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 \k<name> Named backreference
251 \N{name} Named unicode character, or unicode escape
252 \x12 Hexadecimal escape sequence
253 \x{1234} Long hexadecimal escape sequence
255 A C<\w> matches a single alphanumeric character (an alphabetic
256 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
257 to match a string of Perl-identifier characters (which isn't the same
258 as matching an English word). If C<use locale> is in effect, the list
259 of alphabetic characters generated by C<\w> is taken from the current
260 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
261 C<\d>, and C<\D> within character classes, but if you try to use them
262 as endpoints of a range, that's not a range, the "-" is understood
263 literally. If Unicode is in effect, C<\s> matches also "\x{85}",
264 "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about
265 C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general.
266 You can define your own C<\p> and C<\P> properties, see L<perlunicode>.
269 The POSIX character class syntax
274 is also available. Note that the C<[> and C<]> braces are I<literal>;
275 they must always be used within a character class expression.
278 $string =~ /[[:alpha:]]/;
280 # this is not, and will generate a warning:
281 $string =~ /[:alpha:]/;
283 The available classes and their backslash equivalents (if available) are
286 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
287 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
308 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
312 Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
313 also the (very rare) "vertical tabulator", "\ck", chr(11).
317 A Perl extension, see above.
321 For example use C<[:upper:]> to match all the uppercase characters.
322 Note that the C<[]> are part of the C<[::]> construct, not part of the
323 whole character class. For example:
327 matches zero, one, any alphabetic character, and the percentage sign.
329 The following equivalences to Unicode \p{} constructs and equivalent
330 backslash character classes (if available), will hold:
331 X<character class> X<\p> X<\p{}>
333 [[:...:]] \p{...} backslash
351 For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent.
353 If the C<utf8> pragma is not used but the C<locale> pragma is, the
354 classes correlate with the usual isalpha(3) interface (except for
357 The assumedly non-obviously named classes are:
364 Any control character. Usually characters that don't produce output as
365 such but instead control the terminal somehow: for example newline and
366 backspace are control characters. All characters with ord() less than
367 32 are most often classified as control characters (assuming ASCII,
368 the ISO Latin character sets, and Unicode), as is the character with
369 the ord() value of 127 (C<DEL>).
374 Any alphanumeric or punctuation (special) character.
379 Any alphanumeric or punctuation (special) character or the space character.
384 Any punctuation (special) character.
389 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
390 work just fine) it is included for completeness.
394 You can negate the [::] character classes by prefixing the class name
395 with a '^'. This is a Perl extension. For example:
396 X<character class, negation>
398 POSIX traditional Unicode
400 [[:^digit:]] \D \P{IsDigit}
401 [[:^space:]] \S \P{IsSpace}
402 [[:^word:]] \W \P{IsWord}
404 Perl respects the POSIX standard in that POSIX character classes are
405 only supported within a character class. The POSIX character classes
406 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
407 use them will cause an error.
411 Perl defines the following zero-width assertions:
412 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
413 X<regexp, zero-width assertion>
414 X<regular expression, zero-width assertion>
415 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
417 \b Match a word boundary
418 \B Match a non-(word boundary)
419 \A Match only at beginning of string
420 \Z Match only at end of string, or before newline at the end
421 \z Match only at end of string
422 \G Match only at pos() (e.g. at the end-of-match position
425 A word boundary (C<\b>) is a spot between two characters
426 that has a C<\w> on one side of it and a C<\W> on the other side
427 of it (in either order), counting the imaginary characters off the
428 beginning and end of the string as matching a C<\W>. (Within
429 character classes C<\b> represents backspace rather than a word
430 boundary, just as it normally does in any double-quoted string.)
431 The C<\A> and C<\Z> are just like "^" and "$", except that they
432 won't match multiple times when the C</m> modifier is used, while
433 "^" and "$" will match at every internal line boundary. To match
434 the actual end of the string and not ignore an optional trailing
436 X<\b> X<\A> X<\Z> X<\z> X</m>
438 The C<\G> assertion can be used to chain global matches (using
439 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
440 It is also useful when writing C<lex>-like scanners, when you have
441 several patterns that you want to match against consequent substrings
442 of your string, see the previous reference. The actual location
443 where C<\G> will match can also be influenced by using C<pos()> as
444 an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully
445 supported when anchored to the start of the pattern; while it
446 is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some
447 such uses (C</.\G/g>, for example) currently cause problems, and
448 it is recommended that you avoid such usage for now.
451 =head3 Capture buffers
453 The bracketing construct C<( ... )> creates capture buffers. To
454 refer to the digit'th buffer use \<digit> within the
455 match. Outside the match use "$" instead of "\". (The
456 \<digit> notation works in certain circumstances outside
457 the match. See the warning below about \1 vs $1 for details.)
458 Referring back to another part of the match is called a
460 X<regex, capture buffer> X<regexp, capture buffer>
461 X<regular expression, capture buffer> X<backreference>
463 There is no limit to the number of captured substrings that you may
464 use. However Perl also uses \10, \11, etc. as aliases for \010,
465 \011, etc. (Recall that 0 means octal, so \011 is the character at
466 number 9 in your coded character set; which would be the 10th character,
467 a horizontal tab under ASCII.) Perl resolves this
468 ambiguity by interpreting \10 as a backreference only if at least 10
469 left parentheses have opened before it. Likewise \11 is a
470 backreference only if at least 11 left parentheses have opened
471 before it. And so on. \1 through \9 are always interpreted as
474 Additionally, as of Perl 5.10 you may use named capture buffers and named
475 backreferences. The notation is C<< (?<name>...) >> and C<< \k<name> >>
476 (you may also use single quotes instead of angle brackets to quote the
477 name). The only difference with named capture buffers and unnamed ones is
478 that multiple buffers may have the same name and that the contents of
479 named capture buffers is available via the C<%+> hash. When multiple
480 groups share the same name C<$+{name}> and C<< \k<name> >> refer to the
481 leftmost defined group, thus it's possible to do things with named capture
482 buffers that would otherwise require C<(??{})> code to accomplish. Named
483 capture buffers are numbered just as normal capture buffers are and may be
484 referenced via the magic numeric variables or via numeric backreferences
489 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
491 /(.)\1/ # find first doubled char
492 and print "'$1' is the first doubled character\n";
494 /(?<char>.)\k<char>/ # ... a different way
495 and print "'$+{char}' is the first doubled character\n";
497 /(?<char>.)\1/ # ... mix and match
498 and print "'$1' is the first doubled character\n";
500 if (/Time: (..):(..):(..)/) { # parse out values
506 Several special variables also refer back to portions of the previous
507 match. C<$+> returns whatever the last bracket match matched.
