{Lu} always matches only upper case letters.
401 The \X escape matches any number of Unicode characters that form an
402 extended Unicode sequence. \X is equivalent to
406 That is, it matches a character without the "mark" property, followed
407 by zero or more characters with the "mark" property, and treats the
408 sequence as an atomic group (see below). Characters with the "mark"
409 property are typically accents that affect the preceding character.
411 Matching characters by Unicode property is not fast, because PCRE has
412 to search a structure that contains data for over fifteen thousand
413 characters. That is why the traditional escape sequences such as \d and
414 \w do not use Unicode properties in PCRE.
418 The final use of backslash is for certain simple assertions. An asser-
419 tion specifies a condition that has to be met at a particular point in
420 a match, without consuming any characters from the subject string. The
421 use of subpatterns for more complicated assertions is described below.
422 The backslashed assertions are:
424 \b matches at a word boundary
425 \B matches when not at a word boundary
426 \A matches at the start of the subject
427 \Z matches at the end of the subject
428 also matches before a newline at the end of the subject
429 \z matches only at the end of the subject
430 \G matches at the first matching position in the subject
432 These assertions may not appear in character classes (but note that \b
433 has a different meaning, namely the backspace character, inside a char-
436 A word boundary is a position in the subject string where the current
437 character and the previous character do not both match \w or \W (i.e.
438 one matches \w and the other matches \W), or the start or end of the
439 string if the first or last character matches \w, respectively.
441 The \A, \Z, and \z assertions differ from the traditional circumflex
442 and dollar (described in the next section) in that they only ever match
443 at the very start and end of the subject string, whatever options are
444 set. Thus, they are independent of multiline mode. These three asser-
445 tions are not affected by the PCRE_NOTBOL or PCRE_NOTEOL options, which
446 affect only the behaviour of the circumflex and dollar metacharacters.
447 However, if the startoffset argument of pcre_exec() is non-zero, indi-
448 cating that matching is to start at a point other than the beginning of
449 the subject, \A can never match. The difference between \Z and \z is
450 that \Z matches before a newline at the end of the string as well as at
451 the very end, whereas \z matches only at the end.
453 The \G assertion is true only when the current matching position is at
454 the start point of the match, as specified by the startoffset argument
455 of pcre_exec(). It differs from \A when the value of startoffset is
456 non-zero. By calling pcre_exec() multiple times with appropriate argu-
457 ments, you can mimic Perl's /g option, and it is in this kind of imple-
458 mentation where \G can be useful.
460 Note, however, that PCRE's interpretation of \G, as the start of the
461 current match, is subtly different from Perl's, which defines it as the
462 end of the previous match. In Perl, these can be different when the
463 previously matched string was empty. Because PCRE does just one match
464 at a time, it cannot reproduce this behaviour.
466 If all the alternatives of a pattern begin with \G, the expression is
467 anchored to the starting match position, and the "anchored" flag is set
468 in the compiled regular expression.
471 CIRCUMFLEX AND DOLLAR
473 Outside a character class, in the default matching mode, the circumflex
474 character is an assertion that is true only if the current matching
475 point is at the start of the subject string. If the startoffset argu-
476 ment of pcre_exec() is non-zero, circumflex can never match if the
477 PCRE_MULTILINE option is unset. Inside a character class, circumflex
478 has an entirely different meaning (see below).
480 Circumflex need not be the first character of the pattern if a number
481 of alternatives are involved, but it should be the first thing in each
482 alternative in which it appears if the pattern is ever to match that
483 branch. If all possible alternatives start with a circumflex, that is,
484 if the pattern is constrained to match only at the start of the sub-
485 ject, it is said to be an "anchored" pattern. (There are also other
486 constructs that can cause a pattern to be anchored.)
488 A dollar character is an assertion that is true only if the current
489 matching point is at the end of the subject string, or immediately
490 before a newline at the end of the string (by default). Dollar need not
491 be the last character of the pattern if a number of alternatives are
492 involved, but it should be the last item in any branch in which it
493 appears. Dollar has no special meaning in a character class.
495 The meaning of dollar can be changed so that it matches only at the
496 very end of the string, by setting the PCRE_DOLLAR_ENDONLY option at
497 compile time. This does not affect the \Z assertion.
499 The meanings of the circumflex and dollar characters are changed if the
500 PCRE_MULTILINE option is set. When this is the case, a circumflex
501 matches immediately after internal newlines as well as at the start of
502 the subject string. It does not match after a newline that ends the
503 string. A dollar matches before any newlines in the string, as well as
504 at the very end, when PCRE_MULTILINE is set. When newline is specified
505 as the two-character sequence CRLF, isolated CR and LF characters do
506 not indicate newlines.
508 For example, the pattern /^abc$/ matches the subject string "def\nabc"
509 (where \n represents a newline) in multiline mode, but not otherwise.
510 Consequently, patterns that are anchored in single line mode because
511 all branches start with ^ are not anchored in multiline mode, and a
512 match for circumflex is possible when the startoffset argument of
513 pcre_exec() is non-zero. The PCRE_DOLLAR_ENDONLY option is ignored if
514 PCRE_MULTILINE is set.
516 Note that the sequences \A, \Z, and \z can be used to match the start
517 and end of the subject in both modes, and if all branches of a pattern
518 start with \A it is always anchored, whether or not PCRE_MULTILINE is
522 FULL STOP (PERIOD, DOT)
524 Outside a character class, a dot in the pattern matches any one charac-
525 ter in the subject string except (by default) a character that signi-
526 fies the end of a line. In UTF-8 mode, the matched character may be
527 more than one byte long.
529 When a line ending is defined as a single character, dot never matches
530 that character; when the two-character sequence CRLF is used, dot does
531 not match CR if it is immediately followed by LF, but otherwise it
532 matches all characters (including isolated CRs and LFs). When any Uni-
533 code line endings are being recognized, dot does not match CR or LF or
534 any of the other line ending characters.
