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
258 \K Keep the stuff left of the \K, don't include it in $&
259 \v Shortcut for (*PRUNE)
260 \V Shortcut for (*SKIP)
262 A C<\w> matches a single alphanumeric character (an alphabetic
263 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
264 to match a string of Perl-identifier characters (which isn't the same
265 as matching an English word). If C<use locale> is in effect, the list
266 of alphabetic characters generated by C<\w> is taken from the current
267 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
268 C<\d>, and C<\D> within character classes, but if you try to use them
269 as endpoints of a range, that's not a range, the "-" is understood
270 literally. If Unicode is in effect, C<\s> matches also "\x{85}",
271 "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about
272 C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general.
273 You can define your own C<\p> and C<\P> properties, see L<perlunicode>.
276 The POSIX character class syntax
281 is also available. Note that the C<[> and C<]> braces are I<literal>;
282 they must always be used within a character class expression.
285 $string =~ /[[:alpha:]]/;
287 # this is not, and will generate a warning:
288 $string =~ /[:alpha:]/;
290 The available classes and their backslash equivalents (if available) are
293 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
294 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
315 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
319 Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
320 also the (very rare) "vertical tabulator", "\ck", chr(11).
324 A Perl extension, see above.
328 For example use C<[:upper:]> to match all the uppercase characters.
329 Note that the C<[]> are part of the C<[::]> construct, not part of the
330 whole character class. For example:
334 matches zero, one, any alphabetic character, and the percentage sign.
336 The following equivalences to Unicode \p{} constructs and equivalent
337 backslash character classes (if available), will hold:
338 X<character class> X<\p> X<\p{}>
340 [[:...:]] \p{...} backslash
358 For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent.
360 If the C<utf8> pragma is not used but the C<locale> pragma is, the
361 classes correlate with the usual isalpha(3) interface (except for
364 The assumedly non-obviously named classes are:
371 Any control character. Usually characters that don't produce output as
372 such but instead control the terminal somehow: for example newline and
373 backspace are control characters. All characters with ord() less than
374 32 are most often classified as control characters (assuming ASCII,
375 the ISO Latin character sets, and Unicode), as is the character with
376 the ord() value of 127 (C<DEL>).
381 Any alphanumeric or punctuation (special) character.
386 Any alphanumeric or punctuation (special) character or the space character.
391 Any punctuation (special) character.
396 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
397 work just fine) it is included for completeness.
401 You can negate the [::] character classes by prefixing the class name
402 with a '^'. This is a Perl extension. For example:
403 X<character class, negation>
405 POSIX traditional Unicode
407 [[:^digit:]] \D \P{IsDigit}
408 [[:^space:]] \S \P{IsSpace}
409 [[:^word:]] \W \P{IsWord}
411 Perl respects the POSIX standard in that POSIX character classes are
412 only supported within a character class. The POSIX character classes
413 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
414 use them will cause an error.
418 Perl defines the following zero-width assertions:
419 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
420 X<regexp, zero-width assertion>
421 X<regular expression, zero-width assertion>
422 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
424 \b Match a word boundary
425 \B Match a non-(word boundary)
426 \A Match only at beginning of string
427 \Z Match only at end of string, or before newline at the end
428 \z Match only at end of string
429 \G Match only at pos() (e.g. at the end-of-match position
432 A word boundary (C<\b>) is a spot between two characters
433 that has a C<\w> on one side of it and a C<\W> on the other side
434 of it (in either order), counting the imaginary characters off the
435 beginning and end of the string as matching a C<\W>. (Within
436 character classes C<\b> represents backspace rather than a word
437 boundary, just as it normally does in any double-quoted string.)
438 The C<\A> and C<\Z> are just like "^" and "$", except that they
439 won't match multiple times when the C</m> modifier is used, while
440 "^" and "$" will match at every internal line boundary. To match
441 the actual end of the string and not ignore an optional trailing
443 X<\b> X<\A> X<\Z> X<\z> X</m>
445 The C<\G> assertion can be used to chain global matches (using
446 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
447 It is also useful when writing C<lex>-like scanners, when you have
448 several patterns that you want to match against consequent substrings
449 of your string, see the previous reference. The actual location
450 where C<\G> will match can also be influenced by using C<pos()> as
451 an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
452 matches is modified somewhat, in that contents to the left of C<\G> is
453 not counted when determining the length of the match. Thus the following
454 will not match forever:
463 It will print 'A' and then terminate, as it considers the match to
464 be zero-width, and thus will not match at the same position twice in a
467 It is worth noting that C<\G> improperly used can result in an infinite
468 loop. Take care when using patterns that include C<\G> in an alternation.
470 =head3 Capture buffers
472 The bracketing construct C<( ... )> creates capture buffers. To
473 refer to the digit'th buffer use \<digit> within the
474 match. Outside the match use "$" instead of "\". (The
475 \<digit> notation works in certain circumstances outside
476 the match. See the warning below about \1 vs $1 for details.)
477 Referring back to another part of the match is called a
479 X<regex, capture buffer> X<regexp, capture buffer>
480 X<regular expression, capture buffer> X<backreference>
482 There is no limit to the number of captured substrings that you may
483 use. However Perl also uses \10, \11, etc. as aliases for \010,
484 \011, etc. (Recall that 0 means octal, so \011 is the character at
485 number 9 in your coded character set; which would be the 10th character,
486 a horizontal tab under ASCII.) Perl resolves this
487 ambiguity by interpreting \10 as a backreference only if at least 10
488 left parentheses have opened before it. Likewise \11 is a
489 backreference only if at least 11 left parentheses have opened
490 before it. And so on. \1 through \9 are always interpreted as
493 X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
494 In order to provide a safer and easier way to construct patterns using
495 backrefs, in Perl 5.10 the C<\g{N}> notation is provided. The curly
496 brackets are optional, however omitting them is less safe as the meaning
497 of the pattern can be changed by text (such as digits) following it.
498 When N is a positive integer the C<\g{N}> notation is exactly equivalent
499 to using normal backreferences. When N is a negative integer then it is
500 a relative backreference referring to the previous N'th capturing group.
501 When the bracket form is used and N is not an integer, it is treated as a
502 reference to a named buffer.
504 Thus C<\g{-1}> refers to the last buffer, C<\g{-2}> refers to the
505 buffer before that. For example:
511 \g{-1} # backref to buffer 3
512 \g{-3} # backref to buffer 1
516 and would match the same as C</(Y) ( (X) \3 \1 )/x>.
518 Additionally, as of Perl 5.10 you may use named capture buffers and named
519 backreferences. The notation is C<< (?<name>...) >> to declare and C<< \k<name> >>
520 to reference. You may also use single quotes instead of angle brackets to quote the
521 name; and you may use the bracketed C<< \g{name} >> back reference syntax.