508 C<$&> returns the entire matched string. (At one point C<$0> did
509 also, but now it returns the name of the program.) C<$`> returns
510 everything before the matched string. C<$'> returns everything
511 after the matched string. And C<$^N> contains whatever was matched by
512 the most-recently closed group (submatch). C<$^N> can be used in
513 extended patterns (see below), for example to assign a submatch to a
515 X<$+> X<$^N> X<$&> X<$`> X<$'>
517 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
518 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
519 until the end of the enclosing block or until the next successful
520 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
521 X<$+> X<$^N> X<$&> X<$`> X<$'>
522 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
525 B<NOTE>: failed matches in Perl do not reset the match variables,
526 which makes it easier to write code that tests for a series of more
527 specific cases and remembers the best match.
529 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
530 C<$'> anywhere in the program, it has to provide them for every
531 pattern match. This may substantially slow your program. Perl
532 uses the same mechanism to produce $1, $2, etc, so you also pay a
533 price for each pattern that contains capturing parentheses. (To
534 avoid this cost while retaining the grouping behaviour, use the
535 extended regular expression C<(?: ... )> instead.) But if you never
536 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
537 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
538 if you can, but if you can't (and some algorithms really appreciate
539 them), once you've used them once, use them at will, because you've
540 already paid the price. As of 5.005, C<$&> is not so costly as the
544 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
545 C<\w>, C<\n>. Unlike some other regular expression languages, there
546 are no backslashed symbols that aren't alphanumeric. So anything
547 that looks like \\, \(, \), \<, \>, \{, or \} is always
548 interpreted as a literal character, not a metacharacter. This was
549 once used in a common idiom to disable or quote the special meanings
550 of regular expression metacharacters in a string that you want to
551 use for a pattern. Simply quote all non-"word" characters:
553 $pattern =~ s/(\W)/\\$1/g;
555 (If C<use locale> is set, then this depends on the current locale.)
556 Today it is more common to use the quotemeta() function or the C<\Q>
557 metaquoting escape sequence to disable all metacharacters' special
560 /$unquoted\Q$quoted\E$unquoted/
562 Beware that if you put literal backslashes (those not inside
563 interpolated variables) between C<\Q> and C<\E>, double-quotish
564 backslash interpolation may lead to confusing results. If you
565 I<need> to use literal backslashes within C<\Q...\E>,
566 consult L<perlop/"Gory details of parsing quoted constructs">.
568 =head2 Extended Patterns
570 Perl also defines a consistent extension syntax for features not
571 found in standard tools like B<awk> and B<lex>. The syntax is a
572 pair of parentheses with a question mark as the first thing within
573 the parentheses. The character after the question mark indicates
576 The stability of these extensions varies widely. Some have been
577 part of the core language for many years. Others are experimental
578 and may change without warning or be completely removed. Check
579 the documentation on an individual feature to verify its current
582 A question mark was chosen for this and for the minimal-matching
583 construct because 1) question marks are rare in older regular
584 expressions, and 2) whenever you see one, you should stop and
585 "question" exactly what is going on. That's psychology...
592 A comment. The text is ignored. If the C</x> modifier enables
593 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
594 the comment as soon as it sees a C<)>, so there is no way to put a literal
597 =item C<(?imsx-imsx)>
600 One or more embedded pattern-match modifiers, to be turned on (or
601 turned off, if preceded by C<->) for the remainder of the pattern or
602 the remainder of the enclosing pattern group (if any). This is
603 particularly useful for dynamic patterns, such as those read in from a
604 configuration file, read in as an argument, are specified in a table
605 somewhere, etc. Consider the case that some of which want to be case
606 sensitive and some do not. The case insensitive ones need to include
607 merely C<(?i)> at the front of the pattern. For example:
610 if ( /$pattern/i ) { }
614 $pattern = "(?i)foobar";
615 if ( /$pattern/ ) { }
617 These modifiers are restored at the end of the enclosing group. For example,
621 will match a repeated (I<including the case>!) word C<blah> in any
622 case, assuming C<x> modifier, and no C<i> modifier outside this
628 =item C<(?imsx-imsx:pattern)>
630 This is for clustering, not capturing; it groups subexpressions like
631 "()", but doesn't make backreferences as "()" does. So
633 @fields = split(/\b(?:a|b|c)\b/)
637 @fields = split(/\b(a|b|c)\b/)
639 but doesn't spit out extra fields. It's also cheaper not to capture
640 characters if you don't need to.
642 Any letters between C<?> and C<:> act as flags modifiers as with
643 C<(?imsx-imsx)>. For example,
645 /(?s-i:more.*than).*million/i
647 is equivalent to the more verbose
649 /(?:(?s-i)more.*than).*million/i
652 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
654 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
655 matches a word followed by a tab, without including the tab in C<$&>.
658 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
660 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
661 matches any occurrence of "foo" that isn't followed by "bar". Note
662 however that look-ahead and look-behind are NOT the same thing. You cannot
663 use this for look-behind.
665 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
666 will not do what you want. That's because the C<(?!foo)> is just saying that
667 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
668 match. You would have to do something like C</(?!foo)...bar/> for that. We
669 say "like" because there's the case of your "bar" not having three characters
670 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
671 Sometimes it's still easier just to say:
673 if (/bar/ && $` !~ /foo$/)
675 For look-behind see below.
677 =item C<(?<=pattern)>
678 X<(?<=)> X<look-behind, positive> X<lookbehind, positive>
680 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
681 matches a word that follows a tab, without including the tab in C<$&>.