536 The behaviour of dot with regard to newlines can be changed. If the
537 PCRE_DOTALL option is set, a dot matches any one character, without
538 exception. If the two-character sequence CRLF is present in the subject
539 string, it takes two dots to match it.
541 The handling of dot is entirely independent of the handling of circum-
542 flex and dollar, the only relationship being that they both involve
543 newlines. Dot has no special meaning in a character class.
546 MATCHING A SINGLE BYTE
548 Outside a character class, the escape sequence \C matches any one byte,
549 both in and out of UTF-8 mode. Unlike a dot, it always matches any
550 line-ending characters. The feature is provided in Perl in order to
551 match individual bytes in UTF-8 mode. Because it breaks up UTF-8 char-
552 acters into individual bytes, what remains in the string may be a mal-
553 formed UTF-8 string. For this reason, the \C escape sequence is best
556 PCRE does not allow \C to appear in lookbehind assertions (described
557 below), because in UTF-8 mode this would make it impossible to calcu-
558 late the length of the lookbehind.
561 SQUARE BRACKETS AND CHARACTER CLASSES
563 An opening square bracket introduces a character class, terminated by a
564 closing square bracket. A closing square bracket on its own is not spe-
565 cial. If a closing square bracket is required as a member of the class,
566 it should be the first data character in the class (after an initial
567 circumflex, if present) or escaped with a backslash.
569 A character class matches a single character in the subject. In UTF-8
570 mode, the character may occupy more than one byte. A matched character
571 must be in the set of characters defined by the class, unless the first
572 character in the class definition is a circumflex, in which case the
573 subject character must not be in the set defined by the class. If a
574 circumflex is actually required as a member of the class, ensure it is
575 not the first character, or escape it with a backslash.
577 For example, the character class [aeiou] matches any lower case vowel,
578 while [^aeiou] matches any character that is not a lower case vowel.
579 Note that a circumflex is just a convenient notation for specifying the
580 characters that are in the class by enumerating those that are not. A
581 class that starts with a circumflex is not an assertion: it still con-
582 sumes a character from the subject string, and therefore it fails if
583 the current pointer is at the end of the string.
585 In UTF-8 mode, characters with values greater than 255 can be included
586 in a class as a literal string of bytes, or by using the \x{ escaping
589 When caseless matching is set, any letters in a class represent both
590 their upper case and lower case versions, so for example, a caseless
591 [aeiou] matches "A" as well as "a", and a caseless [^aeiou] does not
592 match "A", whereas a caseful version would. In UTF-8 mode, PCRE always
593 understands the concept of case for characters whose values are less
594 than 128, so caseless matching is always possible. For characters with
595 higher values, the concept of case is supported if PCRE is compiled
596 with Unicode property support, but not otherwise. If you want to use
597 caseless matching for characters 128 and above, you must ensure that
598 PCRE is compiled with Unicode property support as well as with UTF-8
601 Characters that might indicate line breaks are never treated in any
602 special way when matching character classes, whatever line-ending
603 sequence is in use, and whatever setting of the PCRE_DOTALL and
604 PCRE_MULTILINE options is used. A class such as [^a] always matches one
607 The minus (hyphen) character can be used to specify a range of charac-
608 ters in a character class. For example, [d-m] matches any letter
609 between d and m, inclusive. If a minus character is required in a
610 class, it must be escaped with a backslash or appear in a position
611 where it cannot be interpreted as indicating a range, typically as the
612 first or last character in the class.
614 It is not possible to have the literal character "]" as the end charac-
615 ter of a range. A pattern such as [W-]46] is interpreted as a class of
616 two characters ("W" and "-") followed by a literal string "46]", so it
617 would match "W46]" or "-46]". However, if the "]" is escaped with a
618 backslash it is interpreted as the end of range, so [W-\]46] is inter-
619 preted as a class containing a range followed by two other characters.
620 The octal or hexadecimal representation of "]" can also be used to end
623 Ranges operate in the collating sequence of character values. They can
624 also be used for characters specified numerically, for example
625 [\000-\037]. In UTF-8 mode, ranges can include characters whose values
626 are greater than 255, for example [\x{100}-\x{2ff}].
628 If a range that includes letters is used when caseless matching is set,
629 it matches the letters in either case. For example, [W-c] is equivalent
630 to [][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if
631 character tables for the "fr_FR" locale are in use, [\xc8-\xcb] matches
632 accented E characters in both cases. In UTF-8 mode, PCRE supports the
633 concept of case for characters with values greater than 128 only when
634 it is compiled with Unicode property support.
636 The character types \d, \D, \p, \P, \s, \S, \w, and \W may also appear
637 in a character class, and add the characters that they match to the
638 class. For example, [\dABCDEF] matches any hexadecimal digit. A circum-
639 flex can conveniently be used with the upper case character types to
640 specify a more restricted set of characters than the matching lower
641 case type. For example, the class [^\W_] matches any letter or digit,
644 The only metacharacters that are recognized in character classes are
645 backslash, hyphen (only where it can be interpreted as specifying a
646 range), circumflex (only at the start), opening square bracket (only
647 when it can be interpreted as introducing a POSIX class name - see the
648 next section), and the terminating closing square bracket. However,
649 escaping other non-alphanumeric characters does no harm.
652 POSIX CHARACTER CLASSES
654 Perl supports the POSIX notation for character classes. This uses names
655 enclosed by [: and :] within the enclosing square brackets. PCRE also
656 supports this notation. For example,
660 matches "0", "1", any alphabetic character, or "%". The supported class
663 alnum letters and digits
665 ascii character codes 0 - 127
666 blank space or tab only
667 cntrl control characters
668 digit decimal digits (same as \d)
669 graph printing characters, excluding space
670 lower lower case letters
671 print printing characters, including space
672 punct printing characters, excluding letters and digits
673 space white space (not quite the same as \s)
674 upper upper case letters
675 word "word" characters (same as \w)
676 xdigit hexadecimal digits
678 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13),
679 and space (32). Notice that this list includes the VT character (code
680 11). This makes "space" different to \s, which does not include VT (for
683 The name "word" is a Perl extension, and "blank" is a GNU extension
684 from Perl 5.8. Another Perl extension is negation, which is indicated
685 by a ^ character after the colon. For example,
689 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the
690 POSIX syntax [.ch.] and [=ch=] where "ch" is a "collating element", but
691 these are not supported, and an error is given if they are encountered.
