2 X<regular expression> X<regex> X<regexp>
4 perlre - Perl regular expressions
8 This page describes the syntax of regular expressions in Perl.
10 If you haven't used regular expressions before, a quick-start
11 introduction is available in L<perlrequick>, and a longer tutorial
12 introduction is available in L<perlretut>.
14 For reference on how regular expressions are used in matching
15 operations, plus various examples of the same, see discussions of
16 C<m//>, C<s///>, C<qr//> and C<??> in L<perlop/"Regexp Quote-Like
19 Matching operations can have various modifiers. Modifiers
20 that relate to the interpretation of the regular expression inside
21 are listed below. Modifiers that alter the way a regular expression
22 is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and
23 L<perlop/"Gory details of parsing quoted constructs">.
28 X</i> X<regex, case-insensitive> X<regexp, case-insensitive>
29 X<regular expression, case-insensitive>
31 Do case-insensitive pattern matching.
33 If C<use locale> is in effect, the case map is taken from the current
34 locale. See L<perllocale>.
37 X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline>
39 Treat string as multiple lines. That is, change "^" and "$" from matching
40 the start or end of the string to matching the start or end of any
41 line anywhere within the string.
44 X</s> X<regex, single-line> X<regexp, single-line>
45 X<regular expression, single-line>
47 Treat string as single line. That is, change "." to match any character
48 whatsoever, even a newline, which normally it would not match.
50 Used together, as /ms, they let the "." match any character whatsoever,
51 while still allowing "^" and "$" to match, respectively, just after
52 and just before newlines within the string.
57 Extend your pattern's legibility by permitting whitespace and comments.
61 These are usually written as "the C</x> modifier", even though the delimiter
62 in question might not really be a slash. Any of these
63 modifiers may also be embedded within the regular expression itself using
64 the C<(?...)> construct. See below.
66 The C</x> modifier itself needs a little more explanation. It tells
67 the regular expression parser to ignore whitespace that is neither
68 backslashed nor within a character class. You can use this to break up
69 your regular expression into (slightly) more readable parts. The C<#>
70 character is also treated as a metacharacter introducing a comment,
71 just as in ordinary Perl code. This also means that if you want real
72 whitespace or C<#> characters in the pattern (outside a character
73 class, where they are unaffected by C</x>), then you'll either have to
74 escape them (using backslashes or C<\Q...\E>) or encode them using octal
75 or hex escapes. Taken together, these features go a long way towards
76 making Perl's regular expressions more readable. Note that you have to
77 be careful not to include the pattern delimiter in the comment--perl has
78 no way of knowing you did not intend to close the pattern early. See
79 the C-comment deletion code in L<perlop>. Also note that anything inside
80 a C<\Q...\E> stays unaffected by C</x>.
83 =head2 Regular Expressions
87 The patterns used in Perl pattern matching derive from supplied in
88 the Version 8 regex routines. (The routines are derived
89 (distantly) from Henry Spencer's freely redistributable reimplementation
90 of the V8 routines.) See L<Version 8 Regular Expressions> for
93 In particular the following metacharacters have their standard I<egrep>-ish
96 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
99 \ Quote the next metacharacter
100 ^ Match the beginning of the line
101 . Match any character (except newline)
102 $ Match the end of the line (or before newline at the end)
107 By default, the "^" character is guaranteed to match only the
108 beginning of the string, the "$" character only the end (or before the
109 newline at the end), and Perl does certain optimizations with the
110 assumption that the string contains only one line. Embedded newlines
111 will not be matched by "^" or "$". You may, however, wish to treat a
112 string as a multi-line buffer, such that the "^" will match after any
113 newline within the string (except if the newline is the last character in
114 the string), and "$" will match before any newline. At the
115 cost of a little more overhead, you can do this by using the /m modifier
116 on the pattern match operator. (Older programs did this by setting C<$*>,
117 but this practice has been removed in perl 5.9.)
120 To simplify multi-line substitutions, the "." character never matches a
121 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
122 the string is a single line--even if it isn't.
127 The following standard quantifiers are recognized:
128 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
130 * Match 0 or more times
131 + Match 1 or more times
133 {n} Match exactly n times
134 {n,} Match at least n times
135 {n,m} Match at least n but not more than m times
137 (If a curly bracket occurs in any other context, it is treated
138 as a regular character. In particular, the lower bound
139 is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+"
140 modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited
141 to integral values less than a preset limit defined when perl is built.
142 This is usually 32766 on the most common platforms. The actual limit can
143 be seen in the error message generated by code such as this:
145 $_ **= $_ , / {$_} / for 2 .. 42;
147 By default, a quantified subpattern is "greedy", that is, it will match as
148 many times as possible (given a particular starting location) while still
149 allowing the rest of the pattern to match. If you want it to match the
150 minimum number of times possible, follow the quantifier with a "?". Note
151 that the meanings don't change, just the "greediness":
152 X<metacharacter> X<greedy> X<greedyness>
153 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
155 *? Match 0 or more times
156 +? Match 1 or more times
158 {n}? Match exactly n times
159 {n,}? Match at least n times
160 {n,m}? Match at least n but not more than m times
162 By default, when a quantified subpattern does not allow the rest of the
163 overall pattern to match, Perl will backtrack. However, this behaviour is
164 sometimes undesirable. Thus Perl provides the "possesive" quantifier form
167 *+ Match 0 or more times and give nothing back
168 ++ Match 1 or more times and give nothing back
169 ?+ Match 0 or 1 time and give nothing back
170 {n}+ Match exactly n times and give nothing back (redundant)
171 {n,}+ Match at least n times and give nothing back
172 {n,m}+ Match at least n but not more than m times and give nothing back
178 will never match, as the C<a++> will gobble up all the C<a>'s in the
179 string and won't leave any for the remaining part of the pattern. This
180 feature can be extremely useful to give perl hints about where it
181 shouldn't backtrack. For instance, the typical "match a double-quoted
182 string" problem can be most efficiently performed when written as:
184 /"(?:[^"\\]++|\\.)*+"/
186 as we know that if the final quote does not match, bactracking will not
187 help. See the independent subexpression C<< (?>...) >> for more details;
188 possessive quantifiers are just syntactic sugar for that construct. For
189 instance the above example could also be written as follows:
191 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
193 =head3 Escape sequences
195 Because patterns are processed as double quoted strings, the following
197 X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q>
198 X<\0> X<\c> X<\N> X<\x>
204 \a alarm (bell) (BEL)
205 \e escape (think troff) (ESC)
206 \033 octal char (think of a PDP-11)
208 \x{263a} wide hex char (Unicode SMILEY)
211 \l lowercase next char (think vi)
212 \u uppercase next char (think vi)
213 \L lowercase till \E (think vi)
214 \U uppercase till \E (think vi)
215 \E end case modification (think vi)
216 \Q quote (disable) pattern metacharacters till \E
218 If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u>
219 and C<\U> is taken from the current locale. See L<perllocale>. For
220 documentation of C<\N{name}>, see L<charnames>.
222 You cannot include a literal C<$> or C<@> within a C<\Q> sequence.
223 An unescaped C<$> or C<@> interpolates the corresponding variable,
224 while escaping will cause the literal string C<\$> to be matched.
225 You'll need to write something like C<m/\Quser\E\@\Qhost/>.
227 =head3 Character classes
229 In addition, Perl defines the following:
231 X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C>
232 X<word> X<whitespace>
234 \w Match a "word" character (alphanumeric plus "_")
235 \W Match a non-"word" character
236 \s Match a whitespace character
237 \S Match a non-whitespace character
238 \d Match a digit character
239 \D Match a non-digit character
240 \pP Match P, named property. Use \p{Prop} for longer names.
