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>), that you'll either have to
74 escape them or encode them using octal or hex escapes. Taken together,
75 these features go a long way towards making Perl's regular expressions
76 more readable. Note that you have to be careful not to include the
77 pattern delimiter in the comment--perl has no way of knowing you did
78 not intend to close the pattern early. See the C-comment deletion code
82 =head2 Regular Expressions
84 The patterns used in Perl pattern matching derive from supplied in
85 the Version 8 regex routines. (The routines are derived
86 (distantly) from Henry Spencer's freely redistributable reimplementation
87 of the V8 routines.) See L<Version 8 Regular Expressions> for
90 In particular the following metacharacters have their standard I<egrep>-ish
93 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
96 \ Quote the next metacharacter
97 ^ Match the beginning of the line
98 . Match any character (except newline)
99 $ Match the end of the line (or before newline at the end)
104 By default, the "^" character is guaranteed to match only the
105 beginning of the string, the "$" character only the end (or before the
106 newline at the end), and Perl does certain optimizations with the
107 assumption that the string contains only one line. Embedded newlines
108 will not be matched by "^" or "$". You may, however, wish to treat a
109 string as a multi-line buffer, such that the "^" will match after any
110 newline within the string, and "$" will match before any newline. At the
111 cost of a little more overhead, you can do this by using the /m modifier
112 on the pattern match operator. (Older programs did this by setting C<$*>,
113 but this practice has been removed in perl 5.9.)
116 To simplify multi-line substitutions, the "." character never matches a
117 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
118 the string is a single line--even if it isn't.
121 The following standard quantifiers are recognized:
122 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
124 * Match 0 or more times
125 + Match 1 or more times
127 {n} Match exactly n times
128 {n,} Match at least n times
129 {n,m} Match at least n but not more than m times
131 (If a curly bracket occurs in any other context, it is treated
132 as a regular character. In particular, the lower bound
133 is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+"
134 modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited
135 to integral values less than a preset limit defined when perl is built.
136 This is usually 32766 on the most common platforms. The actual limit can
137 be seen in the error message generated by code such as this:
139 $_ **= $_ , / {$_} / for 2 .. 42;
141 By default, a quantified subpattern is "greedy", that is, it will match as
142 many times as possible (given a particular starting location) while still
143 allowing the rest of the pattern to match. If you want it to match the
144 minimum number of times possible, follow the quantifier with a "?". Note
145 that the meanings don't change, just the "greediness":
146 X<metacharacter> X<greedy> X<greedyness>
147 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
149 *? Match 0 or more times
150 +? Match 1 or more times
152 {n}? Match exactly n times
153 {n,}? Match at least n times
154 {n,m}? Match at least n but not more than m times
156 Because patterns are processed as double quoted strings, the following
158 X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q>
159 X<\0> X<\c> X<\N> X<\x>
165 \a alarm (bell) (BEL)
166 \e escape (think troff) (ESC)
167 \033 octal char (think of a PDP-11)
169 \x{263a} wide hex char (Unicode SMILEY)
172 \l lowercase next char (think vi)
173 \u uppercase next char (think vi)
174 \L lowercase till \E (think vi)
175 \U uppercase till \E (think vi)
176 \E end case modification (think vi)
177 \Q quote (disable) pattern metacharacters till \E
179 If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u>
180 and C<\U> is taken from the current locale. See L<perllocale>. For
181 documentation of C<\N{name}>, see L<charnames>.
183 You cannot include a literal C<$> or C<@> within a C<\Q> sequence.
184 An unescaped C<$> or C<@> interpolates the corresponding variable,
185 while escaping will cause the literal string C<\$> to be matched.
186 You'll need to write something like C<m/\Quser\E\@\Qhost/>.
188 In addition, Perl defines the following:
190 X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C>
191 X<word> X<whitespace>
193 \w Match a "word" character (alphanumeric plus "_")
194 \W Match a non-"word" character
195 \s Match a whitespace character
196 \S Match a non-whitespace character
197 \d Match a digit character
198 \D Match a non-digit character
199 \pP Match P, named property. Use \p{Prop} for longer names.
201 \X Match eXtended Unicode "combining character sequence",
202 equivalent to (?:\PM\pM*)
203 \C Match a single C char (octet) even under Unicode.