693 In UTF-8 mode, characters with values greater than 128 do not match any
694 of the POSIX character classes.
699 Vertical bar characters are used to separate alternative patterns. For
704 matches either "gilbert" or "sullivan". Any number of alternatives may
705 appear, and an empty alternative is permitted (matching the empty
706 string). The matching process tries each alternative in turn, from left
707 to right, and the first one that succeeds is used. If the alternatives
708 are within a subpattern (defined below), "succeeds" means matching the
709 rest of the main pattern as well as the alternative in the subpattern.
712 INTERNAL OPTION SETTING
714 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
715 PCRE_EXTENDED options can be changed from within the pattern by a
716 sequence of Perl option letters enclosed between "(?" and ")". The
724 For example, (?im) sets caseless, multiline matching. It is also possi-
725 ble to unset these options by preceding the letter with a hyphen, and a
726 combined setting and unsetting such as (?im-sx), which sets PCRE_CASE-
727 LESS and PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED,
728 is also permitted. If a letter appears both before and after the
729 hyphen, the option is unset.
731 When an option change occurs at top level (that is, not inside subpat-
732 tern parentheses), the change applies to the remainder of the pattern
733 that follows. If the change is placed right at the start of a pattern,
734 PCRE extracts it into the global options (and it will therefore show up
735 in data extracted by the pcre_fullinfo() function).
737 An option change within a subpattern (see below for a description of
738 subpatterns) affects only that part of the current pattern that follows
743 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not
744 used). By this means, options can be made to have different settings
745 in different parts of the pattern. Any changes made in one alternative
746 do carry on into subsequent branches within the same subpattern. For
751 matches "ab", "aB", "c", and "C", even though when matching "C" the
752 first branch is abandoned before the option setting. This is because
753 the effects of option settings happen at compile time. There would be
754 some very weird behaviour otherwise.
756 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA
757 can be changed in the same way as the Perl-compatible options by using
758 the characters J, U and X respectively.
763 Subpatterns are delimited by parentheses (round brackets), which can be
764 nested. Turning part of a pattern into a subpattern does two things:
766 1. It localizes a set of alternatives. For example, the pattern
770 matches one of the words "cat", "cataract", or "caterpillar". Without
771 the parentheses, it would match "cataract", "erpillar" or an empty
774 2. It sets up the subpattern as a capturing subpattern. This means
775 that, when the whole pattern matches, that portion of the subject
776 string that matched the subpattern is passed back to the caller via the
777 ovector argument of pcre_exec(). Opening parentheses are counted from
778 left to right (starting from 1) to obtain numbers for the capturing
781 For example, if the string "the red king" is matched against the pat-
784 the ((red|white) (king|queen))
786 the captured substrings are "red king", "red", and "king", and are num-
787 bered 1, 2, and 3, respectively.
789 The fact that plain parentheses fulfil two functions is not always
790 helpful. There are often times when a grouping subpattern is required
791 without a capturing requirement. If an opening parenthesis is followed
792 by a question mark and a colon, the subpattern does not do any captur-
793 ing, and is not counted when computing the number of any subsequent
794 capturing subpatterns. For example, if the string "the white queen" is
795 matched against the pattern
797 the ((?:red|white) (king|queen))
799 the captured substrings are "white queen" and "queen", and are numbered
800 1 and 2. The maximum number of capturing subpatterns is 65535.
802 As a convenient shorthand, if any option settings are required at the
803 start of a non-capturing subpattern, the option letters may appear
804 between the "?" and the ":". Thus the two patterns
807 (?:(?i)saturday|sunday)
809 match exactly the same set of strings. Because alternative branches are
810 tried from left to right, and options are not reset until the end of
811 the subpattern is reached, an option setting in one branch does affect
812 subsequent branches, so the above patterns match "SUNDAY" as well as
818 Identifying capturing parentheses by number is simple, but it can be
819 very hard to keep track of the numbers in complicated regular expres-
820 sions. Furthermore, if an expression is modified, the numbers may
821 change. To help with this difficulty, PCRE supports the naming of sub-
822 patterns. This feature was not added to Perl until release 5.10. Python
823 had the feature earlier, and PCRE introduced it at release 4.0, using
824 the Python syntax. PCRE now supports both the Perl and the Python syn-
827 In PCRE, a subpattern can be named in one of three ways: (?<name>...)
828 or (?'name'...) as in Perl, or (?P<name>...) as in Python. References
829 to capturing parentheses from other parts of the pattern, such as back-
830 references, recursion, and conditions, can be made by name as well as
833 Names consist of up to 32 alphanumeric characters and underscores.
834 Named capturing parentheses are still allocated numbers as well as
835 names, exactly as if the names were not present. The PCRE API provides
836 function calls for extracting the name-to-number translation table from
837 a compiled pattern. There is also a convenience function for extracting
838 a captured substring by name.