682 Works only for fixed-width look-behind.
684 =item C<(?<!pattern)>
685 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
687 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
688 matches any occurrence of "foo" that does not follow "bar". Works
689 only for fixed-width look-behind.
691 =item C<(?'NAME'pattern)>
693 =item C<< (?<NAME>pattern) >>
694 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
696 A named capture buffer. Identical in every respect to normal capturing
697 parens C<()> but for the additional fact that C<%+> may be used after
698 a succesful match to refer to a named buffer. See C<perlvar> for more
699 details on the C<%+> hash.
701 If multiple distinct capture buffers have the same name then the
702 $+{NAME} will refer to the leftmost defined buffer in the match.
704 The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent.
706 B<NOTE:> While the notation of this construct is the same as the similar
707 function in .NET regexes, the behavior is not, in Perl the buffers are
708 numbered sequentially regardless of being named or not. Thus in the
713 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
714 the opposite which is what a .NET regex hacker might expect.
716 Currently NAME is restricted to word chars only. In other words, it
717 must match C</^\w+$/>.
719 =item C<< \k<name> >>
721 =item C<< \k'name' >>
723 Named backreference. Similar to numeric backreferences, except that
724 the group is designated by name and not number. If multiple groups
725 have the same name then it refers to the leftmost defined group in
728 It is an error to refer to a name not defined by a C<(?<NAME>)>
729 earlier in the pattern.
731 Both forms are equivalent.
734 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
736 B<WARNING>: This extended regular expression feature is considered
737 experimental, and may be changed without notice. Code executed that
738 has side effects may not perform identically from version to version
739 due to the effect of future optimisations in the regex engine.
741 This zero-width assertion evaluates any embedded Perl code. It
742 always succeeds, and its C<code> is not interpolated. Currently,
743 the rules to determine where the C<code> ends are somewhat convoluted.
745 This feature can be used together with the special variable C<$^N> to
746 capture the results of submatches in variables without having to keep
747 track of the number of nested parentheses. For example:
749 $_ = "The brown fox jumps over the lazy dog";
750 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
751 print "color = $color, animal = $animal\n";
753 Inside the C<(?{...})> block, C<$_> refers to the string the regular
754 expression is matching against. You can also use C<pos()> to know what is
755 the current position of matching within this string.
757 The C<code> is properly scoped in the following sense: If the assertion
758 is backtracked (compare L<"Backtracking">), all changes introduced after
759 C<local>ization are undone, so that
763 (?{ $cnt = 0 }) # Initialize $cnt.
767 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
771 (?{ $res = $cnt }) # On success copy to non-localized
775 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
776 introduced value, because the scopes that restrict C<local> operators
779 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
780 switch. If I<not> used in this way, the result of evaluation of
781 C<code> is put into the special variable C<$^R>. This happens
782 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
783 inside the same regular expression.
785 The assignment to C<$^R> above is properly localized, so the old
786 value of C<$^R> is restored if the assertion is backtracked; compare
789 Due to an unfortunate implementation issue, the Perl code contained in these
790 blocks is treated as a compile time closure that can have seemingly bizarre
791 consequences when used with lexically scoped variables inside of subroutines
792 or loops. There are various workarounds for this, including simply using
793 global variables instead. If you are using this construct and strange results
794 occur then check for the use of lexically scoped variables.
796 For reasons of security, this construct is forbidden if the regular
797 expression involves run-time interpolation of variables, unless the
798 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
799 variables contain results of C<qr//> operator (see
800 L<perlop/"qr/STRING/imosx">).
802 This restriction is because of the wide-spread and remarkably convenient
803 custom of using run-time determined strings as patterns. For example:
809 Before Perl knew how to execute interpolated code within a pattern,
810 this operation was completely safe from a security point of view,
811 although it could raise an exception from an illegal pattern. If
812 you turn on the C<use re 'eval'>, though, it is no longer secure,
813 so you should only do so if you are also using taint checking.
814 Better yet, use the carefully constrained evaluation within a Safe
815 compartment. See L<perlsec> for details about both these mechanisms.
817 Because perl's regex engine is not currently re-entrant, interpolated
818 code may not invoke the regex engine either directly with C<m//> or C<s///>),
819 or indirectly with functions such as C<split>.
821 =item C<(??{ code })>
823 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
825 B<WARNING>: This extended regular expression feature is considered
826 experimental, and may be changed without notice. Code executed that
827 has side effects may not perform identically from version to version
828 due to the effect of future optimisations in the regex engine.
830 This is a "postponed" regular subexpression. The C<code> is evaluated
831 at run time, at the moment this subexpression may match. The result
832 of evaluation is considered as a regular expression and matched as
833 if it were inserted instead of this construct. Note that this means
834 that the contents of capture buffers defined inside an eval'ed pattern
835 are not available outside of the pattern, and vice versa, there is no
836 way for the inner pattern to refer to a capture buffer defined outside.
839 ('a' x 100)=~/(??{'(.)' x 100})/
841 B<will> match, it will B<not> set $1.
843 The C<code> is not interpolated. As before, the rules to determine
844 where the C<code> ends are currently somewhat convoluted.
846 The following pattern matches a parenthesized group:
851 (?> [^()]+ ) # Non-parens without backtracking
853 (??{ $re }) # Group with matching parens
858 See also C<(?PARNO)> for a different, more efficient way to accomplish
861 Because perl's regex engine is not currently re-entrant, delayed
862 code may not invoke the regex engine either directly with C<m//> or C<s///>),
863 or indirectly with functions such as C<split>.
865 Recursing deeper than 50 times without consuming any input string will
866 result in a fatal error. The maximum depth is compiled into perl, so
867 changing it requires a custom build.
869 =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)>
870 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
871 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
872 X<regex, relative recursion>
874 Similar to C<(??{ code })> except it does not involve compiling any code,
875 instead it treats the contents of a capture buffer as an independent
876 pattern that must match at the current position. Capture buffers
877 contained by the pattern will have the value as determined by the
880 PARNO is a sequence of digits (not starting with 0) whose value reflects
881 the paren-number of the capture buffer to recurse to. C<(?R)> recurses to
882 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
883 C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed
884 to be relative, with negative numbers indicating preceding capture buffers
885 and positive ones following. Thus C<(?-1)> refers to the most recently
886 declared buffer, and C<(?+1)> indicates the next buffer to be declared.