242 \X Match eXtended Unicode "combining character sequence",
243 equivalent to (?:\PM\pM*)
244 \C Match a single C char (octet) even under Unicode.
245 NOTE: breaks up characters into their UTF-8 bytes,
246 so you may end up with malformed pieces of UTF-8.
247 Unsupported in lookbehind.
248 \1 Backreference to a specific group.
249 '1' may actually be any positive integer.
250 \g1 Backreference to a specific or previous group,
251 \g{-1} number may be negative indicating a previous buffer and may
252 optionally be wrapped in curly brackets for safer parsing.
253 \g{name} Named backreference
254 \k<name> Named backreference
255 \N{name} Named unicode character, or unicode escape
256 \x12 Hexadecimal escape sequence
257 \x{1234} Long hexadecimal escape sequence
259 A C<\w> matches a single alphanumeric character (an alphabetic
260 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
261 to match a string of Perl-identifier characters (which isn't the same
262 as matching an English word). If C<use locale> is in effect, the list
263 of alphabetic characters generated by C<\w> is taken from the current
264 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
265 C<\d>, and C<\D> within character classes, but if you try to use them
266 as endpoints of a range, that's not a range, the "-" is understood
267 literally. If Unicode is in effect, C<\s> matches also "\x{85}",
268 "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about
269 C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general.
270 You can define your own C<\p> and C<\P> properties, see L<perlunicode>.
273 The POSIX character class syntax
278 is also available. Note that the C<[> and C<]> braces are I<literal>;
279 they must always be used within a character class expression.
282 $string =~ /[[:alpha:]]/;
284 # this is not, and will generate a warning:
285 $string =~ /[:alpha:]/;
287 The available classes and their backslash equivalents (if available) are
290 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
291 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
312 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
316 Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
317 also the (very rare) "vertical tabulator", "\ck", chr(11).
321 A Perl extension, see above.
325 For example use C<[:upper:]> to match all the uppercase characters.
326 Note that the C<[]> are part of the C<[::]> construct, not part of the
327 whole character class. For example:
331 matches zero, one, any alphabetic character, and the percentage sign.
333 The following equivalences to Unicode \p{} constructs and equivalent
334 backslash character classes (if available), will hold:
335 X<character class> X<\p> X<\p{}>
337 [[:...:]] \p{...} backslash
355 For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent.
357 If the C<utf8> pragma is not used but the C<locale> pragma is, the
358 classes correlate with the usual isalpha(3) interface (except for
361 The assumedly non-obviously named classes are:
368 Any control character. Usually characters that don't produce output as
369 such but instead control the terminal somehow: for example newline and
370 backspace are control characters. All characters with ord() less than
371 32 are most often classified as control characters (assuming ASCII,
372 the ISO Latin character sets, and Unicode), as is the character with
373 the ord() value of 127 (C<DEL>).
378 Any alphanumeric or punctuation (special) character.
383 Any alphanumeric or punctuation (special) character or the space character.
388 Any punctuation (special) character.
393 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
394 work just fine) it is included for completeness.
398 You can negate the [::] character classes by prefixing the class name
399 with a '^'. This is a Perl extension. For example:
400 X<character class, negation>
402 POSIX traditional Unicode
404 [[:^digit:]] \D \P{IsDigit}
405 [[:^space:]] \S \P{IsSpace}
406 [[:^word:]] \W \P{IsWord}
408 Perl respects the POSIX standard in that POSIX character classes are
409 only supported within a character class. The POSIX character classes
410 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
411 use them will cause an error.
415 Perl defines the following zero-width assertions:
416 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
417 X<regexp, zero-width assertion>
418 X<regular expression, zero-width assertion>
419 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
421 \b Match a word boundary
422 \B Match a non-(word boundary)
423 \A Match only at beginning of string
424 \Z Match only at end of string, or before newline at the end
425 \z Match only at end of string
426 \G Match only at pos() (e.g. at the end-of-match position
429 A word boundary (C<\b>) is a spot between two characters
430 that has a C<\w> on one side of it and a C<\W> on the other side
431 of it (in either order), counting the imaginary characters off the
432 beginning and end of the string as matching a C<\W>. (Within
433 character classes C<\b> represents backspace rather than a word
434 boundary, just as it normally does in any double-quoted string.)
435 The C<\A> and C<\Z> are just like "^" and "$", except that they
436 won't match multiple times when the C</m> modifier is used, while
437 "^" and "$" will match at every internal line boundary. To match
438 the actual end of the string and not ignore an optional trailing
440 X<\b> X<\A> X<\Z> X<\z> X</m>
442 The C<\G> assertion can be used to chain global matches (using
443 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
444 It is also useful when writing C<lex>-like scanners, when you have
445 several patterns that you want to match against consequent substrings
446 of your string, see the previous reference. The actual location
447 where C<\G> will match can also be influenced by using C<pos()> as
448 an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
449 matches is modified somewhat, in that contents to the left of C<\G> is
450 not counted when determining the length of the match. Thus the following
451 will not match forever:
460 It will print 'A' and then terminate, as it considers the match to
461 be zero-width, and thus will not match at the same position twice in a
464 It is worth noting that C<\G> improperly used can result in an infinite
465 loop. Take care when using patterns that include C<\G> in an alternation.
467 =head3 Capture buffers
469 The bracketing construct C<( ... )> creates capture buffers. To
470 refer to the digit'th buffer use \<digit> within the
471 match. Outside the match use "$" instead of "\". (The
472 \<digit> notation works in certain circumstances outside
473 the match. See the warning below about \1 vs $1 for details.)
474 Referring back to another part of the match is called a
476 X<regex, capture buffer> X<regexp, capture buffer>
477 X<regular expression, capture buffer> X<backreference>
479 There is no limit to the number of captured substrings that you may
480 use. However Perl also uses \10, \11, etc. as aliases for \010,
481 \011, etc. (Recall that 0 means octal, so \011 is the character at
482 number 9 in your coded character set; which would be the 10th character,
483 a horizontal tab under ASCII.) Perl resolves this
484 ambiguity by interpreting \10 as a backreference only if at least 10
485 left parentheses have opened before it. Likewise \11 is a
486 backreference only if at least 11 left parentheses have opened
487 before it. And so on. \1 through \9 are always interpreted as
490 X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
491 In order to provide a safer and easier way to construct patterns using
492 backrefs, in Perl 5.10 the C<\g{N}> notation is provided. The curly
493 brackets are optional, however omitting them is less safe as the meaning
494 of the pattern can be changed by text (such as digits) following it.
495 When N is a positive integer the C<\g{N}> notation is exactly equivalent
496 to using normal backreferences. When N is a negative integer then it is
497 a relative backreference referring to the previous N'th capturing group.
498 When the bracket form is used and N is not an integer, it is treated as a
499 reference to a named buffer.
501 Thus C<\g{-1}> refers to the last buffer, C<\g{-2}> refers to the
502 buffer before that. For example:
508 \g{-1} # backref to buffer 3
509 \g{-3} # backref to buffer 1
513 and would match the same as C</(Y) ( (X) \3 \1 )/x>.
515 Additionally, as of Perl 5.10 you may use named capture buffers and named
516 backreferences. The notation is C<< (?<name>...) >> to declare and C<< \k<name> >>
517 to reference. You may also use single quotes instead of angle brackets to quote the
518 name; and you may use the bracketed C<< \g{name} >> back reference syntax.