840 By default, a name must be unique within a pattern, but it is possible
841 to relax this constraint by setting the PCRE_DUPNAMES option at compile
842 time. This can be useful for patterns where only one instance of the
843 named parentheses can match. Suppose you want to match the name of a
844 weekday, either as a 3-letter abbreviation or as the full name, and in
845 both cases you want to extract the abbreviation. This pattern (ignoring
846 the line breaks) does the job:
848 (?<DN>Mon|Fri|Sun)(?:day)?|
850 (?<DN>Wed)(?:nesday)?|
851 (?<DN>Thu)(?:rsday)?|
854 There are five capturing substrings, but only one is ever set after a
855 match. The convenience function for extracting the data by name
856 returns the substring for the first (and in this example, the only)
857 subpattern of that name that matched. This saves searching to find
858 which numbered subpattern it was. If you make a reference to a non-
859 unique named subpattern from elsewhere in the pattern, the one that
860 corresponds to the lowest number is used. For further details of the
861 interfaces for handling named subpatterns, see the pcreapi documenta-
867 Repetition is specified by quantifiers, which can follow any of the
870 a literal data character
871 the dot metacharacter
872 the \C escape sequence
873 the \X escape sequence (in UTF-8 mode with Unicode properties)
874 the \R escape sequence
875 an escape such as \d that matches a single character
877 a back reference (see next section)
878 a parenthesized subpattern (unless it is an assertion)
880 The general repetition quantifier specifies a minimum and maximum num-
881 ber of permitted matches, by giving the two numbers in curly brackets
882 (braces), separated by a comma. The numbers must be less than 65536,
883 and the first must be less than or equal to the second. For example:
887 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a
888 special character. If the second number is omitted, but the comma is
889 present, there is no upper limit; if the second number and the comma
890 are both omitted, the quantifier specifies an exact number of required
895 matches at least 3 successive vowels, but may match many more, while
899 matches exactly 8 digits. An opening curly bracket that appears in a
900 position where a quantifier is not allowed, or one that does not match
901 the syntax of a quantifier, is taken as a literal character. For exam-
902 ple, {,6} is not a quantifier, but a literal string of four characters.
904 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to
905 individual bytes. Thus, for example, \x{100}{2} matches two UTF-8 char-
906 acters, each of which is represented by a two-byte sequence. Similarly,
907 when Unicode property support is available, \X{3} matches three Unicode
908 extended sequences, each of which may be several bytes long (and they
909 may be of different lengths).
911 The quantifier {0} is permitted, causing the expression to behave as if
912 the previous item and the quantifier were not present.
914 For convenience, the three most common quantifiers have single-charac-
917 * is equivalent to {0,}
918 + is equivalent to {1,}
919 ? is equivalent to {0,1}
921 It is possible to construct infinite loops by following a subpattern
922 that can match no characters with a quantifier that has no upper limit,
927 Earlier versions of Perl and PCRE used to give an error at compile time
928 for such patterns. However, because there are cases where this can be
929 useful, such patterns are now accepted, but if any repetition of the
930 subpattern does in fact match no characters, the loop is forcibly bro-
933 By default, the quantifiers are "greedy", that is, they match as much
934 as possible (up to the maximum number of permitted times), without
935 causing the rest of the pattern to fail. The classic example of where
936 this gives problems is in trying to match comments in C programs. These
937 appear between /* and */ and within the comment, individual * and /
938 characters may appear. An attempt to match C comments by applying the
945 /* first comment */ not comment /* second comment */
947 fails, because it matches the entire string owing to the greediness of
950 However, if a quantifier is followed by a question mark, it ceases to
951 be greedy, and instead matches the minimum number of times possible, so
956 does the right thing with the C comments. The meaning of the various
957 quantifiers is not otherwise changed, just the preferred number of
958 matches. Do not confuse this use of question mark with its use as a
959 quantifier in its own right. Because it has two uses, it can sometimes
960 appear doubled, as in
964 which matches one digit by preference, but can match two if that is the
965 only way the rest of the pattern matches.
967 If the PCRE_UNGREEDY option is set (an option that is not available in
968 Perl), the quantifiers are not greedy by default, but individual ones
969 can be made greedy by following them with a question mark. In other
970 words, it inverts the default behaviour.
972 When a parenthesized subpattern is quantified with a minimum repeat
973 count that is greater than 1 or with a limited maximum, more memory is
974 required for the compiled pattern, in proportion to the size of the
977 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equiv-
978 alent to Perl's /s) is set, thus allowing the dot to match newlines,
979 the pattern is implicitly anchored, because whatever follows will be
980 tried against every character position in the subject string, so there
981 is no point in retrying the overall match at any position after the
982 first. PCRE normally treats such a pattern as though it were preceded
985 In cases where it is known that the subject string contains no new-
986 lines, it is worth setting PCRE_DOTALL in order to obtain this opti-
987 mization, or alternatively using ^ to indicate anchoring explicitly.
989 However, there is one situation where the optimization cannot be used.
990 When .* is inside capturing parentheses that are the subject of a
991 backreference elsewhere in the pattern, a match at the start may fail
992 where a later one succeeds. Consider, for example:
996 If the subject is "xyz123abc123" the match point is the fourth charac-
997 ter. For this reason, such a pattern is not implicitly anchored.
999 When a capturing subpattern is repeated, the value captured is the sub-
1000 string that matched the final iteration. For example, after
1002 (tweedle[dume]{3}\s*)+
1004 has matched "tweedledum tweedledee" the value of the captured substring
1005 is "tweedledee". However, if there are nested capturing subpatterns,
1006 the corresponding captured values may have been set in previous itera-
1007 tions. For example, after
1011 matches "aba" the value of the second captured substring is "b".
1014 ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS
1016 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1017 repetition, failure of what follows normally causes the repeated item
1018 to be re-evaluated to see if a different number of repeats allows the
1019 rest of the pattern to match. Sometimes it is useful to prevent this,
1020 either to change the nature of the match, or to cause it fail earlier
1021 than it otherwise might, when the author of the pattern knows there is
1022 no point in carrying on.
1024 Consider, for example, the pattern \d+foo when applied to the subject
1029 After matching all 6 digits and then failing to match "foo", the normal
1030 action of the matcher is to try again with only 5 digits matching the
1031 \d+ item, and then with 4, and so on, before ultimately failing.
1032 "Atomic grouping" (a term taken from Jeffrey Friedl's book) provides
1033 the means for specifying that once a subpattern has matched, it is not
1034 to be re-evaluated in this way.
1036 If we use atomic grouping for the previous example, the matcher gives
1037 up immediately on failing to match "foo" the first time. The notation
1038 is a kind of special parenthesis, starting with (?> as in this example:
1042 This kind of parenthesis "locks up" the part of the pattern it con-
1043 tains once it has matched, and a failure further into the pattern is
1044 prevented from backtracking into it. Backtracking past it to previous
1045 items, however, works as normal.