888 The following pattern matches a function foo() which may contain
889 balanced parentheses as the argument.
891 $re = qr{ ( # paren group 1 (full function)
893 ( # paren group 2 (parens)
895 ( # paren group 3 (contents of parens)
897 (?> [^()]+ ) # Non-parens without backtracking
899 (?2) # Recurse to start of paren group 2
907 If the pattern was used as follows
909 'foo(bar(baz)+baz(bop))'=~/$re/
910 and print "\$1 = $1\n",
914 the output produced should be the following:
916 $1 = foo(bar(baz)+baz(bop))
917 $2 = (bar(baz)+baz(bop))
918 $3 = bar(baz)+baz(bop)
920 If there is no corresponding capture buffer defined, then it is a
921 fatal error. Recursing deeper than 50 times without consuming any input
922 string will also result in a fatal error. The maximum depth is compiled
923 into perl, so changing it requires a custom build.
925 The following shows how using negative indexing can make it
926 easier to embed recursive patterns inside of a C<qr//> construct
929 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
930 if (/foo $parens \s+ + \s+ bar $parens/x) {
931 # do something here...
934 B<Note> that this pattern does not behave the same way as the equivalent
935 PCRE or Python construct of the same form. In perl you can backtrack into
936 a recursed group, in PCRE and Python the recursed into group is treated
937 as atomic. Also, modifiers are resolved at compile time, so constructs
938 like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
944 Recurse to a named subpattern. Identical to (?PARNO) except that the
945 parenthesis to recurse to is determined by name. If multiple parens have
946 the same name, then it recurses to the leftmost.
948 It is an error to refer to a name that is not declared somewhere in the
951 =item C<(?(condition)yes-pattern|no-pattern)>
954 =item C<(?(condition)yes-pattern)>
956 Conditional expression. C<(condition)> should be either an integer in
957 parentheses (which is valid if the corresponding pair of parentheses
958 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
959 name in angle brackets or single quotes (which is valid if a buffer
960 with the given name matched), or the special symbol (R) (true when
961 evaluated inside of recursion or eval). Additionally the R may be
962 followed by a number, (which will be true when evaluated when recursing
963 inside of the appropriate group), or by C<&NAME>, in which case it will
964 be true only when evaluated during recursion in the named group.
966 Here's a summary of the possible predicates:
972 Checks if the numbered capturing buffer has matched something.
974 =item (<NAME>) ('NAME')
976 Checks if a buffer with the given name has matched something.
980 Treats the code block as the condition.
984 Checks if the expression has been evaluated inside of recursion.
988 Checks if the expression has been evaluated while executing directly
989 inside of the n-th capture group. This check is the regex equivalent of
991 if ((caller(0))[3] eq 'subname') { ... }
993 In other words, it does not check the full recursion stack.
997 Similar to C<(R1)>, this predicate checks to see if we're executing
998 directly inside of the leftmost group with a given name (this is the same
999 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1000 stack, but only the name of the innermost active recursion.
1004 In this case, the yes-pattern is never directly executed, and no
1005 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1006 See below for details.
1017 matches a chunk of non-parentheses, possibly included in parentheses
1020 A special form is the C<(DEFINE)> predicate, which never executes directly
1021 its yes-pattern, and does not allow a no-pattern. This allows to define
1022 subpatterns which will be executed only by using the recursion mechanism.
1023 This way, you can define a set of regular expression rules that can be
1024 bundled into any pattern you choose.
1026 It is recommended that for this usage you put the DEFINE block at the
1027 end of the pattern, and that you name any subpatterns defined within it.
1029 Also, it's worth noting that patterns defined this way probably will
1030 not be as efficient, as the optimiser is not very clever about
1033 An example of how this might be used is as follows:
1035 /(?<NAME>(&NAME_PAT))(?<ADDR>(&ADDRESS_PAT))
1041 Note that capture buffers matched inside of recursion are not accessible
1042 after the recursion returns, so the extra layer of capturing buffers are
1043 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1044 C<$+{NAME}> would be.
1046 =item C<< (?>pattern) >>
1047 X<backtrack> X<backtracking> X<atomic> X<possessive>
1049 An "independent" subexpression, one which matches the substring
1050 that a I<standalone> C<pattern> would match if anchored at the given
1051 position, and it matches I<nothing other than this substring>. This
1052 construct is useful for optimizations of what would otherwise be
1053 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1054 It may also be useful in places where the "grab all you can, and do not
1055 give anything back" semantic is desirable.
1057 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1058 (anchored at the beginning of string, as above) will match I<all>
1059 characters C<a> at the beginning of string, leaving no C<a> for
1060 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1061 since the match of the subgroup C<a*> is influenced by the following
1062 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1063 C<a*ab> will match fewer characters than a standalone C<a*>, since
1064 this makes the tail match.
1066 An effect similar to C<< (?>pattern) >> may be achieved by writing
1067 C<(?=(pattern))\1>. This matches the same substring as a standalone
1068 C<a+>, and the following C<\1> eats the matched string; it therefore
1069 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1070 (The difference between these two constructs is that the second one
1071 uses a capturing group, thus shifting ordinals of backreferences
1072 in the rest of a regular expression.)
1074 Consider this pattern:
1085 That will efficiently match a nonempty group with matching parentheses
1086 two levels deep or less. However, if there is no such group, it
1087 will take virtually forever on a long string. That's because there
1088 are so many different ways to split a long string into several
1089 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1090 to a subpattern of the above pattern. Consider how the pattern
1091 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1092 seconds, but that each extra letter doubles this time. This
1093 exponential performance will make it appear that your program has
1094 hung. However, a tiny change to this pattern
1098 (?> [^()]+ ) # change x+ above to (?> x+ )
1105 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1106 this yourself would be a productive exercise), but finishes in a fourth
1107 the time when used on a similar string with 1000000 C<a>s. Be aware,
1108 however, that this pattern currently triggers a warning message under
1109 the C<use warnings> pragma or B<-w> switch saying it
1110 C<"matches null string many times in regex">.
1112 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1113 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1114 This was only 4 times slower on a string with 1000000 C<a>s.