519 The only difference between named capture buffers and unnamed ones is
520 that multiple buffers may have the same name and that the contents of
521 named capture buffers are available via the C<%+> hash. When multiple
522 groups share the same name C<$+{name}> and C<< \k<name> >> refer to the
523 leftmost defined group, thus it's possible to do things with named capture
524 buffers that would otherwise require C<(??{})> code to accomplish. Named
525 capture buffers are numbered just as normal capture buffers are and may be
526 referenced via the magic numeric variables or via numeric backreferences
531 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
533 /(.)\1/ # find first doubled char
534 and print "'$1' is the first doubled character\n";
536 /(?<char>.)\k<char>/ # ... a different way
537 and print "'$+{char}' is the first doubled character\n";
539 /(?<char>.)\1/ # ... mix and match
540 and print "'$1' is the first doubled character\n";
542 if (/Time: (..):(..):(..)/) { # parse out values
548 Several special variables also refer back to portions of the previous
549 match. C<$+> returns whatever the last bracket match matched.
550 C<$&> returns the entire matched string. (At one point C<$0> did
551 also, but now it returns the name of the program.) C<$`> returns
552 everything before the matched string. C<$'> returns everything
553 after the matched string. And C<$^N> contains whatever was matched by
554 the most-recently closed group (submatch). C<$^N> can be used in
555 extended patterns (see below), for example to assign a submatch to a
557 X<$+> X<$^N> X<$&> X<$`> X<$'>
559 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
560 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
561 until the end of the enclosing block or until the next successful
562 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
563 X<$+> X<$^N> X<$&> X<$`> X<$'>
564 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
567 B<NOTE>: failed matches in Perl do not reset the match variables,
568 which makes it easier to write code that tests for a series of more
569 specific cases and remembers the best match.
571 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
572 C<$'> anywhere in the program, it has to provide them for every
573 pattern match. This may substantially slow your program. Perl
574 uses the same mechanism to produce $1, $2, etc, so you also pay a
575 price for each pattern that contains capturing parentheses. (To
576 avoid this cost while retaining the grouping behaviour, use the
577 extended regular expression C<(?: ... )> instead.) But if you never
578 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
579 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
580 if you can, but if you can't (and some algorithms really appreciate
581 them), once you've used them once, use them at will, because you've
582 already paid the price. As of 5.005, C<$&> is not so costly as the
586 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
587 C<\w>, C<\n>. Unlike some other regular expression languages, there
588 are no backslashed symbols that aren't alphanumeric. So anything
589 that looks like \\, \(, \), \<, \>, \{, or \} is always
590 interpreted as a literal character, not a metacharacter. This was
591 once used in a common idiom to disable or quote the special meanings
592 of regular expression metacharacters in a string that you want to
593 use for a pattern. Simply quote all non-"word" characters:
595 $pattern =~ s/(\W)/\\$1/g;
597 (If C<use locale> is set, then this depends on the current locale.)
598 Today it is more common to use the quotemeta() function or the C<\Q>
599 metaquoting escape sequence to disable all metacharacters' special
602 /$unquoted\Q$quoted\E$unquoted/
604 Beware that if you put literal backslashes (those not inside
605 interpolated variables) between C<\Q> and C<\E>, double-quotish
606 backslash interpolation may lead to confusing results. If you
607 I<need> to use literal backslashes within C<\Q...\E>,
608 consult L<perlop/"Gory details of parsing quoted constructs">.
610 =head2 Extended Patterns
612 Perl also defines a consistent extension syntax for features not
613 found in standard tools like B<awk> and B<lex>. The syntax is a
614 pair of parentheses with a question mark as the first thing within
615 the parentheses. The character after the question mark indicates
618 The stability of these extensions varies widely. Some have been
619 part of the core language for many years. Others are experimental
620 and may change without warning or be completely removed. Check
621 the documentation on an individual feature to verify its current
624 A question mark was chosen for this and for the minimal-matching
625 construct because 1) question marks are rare in older regular
626 expressions, and 2) whenever you see one, you should stop and
627 "question" exactly what is going on. That's psychology...
634 A comment. The text is ignored. If the C</x> modifier enables
635 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
636 the comment as soon as it sees a C<)>, so there is no way to put a literal
639 =item C<(?imsx-imsx)>
642 One or more embedded pattern-match modifiers, to be turned on (or
643 turned off, if preceded by C<->) for the remainder of the pattern or
644 the remainder of the enclosing pattern group (if any). This is
645 particularly useful for dynamic patterns, such as those read in from a
646 configuration file, read in as an argument, are specified in a table
647 somewhere, etc. Consider the case that some of which want to be case
648 sensitive and some do not. The case insensitive ones need to include
649 merely C<(?i)> at the front of the pattern. For example:
652 if ( /$pattern/i ) { }
656 $pattern = "(?i)foobar";
657 if ( /$pattern/ ) { }
659 These modifiers are restored at the end of the enclosing group. For example,
663 will match a repeated (I<including the case>!) word C<blah> in any
664 case, assuming C<x> modifier, and no C<i> modifier outside this
670 =item C<(?imsx-imsx:pattern)>
672 This is for clustering, not capturing; it groups subexpressions like
673 "()", but doesn't make backreferences as "()" does. So
675 @fields = split(/\b(?:a|b|c)\b/)
679 @fields = split(/\b(a|b|c)\b/)
681 but doesn't spit out extra fields. It's also cheaper not to capture
682 characters if you don't need to.
684 Any letters between C<?> and C<:> act as flags modifiers as with
685 C<(?imsx-imsx)>. For example,
687 /(?s-i:more.*than).*million/i
689 is equivalent to the more verbose
691 /(?:(?s-i)more.*than).*million/i
694 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
696 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
697 matches a word followed by a tab, without including the tab in C<$&>.
700 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
702 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
703 matches any occurrence of "foo" that isn't followed by "bar". Note
704 however that look-ahead and look-behind are NOT the same thing. You cannot
705 use this for look-behind.
707 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
708 will not do what you want. That's because the C<(?!foo)> is just saying that
709 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
710 match. You would have to do something like C</(?!foo)...bar/> for that. We
711 say "like" because there's the case of your "bar" not having three characters
712 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
713 Sometimes it's still easier just to say:
715 if (/bar/ && $` !~ /foo$/)
717 For look-behind see below.
719 =item C<(?<=pattern)>
720 X<(?<=)> X<look-behind, positive> X<lookbehind, positive>
722 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
723 matches a word that follows a tab, without including the tab in C<$&>.
724 Works only for fixed-width look-behind.
726 =item C<(?<!pattern)>
727 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
729 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
730 matches any occurrence of "foo" that does not follow "bar". Works
731 only for fixed-width look-behind.
733 =item C<(?'NAME'pattern)>
735 =item C<< (?<NAME>pattern) >>
736 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
738 A named capture buffer. Identical in every respect to normal capturing
739 parens C<()> but for the additional fact that C<%+> may be used after
740 a succesful match to refer to a named buffer. See C<perlvar> for more
741 details on the C<%+> hash.
743 If multiple distinct capture buffers have the same name then the
744 $+{NAME} will refer to the leftmost defined buffer in the match.
746 The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent.
748 B<NOTE:> While the notation of this construct is the same as the similar
749 function in .NET regexes, the behavior is not, in Perl the buffers are
750 numbered sequentially regardless of being named or not. Thus in the
755 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
756 the opposite which is what a .NET regex hacker might expect.