1047 An alternative description is that a subpattern of this type matches
1048 the string of characters that an identical standalone pattern would
1049 match, if anchored at the current point in the subject string.
1051 Atomic grouping subpatterns are not capturing subpatterns. Simple cases
1052 such as the above example can be thought of as a maximizing repeat that
1053 must swallow everything it can. So, while both \d+ and \d+? are pre-
1054 pared to adjust the number of digits they match in order to make the
1055 rest of the pattern match, (?>\d+) can only match an entire sequence of
1058 Atomic groups in general can of course contain arbitrarily complicated
1059 subpatterns, and can be nested. However, when the subpattern for an
1060 atomic group is just a single repeated item, as in the example above, a
1061 simpler notation, called a "possessive quantifier" can be used. This
1062 consists of an additional + character following a quantifier. Using
1063 this notation, the previous example can be rewritten as
1067 Possessive quantifiers are always greedy; the setting of the
1068 PCRE_UNGREEDY option is ignored. They are a convenient notation for the
1069 simpler forms of atomic group. However, there is no difference in the
1070 meaning of a possessive quantifier and the equivalent atomic group,
1071 though there may be a performance difference; possessive quantifiers
1072 should be slightly faster.
1074 The possessive quantifier syntax is an extension to the Perl 5.8 syn-
1075 tax. Jeffrey Friedl originated the idea (and the name) in the first
1076 edition of his book. Mike McCloskey liked it, so implemented it when he
1077 built Sun's Java package, and PCRE copied it from there. It ultimately
1078 found its way into Perl at release 5.10.
1080 PCRE has an optimization that automatically "possessifies" certain sim-
1081 ple pattern constructs. For example, the sequence A+B is treated as
1082 A++B because there is no point in backtracking into a sequence of A's
1085 When a pattern contains an unlimited repeat inside a subpattern that
1086 can itself be repeated an unlimited number of times, the use of an
1087 atomic group is the only way to avoid some failing matches taking a
1088 very long time indeed. The pattern
1092 matches an unlimited number of substrings that either consist of non-
1093 digits, or digits enclosed in <>, followed by either ! or ?. When it
1094 matches, it runs quickly. However, if it is applied to
1096 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1098 it takes a long time before reporting failure. This is because the
1099 string can be divided between the internal \D+ repeat and the external
1100 * repeat in a large number of ways, and all have to be tried. (The
1101 example uses [!?] rather than a single character at the end, because
1102 both PCRE and Perl have an optimization that allows for fast failure
1103 when a single character is used. They remember the last single charac-
1104 ter that is required for a match, and fail early if it is not present
1105 in the string.) If the pattern is changed so that it uses an atomic
1108 ((?>\D+)|<\d+>)*[!?]
1110 sequences of non-digits cannot be broken, and failure happens quickly.
1115 Outside a character class, a backslash followed by a digit greater than
1116 0 (and possibly further digits) is a back reference to a capturing sub-
1117 pattern earlier (that is, to its left) in the pattern, provided there
1118 have been that many previous capturing left parentheses.
1120 However, if the decimal number following the backslash is less than 10,
1121 it is always taken as a back reference, and causes an error only if
1122 there are not that many capturing left parentheses in the entire pat-
1123 tern. In other words, the parentheses that are referenced need not be
1124 to the left of the reference for numbers less than 10. A "forward back
1125 reference" of this type can make sense when a repetition is involved
1126 and the subpattern to the right has participated in an earlier itera-
1129 It is not possible to have a numerical "forward back reference" to a
1130 subpattern whose number is 10 or more using this syntax because a
1131 sequence such as \50 is interpreted as a character defined in octal.
1132 See the subsection entitled "Non-printing characters" above for further
1133 details of the handling of digits following a backslash. There is no
1134 such problem when named parentheses are used. A back reference to any
1135 subpattern is possible using named parentheses (see below).
1137 Another way of avoiding the ambiguity inherent in the use of digits
1138 following a backslash is to use the \g escape sequence, which is a fea-
1139 ture introduced in Perl 5.10. This escape must be followed by a posi-
1140 tive or a negative number, optionally enclosed in braces. These exam-
1141 ples are all identical:
1147 A positive number specifies an absolute reference without the ambiguity
1148 that is present in the older syntax. It is also useful when literal
1149 digits follow the reference. A negative number is a relative reference.
1150 Consider this example:
1154 The sequence \g{-1} is a reference to the most recently started captur-
1155 ing subpattern before \g, that is, is it equivalent to \2. Similarly,
1156 \g{-2} would be equivalent to \1. The use of relative references can be
1157 helpful in long patterns, and also in patterns that are created by
1158 joining together fragments that contain references within themselves.
1160 A back reference matches whatever actually matched the capturing sub-
1161 pattern in the current subject string, rather than anything matching
1162 the subpattern itself (see "Subpatterns as subroutines" below for a way
1163 of doing that). So the pattern
1165 (sens|respons)e and \1ibility
1167 matches "sense and sensibility" and "response and responsibility", but
1168 not "sense and responsibility". If caseful matching is in force at the
1169 time of the back reference, the case of letters is relevant. For exam-
1174 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the
1175 original capturing subpattern is matched caselessly.
1177 Back references to named subpatterns use the Perl syntax \k<name> or
1178 \k'name' or the Python syntax (?P=name). We could rewrite the above
1179 example in either of the following ways:
1181 (?<p1>(?i)rah)\s+\k<p1>
1182 (?P<p1>(?i)rah)\s+(?P=p1)
1184 A subpattern that is referenced by name may appear in the pattern
1185 before or after the reference.
1187 There may be more than one back reference to the same subpattern. If a
1188 subpattern has not actually been used in a particular match, any back
1189 references to it always fail. For example, the pattern
1193 always fails if it starts to match "a" rather than "bc". Because there
1194 may be many capturing parentheses in a pattern, all digits following
1195 the backslash are taken as part of a potential back reference number.