1116 The "grab all you can, and do not give anything back" semantic is desirable
1117 in many situations where on the first sight a simple C<()*> looks like
1118 the correct solution. Suppose we parse text with comments being delimited
1119 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1120 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1121 the comment delimiter, because it may "give up" some whitespace if
1122 the remainder of the pattern can be made to match that way. The correct
1123 answer is either one of these:
1128 For example, to grab non-empty comments into $1, one should use either
1131 / (?> \# [ \t]* ) ( .+ ) /x;
1132 / \# [ \t]* ( [^ \t] .* ) /x;
1134 Which one you pick depends on which of these expressions better reflects
1135 the above specification of comments.
1137 In some literature this construct is called "atomic matching" or
1138 "possessive matching".
1140 Possessive quantifiers are equivalent to putting the item they are applied
1141 to inside of one of these constructs. The following equivalences apply:
1143 Quantifier Form Bracketing Form
1144 --------------- ---------------
1148 PAT{min,max}+ (?>PAT{min,max})
1152 =head2 Special Backtracking Control Verbs
1154 B<WARNING:> These patterns are experimental and subject to change or
1155 removal in a future version of perl. Their usage in production code should
1156 be noted to avoid problems during upgrades.
1158 These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1159 otherwise stated the ARG argument is optional; in some cases, it is
1162 Any pattern containing a special backtracking verb that allows an argument
1163 has the special behaviour that when executed it sets the current packages'
1164 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1167 On failure, the C<$REGERROR> variable will be set to the ARG value of the
1168 verb pattern, if the verb was involved in the failure of the match. If the
1169 ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1170 name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1171 none. Also, the C<$REGMARK> variable will be set to FALSE.
1173 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1174 the C<$REGMARK> variable will be set to the name of the last
1175 C<(*MARK:NAME)> pattern executed. See the explanation for the
1176 C<(*MARK:NAME)> verb below for more details.
1178 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1179 and most other regex related variables. They are not local to a scope, nor
1180 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1181 Use C<local> to localize changes to them to a specific scope if necessary.
1183 If a pattern does not contain a special backtracking verb that allows an
1184 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
1188 =item Verbs that take an argument
1192 =item C<(*PRUNE)> C<(*PRUNE:NAME)>
1193 X<(*PRUNE)> X<(*PRUNE:NAME)>
1195 This zero-width pattern prunes the backtracking tree at the current point
1196 when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1197 where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1198 A may backtrack as necessary to match. Once it is reached, matching
1199 continues in B, which may also backtrack as necessary; however, should B
1200 not match, then no further backtracking will take place, and the pattern
1201 will fail outright at the current starting position.
1203 The following example counts all the possible matching strings in a
1204 pattern (without actually matching any of them).
1206 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1207 print "Count=$count\n";
1222 If we add a C<(*PRUNE)> before the count like the following
1224 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
1225 print "Count=$count\n";
1227 we prevent backtracking and find the count of the longest matching
1228 at each matching startpoint like so:
1235 Any number of C<(*PRUNE)> assertions may be used in a pattern.
1237 See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1238 control backtracking. In some cases, the use of C<(*PRUNE)> can be
1239 replaced with a C<< (?>pattern) >> with no functional difference; however,
1240 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1241 C<< (?>pattern) >> alone.
1244 =item C<(*SKIP)> C<(*SKIP:NAME)>
1247 This zero-width pattern is similar to C<(*PRUNE)>, except that on
1248 failure it also signifies that whatever text that was matched leading up
1249 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1250 of this pattern. This effectively means that the regex engine "skips" forward
1251 to this position on failure and tries to match again, (assuming that
1252 there is sufficient room to match).
1254 The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1255 C<(*MARK:NAME)> was encountered while matching, then it is that position
1256 which is used as the "skip point". If no C<(*MARK)> of that name was
1257 encountered, then the C<(*SKIP)> operator has no effect. When used
1258 without a name the "skip point" is where the match point was when
1259 executing the (*SKIP) pattern.
1261 Compare the following to the examples in C<(*PRUNE)>, note the string
1264 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
1265 print "Count=$count\n";
1273 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
1274 executed, the next startpoint will be where the cursor was when the
1275 C<(*SKIP)> was executed.
1277 As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
1279 =item C<(*MARK:NAME)> C<(*:NAME)>
1280 X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)>
1282 This zero-width pattern can be used to mark the point reached in a string
1283 when a certain part of the pattern has been successfully matched. This
1284 mark may be given a name. A later C<(*SKIP)> pattern will then skip
1285 forward to that point if backtracked into on failure. Any number of
1286 C<(*MARK)> patterns are allowed, and the NAME portion is optional and may
1289 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1290 can be used to "label" a pattern branch, so that after matching, the
1291 program can determine which branches of the pattern were involved in the
1294 When a match is successful, the C<$REGMARK> variable will be set to the
1295 name of the most recently executed C<(*MARK:NAME)> that was involved
1298 This can be used to determine which branch of a pattern was matched
1299 without using a seperate capture buffer for each branch, which in turn
1300 can result in a performance improvement, as perl cannot optimize
1301 C</(?:(x)|(y)|(z))/> as efficiently as something like
1302 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1304 When a match has failed, and unless another verb has been involved in
1305 failing the match and has provided its own name to use, the C<$REGERROR>
1306 variable will be set to the name of the most recently executed
1309 See C<(*SKIP)> for more details.
1311 =item C<(*THEN)> C<(*THEN:NAME)>
1313 This is similar to the "cut group" operator C<::> from Perl6. Like
1314 C<(*PRUNE)>, this verb always matches, and when backtracked into on
1315 failure, it causes the regex engine to try the next alternation in the
1316 innermost enclosing group (capturing or otherwise).
1318 Its name comes from the observation that this operation combined with the
1319 alternation operator (C<|>) can be used to create what is essentially a
1320 pattern-based if/then/else block:
1322 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
1324 Note that if this operator is used and NOT inside of an alternation then
1325 it acts exactly like the C<(*PRUNE)> operator.