758 Currently NAME is restricted to simple identifiers only.
759 In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
760 its Unicode extension (see L<utf8>),
761 though it isn't extended by the locale (see L<perllocale>).
763 B<NOTE:> In order to make things easier for programmers with experience
764 with the Python or PCRE regex engines the pattern C<< (?P<NAME>pattern) >>
765 maybe be used instead of C<< (?<NAME>pattern) >>; however this form does not
766 support the use of single quotes as a delimiter for the name. This is
767 only available in Perl 5.10 or later.
769 =item C<< \k<NAME> >>
771 =item C<< \k'NAME' >>
773 Named backreference. Similar to numeric backreferences, except that
774 the group is designated by name and not number. If multiple groups
775 have the same name then it refers to the leftmost defined group in
778 It is an error to refer to a name not defined by a C<(?<NAME>)>
779 earlier in the pattern.
781 Both forms are equivalent.
783 B<NOTE:> In order to make things easier for programmers with experience
784 with the Python or PCRE regex engines the pattern C<< (?P=NAME) >>
785 maybe be used instead of C<< \k<NAME> >> in Perl 5.10 or later.
788 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
790 B<WARNING>: This extended regular expression feature is considered
791 experimental, and may be changed without notice. Code executed that
792 has side effects may not perform identically from version to version
793 due to the effect of future optimisations in the regex engine.
795 This zero-width assertion evaluates any embedded Perl code. It
796 always succeeds, and its C<code> is not interpolated. Currently,
797 the rules to determine where the C<code> ends are somewhat convoluted.
799 This feature can be used together with the special variable C<$^N> to
800 capture the results of submatches in variables without having to keep
801 track of the number of nested parentheses. For example:
803 $_ = "The brown fox jumps over the lazy dog";
804 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
805 print "color = $color, animal = $animal\n";
807 Inside the C<(?{...})> block, C<$_> refers to the string the regular
808 expression is matching against. You can also use C<pos()> to know what is
809 the current position of matching within this string.
811 The C<code> is properly scoped in the following sense: If the assertion
812 is backtracked (compare L<"Backtracking">), all changes introduced after
813 C<local>ization are undone, so that
817 (?{ $cnt = 0 }) # Initialize $cnt.
821 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
825 (?{ $res = $cnt }) # On success copy to non-localized
829 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
830 introduced value, because the scopes that restrict C<local> operators
833 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
834 switch. If I<not> used in this way, the result of evaluation of
835 C<code> is put into the special variable C<$^R>. This happens
836 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
837 inside the same regular expression.
839 The assignment to C<$^R> above is properly localized, so the old
840 value of C<$^R> is restored if the assertion is backtracked; compare
843 Due to an unfortunate implementation issue, the Perl code contained in these
844 blocks is treated as a compile time closure that can have seemingly bizarre
845 consequences when used with lexically scoped variables inside of subroutines
846 or loops. There are various workarounds for this, including simply using
847 global variables instead. If you are using this construct and strange results
848 occur then check for the use of lexically scoped variables.
850 For reasons of security, this construct is forbidden if the regular
851 expression involves run-time interpolation of variables, unless the
852 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
853 variables contain results of C<qr//> operator (see
854 L<perlop/"qr/STRING/imosx">).
856 This restriction is because of the wide-spread and remarkably convenient
857 custom of using run-time determined strings as patterns. For example:
863 Before Perl knew how to execute interpolated code within a pattern,
864 this operation was completely safe from a security point of view,
865 although it could raise an exception from an illegal pattern. If
866 you turn on the C<use re 'eval'>, though, it is no longer secure,
867 so you should only do so if you are also using taint checking.
868 Better yet, use the carefully constrained evaluation within a Safe
869 compartment. See L<perlsec> for details about both these mechanisms.
871 Because perl's regex engine is not currently re-entrant, interpolated
872 code may not invoke the regex engine either directly with C<m//> or C<s///>),
873 or indirectly with functions such as C<split>.
875 =item C<(??{ code })>
877 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
879 B<WARNING>: This extended regular expression feature is considered
880 experimental, and may be changed without notice. Code executed that
881 has side effects may not perform identically from version to version
882 due to the effect of future optimisations in the regex engine.
884 This is a "postponed" regular subexpression. The C<code> is evaluated
885 at run time, at the moment this subexpression may match. The result
886 of evaluation is considered as a regular expression and matched as
887 if it were inserted instead of this construct. Note that this means
888 that the contents of capture buffers defined inside an eval'ed pattern
889 are not available outside of the pattern, and vice versa, there is no
890 way for the inner pattern to refer to a capture buffer defined outside.
893 ('a' x 100)=~/(??{'(.)' x 100})/
895 B<will> match, it will B<not> set $1.
897 The C<code> is not interpolated. As before, the rules to determine
898 where the C<code> ends are currently somewhat convoluted.
900 The following pattern matches a parenthesized group:
905 (?> [^()]+ ) # Non-parens without backtracking
907 (??{ $re }) # Group with matching parens
912 See also C<(?PARNO)> for a different, more efficient way to accomplish
915 Because perl's regex engine is not currently re-entrant, delayed
916 code may not invoke the regex engine either directly with C<m//> or C<s///>),
917 or indirectly with functions such as C<split>.
919 Recursing deeper than 50 times without consuming any input string will
920 result in a fatal error. The maximum depth is compiled into perl, so
921 changing it requires a custom build.
923 =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)>
924 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
925 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
926 X<regex, relative recursion>
928 Similar to C<(??{ code })> except it does not involve compiling any code,
929 instead it treats the contents of a capture buffer as an independent
930 pattern that must match at the current position. Capture buffers
931 contained by the pattern will have the value as determined by the
934 PARNO is a sequence of digits (not starting with 0) whose value reflects
935 the paren-number of the capture buffer to recurse to. C<(?R)> recurses to
936 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
937 C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed
938 to be relative, with negative numbers indicating preceding capture buffers
939 and positive ones following. Thus C<(?-1)> refers to the most recently
940 declared buffer, and C<(?+1)> indicates the next buffer to be declared.
941 Note that the counting for relative recursion differs from that of
942 relative backreferences, in that with recursion unclosed buffers B<are>
945 The following pattern matches a function foo() which may contain
946 balanced parentheses as the argument.
948 $re = qr{ ( # paren group 1 (full function)
950 ( # paren group 2 (parens)
952 ( # paren group 3 (contents of parens)
954 (?> [^()]+ ) # Non-parens without backtracking
956 (?2) # Recurse to start of paren group 2
964 If the pattern was used as follows
966 'foo(bar(baz)+baz(bop))'=~/$re/
967 and print "\$1 = $1\n",
971 the output produced should be the following:
973 $1 = foo(bar(baz)+baz(bop))
974 $2 = (bar(baz)+baz(bop))
975 $3 = bar(baz)+baz(bop)
977 If there is no corresponding capture buffer defined, then it is a
978 fatal error. Recursing deeper than 50 times without consuming any input
979 string will also result in a fatal error. The maximum depth is compiled
980 into perl, so changing it requires a custom build.
982 The following shows how using negative indexing can make it
983 easier to embed recursive patterns inside of a C<qr//> construct
986 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
987 if (/foo $parens \s+ + \s+ bar $parens/x) {
988 # do something here...