522 The only difference between named capture buffers and unnamed ones is
523 that multiple buffers may have the same name and that the contents of
524 named capture buffers are available via the C<%+> hash. When multiple
525 groups share the same name C<$+{name}> and C<< \k<name> >> refer to the
526 leftmost defined group, thus it's possible to do things with named capture
527 buffers that would otherwise require C<(??{})> code to accomplish. Named
528 capture buffers are numbered just as normal capture buffers are and may be
529 referenced via the magic numeric variables or via numeric backreferences
534 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
536 /(.)\1/ # find first doubled char
537 and print "'$1' is the first doubled character\n";
539 /(?<char>.)\k<char>/ # ... a different way
540 and print "'$+{char}' is the first doubled character\n";
542 /(?<char>.)\1/ # ... mix and match
543 and print "'$1' is the first doubled character\n";
545 if (/Time: (..):(..):(..)/) { # parse out values
551 Several special variables also refer back to portions of the previous
552 match. C<$+> returns whatever the last bracket match matched.
553 C<$&> returns the entire matched string. (At one point C<$0> did
554 also, but now it returns the name of the program.) C<$`> returns
555 everything before the matched string. C<$'> returns everything
556 after the matched string. And C<$^N> contains whatever was matched by
557 the most-recently closed group (submatch). C<$^N> can be used in
558 extended patterns (see below), for example to assign a submatch to a
560 X<$+> X<$^N> X<$&> X<$`> X<$'>
562 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
563 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
564 until the end of the enclosing block or until the next successful
565 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
566 X<$+> X<$^N> X<$&> X<$`> X<$'>
567 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
570 B<NOTE>: failed matches in Perl do not reset the match variables,
571 which makes it easier to write code that tests for a series of more
572 specific cases and remembers the best match.
574 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
575 C<$'> anywhere in the program, it has to provide them for every
576 pattern match. This may substantially slow your program. Perl
577 uses the same mechanism to produce $1, $2, etc, so you also pay a
578 price for each pattern that contains capturing parentheses. (To
579 avoid this cost while retaining the grouping behaviour, use the
580 extended regular expression C<(?: ... )> instead.) But if you never
581 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
582 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
583 if you can, but if you can't (and some algorithms really appreciate
584 them), once you've used them once, use them at will, because you've
585 already paid the price. As of 5.005, C<$&> is not so costly as the
589 As a workaround for this problem, Perl 5.10 introduces C<${^PREMATCH}>,
590 C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&>
591 and C<$'>, B<except> that they are only guaranteed to be defined after a
592 successful match that was executed with the C</k> (keep-copy) modifier.
593 The use of these variables incurs no global performance penalty, unlike
594 their punctuation char equivalents, however at the trade-off that you
595 have to tell perl when you want to use them.
598 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
599 C<\w>, C<\n>. Unlike some other regular expression languages, there
600 are no backslashed symbols that aren't alphanumeric. So anything
601 that looks like \\, \(, \), \<, \>, \{, or \} is always
602 interpreted as a literal character, not a metacharacter. This was
603 once used in a common idiom to disable or quote the special meanings
604 of regular expression metacharacters in a string that you want to
605 use for a pattern. Simply quote all non-"word" characters:
607 $pattern =~ s/(\W)/\\$1/g;
609 (If C<use locale> is set, then this depends on the current locale.)
610 Today it is more common to use the quotemeta() function or the C<\Q>
611 metaquoting escape sequence to disable all metacharacters' special
614 /$unquoted\Q$quoted\E$unquoted/
616 Beware that if you put literal backslashes (those not inside
617 interpolated variables) between C<\Q> and C<\E>, double-quotish
618 backslash interpolation may lead to confusing results. If you
619 I<need> to use literal backslashes within C<\Q...\E>,
620 consult L<perlop/"Gory details of parsing quoted constructs">.
622 =head2 Extended Patterns
624 Perl also defines a consistent extension syntax for features not
625 found in standard tools like B<awk> and B<lex>. The syntax is a
626 pair of parentheses with a question mark as the first thing within
627 the parentheses. The character after the question mark indicates
630 The stability of these extensions varies widely. Some have been
631 part of the core language for many years. Others are experimental
632 and may change without warning or be completely removed. Check
633 the documentation on an individual feature to verify its current
636 A question mark was chosen for this and for the minimal-matching
637 construct because 1) question marks are rare in older regular
638 expressions, and 2) whenever you see one, you should stop and
639 "question" exactly what is going on. That's psychology...
646 A comment. The text is ignored. If the C</x> modifier enables
647 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
648 the comment as soon as it sees a C<)>, so there is no way to put a literal
651 =item C<(?kimsx-imsx)>
654 One or more embedded pattern-match modifiers, to be turned on (or
655 turned off, if preceded by C<->) for the remainder of the pattern or
656 the remainder of the enclosing pattern group (if any). This is
657 particularly useful for dynamic patterns, such as those read in from a
658 configuration file, read in as an argument, are specified in a table
659 somewhere, etc. Consider the case that some of which want to be case
660 sensitive and some do not. The case insensitive ones need to include
661 merely C<(?i)> at the front of the pattern. For example:
664 if ( /$pattern/i ) { }
668 $pattern = "(?i)foobar";
669 if ( /$pattern/ ) { }
671 These modifiers are restored at the end of the enclosing group. For example,
675 will match a repeated (I<including the case>!) word C<blah> in any
676 case, assuming C<x> modifier, and no C<i> modifier outside this
679 Note that the C<k> modifier is special in that it can only be enabled,
680 not disabled, and that its presence anywhere in a pattern has a global
681 effect. Thus C<(?-k)> and C<(?-k:...)> are meaningless and will warn
682 when executed under C<use warnings>.
687 =item C<(?imsx-imsx:pattern)>
689 This is for clustering, not capturing; it groups subexpressions like
690 "()", but doesn't make backreferences as "()" does. So
692 @fields = split(/\b(?:a|b|c)\b/)
696 @fields = split(/\b(a|b|c)\b/)
698 but doesn't spit out extra fields. It's also cheaper not to capture
699 characters if you don't need to.
701 Any letters between C<?> and C<:> act as flags modifiers as with
702 C<(?imsx-imsx)>. For example,
704 /(?s-i:more.*than).*million/i
706 is equivalent to the more verbose
708 /(?:(?s-i)more.*than).*million/i
710 =item Look-Around Assertions
711 X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
713 Look-around assertions are zero width patterns which match a specific
714 pattern without including it in C<$&>. Positive assertions match when
715 their subpattern matches, negative assertions match when their subpattern
716 fails. Look-behind matches text up to the current match position,
717 look-ahead matches text following the current match position.