1196 If the pattern continues with a digit character, some delimiter must be
1197 used to terminate the back reference. If the PCRE_EXTENDED option is
1198 set, this can be whitespace. Otherwise an empty comment (see "Com-
1199 ments" below) can be used.
1201 A back reference that occurs inside the parentheses to which it refers
1202 fails when the subpattern is first used, so, for example, (a\1) never
1203 matches. However, such references can be useful inside repeated sub-
1204 patterns. For example, the pattern
1208 matches any number of "a"s and also "aba", "ababbaa" etc. At each iter-
1209 ation of the subpattern, the back reference matches the character
1210 string corresponding to the previous iteration. In order for this to
1211 work, the pattern must be such that the first iteration does not need
1212 to match the back reference. This can be done using alternation, as in
1213 the example above, or by a quantifier with a minimum of zero.
1218 An assertion is a test on the characters following or preceding the
1219 current matching point that does not actually consume any characters.
1220 The simple assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are
1223 More complicated assertions are coded as subpatterns. There are two
1224 kinds: those that look ahead of the current position in the subject
1225 string, and those that look behind it. An assertion subpattern is
1226 matched in the normal way, except that it does not cause the current
1227 matching position to be changed.
1229 Assertion subpatterns are not capturing subpatterns, and may not be
1230 repeated, because it makes no sense to assert the same thing several
1231 times. If any kind of assertion contains capturing subpatterns within
1232 it, these are counted for the purposes of numbering the capturing sub-
1233 patterns in the whole pattern. However, substring capturing is carried
1234 out only for positive assertions, because it does not make sense for
1235 negative assertions.
1237 Lookahead assertions
1239 Lookahead assertions start with (?= for positive assertions and (?! for
1240 negative assertions. For example,
1244 matches a word followed by a semicolon, but does not include the semi-
1245 colon in the match, and
1249 matches any occurrence of "foo" that is not followed by "bar". Note
1250 that the apparently similar pattern
1254 does not find an occurrence of "bar" that is preceded by something
1255 other than "foo"; it finds any occurrence of "bar" whatsoever, because
1256 the assertion (?!foo) is always true when the next three characters are
1257 "bar". A lookbehind assertion is needed to achieve the other effect.
1259 If you want to force a matching failure at some point in a pattern, the
1260 most convenient way to do it is with (?!) because an empty string
1261 always matches, so an assertion that requires there not to be an empty
1262 string must always fail.
1264 Lookbehind assertions
1266 Lookbehind assertions start with (?<= for positive assertions and (?<!
1267 for negative assertions. For example,
1271 does find an occurrence of "bar" that is not preceded by "foo". The
1272 contents of a lookbehind assertion are restricted such that all the
1273 strings it matches must have a fixed length. However, if there are sev-
1274 eral top-level alternatives, they do not all have to have the same
1283 causes an error at compile time. Branches that match different length
1284 strings are permitted only at the top level of a lookbehind assertion.
1285 This is an extension compared with Perl (at least for 5.8), which
1286 requires all branches to match the same length of string. An assertion
1291 is not permitted, because its single top-level branch can match two
1292 different lengths, but it is acceptable if rewritten to use two top-
1297 The implementation of lookbehind assertions is, for each alternative,
1298 to temporarily move the current position back by the fixed length and
1299 then try to match. If there are insufficient characters before the cur-
1300 rent position, the assertion fails.
1302 PCRE does not allow the \C escape (which matches a single byte in UTF-8
1303 mode) to appear in lookbehind assertions, because it makes it impossi-
1304 ble to calculate the length of the lookbehind. The \X and \R escapes,
1305 which can match different numbers of bytes, are also not permitted.
1307 Possessive quantifiers can be used in conjunction with lookbehind
1308 assertions to specify efficient matching at the end of the subject
1309 string. Consider a simple pattern such as
1313 when applied to a long string that does not match. Because matching
1314 proceeds from left to right, PCRE will look for each "a" in the subject
1315 and then see if what follows matches the rest of the pattern. If the
1316 pattern is specified as
1320 the initial .* matches the entire string at first, but when this fails
1321 (because there is no following "a"), it backtracks to match all but the
1322 last character, then all but the last two characters, and so on. Once
1323 again the search for "a" covers the entire string, from right to left,
1324 so we are no better off. However, if the pattern is written as
1328 there can be no backtracking for the .*+ item; it can match only the
1329 entire string. The subsequent lookbehind assertion does a single test
1330 on the last four characters. If it fails, the match fails immediately.
1331 For long strings, this approach makes a significant difference to the
1334 Using multiple assertions
1336 Several assertions (of any sort) may occur in succession. For example,
1338 (?<=\d{3})(?<!999)foo
1340 matches "foo" preceded by three digits that are not "999". Notice that
1341 each of the assertions is applied independently at the same point in
1342 the subject string. First there is a check that the previous three
1343 characters are all digits, and then there is a check that the same
1344 three characters are not "999". This pattern does not match "foo" pre-
1345 ceded by six characters, the first of which are digits and the last
1346 three of which are not "999". For example, it doesn't match "123abc-
1347 foo". A pattern to do that is
1349 (?<=\d{3}...)(?<!999)foo
1351 This time the first assertion looks at the preceding six characters,
1352 checking that the first three are digits, and then the second assertion
1353 checks that the preceding three characters are not "999".
1355 Assertions can be nested in any combination. For example,
1359 matches an occurrence of "baz" that is preceded by "bar" which in turn
1360 is not preceded by "foo", while
1362 (?<=\d{3}(?!999)...)foo
1364 is another pattern that matches "foo" preceded by three digits and any
1365 three characters that are not "999".
1368 CONDITIONAL SUBPATTERNS
1370 It is possible to cause the matching process to obey a subpattern con-
1371 ditionally or to choose between two alternative subpatterns, depending
1372 on the result of an assertion, or whether a previous capturing subpat-
1373 tern matched or not. The two possible forms of conditional subpattern
1376 (?(condition)yes-pattern)
1377 (?(condition)yes-pattern|no-pattern)
1379 If the condition is satisfied, the yes-pattern is used; otherwise the
1380 no-pattern (if present) is used. If there are more than two alterna-
1381 tives in the subpattern, a compile-time error occurs.