1335 / ( A (*THEN) B | C (*THEN) D ) /
1339 / ( A (*PRUNE) B | C (*PRUNE) D ) /
1341 as after matching the A but failing on the B the C<(*THEN)> verb will
1342 backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
1347 This is the Perl6 "commit pattern" C<< <commit> >> or C<:::>. It's a
1348 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
1349 into on failure it causes the match to fail outright. No further attempts
1350 to find a valid match by advancing the start pointer will occur again.
1353 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
1354 print "Count=$count\n";
1361 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1362 does not match, the regex engine will not try any further matching on the
1367 =item Verbs without an argument
1371 =item C<(*FAIL)> C<(*F)>
1374 This pattern matches nothing and always fails. It can be used to force the
1375 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1376 fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
1378 It is probably useful only when combined with C<(?{})> or C<(??{})>.
1383 B<WARNING:> This feature is highly experimental. It is not recommended
1384 for production code.
1386 This pattern matches nothing and causes the end of successful matching at
1387 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1388 whether there is actually more to match in the string. When inside of a
1389 nested pattern, such as recursion or a dynamically generated subbpattern
1390 via C<(??{})>, only the innermost pattern is ended immediately.
1392 If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are
1393 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1396 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
1398 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
1399 be set. If another branch in the inner parens were matched, such as in the
1400 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
1407 X<backtrack> X<backtracking>
1409 NOTE: This section presents an abstract approximation of regular
1410 expression behavior. For a more rigorous (and complicated) view of
1411 the rules involved in selecting a match among possible alternatives,
1412 see L<Combining pieces together>.
1414 A fundamental feature of regular expression matching involves the
1415 notion called I<backtracking>, which is currently used (when needed)
1416 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
1417 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1418 internally, but the general principle outlined here is valid.
1420 For a regular expression to match, the I<entire> regular expression must
1421 match, not just part of it. So if the beginning of a pattern containing a
1422 quantifier succeeds in a way that causes later parts in the pattern to
1423 fail, the matching engine backs up and recalculates the beginning
1424 part--that's why it's called backtracking.
1426 Here is an example of backtracking: Let's say you want to find the
1427 word following "foo" in the string "Food is on the foo table.":
1429 $_ = "Food is on the foo table.";
1430 if ( /\b(foo)\s+(\w+)/i ) {
1431 print "$2 follows $1.\n";
1434 When the match runs, the first part of the regular expression (C<\b(foo)>)
1435 finds a possible match right at the beginning of the string, and loads up
1436 $1 with "Foo". However, as soon as the matching engine sees that there's
1437 no whitespace following the "Foo" that it had saved in $1, it realizes its
1438 mistake and starts over again one character after where it had the
1439 tentative match. This time it goes all the way until the next occurrence
1440 of "foo". The complete regular expression matches this time, and you get
1441 the expected output of "table follows foo."
1443 Sometimes minimal matching can help a lot. Imagine you'd like to match
1444 everything between "foo" and "bar". Initially, you write something
1447 $_ = "The food is under the bar in the barn.";
1448 if ( /foo(.*)bar/ ) {
1452 Which perhaps unexpectedly yields:
1454 got <d is under the bar in the >
1456 That's because C<.*> was greedy, so you get everything between the
1457 I<first> "foo" and the I<last> "bar". Here it's more effective
1458 to use minimal matching to make sure you get the text between a "foo"
1459 and the first "bar" thereafter.
1461 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1462 got <d is under the >
1464 Here's another example: let's say you'd like to match a number at the end
1465 of a string, and you also want to keep the preceding part of the match.
1468 $_ = "I have 2 numbers: 53147";
1469 if ( /(.*)(\d*)/ ) { # Wrong!
1470 print "Beginning is <$1>, number is <$2>.\n";
1473 That won't work at all, because C<.*> was greedy and gobbled up the
1474 whole string. As C<\d*> can match on an empty string the complete
1475 regular expression matched successfully.
1477 Beginning is <I have 2 numbers: 53147>, number is <>.
1479 Here are some variants, most of which don't work:
1481 $_ = "I have 2 numbers: 53147";
1494 printf "%-12s ", $pat;
1496 print "<$1> <$2>\n";
1502 That will print out:
1504 (.*)(\d*) <I have 2 numbers: 53147> <>
1505 (.*)(\d+) <I have 2 numbers: 5314> <7>
1507 (.*?)(\d+) <I have > <2>
1508 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1509 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1510 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1511 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1513 As you see, this can be a bit tricky. It's important to realize that a
1514 regular expression is merely a set of assertions that gives a definition
1515 of success. There may be 0, 1, or several different ways that the
1516 definition might succeed against a particular string. And if there are
1517 multiple ways it might succeed, you need to understand backtracking to
1518 know which variety of success you will achieve.
1520 When using look-ahead assertions and negations, this can all get even
1521 trickier. Imagine you'd like to find a sequence of non-digits not
1522 followed by "123". You might try to write that as
1525 if ( /^\D*(?!123)/ ) { # Wrong!
1526 print "Yup, no 123 in $_\n";
1529 But that isn't going to match; at least, not the way you're hoping. It
1530 claims that there is no 123 in the string. Here's a clearer picture of
1531 why that pattern matches, contrary to popular expectations:
1536 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1537 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1539 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1540 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1548 You might have expected test 3 to fail because it seems to a more
1549 general purpose version of test 1. The important difference between
1550 them is that test 3 contains a quantifier (C<\D*>) and so can use
1551 backtracking, whereas test 1 will not. What's happening is
1552 that you've asked "Is it true that at the start of $x, following 0 or more
1553 non-digits, you have something that's not 123?" If the pattern matcher had
1554 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1557 The search engine will initially match C<\D*> with "ABC". Then it will
1558 try to match C<(?!123> with "123", which fails. But because
1559 a quantifier (C<\D*>) has been used in the regular expression, the
1560 search engine can backtrack and retry the match differently
1561 in the hope of matching the complete regular expression.
1563 The pattern really, I<really> wants to succeed, so it uses the
1564 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1565 time. Now there's indeed something following "AB" that is not
1566 "123". It's "C123", which suffices.