722 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
724 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
725 matches a word followed by a tab, without including the tab in C<$&>.
728 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
730 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
731 matches any occurrence of "foo" that isn't followed by "bar". Note
732 however that look-ahead and look-behind are NOT the same thing. You cannot
733 use this for look-behind.
735 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
736 will not do what you want. That's because the C<(?!foo)> is just saying that
737 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
738 match. You would have to do something like C</(?!foo)...bar/> for that. We
739 say "like" because there's the case of your "bar" not having three characters
740 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
741 Sometimes it's still easier just to say:
743 if (/bar/ && $` !~ /foo$/)
745 For look-behind see below.
747 =item C<(?<=pattern)> C<\K>
748 X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K>
750 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
751 matches a word that follows a tab, without including the tab in C<$&>.
752 Works only for fixed-width look-behind.
754 There is a special form of this construct, called C<\K>, which causes the
755 regex engine to "keep" everything it had matched prior to the C<\K> and
756 not include it in C<$&>. This effectively provides variable length
757 look-behind. The use of C<\K> inside of another look-around assertion
758 is allowed, but the behaviour is currently not well defined.
760 For various reasons C<\K> may be signifigantly more efficient than the
761 equivalent C<< (?<=...) >> construct, and it is especially useful in
762 situations where you want to efficiently remove something following
763 something else in a string. For instance
767 can be rewritten as the much more efficient
771 =item C<(?<!pattern)>
772 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
774 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
775 matches any occurrence of "foo" that does not follow "bar". Works
776 only for fixed-width look-behind.
780 =item C<(?'NAME'pattern)>
782 =item C<< (?<NAME>pattern) >>
783 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
785 A named capture buffer. Identical in every respect to normal capturing
786 parens C<()> but for the additional fact that C<%+> may be used after
787 a succesful match to refer to a named buffer. See C<perlvar> for more
788 details on the C<%+> hash.
790 If multiple distinct capture buffers have the same name then the
791 $+{NAME} will refer to the leftmost defined buffer in the match.
793 The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent.
795 B<NOTE:> While the notation of this construct is the same as the similar
796 function in .NET regexes, the behavior is not, in Perl the buffers are
797 numbered sequentially regardless of being named or not. Thus in the
802 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
803 the opposite which is what a .NET regex hacker might expect.
805 Currently NAME is restricted to simple identifiers only.
806 In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
807 its Unicode extension (see L<utf8>),
808 though it isn't extended by the locale (see L<perllocale>).
810 B<NOTE:> In order to make things easier for programmers with experience
811 with the Python or PCRE regex engines the pattern C<< (?PE<lt>NAMEE<gt>pattern) >>
812 maybe be used instead of C<< (?<NAME>pattern) >>; however this form does not
813 support the use of single quotes as a delimiter for the name. This is
814 only available in Perl 5.10 or later.
816 =item C<< \k<NAME> >>
818 =item C<< \k'NAME' >>
820 Named backreference. Similar to numeric backreferences, except that
821 the group is designated by name and not number. If multiple groups
822 have the same name then it refers to the leftmost defined group in
825 It is an error to refer to a name not defined by a C<(?<NAME>)>
826 earlier in the pattern.
828 Both forms are equivalent.
830 B<NOTE:> In order to make things easier for programmers with experience
831 with the Python or PCRE regex engines the pattern C<< (?P=NAME) >>
832 maybe be used instead of C<< \k<NAME> >> in Perl 5.10 or later.
835 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
837 B<WARNING>: This extended regular expression feature is considered
838 experimental, and may be changed without notice. Code executed that
839 has side effects may not perform identically from version to version
840 due to the effect of future optimisations in the regex engine.
842 This zero-width assertion evaluates any embedded Perl code. It
843 always succeeds, and its C<code> is not interpolated. Currently,
844 the rules to determine where the C<code> ends are somewhat convoluted.
846 This feature can be used together with the special variable C<$^N> to
847 capture the results of submatches in variables without having to keep
848 track of the number of nested parentheses. For example:
850 $_ = "The brown fox jumps over the lazy dog";
851 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
852 print "color = $color, animal = $animal\n";
854 Inside the C<(?{...})> block, C<$_> refers to the string the regular
855 expression is matching against. You can also use C<pos()> to know what is
856 the current position of matching within this string.
858 The C<code> is properly scoped in the following sense: If the assertion
859 is backtracked (compare L<"Backtracking">), all changes introduced after
860 C<local>ization are undone, so that
864 (?{ $cnt = 0 }) # Initialize $cnt.
868 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
872 (?{ $res = $cnt }) # On success copy to non-localized
876 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
877 introduced value, because the scopes that restrict C<local> operators
880 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
881 switch. If I<not> used in this way, the result of evaluation of
882 C<code> is put into the special variable C<$^R>. This happens
883 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
884 inside the same regular expression.
886 The assignment to C<$^R> above is properly localized, so the old
887 value of C<$^R> is restored if the assertion is backtracked; compare
890 Due to an unfortunate implementation issue, the Perl code contained in these
891 blocks is treated as a compile time closure that can have seemingly bizarre
892 consequences when used with lexically scoped variables inside of subroutines
893 or loops. There are various workarounds for this, including simply using
894 global variables instead. If you are using this construct and strange results
895 occur then check for the use of lexically scoped variables.
897 For reasons of security, this construct is forbidden if the regular
898 expression involves run-time interpolation of variables, unless the
899 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
900 variables contain results of C<qr//> operator (see
901 L<perlop/"qr/STRING/imosx">).
903 This restriction is because of the wide-spread and remarkably convenient
904 custom of using run-time determined strings as patterns. For example:
910 Before Perl knew how to execute interpolated code within a pattern,
911 this operation was completely safe from a security point of view,
912 although it could raise an exception from an illegal pattern. If
913 you turn on the C<use re 'eval'>, though, it is no longer secure,
914 so you should only do so if you are also using taint checking.
915 Better yet, use the carefully constrained evaluation within a Safe
916 compartment. See L<perlsec> for details about both these mechanisms.
918 Because perl's regex engine is not currently re-entrant, interpolated
919 code may not invoke the regex engine either directly with C<m//> or C<s///>),
920 or indirectly with functions such as C<split>.
922 =item C<(??{ code })>
924 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
926 B<WARNING>: This extended regular expression feature is considered
927 experimental, and may be changed without notice. Code executed that
928 has side effects may not perform identically from version to version
929 due to the effect of future optimisations in the regex engine.