1383 There are four kinds of condition: references to subpatterns, refer-
1384 ences to recursion, a pseudo-condition called DEFINE, and assertions.
1386 Checking for a used subpattern by number
1388 If the text between the parentheses consists of a sequence of digits,
1389 the condition is true if the capturing subpattern of that number has
1392 Consider the following pattern, which contains non-significant white
1393 space to make it more readable (assume the PCRE_EXTENDED option) and to
1394 divide it into three parts for ease of discussion:
1396 ( \( )? [^()]+ (?(1) \) )
1398 The first part matches an optional opening parenthesis, and if that
1399 character is present, sets it as the first captured substring. The sec-
1400 ond part matches one or more characters that are not parentheses. The
1401 third part is a conditional subpattern that tests whether the first set
1402 of parentheses matched or not. If they did, that is, if subject started
1403 with an opening parenthesis, the condition is true, and so the yes-pat-
1404 tern is executed and a closing parenthesis is required. Otherwise,
1405 since no-pattern is not present, the subpattern matches nothing. In
1406 other words, this pattern matches a sequence of non-parentheses,
1407 optionally enclosed in parentheses.
1409 Checking for a used subpattern by name
1411 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a
1412 used subpattern by name. For compatibility with earlier versions of
1413 PCRE, which had this facility before Perl, the syntax (?(name)...) is
1414 also recognized. However, there is a possible ambiguity with this syn-
1415 tax, because subpattern names may consist entirely of digits. PCRE
1416 looks first for a named subpattern; if it cannot find one and the name
1417 consists entirely of digits, PCRE looks for a subpattern of that num-
1418 ber, which must be greater than zero. Using subpattern names that con-
1419 sist entirely of digits is not recommended.
1421 Rewriting the above example to use a named subpattern gives this:
1423 (?<OPEN> \( )? [^()]+ (?(<OPEN>) \) )
1426 Checking for pattern recursion
1428 If the condition is the string (R), and there is no subpattern with the
1429 name R, the condition is true if a recursive call to the whole pattern
1430 or any subpattern has been made. If digits or a name preceded by amper-
1431 sand follow the letter R, for example:
1433 (?(R3)...) or (?(R&name)...)
1435 the condition is true if the most recent recursion is into the subpat-
1436 tern whose number or name is given. This condition does not check the
1437 entire recursion stack.
1439 At "top level", all these recursion test conditions are false. Recur-
1440 sive patterns are described below.
1442 Defining subpatterns for use by reference only
1444 If the condition is the string (DEFINE), and there is no subpattern
1445 with the name DEFINE, the condition is always false. In this case,
1446 there may be only one alternative in the subpattern. It is always
1447 skipped if control reaches this point in the pattern; the idea of
1448 DEFINE is that it can be used to define "subroutines" that can be ref-
1449 erenced from elsewhere. (The use of "subroutines" is described below.)
1450 For example, a pattern to match an IPv4 address could be written like
1451 this (ignore whitespace and line breaks):
1453 (?(DEFINE) (?<byte> 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) )
1454 \b (?&byte) (\.(?&byte)){3} \b
1456 The first part of the pattern is a DEFINE group inside which a another
1457 group named "byte" is defined. This matches an individual component of
1458 an IPv4 address (a number less than 256). When matching takes place,
1459 this part of the pattern is skipped because DEFINE acts like a false
1462 The rest of the pattern uses references to the named group to match the
1463 four dot-separated components of an IPv4 address, insisting on a word
1464 boundary at each end.
1466 Assertion conditions
1468 If the condition is not in any of the above formats, it must be an
1469 assertion. This may be a positive or negative lookahead or lookbehind
1470 assertion. Consider this pattern, again containing non-significant
1471 white space, and with the two alternatives on the second line:
1474 \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )
1476 The condition is a positive lookahead assertion that matches an
1477 optional sequence of non-letters followed by a letter. In other words,
1478 it tests for the presence of at least one letter in the subject. If a
1479 letter is found, the subject is matched against the first alternative;
1480 otherwise it is matched against the second. This pattern matches
1481 strings in one of the two forms dd-aaa-dd or dd-dd-dd, where aaa are
1482 letters and dd are digits.
1487 The sequence (?# marks the start of a comment that continues up to the
1488 next closing parenthesis. Nested parentheses are not permitted. The
1489 characters that make up a comment play no part in the pattern matching
1492 If the PCRE_EXTENDED option is set, an unescaped # character outside a
1493 character class introduces a comment that continues to immediately
1494 after the next newline in the pattern.
1499 Consider the problem of matching a string in parentheses, allowing for
1500 unlimited nested parentheses. Without the use of recursion, the best
1501 that can be done is to use a pattern that matches up to some fixed
1502 depth of nesting. It is not possible to handle an arbitrary nesting
1505 For some time, Perl has provided a facility that allows regular expres-
1506 sions to recurse (amongst other things). It does this by interpolating
1507 Perl code in the expression at run time, and the code can refer to the
1508 expression itself. A Perl pattern using code interpolation to solve the
1509 parentheses problem can be created like this:
1511 $re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x;
1513 The (?p{...}) item interpolates Perl code at run time, and in this case
1514 refers recursively to the pattern in which it appears.
1516 Obviously, PCRE cannot support the interpolation of Perl code. Instead,
1517 it supports special syntax for recursion of the entire pattern, and
1518 also for individual subpattern recursion. After its introduction in
1519 PCRE and Python, this kind of recursion was introduced into Perl at
1522 A special item that consists of (? followed by a number greater than
1523 zero and a closing parenthesis is a recursive call of the subpattern of
1524 the given number, provided that it occurs inside that subpattern. (If
1525 not, it is a "subroutine" call, which is described in the next sec-
1526 tion.) The special item (?R) or (?0) is a recursive call of the entire
1529 In PCRE (like Python, but unlike Perl), a recursive subpattern call is
1530 always treated as an atomic group. That is, once it has matched some of
1531 the subject string, it is never re-entered, even if it contains untried
1532 alternatives and there is a subsequent matching failure.