1568 We can deal with this by using both an assertion and a negation.
1569 We'll say that the first part in $1 must be followed both by a digit
1570 and by something that's not "123". Remember that the look-aheads
1571 are zero-width expressions--they only look, but don't consume any
1572 of the string in their match. So rewriting this way produces what
1573 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1575 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1576 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1580 In other words, the two zero-width assertions next to each other work as though
1581 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1582 matches only if you're at the beginning of the line AND the end of the
1583 line simultaneously. The deeper underlying truth is that juxtaposition in
1584 regular expressions always means AND, except when you write an explicit OR
1585 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1586 although the attempted matches are made at different positions because "a"
1587 is not a zero-width assertion, but a one-width assertion.
1589 B<WARNING>: particularly complicated regular expressions can take
1590 exponential time to solve because of the immense number of possible
1591 ways they can use backtracking to try match. For example, without
1592 internal optimizations done by the regular expression engine, this will
1593 take a painfully long time to run:
1595 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1597 And if you used C<*>'s in the internal groups instead of limiting them
1598 to 0 through 5 matches, then it would take forever--or until you ran
1599 out of stack space. Moreover, these internal optimizations are not
1600 always applicable. For example, if you put C<{0,5}> instead of C<*>
1601 on the external group, no current optimization is applicable, and the
1602 match takes a long time to finish.
1604 A powerful tool for optimizing such beasts is what is known as an
1605 "independent group",
1606 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1607 zero-length look-ahead/look-behind assertions will not backtrack to make
1608 the tail match, since they are in "logical" context: only
1609 whether they match is considered relevant. For an example
1610 where side-effects of look-ahead I<might> have influenced the
1611 following match, see L<C<< (?>pattern) >>>.
1613 =head2 Version 8 Regular Expressions
1614 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1616 In case you're not familiar with the "regular" Version 8 regex
1617 routines, here are the pattern-matching rules not described above.
1619 Any single character matches itself, unless it is a I<metacharacter>
1620 with a special meaning described here or above. You can cause
1621 characters that normally function as metacharacters to be interpreted
1622 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1623 character; "\\" matches a "\"). A series of characters matches that
1624 series of characters in the target string, so the pattern C<blurfl>
1625 would match "blurfl" in the target string.
1627 You can specify a character class, by enclosing a list of characters
1628 in C<[]>, which will match any character from the list. If the
1629 first character after the "[" is "^", the class matches any character not
1630 in the list. Within a list, the "-" character specifies a
1631 range, so that C<a-z> represents all characters between "a" and "z",
1632 inclusive. If you want either "-" or "]" itself to be a member of a
1633 class, put it at the start of the list (possibly after a "^"), or
1634 escape it with a backslash. "-" is also taken literally when it is
1635 at the end of the list, just before the closing "]". (The
1636 following all specify the same class of three characters: C<[-az]>,
1637 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1638 specifies a class containing twenty-six characters, even on EBCDIC-based
1639 character sets.) Also, if you try to use the character
1640 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1641 a range, the "-" is understood literally.
1643 Note also that the whole range idea is rather unportable between
1644 character sets--and even within character sets they may cause results
1645 you probably didn't expect. A sound principle is to use only ranges
1646 that begin from and end at either alphabets of equal case ([a-e],
1647 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1648 spell out the character sets in full.
1650 Characters may be specified using a metacharacter syntax much like that
1651 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1652 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1653 of octal digits, matches the character whose coded character set value
1654 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1655 matches the character whose numeric value is I<nn>. The expression \cI<x>
1656 matches the character control-I<x>. Finally, the "." metacharacter
1657 matches any character except "\n" (unless you use C</s>).
1659 You can specify a series of alternatives for a pattern using "|" to
1660 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1661 or "foe" in the target string (as would C<f(e|i|o)e>). The
1662 first alternative includes everything from the last pattern delimiter
1663 ("(", "[", or the beginning of the pattern) up to the first "|", and
1664 the last alternative contains everything from the last "|" to the next
1665 pattern delimiter. That's why it's common practice to include
1666 alternatives in parentheses: to minimize confusion about where they
1669 Alternatives are tried from left to right, so the first
1670 alternative found for which the entire expression matches, is the one that
1671 is chosen. This means that alternatives are not necessarily greedy. For
1672 example: when matching C<foo|foot> against "barefoot", only the "foo"
1673 part will match, as that is the first alternative tried, and it successfully
1674 matches the target string. (This might not seem important, but it is
1675 important when you are capturing matched text using parentheses.)
1677 Also remember that "|" is interpreted as a literal within square brackets,
1678 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1680 Within a pattern, you may designate subpatterns for later reference
1681 by enclosing them in parentheses, and you may refer back to the
1682 I<n>th subpattern later in the pattern using the metacharacter
1683 \I<n>. Subpatterns are numbered based on the left to right order
1684 of their opening parenthesis. A backreference matches whatever
1685 actually matched the subpattern in the string being examined, not
1686 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1687 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1688 1 matched "0x", even though the rule C<0|0x> could potentially match
1689 the leading 0 in the second number.
1691 =head2 Warning on \1 vs $1
1693 Some people get too used to writing things like:
1695 $pattern =~ s/(\W)/\\\1/g;
1697 This is grandfathered for the RHS of a substitute to avoid shocking the
1698 B<sed> addicts, but it's a dirty habit to get into. That's because in
1699 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1700 the usual double-quoted string means a control-A. The customary Unix
1701 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1702 of doing that, you get yourself into trouble if you then add an C</e>
1705 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1711 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1712 C<${1}000>. The operation of interpolation should not be confused
1713 with the operation of matching a backreference. Certainly they mean two
1714 different things on the I<left> side of the C<s///>.
1716 =head2 Repeated patterns matching zero-length substring
1718 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1720 Regular expressions provide a terse and powerful programming language. As
1721 with most other power tools, power comes together with the ability
1724 A common abuse of this power stems from the ability to make infinite
1725 loops using regular expressions, with something as innocuous as:
1727 'foo' =~ m{ ( o? )* }x;
1729 The C<o?> can match at the beginning of C<'foo'>, and since the position
1730 in the string is not moved by the match, C<o?> would match again and again
1731 because of the C<*> modifier. Another common way to create a similar cycle
1732 is with the looping modifier C<//g>:
1734 @matches = ( 'foo' =~ m{ o? }xg );
1738 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1740 or the loop implied by split().