931 This is a "postponed" regular subexpression. The C<code> is evaluated
932 at run time, at the moment this subexpression may match. The result
933 of evaluation is considered as a regular expression and matched as
934 if it were inserted instead of this construct. Note that this means
935 that the contents of capture buffers defined inside an eval'ed pattern
936 are not available outside of the pattern, and vice versa, there is no
937 way for the inner pattern to refer to a capture buffer defined outside.
940 ('a' x 100)=~/(??{'(.)' x 100})/
942 B<will> match, it will B<not> set $1.
944 The C<code> is not interpolated. As before, the rules to determine
945 where the C<code> ends are currently somewhat convoluted.
947 The following pattern matches a parenthesized group:
952 (?> [^()]+ ) # Non-parens without backtracking
954 (??{ $re }) # Group with matching parens
959 See also C<(?PARNO)> for a different, more efficient way to accomplish
962 Because perl's regex engine is not currently re-entrant, delayed
963 code may not invoke the regex engine either directly with C<m//> or C<s///>),
964 or indirectly with functions such as C<split>.
966 Recursing deeper than 50 times without consuming any input string will
967 result in a fatal error. The maximum depth is compiled into perl, so
968 changing it requires a custom build.
970 =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)>
971 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
972 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
973 X<regex, relative recursion>
975 Similar to C<(??{ code })> except it does not involve compiling any code,
976 instead it treats the contents of a capture buffer as an independent
977 pattern that must match at the current position. Capture buffers
978 contained by the pattern will have the value as determined by the
981 PARNO is a sequence of digits (not starting with 0) whose value reflects
982 the paren-number of the capture buffer to recurse to. C<(?R)> recurses to
983 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
984 C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed
985 to be relative, with negative numbers indicating preceding capture buffers
986 and positive ones following. Thus C<(?-1)> refers to the most recently
987 declared buffer, and C<(?+1)> indicates the next buffer to be declared.
988 Note that the counting for relative recursion differs from that of
989 relative backreferences, in that with recursion unclosed buffers B<are>
992 The following pattern matches a function foo() which may contain
993 balanced parentheses as the argument.
995 $re = qr{ ( # paren group 1 (full function)
997 ( # paren group 2 (parens)
999 ( # paren group 3 (contents of parens)
1001 (?> [^()]+ ) # Non-parens without backtracking
1003 (?2) # Recurse to start of paren group 2
1011 If the pattern was used as follows
1013 'foo(bar(baz)+baz(bop))'=~/$re/
1014 and print "\$1 = $1\n",
1018 the output produced should be the following:
1020 $1 = foo(bar(baz)+baz(bop))
1021 $2 = (bar(baz)+baz(bop))
1022 $3 = bar(baz)+baz(bop)
1024 If there is no corresponding capture buffer defined, then it is a
1025 fatal error. Recursing deeper than 50 times without consuming any input
1026 string will also result in a fatal error. The maximum depth is compiled
1027 into perl, so changing it requires a custom build.
1029 The following shows how using negative indexing can make it
1030 easier to embed recursive patterns inside of a C<qr//> construct
1033 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
1034 if (/foo $parens \s+ + \s+ bar $parens/x) {
1035 # do something here...
1038 B<Note> that this pattern does not behave the same way as the equivalent
1039 PCRE or Python construct of the same form. In perl you can backtrack into
1040 a recursed group, in PCRE and Python the recursed into group is treated
1041 as atomic. Also, modifiers are resolved at compile time, so constructs
1042 like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
1048 Recurse to a named subpattern. Identical to (?PARNO) except that the
1049 parenthesis to recurse to is determined by name. If multiple parens have
1050 the same name, then it recurses to the leftmost.
1052 It is an error to refer to a name that is not declared somewhere in the
1055 B<NOTE:> In order to make things easier for programmers with experience
1056 with the Python or PCRE regex engines the pattern C<< (?P>NAME) >>
1057 maybe be used instead of C<< (?&NAME) >> as of Perl 5.10.
1059 =item C<(?(condition)yes-pattern|no-pattern)>
1062 =item C<(?(condition)yes-pattern)>
1064 Conditional expression. C<(condition)> should be either an integer in
1065 parentheses (which is valid if the corresponding pair of parentheses
1066 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
1067 name in angle brackets or single quotes (which is valid if a buffer
1068 with the given name matched), or the special symbol (R) (true when
1069 evaluated inside of recursion or eval). Additionally the R may be
1070 followed by a number, (which will be true when evaluated when recursing
1071 inside of the appropriate group), or by C<&NAME>, in which case it will
1072 be true only when evaluated during recursion in the named group.
1074 Here's a summary of the possible predicates:
1080 Checks if the numbered capturing buffer has matched something.
1082 =item (<NAME>) ('NAME')
1084 Checks if a buffer with the given name has matched something.
1088 Treats the code block as the condition.
1092 Checks if the expression has been evaluated inside of recursion.
1096 Checks if the expression has been evaluated while executing directly
1097 inside of the n-th capture group. This check is the regex equivalent of
1099 if ((caller(0))[3] eq 'subname') { ... }
1101 In other words, it does not check the full recursion stack.
1105 Similar to C<(R1)>, this predicate checks to see if we're executing
1106 directly inside of the leftmost group with a given name (this is the same
1107 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1108 stack, but only the name of the innermost active recursion.
1112 In this case, the yes-pattern is never directly executed, and no
1113 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1114 See below for details.
1125 matches a chunk of non-parentheses, possibly included in parentheses
1128 A special form is the C<(DEFINE)> predicate, which never executes directly
1129 its yes-pattern, and does not allow a no-pattern. This allows to define
1130 subpatterns which will be executed only by using the recursion mechanism.
1131 This way, you can define a set of regular expression rules that can be
1132 bundled into any pattern you choose.
1134 It is recommended that for this usage you put the DEFINE block at the
1135 end of the pattern, and that you name any subpatterns defined within it.
1137 Also, it's worth noting that patterns defined this way probably will
1138 not be as efficient, as the optimiser is not very clever about
1141 An example of how this might be used is as follows:
1143 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
1149 Note that capture buffers matched inside of recursion are not accessible
1150 after the recursion returns, so the extra layer of capturing buffers are
1151 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1152 C<$+{NAME}> would be.
1154 =item C<< (?>pattern) >>
1155 X<backtrack> X<backtracking> X<atomic> X<possessive>
1157 An "independent" subexpression, one which matches the substring
1158 that a I<standalone> C<pattern> would match if anchored at the given
1159 position, and it matches I<nothing other than this substring>. This
1160 construct is useful for optimizations of what would otherwise be
1161 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1162 It may also be useful in places where the "grab all you can, and do not
1163 give anything back" semantic is desirable.