1534 This PCRE pattern solves the nested parentheses problem (assume the
1535 PCRE_EXTENDED option is set so that white space is ignored):
1537 \( ( (?>[^()]+) | (?R) )* \)
1539 First it matches an opening parenthesis. Then it matches any number of
1540 substrings which can either be a sequence of non-parentheses, or a
1541 recursive match of the pattern itself (that is, a correctly parenthe-
1542 sized substring). Finally there is a closing parenthesis.
1544 If this were part of a larger pattern, you would not want to recurse
1545 the entire pattern, so instead you could use this:
1547 ( \( ( (?>[^()]+) | (?1) )* \) )
1549 We have put the pattern into parentheses, and caused the recursion to
1550 refer to them instead of the whole pattern. In a larger pattern, keep-
1551 ing track of parenthesis numbers can be tricky. It may be more conve-
1552 nient to use named parentheses instead. The Perl syntax for this is
1553 (?&name); PCRE's earlier syntax (?P>name) is also supported. We could
1554 rewrite the above example as follows:
1556 (?<pn> \( ( (?>[^()]+) | (?&pn) )* \) )
1558 If there is more than one subpattern with the same name, the earliest
1559 one is used. This particular example pattern contains nested unlimited
1560 repeats, and so the use of atomic grouping for matching strings of non-
1561 parentheses is important when applying the pattern to strings that do
1562 not match. For example, when this pattern is applied to
1564 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
1566 it yields "no match" quickly. However, if atomic grouping is not used,
1567 the match runs for a very long time indeed because there are so many
1568 different ways the + and * repeats can carve up the subject, and all
1569 have to be tested before failure can be reported.
1571 At the end of a match, the values set for any capturing subpatterns are
1572 those from the outermost level of the recursion at which the subpattern
1573 value is set. If you want to obtain intermediate values, a callout
1574 function can be used (see below and the pcrecallout documentation). If
1575 the pattern above is matched against
1579 the value for the capturing parentheses is "ef", which is the last
1580 value taken on at the top level. If additional parentheses are added,
1583 \( ( ( (?>[^()]+) | (?R) )* ) \)
1587 the string they capture is "ab(cd)ef", the contents of the top level
1588 parentheses. If there are more than 15 capturing parentheses in a pat-
1589 tern, PCRE has to obtain extra memory to store data during a recursion,
1590 which it does by using pcre_malloc, freeing it via pcre_free after-
1591 wards. If no memory can be obtained, the match fails with the
1592 PCRE_ERROR_NOMEMORY error.
1594 Do not confuse the (?R) item with the condition (R), which tests for
1595 recursion. Consider this pattern, which matches text in angle brack-
1596 ets, allowing for arbitrary nesting. Only digits are allowed in nested
1597 brackets (that is, when recursing), whereas any characters are permit-
1598 ted at the outer level.
1600 < (?: (?(R) \d++ | [^<>]*+) | (?R)) * >
1602 In this pattern, (?(R) is the start of a conditional subpattern, with
1603 two different alternatives for the recursive and non-recursive cases.
1604 The (?R) item is the actual recursive call.
1607 SUBPATTERNS AS SUBROUTINES
1609 If the syntax for a recursive subpattern reference (either by number or
1610 by name) is used outside the parentheses to which it refers, it oper-
1611 ates like a subroutine in a programming language. The "called" subpat-
1612 tern may be defined before or after the reference. An earlier example
1613 pointed out that the pattern
1615 (sens|respons)e and \1ibility
1617 matches "sense and sensibility" and "response and responsibility", but
1618 not "sense and responsibility". If instead the pattern
1620 (sens|respons)e and (?1)ibility
1622 is used, it does match "sense and responsibility" as well as the other
1623 two strings. Another example is given in the discussion of DEFINE
1626 Like recursive subpatterns, a "subroutine" call is always treated as an
1627 atomic group. That is, once it has matched some of the subject string,
1628 it is never re-entered, even if it contains untried alternatives and
1629 there is a subsequent matching failure.
1631 When a subpattern is used as a subroutine, processing options such as
1632 case-independence are fixed when the subpattern is defined. They cannot
1633 be changed for different calls. For example, consider this pattern:
1637 It matches "abcabc". It does not match "abcABC" because the change of
1638 processing option does not affect the called subpattern.
1643 Perl has a feature whereby using the sequence (?{...}) causes arbitrary
1644 Perl code to be obeyed in the middle of matching a regular expression.
1645 This makes it possible, amongst other things, to extract different sub-
1646 strings that match the same pair of parentheses when there is a repeti-
1649 PCRE provides a similar feature, but of course it cannot obey arbitrary
1650 Perl code. The feature is called "callout". The caller of PCRE provides
1651 an external function by putting its entry point in the global variable
1652 pcre_callout. By default, this variable contains NULL, which disables
1655 Within a regular expression, (?C) indicates the points at which the
1656 external function is to be called. If you want to identify different
1657 callout points, you can put a number less than 256 after the letter C.
1658 The default value is zero. For example, this pattern has two callout
1663 If the PCRE_AUTO_CALLOUT flag is passed to pcre_compile(), callouts are
1664 automatically installed before each item in the pattern. They are all
1667 During matching, when PCRE reaches a callout point (and pcre_callout is
1668 set), the external function is called. It is provided with the number
1669 of the callout, the position in the pattern, and, optionally, one item
1670 of data originally supplied by the caller of pcre_exec(). The callout
1671 function may cause matching to proceed, to backtrack, or to fail alto-
1672 gether. A complete description of the interface to the callout function
1673 is given in the pcrecallout documentation.
1678 pcreapi(3), pcrecallout(3), pcrematching(3), pcre(3).
1680 Last updated: 06 December 2006
1681 Copyright (c) 1997-2006 University of Cambridge.
git://git.exim.org / exim.git / blob