991 B<Note> that this pattern does not behave the same way as the equivalent
992 PCRE or Python construct of the same form. In perl you can backtrack into
993 a recursed group, in PCRE and Python the recursed into group is treated
994 as atomic. Also, modifiers are resolved at compile time, so constructs
995 like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
1001 Recurse to a named subpattern. Identical to (?PARNO) except that the
1002 parenthesis to recurse to is determined by name. If multiple parens have
1003 the same name, then it recurses to the leftmost.
1005 It is an error to refer to a name that is not declared somewhere in the
1008 B<NOTE:> In order to make things easier for programmers with experience
1009 with the Python or PCRE regex engines the pattern C<< (?P>NAME) >>
1010 maybe be used instead of C<< (?&NAME) >> as of Perl 5.10.
1012 =item C<(?(condition)yes-pattern|no-pattern)>
1015 =item C<(?(condition)yes-pattern)>
1017 Conditional expression. C<(condition)> should be either an integer in
1018 parentheses (which is valid if the corresponding pair of parentheses
1019 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
1020 name in angle brackets or single quotes (which is valid if a buffer
1021 with the given name matched), or the special symbol (R) (true when
1022 evaluated inside of recursion or eval). Additionally the R may be
1023 followed by a number, (which will be true when evaluated when recursing
1024 inside of the appropriate group), or by C<&NAME>, in which case it will
1025 be true only when evaluated during recursion in the named group.
1027 Here's a summary of the possible predicates:
1033 Checks if the numbered capturing buffer has matched something.
1035 =item (<NAME>) ('NAME')
1037 Checks if a buffer with the given name has matched something.
1041 Treats the code block as the condition.
1045 Checks if the expression has been evaluated inside of recursion.
1049 Checks if the expression has been evaluated while executing directly
1050 inside of the n-th capture group. This check is the regex equivalent of
1052 if ((caller(0))[3] eq 'subname') { ... }
1054 In other words, it does not check the full recursion stack.
1058 Similar to C<(R1)>, this predicate checks to see if we're executing
1059 directly inside of the leftmost group with a given name (this is the same
1060 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1061 stack, but only the name of the innermost active recursion.
1065 In this case, the yes-pattern is never directly executed, and no
1066 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1067 See below for details.
1078 matches a chunk of non-parentheses, possibly included in parentheses
1081 A special form is the C<(DEFINE)> predicate, which never executes directly
1082 its yes-pattern, and does not allow a no-pattern. This allows to define
1083 subpatterns which will be executed only by using the recursion mechanism.
1084 This way, you can define a set of regular expression rules that can be
1085 bundled into any pattern you choose.
1087 It is recommended that for this usage you put the DEFINE block at the
1088 end of the pattern, and that you name any subpatterns defined within it.
1090 Also, it's worth noting that patterns defined this way probably will
1091 not be as efficient, as the optimiser is not very clever about
1094 An example of how this might be used is as follows:
1096 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
1102 Note that capture buffers matched inside of recursion are not accessible
1103 after the recursion returns, so the extra layer of capturing buffers are
1104 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1105 C<$+{NAME}> would be.
1107 =item C<< (?>pattern) >>
1108 X<backtrack> X<backtracking> X<atomic> X<possessive>
1110 An "independent" subexpression, one which matches the substring
1111 that a I<standalone> C<pattern> would match if anchored at the given
1112 position, and it matches I<nothing other than this substring>. This
1113 construct is useful for optimizations of what would otherwise be
1114 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1115 It may also be useful in places where the "grab all you can, and do not
1116 give anything back" semantic is desirable.
1118 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1119 (anchored at the beginning of string, as above) will match I<all>
1120 characters C<a> at the beginning of string, leaving no C<a> for
1121 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1122 since the match of the subgroup C<a*> is influenced by the following
1123 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1124 C<a*ab> will match fewer characters than a standalone C<a*>, since
1125 this makes the tail match.
1127 An effect similar to C<< (?>pattern) >> may be achieved by writing
1128 C<(?=(pattern))\1>. This matches the same substring as a standalone
1129 C<a+>, and the following C<\1> eats the matched string; it therefore
1130 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1131 (The difference between these two constructs is that the second one
1132 uses a capturing group, thus shifting ordinals of backreferences
1133 in the rest of a regular expression.)
1135 Consider this pattern:
1146 That will efficiently match a nonempty group with matching parentheses
1147 two levels deep or less. However, if there is no such group, it
1148 will take virtually forever on a long string. That's because there
1149 are so many different ways to split a long string into several
1150 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1151 to a subpattern of the above pattern. Consider how the pattern
1152 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1153 seconds, but that each extra letter doubles this time. This
1154 exponential performance will make it appear that your program has
1155 hung. However, a tiny change to this pattern
1159 (?> [^()]+ ) # change x+ above to (?> x+ )
1166 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1167 this yourself would be a productive exercise), but finishes in a fourth
1168 the time when used on a similar string with 1000000 C<a>s. Be aware,
1169 however, that this pattern currently triggers a warning message under
1170 the C<use warnings> pragma or B<-w> switch saying it
1171 C<"matches null string many times in regex">.
1173 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1174 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1175 This was only 4 times slower on a string with 1000000 C<a>s.
1177 The "grab all you can, and do not give anything back" semantic is desirable
1178 in many situations where on the first sight a simple C<()*> looks like
1179 the correct solution. Suppose we parse text with comments being delimited
1180 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1181 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1182 the comment delimiter, because it may "give up" some whitespace if
1183 the remainder of the pattern can be made to match that way. The correct
1184 answer is either one of these:
1189 For example, to grab non-empty comments into $1, one should use either
1192 / (?> \# [ \t]* ) ( .+ ) /x;
1193 / \# [ \t]* ( [^ \t] .* ) /x;
1195 Which one you pick depends on which of these expressions better reflects
1196 the above specification of comments.
1198 In some literature this construct is called "atomic matching" or
1199 "possessive matching".
1201 Possessive quantifiers are equivalent to putting the item they are applied
1202 to inside of one of these constructs. The following equivalences apply:
1204 Quantifier Form Bracketing Form
1205 --------------- ---------------
1209 PAT{min,max}+ (?>PAT{min,max})
1213 =head2 Special Backtracking Control Verbs
1215 B<WARNING:> These patterns are experimental and subject to change or
1216 removal in a future version of perl. Their usage in production code should
1217 be noted to avoid problems during upgrades.
1219 These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1220 otherwise stated the ARG argument is optional; in some cases, it is
1223 Any pattern containing a special backtracking verb that allows an argument
1224 has the special behaviour that when executed it sets the current packages'
1225 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1228 On failure, the C<$REGERROR> variable will be set to the ARG value of the
1229 verb pattern, if the verb was involved in the failure of the match. If the
1230 ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1231 name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1232 none. Also, the C<$REGMARK> variable will be set to FALSE.
1234 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1235 the C<$REGMARK> variable will be set to the name of the last
1236 C<(*MARK:NAME)> pattern executed. See the explanation for the
1237 C<(*MARK:NAME)> verb below for more details.
1239 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1240 and most other regex related variables. They are not local to a scope, nor
1241 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1242 Use C<local> to localize changes to them to a specific scope if necessary.
1244 If a pattern does not contain a special backtracking verb that allows an
1245 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
1249 =item Verbs that take an argument
1253 =item C<(*PRUNE)> C<(*PRUNE:NAME)>
1254 X<(*PRUNE)> X<(*PRUNE:NAME)>
1256 This zero-width pattern prunes the backtracking tree at the current point
1257 when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1258 where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1259 A may backtrack as necessary to match. Once it is reached, matching
1260 continues in B, which may also backtrack as necessary; however, should B
1261 not match, then no further backtracking will take place, and the pattern
1262 will fail outright at the current starting position.