1165 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1166 (anchored at the beginning of string, as above) will match I<all>
1167 characters C<a> at the beginning of string, leaving no C<a> for
1168 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1169 since the match of the subgroup C<a*> is influenced by the following
1170 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1171 C<a*ab> will match fewer characters than a standalone C<a*>, since
1172 this makes the tail match.
1174 An effect similar to C<< (?>pattern) >> may be achieved by writing
1175 C<(?=(pattern))\1>. This matches the same substring as a standalone
1176 C<a+>, and the following C<\1> eats the matched string; it therefore
1177 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1178 (The difference between these two constructs is that the second one
1179 uses a capturing group, thus shifting ordinals of backreferences
1180 in the rest of a regular expression.)
1182 Consider this pattern:
1193 That will efficiently match a nonempty group with matching parentheses
1194 two levels deep or less. However, if there is no such group, it
1195 will take virtually forever on a long string. That's because there
1196 are so many different ways to split a long string into several
1197 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1198 to a subpattern of the above pattern. Consider how the pattern
1199 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1200 seconds, but that each extra letter doubles this time. This
1201 exponential performance will make it appear that your program has
1202 hung. However, a tiny change to this pattern
1206 (?> [^()]+ ) # change x+ above to (?> x+ )
1213 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1214 this yourself would be a productive exercise), but finishes in a fourth
1215 the time when used on a similar string with 1000000 C<a>s. Be aware,
1216 however, that this pattern currently triggers a warning message under
1217 the C<use warnings> pragma or B<-w> switch saying it
1218 C<"matches null string many times in regex">.
1220 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1221 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1222 This was only 4 times slower on a string with 1000000 C<a>s.
1224 The "grab all you can, and do not give anything back" semantic is desirable
1225 in many situations where on the first sight a simple C<()*> looks like
1226 the correct solution. Suppose we parse text with comments being delimited
1227 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1228 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1229 the comment delimiter, because it may "give up" some whitespace if
1230 the remainder of the pattern can be made to match that way. The correct
1231 answer is either one of these:
1236 For example, to grab non-empty comments into $1, one should use either
1239 / (?> \# [ \t]* ) ( .+ ) /x;
1240 / \# [ \t]* ( [^ \t] .* ) /x;
1242 Which one you pick depends on which of these expressions better reflects
1243 the above specification of comments.
1245 In some literature this construct is called "atomic matching" or
1246 "possessive matching".
1248 Possessive quantifiers are equivalent to putting the item they are applied
1249 to inside of one of these constructs. The following equivalences apply:
1251 Quantifier Form Bracketing Form
1252 --------------- ---------------
1256 PAT{min,max}+ (?>PAT{min,max})
1260 =head2 Special Backtracking Control Verbs
1262 B<WARNING:> These patterns are experimental and subject to change or
1263 removal in a future version of perl. Their usage in production code should
1264 be noted to avoid problems during upgrades.
1266 These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1267 otherwise stated the ARG argument is optional; in some cases, it is
1270 Any pattern containing a special backtracking verb that allows an argument
1271 has the special behaviour that when executed it sets the current packages'
1272 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1275 On failure, the C<$REGERROR> variable will be set to the ARG value of the
1276 verb pattern, if the verb was involved in the failure of the match. If the
1277 ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1278 name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1279 none. Also, the C<$REGMARK> variable will be set to FALSE.
1281 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1282 the C<$REGMARK> variable will be set to the name of the last
1283 C<(*MARK:NAME)> pattern executed. See the explanation for the
1284 C<(*MARK:NAME)> verb below for more details.
1286 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1287 and most other regex related variables. They are not local to a scope, nor
1288 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1289 Use C<local> to localize changes to them to a specific scope if necessary.
1291 If a pattern does not contain a special backtracking verb that allows an
1292 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
1296 =item Verbs that take an argument
1300 =item C<(*PRUNE)> C<(*PRUNE:NAME)>
1301 X<(*PRUNE)> X<(*PRUNE:NAME)> X<\v>
1303 This zero-width pattern prunes the backtracking tree at the current point
1304 when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1305 where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1306 A may backtrack as necessary to match. Once it is reached, matching
1307 continues in B, which may also backtrack as necessary; however, should B
1308 not match, then no further backtracking will take place, and the pattern
1309 will fail outright at the current starting position.
1311 As a shortcut, X<\v> is exactly equivalent to C<(*PRUNE)>.
1313 The following example counts all the possible matching strings in a
1314 pattern (without actually matching any of them).
1316 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1317 print "Count=$count\n";
1332 If we add a C<(*PRUNE)> before the count like the following
1334 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
1335 print "Count=$count\n";
1337 we prevent backtracking and find the count of the longest matching
1338 at each matching startpoint like so:
1345 Any number of C<(*PRUNE)> assertions may be used in a pattern.
1347 See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1348 control backtracking. In some cases, the use of C<(*PRUNE)> can be
1349 replaced with a C<< (?>pattern) >> with no functional difference; however,
1350 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1351 C<< (?>pattern) >> alone.
1354 =item C<(*SKIP)> C<(*SKIP:NAME)>
1357 This zero-width pattern is similar to C<(*PRUNE)>, except that on
1358 failure it also signifies that whatever text that was matched leading up
1359 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1360 of this pattern. This effectively means that the regex engine "skips" forward
1361 to this position on failure and tries to match again, (assuming that
1362 there is sufficient room to match).
1364 As a shortcut X<\V> is exactly equivalent to C<(*SKIP)>.
1366 The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1367 C<(*MARK:NAME)> was encountered while matching, then it is that position
1368 which is used as the "skip point". If no C<(*MARK)> of that name was
1369 encountered, then the C<(*SKIP)> operator has no effect. When used
1370 without a name the "skip point" is where the match point was when
1371 executing the (*SKIP) pattern.
1373 Compare the following to the examples in C<(*PRUNE)>, note the string
1376 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
1377 print "Count=$count\n";
1385 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
1386 executed, the next startpoint will be where the cursor was when the
1387 C<(*SKIP)> was executed.
1389 =item C<(*MARK:NAME)> C<(*:NAME)>
1390 X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)>
1392 This zero-width pattern can be used to mark the point reached in a string
1393 when a certain part of the pattern has been successfully matched. This
1394 mark may be given a name. A later C<(*SKIP)> pattern will then skip
1395 forward to that point if backtracked into on failure. Any number of
1396 C<(*MARK)> patterns are allowed, and the NAME portion is optional and may
1399 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1400 can be used to "label" a pattern branch, so that after matching, the
1401 program can determine which branches of the pattern were involved in the
1404 When a match is successful, the C<$REGMARK> variable will be set to the
1405 name of the most recently executed C<(*MARK:NAME)> that was involved
1408 This can be used to determine which branch of a pattern was matched
1409 without using a seperate capture buffer for each branch, which in turn
1410 can result in a performance improvement, as perl cannot optimize
1411 C</(?:(x)|(y)|(z))/> as efficiently as something like
1412 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1414 When a match has failed, and unless another verb has been involved in
1415 failing the match and has provided its own name to use, the C<$REGERROR>
1416 variable will be set to the name of the most recently executed
1419 See C<(*SKIP)> for more details.
1421 As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
1423 =item C<(*THEN)> C<(*THEN:NAME)>
1425 This is similar to the "cut group" operator C<::> from Perl6. Like
1426 C<(*PRUNE)>, this verb always matches, and when backtracked into on
1427 failure, it causes the regex engine to try the next alternation in the
1428 innermost enclosing group (capturing or otherwise).