1742 However, long experience has shown that many programming tasks may
1743 be significantly simplified by using repeated subexpressions that
1744 may match zero-length substrings. Here's a simple example being:
1746 @chars = split //, $string; # // is not magic in split
1747 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1749 Thus Perl allows such constructs, by I<forcefully breaking
1750 the infinite loop>. The rules for this are different for lower-level
1751 loops given by the greedy modifiers C<*+{}>, and for higher-level
1752 ones like the C</g> modifier or split() operator.
1754 The lower-level loops are I<interrupted> (that is, the loop is
1755 broken) when Perl detects that a repeated expression matched a
1756 zero-length substring. Thus
1758 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1760 is made equivalent to
1762 m{ (?: NON_ZERO_LENGTH )*
1767 The higher level-loops preserve an additional state between iterations:
1768 whether the last match was zero-length. To break the loop, the following
1769 match after a zero-length match is prohibited to have a length of zero.
1770 This prohibition interacts with backtracking (see L<"Backtracking">),
1771 and so the I<second best> match is chosen if the I<best> match is of
1779 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1780 match given by non-greedy C<??> is the zero-length match, and the I<second
1781 best> match is what is matched by C<\w>. Thus zero-length matches
1782 alternate with one-character-long matches.
1784 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1785 position one notch further in the string.
1787 The additional state of being I<matched with zero-length> is associated with
1788 the matched string, and is reset by each assignment to pos().
1789 Zero-length matches at the end of the previous match are ignored
1792 =head2 Combining pieces together
1794 Each of the elementary pieces of regular expressions which were described
1795 before (such as C<ab> or C<\Z>) could match at most one substring
1796 at the given position of the input string. However, in a typical regular
1797 expression these elementary pieces are combined into more complicated
1798 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1799 (in these examples C<S> and C<T> are regular subexpressions).
1801 Such combinations can include alternatives, leading to a problem of choice:
1802 if we match a regular expression C<a|ab> against C<"abc">, will it match
1803 substring C<"a"> or C<"ab">? One way to describe which substring is
1804 actually matched is the concept of backtracking (see L<"Backtracking">).
1805 However, this description is too low-level and makes you think
1806 in terms of a particular implementation.
1808 Another description starts with notions of "better"/"worse". All the
1809 substrings which may be matched by the given regular expression can be
1810 sorted from the "best" match to the "worst" match, and it is the "best"
1811 match which is chosen. This substitutes the question of "what is chosen?"
1812 by the question of "which matches are better, and which are worse?".
1814 Again, for elementary pieces there is no such question, since at most
1815 one match at a given position is possible. This section describes the
1816 notion of better/worse for combining operators. In the description
1817 below C<S> and C<T> are regular subexpressions.
1823 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1824 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1825 which can be matched by C<T>.
1827 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1830 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1831 C<B> is better match for C<T> than C<B'>.
1835 When C<S> can match, it is a better match than when only C<T> can match.
1837 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1838 two matches for C<T>.
1840 =item C<S{REPEAT_COUNT}>
1842 Matches as C<SSS...S> (repeated as many times as necessary).
1846 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1848 =item C<S{min,max}?>
1850 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1852 =item C<S?>, C<S*>, C<S+>
1854 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1856 =item C<S??>, C<S*?>, C<S+?>
1858 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1862 Matches the best match for C<S> and only that.
1864 =item C<(?=S)>, C<(?<=S)>
1866 Only the best match for C<S> is considered. (This is important only if
1867 C<S> has capturing parentheses, and backreferences are used somewhere
1868 else in the whole regular expression.)
1870 =item C<(?!S)>, C<(?<!S)>
1872 For this grouping operator there is no need to describe the ordering, since
1873 only whether or not C<S> can match is important.
1875 =item C<(??{ EXPR })>, C<(?PARNO)>
1877 The ordering is the same as for the regular expression which is
1878 the result of EXPR, or the pattern contained by capture buffer PARNO.
1880 =item C<(?(condition)yes-pattern|no-pattern)>
1882 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1883 already determined. The ordering of the matches is the same as for the
1884 chosen subexpression.
1888 The above recipes describe the ordering of matches I<at a given position>.
1889 One more rule is needed to understand how a match is determined for the
1890 whole regular expression: a match at an earlier position is always better
1891 than a match at a later position.
1893 =head2 Creating custom RE engines
1895 Overloaded constants (see L<overload>) provide a simple way to extend
1896 the functionality of the RE engine.
1898 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
1899 matches at boundary between whitespace characters and non-whitespace
1900 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1901 at these positions, so we want to have each C<\Y|> in the place of the
1902 more complicated version. We can create a module C<customre> to do
1910 die "No argument to customre::import allowed" if @_;
1911 overload::constant 'qr' => \&convert;
1914 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1916 # We must also take care of not escaping the legitimate \\Y|
1917 # sequence, hence the presence of '\\' in the conversion rules.
1918 my %rules = ( '\\' => '\\\\',
1919 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
1925 { $rules{$1} or invalid($re,$1) }sgex;
1929 Now C<use customre> enables the new escape in constant regular
1930 expressions, i.e., those without any runtime variable interpolations.
1931 As documented in L<overload>, this conversion will work only over
1932 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
1933 part of this regular expression needs to be converted explicitly
1934 (but only if the special meaning of C<\Y|> should be enabled inside $re):
1939 $re = customre::convert $re;
1944 This document varies from difficult to understand to completely
1945 and utterly opaque. The wandering prose riddled with jargon is
1946 hard to fathom in several places.
1948 This document needs a rewrite that separates the tutorial content
1949 from the reference content.
1957 L<perlop/"Regexp Quote-Like Operators">.
1959 L<perlop/"Gory details of parsing quoted constructs">.
1969 I<Mastering Regular Expressions> by Jeffrey Friedl, published
1970 by O'Reilly and Associates.