1264 The following example counts all the possible matching strings in a
1265 pattern (without actually matching any of them).
1267 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1268 print "Count=$count\n";
1283 If we add a C<(*PRUNE)> before the count like the following
1285 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
1286 print "Count=$count\n";
1288 we prevent backtracking and find the count of the longest matching
1289 at each matching startpoint like so:
1296 Any number of C<(*PRUNE)> assertions may be used in a pattern.
1298 See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1299 control backtracking. In some cases, the use of C<(*PRUNE)> can be
1300 replaced with a C<< (?>pattern) >> with no functional difference; however,
1301 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1302 C<< (?>pattern) >> alone.
1305 =item C<(*SKIP)> C<(*SKIP:NAME)>
1308 This zero-width pattern is similar to C<(*PRUNE)>, except that on
1309 failure it also signifies that whatever text that was matched leading up
1310 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1311 of this pattern. This effectively means that the regex engine "skips" forward
1312 to this position on failure and tries to match again, (assuming that
1313 there is sufficient room to match).
1315 The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1316 C<(*MARK:NAME)> was encountered while matching, then it is that position
1317 which is used as the "skip point". If no C<(*MARK)> of that name was
1318 encountered, then the C<(*SKIP)> operator has no effect. When used
1319 without a name the "skip point" is where the match point was when
1320 executing the (*SKIP) pattern.
1322 Compare the following to the examples in C<(*PRUNE)>, note the string
1325 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
1326 print "Count=$count\n";
1334 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
1335 executed, the next startpoint will be where the cursor was when the
1336 C<(*SKIP)> was executed.
1338 =item C<(*MARK:NAME)> C<(*:NAME)>
1339 X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)>
1341 This zero-width pattern can be used to mark the point reached in a string
1342 when a certain part of the pattern has been successfully matched. This
1343 mark may be given a name. A later C<(*SKIP)> pattern will then skip
1344 forward to that point if backtracked into on failure. Any number of
1345 C<(*MARK)> patterns are allowed, and the NAME portion is optional and may
1348 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1349 can be used to "label" a pattern branch, so that after matching, the
1350 program can determine which branches of the pattern were involved in the
1353 When a match is successful, the C<$REGMARK> variable will be set to the
1354 name of the most recently executed C<(*MARK:NAME)> that was involved
1357 This can be used to determine which branch of a pattern was matched
1358 without using a seperate capture buffer for each branch, which in turn
1359 can result in a performance improvement, as perl cannot optimize
1360 C</(?:(x)|(y)|(z))/> as efficiently as something like
1361 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1363 When a match has failed, and unless another verb has been involved in
1364 failing the match and has provided its own name to use, the C<$REGERROR>
1365 variable will be set to the name of the most recently executed
1368 See C<(*SKIP)> for more details.
1370 As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
1372 =item C<(*THEN)> C<(*THEN:NAME)>
1374 This is similar to the "cut group" operator C<::> from Perl6. Like
1375 C<(*PRUNE)>, this verb always matches, and when backtracked into on
1376 failure, it causes the regex engine to try the next alternation in the
1377 innermost enclosing group (capturing or otherwise).
1379 Its name comes from the observation that this operation combined with the
1380 alternation operator (C<|>) can be used to create what is essentially a
1381 pattern-based if/then/else block:
1383 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
1385 Note that if this operator is used and NOT inside of an alternation then
1386 it acts exactly like the C<(*PRUNE)> operator.
1396 / ( A (*THEN) B | C (*THEN) D ) /
1400 / ( A (*PRUNE) B | C (*PRUNE) D ) /
1402 as after matching the A but failing on the B the C<(*THEN)> verb will
1403 backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
1408 This is the Perl6 "commit pattern" C<< <commit> >> or C<:::>. It's a
1409 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
1410 into on failure it causes the match to fail outright. No further attempts
1411 to find a valid match by advancing the start pointer will occur again.
1414 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
1415 print "Count=$count\n";
1422 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1423 does not match, the regex engine will not try any further matching on the
1428 =item Verbs without an argument
1432 =item C<(*FAIL)> C<(*F)>
1435 This pattern matches nothing and always fails. It can be used to force the
1436 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1437 fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
1439 It is probably useful only when combined with C<(?{})> or C<(??{})>.
1444 B<WARNING:> This feature is highly experimental. It is not recommended
1445 for production code.
1447 This pattern matches nothing and causes the end of successful matching at
1448 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1449 whether there is actually more to match in the string. When inside of a
1450 nested pattern, such as recursion or a dynamically generated subbpattern
1451 via C<(??{})>, only the innermost pattern is ended immediately.
1453 If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are
1454 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1457 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
1459 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
1460 be set. If another branch in the inner parens were matched, such as in the
1461 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
1468 X<backtrack> X<backtracking>
1470 NOTE: This section presents an abstract approximation of regular
1471 expression behavior. For a more rigorous (and complicated) view of
1472 the rules involved in selecting a match among possible alternatives,
1473 see L<Combining pieces together>.
1475 A fundamental feature of regular expression matching involves the
1476 notion called I<backtracking>, which is currently used (when needed)
1477 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
1478 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1479 internally, but the general principle outlined here is valid.
1481 For a regular expression to match, the I<entire> regular expression must
1482 match, not just part of it. So if the beginning of a pattern containing a
1483 quantifier succeeds in a way that causes later parts in the pattern to
1484 fail, the matching engine backs up and recalculates the beginning
1485 part--that's why it's called backtracking.
1487 Here is an example of backtracking: Let's say you want to find the
1488 word following "foo" in the string "Food is on the foo table.":
1490 $_ = "Food is on the foo table.";
1491 if ( /\b(foo)\s+(\w+)/i ) {
1492 print "$2 follows $1.\n";
1495 When the match runs, the first part of the regular expression (C<\b(foo)>)
1496 finds a possible match right at the beginning of the string, and loads up
1497 $1 with "Foo". However, as soon as the matching engine sees that there's
1498 no whitespace following the "Foo" that it had saved in $1, it realizes its
1499 mistake and starts over again one character after where it had the
1500 tentative match. This time it goes all the way until the next occurrence
1501 of "foo". The complete regular expression matches this time, and you get
1502 the expected output of "table follows foo."
1504 Sometimes minimal matching can help a lot. Imagine you'd like to match
1505 everything between "foo" and "bar". Initially, you write something
1508 $_ = "The food is under the bar in the barn.";
1509 if ( /foo(.*)bar/ ) {
1513 Which perhaps unexpectedly yields:
1515 got <d is under the bar in the >
1517 That's because C<.*> was greedy, so you get everything between the
1518 I<first> "foo" and the I<last> "bar". Here it's more effective
1519 to use minimal matching to make sure you get the text between a "foo"
1520 and the first "bar" thereafter.
1522 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1523 got <d is under the >
1525 Here's another example: let's say you'd like to match a number at the end
1526 of a string, and you also want to keep the preceding part of the match.
1529 $_ = "I have 2 numbers: 53147";
1530 if ( /(.*)(\d*)/ ) { # Wrong!
1531 print "Beginning is <$1>, number is <$2>.\n";
1534 That won't work at all, because C<.*> was greedy and gobbled up the
1535 whole string. As C<\d*> can match on an empty string the complete
1536 regular expression matched successfully.