1430 Its name comes from the observation that this operation combined with the
1431 alternation operator (C<|>) can be used to create what is essentially a
1432 pattern-based if/then/else block:
1434 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
1436 Note that if this operator is used and NOT inside of an alternation then
1437 it acts exactly like the C<(*PRUNE)> operator.
1447 / ( A (*THEN) B | C (*THEN) D ) /
1451 / ( A (*PRUNE) B | C (*PRUNE) D ) /
1453 as after matching the A but failing on the B the C<(*THEN)> verb will
1454 backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
1459 This is the Perl6 "commit pattern" C<< <commit> >> or C<:::>. It's a
1460 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
1461 into on failure it causes the match to fail outright. No further attempts
1462 to find a valid match by advancing the start pointer will occur again.
1465 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
1466 print "Count=$count\n";
1473 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1474 does not match, the regex engine will not try any further matching on the
1479 =item Verbs without an argument
1483 =item C<(*FAIL)> C<(*F)>
1486 This pattern matches nothing and always fails. It can be used to force the
1487 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1488 fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
1490 It is probably useful only when combined with C<(?{})> or C<(??{})>.
1495 B<WARNING:> This feature is highly experimental. It is not recommended
1496 for production code.
1498 This pattern matches nothing and causes the end of successful matching at
1499 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1500 whether there is actually more to match in the string. When inside of a
1501 nested pattern, such as recursion or a dynamically generated subbpattern
1502 via C<(??{})>, only the innermost pattern is ended immediately.
1504 If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are
1505 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1508 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
1510 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
1511 be set. If another branch in the inner parens were matched, such as in the
1512 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
1519 X<backtrack> X<backtracking>
1521 NOTE: This section presents an abstract approximation of regular
1522 expression behavior. For a more rigorous (and complicated) view of
1523 the rules involved in selecting a match among possible alternatives,
1524 see L<Combining pieces together>.
1526 A fundamental feature of regular expression matching involves the
1527 notion called I<backtracking>, which is currently used (when needed)
1528 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
1529 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1530 internally, but the general principle outlined here is valid.
1532 For a regular expression to match, the I<entire> regular expression must
1533 match, not just part of it. So if the beginning of a pattern containing a
1534 quantifier succeeds in a way that causes later parts in the pattern to
1535 fail, the matching engine backs up and recalculates the beginning
1536 part--that's why it's called backtracking.
1538 Here is an example of backtracking: Let's say you want to find the
1539 word following "foo" in the string "Food is on the foo table.":
1541 $_ = "Food is on the foo table.";
1542 if ( /\b(foo)\s+(\w+)/i ) {
1543 print "$2 follows $1.\n";
1546 When the match runs, the first part of the regular expression (C<\b(foo)>)
1547 finds a possible match right at the beginning of the string, and loads up
1548 $1 with "Foo". However, as soon as the matching engine sees that there's
1549 no whitespace following the "Foo" that it had saved in $1, it realizes its
1550 mistake and starts over again one character after where it had the
1551 tentative match. This time it goes all the way until the next occurrence
1552 of "foo". The complete regular expression matches this time, and you get
1553 the expected output of "table follows foo."
1555 Sometimes minimal matching can help a lot. Imagine you'd like to match
1556 everything between "foo" and "bar". Initially, you write something
1559 $_ = "The food is under the bar in the barn.";
1560 if ( /foo(.*)bar/ ) {
1564 Which perhaps unexpectedly yields:
1566 got <d is under the bar in the >
1568 That's because C<.*> was greedy, so you get everything between the
1569 I<first> "foo" and the I<last> "bar". Here it's more effective
1570 to use minimal matching to make sure you get the text between a "foo"
1571 and the first "bar" thereafter.
1573 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1574 got <d is under the >
1576 Here's another example: let's say you'd like to match a number at the end
1577 of a string, and you also want to keep the preceding part of the match.
1580 $_ = "I have 2 numbers: 53147";
1581 if ( /(.*)(\d*)/ ) { # Wrong!
1582 print "Beginning is <$1>, number is <$2>.\n";
1585 That won't work at all, because C<.*> was greedy and gobbled up the
1586 whole string. As C<\d*> can match on an empty string the complete
1587 regular expression matched successfully.
1589 Beginning is <I have 2 numbers: 53147>, number is <>.
1591 Here are some variants, most of which don't work:
1593 $_ = "I have 2 numbers: 53147";
1606 printf "%-12s ", $pat;
1608 print "<$1> <$2>\n";
1614 That will print out:
1616 (.*)(\d*) <I have 2 numbers: 53147> <>
1617 (.*)(\d+) <I have 2 numbers: 5314> <7>
1619 (.*?)(\d+) <I have > <2>
1620 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1621 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1622 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1623 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1625 As you see, this can be a bit tricky. It's important to realize that a
1626 regular expression is merely a set of assertions that gives a definition
1627 of success. There may be 0, 1, or several different ways that the
1628 definition might succeed against a particular string. And if there are
1629 multiple ways it might succeed, you need to understand backtracking to
1630 know which variety of success you will achieve.
1632 When using look-ahead assertions and negations, this can all get even
1633 trickier. Imagine you'd like to find a sequence of non-digits not
1634 followed by "123". You might try to write that as
1637 if ( /^\D*(?!123)/ ) { # Wrong!
1638 print "Yup, no 123 in $_\n";
1641 But that isn't going to match; at least, not the way you're hoping. It
1642 claims that there is no 123 in the string. Here's a clearer picture of
1643 why that pattern matches, contrary to popular expectations:
1648 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1649 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1651 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1652 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1660 You might have expected test 3 to fail because it seems to a more
1661 general purpose version of test 1. The important difference between
1662 them is that test 3 contains a quantifier (C<\D*>) and so can use
1663 backtracking, whereas test 1 will not. What's happening is
1664 that you've asked "Is it true that at the start of $x, following 0 or more
1665 non-digits, you have something that's not 123?" If the pattern matcher had
1666 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1669 The search engine will initially match C<\D*> with "ABC". Then it will
1670 try to match C<(?!123> with "123", which fails. But because
1671 a quantifier (C<\D*>) has been used in the regular expression, the
1672 search engine can backtrack and retry the match differently
1673 in the hope of matching the complete regular expression.