1538 Beginning is <I have 2 numbers: 53147>, number is <>.
1540 Here are some variants, most of which don't work:
1542 $_ = "I have 2 numbers: 53147";
1555 printf "%-12s ", $pat;
1557 print "<$1> <$2>\n";
1563 That will print out:
1565 (.*)(\d*) <I have 2 numbers: 53147> <>
1566 (.*)(\d+) <I have 2 numbers: 5314> <7>
1568 (.*?)(\d+) <I have > <2>
1569 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1570 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1571 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1572 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1574 As you see, this can be a bit tricky. It's important to realize that a
1575 regular expression is merely a set of assertions that gives a definition
1576 of success. There may be 0, 1, or several different ways that the
1577 definition might succeed against a particular string. And if there are
1578 multiple ways it might succeed, you need to understand backtracking to
1579 know which variety of success you will achieve.
1581 When using look-ahead assertions and negations, this can all get even
1582 trickier. Imagine you'd like to find a sequence of non-digits not
1583 followed by "123". You might try to write that as
1586 if ( /^\D*(?!123)/ ) { # Wrong!
1587 print "Yup, no 123 in $_\n";
1590 But that isn't going to match; at least, not the way you're hoping. It
1591 claims that there is no 123 in the string. Here's a clearer picture of
1592 why that pattern matches, contrary to popular expectations:
1597 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1598 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1600 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1601 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1609 You might have expected test 3 to fail because it seems to a more
1610 general purpose version of test 1. The important difference between
1611 them is that test 3 contains a quantifier (C<\D*>) and so can use
1612 backtracking, whereas test 1 will not. What's happening is
1613 that you've asked "Is it true that at the start of $x, following 0 or more
1614 non-digits, you have something that's not 123?" If the pattern matcher had
1615 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1618 The search engine will initially match C<\D*> with "ABC". Then it will
1619 try to match C<(?!123> with "123", which fails. But because
1620 a quantifier (C<\D*>) has been used in the regular expression, the
1621 search engine can backtrack and retry the match differently
1622 in the hope of matching the complete regular expression.
1624 The pattern really, I<really> wants to succeed, so it uses the
1625 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1626 time. Now there's indeed something following "AB" that is not
1627 "123". It's "C123", which suffices.
1629 We can deal with this by using both an assertion and a negation.
1630 We'll say that the first part in $1 must be followed both by a digit
1631 and by something that's not "123". Remember that the look-aheads
1632 are zero-width expressions--they only look, but don't consume any
1633 of the string in their match. So rewriting this way produces what
1634 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1636 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1637 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1641 In other words, the two zero-width assertions next to each other work as though
1642 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1643 matches only if you're at the beginning of the line AND the end of the
1644 line simultaneously. The deeper underlying truth is that juxtaposition in
1645 regular expressions always means AND, except when you write an explicit OR
1646 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1647 although the attempted matches are made at different positions because "a"
1648 is not a zero-width assertion, but a one-width assertion.
1650 B<WARNING>: particularly complicated regular expressions can take
1651 exponential time to solve because of the immense number of possible
1652 ways they can use backtracking to try match. For example, without
1653 internal optimizations done by the regular expression engine, this will
1654 take a painfully long time to run:
1656 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1658 And if you used C<*>'s in the internal groups instead of limiting them
1659 to 0 through 5 matches, then it would take forever--or until you ran
1660 out of stack space. Moreover, these internal optimizations are not
1661 always applicable. For example, if you put C<{0,5}> instead of C<*>
1662 on the external group, no current optimization is applicable, and the
1663 match takes a long time to finish.
1665 A powerful tool for optimizing such beasts is what is known as an
1666 "independent group",
1667 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1668 zero-length look-ahead/look-behind assertions will not backtrack to make
1669 the tail match, since they are in "logical" context: only
1670 whether they match is considered relevant. For an example
1671 where side-effects of look-ahead I<might> have influenced the
1672 following match, see L<C<< (?>pattern) >>>.
1674 =head2 Version 8 Regular Expressions
1675 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1677 In case you're not familiar with the "regular" Version 8 regex
1678 routines, here are the pattern-matching rules not described above.
1680 Any single character matches itself, unless it is a I<metacharacter>
1681 with a special meaning described here or above. You can cause
1682 characters that normally function as metacharacters to be interpreted
1683 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1684 character; "\\" matches a "\"). A series of characters matches that
1685 series of characters in the target string, so the pattern C<blurfl>
1686 would match "blurfl" in the target string.
1688 You can specify a character class, by enclosing a list of characters
1689 in C<[]>, which will match any character from the list. If the
1690 first character after the "[" is "^", the class matches any character not
1691 in the list. Within a list, the "-" character specifies a
1692 range, so that C<a-z> represents all characters between "a" and "z",
1693 inclusive. If you want either "-" or "]" itself to be a member of a
1694 class, put it at the start of the list (possibly after a "^"), or
1695 escape it with a backslash. "-" is also taken literally when it is
1696 at the end of the list, just before the closing "]". (The
1697 following all specify the same class of three characters: C<[-az]>,
1698 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1699 specifies a class containing twenty-six characters, even on EBCDIC-based
1700 character sets.) Also, if you try to use the character
1701 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1702 a range, the "-" is understood literally.
1704 Note also that the whole range idea is rather unportable between
1705 character sets--and even within character sets they may cause results
1706 you probably didn't expect. A sound principle is to use only ranges
1707 that begin from and end at either alphabets of equal case ([a-e],
1708 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1709 spell out the character sets in full.
1711 Characters may be specified using a metacharacter syntax much like that
1712 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1713 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1714 of octal digits, matches the character whose coded character set value
1715 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1716 matches the character whose numeric value is I<nn>. The expression \cI<x>
1717 matches the character control-I<x>. Finally, the "." metacharacter
1718 matches any character except "\n" (unless you use C</s>).
1720 You can specify a series of alternatives for a pattern using "|" to
1721 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1722 or "foe" in the target string (as would C<f(e|i|o)e>). The
1723 first alternative includes everything from the last pattern delimiter
1724 ("(", "[", or the beginning of the pattern) up to the first "|", and
1725 the last alternative contains everything from the last "|" to the next
1726 pattern delimiter. That's why it's common practice to include
1727 alternatives in parentheses: to minimize confusion about where they
1730 Alternatives are tried from left to right, so the first
1731 alternative found for which the entire expression matches, is the one that
1732 is chosen. This means that alternatives are not necessarily greedy. For
1733 example: when matching C<foo|foot> against "barefoot", only the "foo"
1734 part will match, as that is the first alternative tried, and it successfully
1735 matches the target string. (This might not seem important, but it is
1736 important when you are capturing matched text using parentheses.)
1738 Also remember that "|" is interpreted as a literal within square brackets,
1739 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1741 Within a pattern, you may designate subpatterns for later reference
1742 by enclosing them in parentheses, and you may refer back to the
1743 I<n>th subpattern later in the pattern using the metacharacter
1744 \I<n>. Subpatterns are numbered based on the left to right order
1745 of their opening parenthesis. A backreference matches whatever
1746 actually matched the subpattern in the string being examined, not
1747 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1748 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1749 1 matched "0x", even though the rule C<0|0x> could potentially match
1750 the leading 0 in the second number.
1752 =head2 Warning on \1 vs $1
1754 Some people get too used to writing things like:
1756 $pattern =~ s/(\W)/\\\1/g;
1758 This is grandfathered for the RHS of a substitute to avoid shocking the
1759 B<sed> addicts, but it's a dirty habit to get into. That's because in
1760 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1761 the usual double-quoted string means a control-A. The customary Unix
1762 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1763 of doing that, you get yourself into trouble if you then add an C</e>
1766 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1772 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1773 C<${1}000>. The operation of interpolation should not be confused
1774 with the operation of matching a backreference. Certainly they mean two
1775 different things on the I<left> side of the C<s///>.