1675 The pattern really, I<really> wants to succeed, so it uses the
1676 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1677 time. Now there's indeed something following "AB" that is not
1678 "123". It's "C123", which suffices.
1680 We can deal with this by using both an assertion and a negation.
1681 We'll say that the first part in $1 must be followed both by a digit
1682 and by something that's not "123". Remember that the look-aheads
1683 are zero-width expressions--they only look, but don't consume any
1684 of the string in their match. So rewriting this way produces what
1685 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1687 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1688 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1692 In other words, the two zero-width assertions next to each other work as though
1693 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1694 matches only if you're at the beginning of the line AND the end of the
1695 line simultaneously. The deeper underlying truth is that juxtaposition in
1696 regular expressions always means AND, except when you write an explicit OR
1697 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1698 although the attempted matches are made at different positions because "a"
1699 is not a zero-width assertion, but a one-width assertion.
1701 B<WARNING>: particularly complicated regular expressions can take
1702 exponential time to solve because of the immense number of possible
1703 ways they can use backtracking to try match. For example, without
1704 internal optimizations done by the regular expression engine, this will
1705 take a painfully long time to run:
1707 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1709 And if you used C<*>'s in the internal groups instead of limiting them
1710 to 0 through 5 matches, then it would take forever--or until you ran
1711 out of stack space. Moreover, these internal optimizations are not
1712 always applicable. For example, if you put C<{0,5}> instead of C<*>
1713 on the external group, no current optimization is applicable, and the
1714 match takes a long time to finish.
1716 A powerful tool for optimizing such beasts is what is known as an
1717 "independent group",
1718 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1719 zero-length look-ahead/look-behind assertions will not backtrack to make
1720 the tail match, since they are in "logical" context: only
1721 whether they match is considered relevant. For an example
1722 where side-effects of look-ahead I<might> have influenced the
1723 following match, see L<C<< (?>pattern) >>>.
1725 =head2 Version 8 Regular Expressions
1726 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1728 In case you're not familiar with the "regular" Version 8 regex
1729 routines, here are the pattern-matching rules not described above.
1731 Any single character matches itself, unless it is a I<metacharacter>
1732 with a special meaning described here or above. You can cause
1733 characters that normally function as metacharacters to be interpreted
1734 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1735 character; "\\" matches a "\"). A series of characters matches that
1736 series of characters in the target string, so the pattern C<blurfl>
1737 would match "blurfl" in the target string.
1739 You can specify a character class, by enclosing a list of characters
1740 in C<[]>, which will match any character from the list. If the
1741 first character after the "[" is "^", the class matches any character not
1742 in the list. Within a list, the "-" character specifies a
1743 range, so that C<a-z> represents all characters between "a" and "z",
1744 inclusive. If you want either "-" or "]" itself to be a member of a
1745 class, put it at the start of the list (possibly after a "^"), or
1746 escape it with a backslash. "-" is also taken literally when it is
1747 at the end of the list, just before the closing "]". (The
1748 following all specify the same class of three characters: C<[-az]>,
1749 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1750 specifies a class containing twenty-six characters, even on EBCDIC-based
1751 character sets.) Also, if you try to use the character
1752 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1753 a range, the "-" is understood literally.
1755 Note also that the whole range idea is rather unportable between
1756 character sets--and even within character sets they may cause results
1757 you probably didn't expect. A sound principle is to use only ranges
1758 that begin from and end at either alphabets of equal case ([a-e],
1759 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1760 spell out the character sets in full.
1762 Characters may be specified using a metacharacter syntax much like that
1763 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1764 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1765 of octal digits, matches the character whose coded character set value
1766 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1767 matches the character whose numeric value is I<nn>. The expression \cI<x>
1768 matches the character control-I<x>. Finally, the "." metacharacter
1769 matches any character except "\n" (unless you use C</s>).
1771 You can specify a series of alternatives for a pattern using "|" to
1772 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1773 or "foe" in the target string (as would C<f(e|i|o)e>). The
1774 first alternative includes everything from the last pattern delimiter
1775 ("(", "[", or the beginning of the pattern) up to the first "|", and
1776 the last alternative contains everything from the last "|" to the next
1777 pattern delimiter. That's why it's common practice to include
1778 alternatives in parentheses: to minimize confusion about where they
1781 Alternatives are tried from left to right, so the first
1782 alternative found for which the entire expression matches, is the one that
1783 is chosen. This means that alternatives are not necessarily greedy. For
1784 example: when matching C<foo|foot> against "barefoot", only the "foo"
1785 part will match, as that is the first alternative tried, and it successfully
1786 matches the target string. (This might not seem important, but it is
1787 important when you are capturing matched text using parentheses.)
1789 Also remember that "|" is interpreted as a literal within square brackets,
1790 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1792 Within a pattern, you may designate subpatterns for later reference
1793 by enclosing them in parentheses, and you may refer back to the
1794 I<n>th subpattern later in the pattern using the metacharacter
1795 \I<n>. Subpatterns are numbered based on the left to right order
1796 of their opening parenthesis. A backreference matches whatever
1797 actually matched the subpattern in the string being examined, not
1798 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1799 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1800 1 matched "0x", even though the rule C<0|0x> could potentially match
1801 the leading 0 in the second number.
1803 =head2 Warning on \1 vs $1
1805 Some people get too used to writing things like:
1807 $pattern =~ s/(\W)/\\\1/g;
1809 This is grandfathered for the RHS of a substitute to avoid shocking the
1810 B<sed> addicts, but it's a dirty habit to get into. That's because in
1811 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1812 the usual double-quoted string means a control-A. The customary Unix
1813 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1814 of doing that, you get yourself into trouble if you then add an C</e>
1817 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1823 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1824 C<${1}000>. The operation of interpolation should not be confused
1825 with the operation of matching a backreference. Certainly they mean two
1826 different things on the I<left> side of the C<s///>.
1828 =head2 Repeated patterns matching zero-length substring
1830 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1832 Regular expressions provide a terse and powerful programming language. As
1833 with most other power tools, power comes together with the ability
1836 A common abuse of this power stems from the ability to make infinite
1837 loops using regular expressions, with something as innocuous as:
1839 'foo' =~ m{ ( o? )* }x;
1841 The C<o?> can match at the beginning of C<'foo'>, and since the position
1842 in the string is not moved by the match, C<o?> would match again and again
1843 because of the C<*> modifier. Another common way to create a similar cycle
1844 is with the looping modifier C<//g>:
1846 @matches = ( 'foo' =~ m{ o? }xg );
1850 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1852 or the loop implied by split().
1854 However, long experience has shown that many programming tasks may
1855 be significantly simplified by using repeated subexpressions that
1856 may match zero-length substrings. Here's a simple example being:
1858 @chars = split //, $string; # // is not magic in split
1859 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1861 Thus Perl allows such constructs, by I<forcefully breaking
1862 the infinite loop>. The rules for this are different for lower-level
1863 loops given by the greedy modifiers C<*+{}>, and for higher-level
1864 ones like the C</g> modifier or split() operator.
1866 The lower-level loops are I<interrupted> (that is, the loop is
1867 broken) when Perl detects that a repeated expression matched a
1868 zero-length substring. Thus
1870 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1872 is made equivalent to
1874 m{ (?: NON_ZERO_LENGTH )*
1879 The higher level-loops preserve an additional state between iterations:
1880 whether the last match was zero-length. To break the loop, the following
1881 match after a zero-length match is prohibited to have a length of zero.
1882 This prohibition interacts with backtracking (see L<"Backtracking">),
1883 and so the I<second best> match is chosen if the I<best> match is of
1891 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1892 match given by non-greedy C<??> is the zero-length match, and the I<second
1893 best> match is what is matched by C<\w>. Thus zero-length matches
1894 alternate with one-character-long matches.
1896 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1897 position one notch further in the string.
1899 The additional state of being I<matched with zero-length> is associated with
1900 the matched string, and is reset by each assignment to pos().
1901 Zero-length matches at the end of the previous match are ignored
1904 =head2 Combining pieces together
1906 Each of the elementary pieces of regular expressions which were described
1907 before (such as C<ab> or C<\Z>) could match at most one substring
1908 at the given position of the input string. However, in a typical regular
1909 expression these elementary pieces are combined into more complicated
1910 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1911 (in these examples C<S> and C<T> are regular subexpressions).
1913 Such combinations can include alternatives, leading to a problem of choice:
1914 if we match a regular expression C<a|ab> against C<"abc">, will it match
1915 substring C<"a"> or C<"ab">? One way to describe which substring is
1916 actually matched is the concept of backtracking (see L<"Backtracking">).
1917 However, this description is too low-level and makes you think
1918 in terms of a particular implementation.
1920 Another description starts with notions of "better"/"worse". All the
1921 substrings which may be matched by the given regular expression can be
1922 sorted from the "best" match to the "worst" match, and it is the "best"
1923 match which is chosen. This substitutes the question of "what is chosen?"
1924 by the question of "which matches are better, and which are worse?".
1926 Again, for elementary pieces there is no such question, since at most
1927 one match at a given position is possible. This section describes the
1928 notion of better/worse for combining operators. In the description
1929 below C<S> and C<T> are regular subexpressions.
1935 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1936 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1937 which can be matched by C<T>.
1939 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1942 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1943 C<B> is better match for C<T> than C<B'>.
1947 When C<S> can match, it is a better match than when only C<T> can match.
1949 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1950 two matches for C<T>.
1952 =item C<S{REPEAT_COUNT}>
1954 Matches as C<SSS...S> (repeated as many times as necessary).
1958 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1960 =item C<S{min,max}?>
1962 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1964 =item C<S?>, C<S*>, C<S+>
1966 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1968 =item C<S??>, C<S*?>, C<S+?>
1970 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1974 Matches the best match for C<S> and only that.
1976 =item C<(?=S)>, C<(?<=S)>
1978 Only the best match for C<S> is considered. (This is important only if
1979 C<S> has capturing parentheses, and backreferences are used somewhere
1980 else in the whole regular expression.)
1982 =item C<(?!S)>, C<(?<!S)>
1984 For this grouping operator there is no need to describe the ordering, since
1985 only whether or not C<S> can match is important.
1987 =item C<(??{ EXPR })>, C<(?PARNO)>
1989 The ordering is the same as for the regular expression which is
1990 the result of EXPR, or the pattern contained by capture buffer PARNO.
1992 =item C<(?(condition)yes-pattern|no-pattern)>
1994 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1995 already determined. The ordering of the matches is the same as for the
1996 chosen subexpression.
2000 The above recipes describe the ordering of matches I<at a given position>.
2001 One more rule is needed to understand how a match is determined for the
2002 whole regular expression: a match at an earlier position is always better
2003 than a match at a later position.
2005 =head2 Creating custom RE engines
2007 Overloaded constants (see L<overload>) provide a simple way to extend
2008 the functionality of the RE engine.
2010 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
2011 matches at boundary between whitespace characters and non-whitespace
2012 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
2013 at these positions, so we want to have each C<\Y|> in the place of the
2014 more complicated version. We can create a module C<customre> to do
2022 die "No argument to customre::import allowed" if @_;
2023 overload::constant 'qr' => \&convert;
2026 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
2028 # We must also take care of not escaping the legitimate \\Y|
2029 # sequence, hence the presence of '\\' in the conversion rules.
2030 my %rules = ( '\\' => '\\\\',
2031 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
2037 { $rules{$1} or invalid($re,$1) }sgex;
2041 Now C<use customre> enables the new escape in constant regular
2042 expressions, i.e., those without any runtime variable interpolations.
2043 As documented in L<overload>, this conversion will work only over
2044 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
2045 part of this regular expression needs to be converted explicitly
2046 (but only if the special meaning of C<\Y|> should be enabled inside $re):
2051 $re = customre::convert $re;
2054 =head1 PCRE/Python Support
2056 As of Perl 5.10 Perl supports several Python/PCRE specific extensions
2057 to the regex syntax. While Perl programmers are encouraged to use the
2058 Perl specific syntax, the following are legal in Perl 5.10:
2062 =item C<< (?PE<lt>NAMEE<gt>pattern) >>
2064 Define a named capture buffer. Equivalent to C<< (?<NAME>pattern) >>.
2066 =item C<< (?P=NAME) >>
2068 Backreference to a named capture buffer. Equivalent to C<< \g{NAME} >>.
2070 =item C<< (?P>NAME) >>
2072 Subroutine call to a named capture buffer. Equivalent to C<< (?&NAME) >>.
2078 This document varies from difficult to understand to completely
2079 and utterly opaque. The wandering prose riddled with jargon is
2080 hard to fathom in several places.
2082 This document needs a rewrite that separates the tutorial content
2083 from the reference content.
2091 L<perlop/"Regexp Quote-Like Operators">.
2093 L<perlop/"Gory details of parsing quoted constructs">.
2103 I<Mastering Regular Expressions> by Jeffrey Friedl, published
2104 by O'Reilly and Associates.