1777 =head2 Repeated patterns matching zero-length substring
1779 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1781 Regular expressions provide a terse and powerful programming language. As
1782 with most other power tools, power comes together with the ability
1785 A common abuse of this power stems from the ability to make infinite
1786 loops using regular expressions, with something as innocuous as:
1788 'foo' =~ m{ ( o? )* }x;
1790 The C<o?> can match at the beginning of C<'foo'>, and since the position
1791 in the string is not moved by the match, C<o?> would match again and again
1792 because of the C<*> modifier. Another common way to create a similar cycle
1793 is with the looping modifier C<//g>:
1795 @matches = ( 'foo' =~ m{ o? }xg );
1799 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1801 or the loop implied by split().
1803 However, long experience has shown that many programming tasks may
1804 be significantly simplified by using repeated subexpressions that
1805 may match zero-length substrings. Here's a simple example being:
1807 @chars = split //, $string; # // is not magic in split
1808 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1810 Thus Perl allows such constructs, by I<forcefully breaking
1811 the infinite loop>. The rules for this are different for lower-level
1812 loops given by the greedy modifiers C<*+{}>, and for higher-level
1813 ones like the C</g> modifier or split() operator.
1815 The lower-level loops are I<interrupted> (that is, the loop is
1816 broken) when Perl detects that a repeated expression matched a
1817 zero-length substring. Thus
1819 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1821 is made equivalent to
1823 m{ (?: NON_ZERO_LENGTH )*
1828 The higher level-loops preserve an additional state between iterations:
1829 whether the last match was zero-length. To break the loop, the following
1830 match after a zero-length match is prohibited to have a length of zero.
1831 This prohibition interacts with backtracking (see L<"Backtracking">),
1832 and so the I<second best> match is chosen if the I<best> match is of
1840 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1841 match given by non-greedy C<??> is the zero-length match, and the I<second
1842 best> match is what is matched by C<\w>. Thus zero-length matches
1843 alternate with one-character-long matches.
1845 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1846 position one notch further in the string.
1848 The additional state of being I<matched with zero-length> is associated with
1849 the matched string, and is reset by each assignment to pos().
1850 Zero-length matches at the end of the previous match are ignored
1853 =head2 Combining pieces together
1855 Each of the elementary pieces of regular expressions which were described
1856 before (such as C<ab> or C<\Z>) could match at most one substring
1857 at the given position of the input string. However, in a typical regular
1858 expression these elementary pieces are combined into more complicated
1859 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1860 (in these examples C<S> and C<T> are regular subexpressions).
1862 Such combinations can include alternatives, leading to a problem of choice:
1863 if we match a regular expression C<a|ab> against C<"abc">, will it match
1864 substring C<"a"> or C<"ab">? One way to describe which substring is
1865 actually matched is the concept of backtracking (see L<"Backtracking">).
1866 However, this description is too low-level and makes you think
1867 in terms of a particular implementation.
1869 Another description starts with notions of "better"/"worse". All the
1870 substrings which may be matched by the given regular expression can be
1871 sorted from the "best" match to the "worst" match, and it is the "best"
1872 match which is chosen. This substitutes the question of "what is chosen?"
1873 by the question of "which matches are better, and which are worse?".
1875 Again, for elementary pieces there is no such question, since at most
1876 one match at a given position is possible. This section describes the
1877 notion of better/worse for combining operators. In the description
1878 below C<S> and C<T> are regular subexpressions.
1884 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1885 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1886 which can be matched by C<T>.
1888 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1891 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1892 C<B> is better match for C<T> than C<B'>.
1896 When C<S> can match, it is a better match than when only C<T> can match.
1898 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1899 two matches for C<T>.
1901 =item C<S{REPEAT_COUNT}>
1903 Matches as C<SSS...S> (repeated as many times as necessary).
1907 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1909 =item C<S{min,max}?>
1911 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1913 =item C<S?>, C<S*>, C<S+>
1915 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1917 =item C<S??>, C<S*?>, C<S+?>
1919 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1923 Matches the best match for C<S> and only that.
1925 =item C<(?=S)>, C<(?<=S)>
1927 Only the best match for C<S> is considered. (This is important only if
1928 C<S> has capturing parentheses, and backreferences are used somewhere
1929 else in the whole regular expression.)
1931 =item C<(?!S)>, C<(?<!S)>
1933 For this grouping operator there is no need to describe the ordering, since
1934 only whether or not C<S> can match is important.
1936 =item C<(??{ EXPR })>, C<(?PARNO)>
1938 The ordering is the same as for the regular expression which is
1939 the result of EXPR, or the pattern contained by capture buffer PARNO.
1941 =item C<(?(condition)yes-pattern|no-pattern)>
1943 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1944 already determined. The ordering of the matches is the same as for the
1945 chosen subexpression.
1949 The above recipes describe the ordering of matches I<at a given position>.
1950 One more rule is needed to understand how a match is determined for the
1951 whole regular expression: a match at an earlier position is always better
1952 than a match at a later position.
1954 =head2 Creating custom RE engines
1956 Overloaded constants (see L<overload>) provide a simple way to extend
1957 the functionality of the RE engine.
1959 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
1960 matches at boundary between whitespace characters and non-whitespace
1961 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1962 at these positions, so we want to have each C<\Y|> in the place of the
1963 more complicated version. We can create a module C<customre> to do
1971 die "No argument to customre::import allowed" if @_;
1972 overload::constant 'qr' => \&convert;
1975 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1977 # We must also take care of not escaping the legitimate \\Y|
1978 # sequence, hence the presence of '\\' in the conversion rules.
1979 my %rules = ( '\\' => '\\\\',
1980 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
1986 { $rules{$1} or invalid($re,$1) }sgex;
1990 Now C<use customre> enables the new escape in constant regular
1991 expressions, i.e., those without any runtime variable interpolations.
1992 As documented in L<overload>, this conversion will work only over
1993 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
1994 part of this regular expression needs to be converted explicitly
1995 (but only if the special meaning of C<\Y|> should be enabled inside $re):
2000 $re = customre::convert $re;
2003 =head1 PCRE/Python Support
2005 As of Perl 5.10 Perl supports several Python/PCRE specific extensions
2006 to the regex syntax. While Perl programmers are encouraged to use the
2007 Perl specific syntax, the following are legal in Perl 5.10:
2011 =item C<< (?P<NAME>pattern) >>
2013 Define a named capture buffer. Equivalent to C<< (?<NAME>pattern) >>.
2015 =item C<< (?P=NAME) >>
2017 Backreference to a named capture buffer. Equivalent to C<< \g{NAME} >>.
2019 =item C<< (?P>NAME) >>
2021 Subroutine call to a named capture buffer. Equivalent to C<< (?&NAME) >>.
2027 This document varies from difficult to understand to completely
2028 and utterly opaque. The wandering prose riddled with jargon is
2029 hard to fathom in several places.
2031 This document needs a rewrite that separates the tutorial content
2032 from the reference content.
2040 L<perlop/"Regexp Quote-Like Operators">.
2042 L<perlop/"Gory details of parsing quoted constructs">.
2052 I<Mastering Regular Expressions> by Jeffrey Friedl, published
2053 by O'Reilly and Associates.