204 NOTE: breaks up characters into their UTF-8 bytes,
205 so you may end up with malformed pieces of UTF-8.
206 Unsupported in lookbehind.
208 A C<\w> matches a single alphanumeric character (an alphabetic
209 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
210 to match a string of Perl-identifier characters (which isn't the same
211 as matching an English word). If C<use locale> is in effect, the list
212 of alphabetic characters generated by C<\w> is taken from the current
213 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
214 C<\d>, and C<\D> within character classes, but if you try to use them
215 as endpoints of a range, that's not a range, the "-" is understood
216 literally. If Unicode is in effect, C<\s> matches also "\x{85}",
217 "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about
218 C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general.
219 You can define your own C<\p> and C<\P> properties, see L<perlunicode>.
222 The POSIX character class syntax
227 is also available. The available classes and their backslash
228 equivalents (if available) are as follows:
230 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
231 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
252 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
256 Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
257 also the (very rare) "vertical tabulator", "\ck", chr(11).
261 A Perl extension, see above.
265 For example use C<[:upper:]> to match all the uppercase characters.
266 Note that the C<[]> are part of the C<[::]> construct, not part of the
267 whole character class. For example:
271 matches zero, one, any alphabetic character, and the percentage sign.
273 The following equivalences to Unicode \p{} constructs and equivalent
274 backslash character classes (if available), will hold:
275 X<character class> X<\p> X<\p{}>
277 [:...:] \p{...} backslash
295 For example C<[:lower:]> and C<\p{IsLower}> are equivalent.
297 If the C<utf8> pragma is not used but the C<locale> pragma is, the
298 classes correlate with the usual isalpha(3) interface (except for
301 The assumedly non-obviously named classes are:
308 Any control character. Usually characters that don't produce output as
309 such but instead control the terminal somehow: for example newline and
310 backspace are control characters. All characters with ord() less than
311 32 are most often classified as control characters (assuming ASCII,
312 the ISO Latin character sets, and Unicode), as is the character with
313 the ord() value of 127 (C<DEL>).
318 Any alphanumeric or punctuation (special) character.
323 Any alphanumeric or punctuation (special) character or the space character.
328 Any punctuation (special) character.
333 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
334 work just fine) it is included for completeness.
338 You can negate the [::] character classes by prefixing the class name
339 with a '^'. This is a Perl extension. For example:
340 X<character class, negation>
342 POSIX traditional Unicode
344 [:^digit:] \D \P{IsDigit}
345 [:^space:] \S \P{IsSpace}
346 [:^word:] \W \P{IsWord}
348 Perl respects the POSIX standard in that POSIX character classes are
349 only supported within a character class. The POSIX character classes
350 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
351 use them will cause an error.
353 Perl defines the following zero-width assertions:
354 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
355 X<regexp, zero-width assertion>
356 X<regular expression, zero-width assertion>
357 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
359 \b Match a word boundary
360 \B Match a non-(word boundary)
361 \A Match only at beginning of string
362 \Z Match only at end of string, or before newline at the end
363 \z Match only at end of string
364 \G Match only at pos() (e.g. at the end-of-match position
367 A word boundary (C<\b>) is a spot between two characters
368 that has a C<\w> on one side of it and a C<\W> on the other side
369 of it (in either order), counting the imaginary characters off the
370 beginning and end of the string as matching a C<\W>. (Within
371 character classes C<\b> represents backspace rather than a word
372 boundary, just as it normally does in any double-quoted string.)
373 The C<\A> and C<\Z> are just like "^" and "$", except that they
374 won't match multiple times when the C</m> modifier is used, while
375 "^" and "$" will match at every internal line boundary. To match
376 the actual end of the string and not ignore an optional trailing
378 X<\b> X<\A> X<\Z> X<\z> X</m>
380 The C<\G> assertion can be used to chain global matches (using
381 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
382 It is also useful when writing C<lex>-like scanners, when you have
383 several patterns that you want to match against consequent substrings
384 of your string, see the previous reference. The actual location
385 where C<\G> will match can also be influenced by using C<pos()> as
386 an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully
387 supported when anchored to the start of the pattern; while it
388 is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some
389 such uses (C</.\G/g>, for example) currently cause problems, and
390 it is recommended that you avoid such usage for now.
393 The bracketing construct C<( ... )> creates capture buffers. To
394 refer to the digit'th buffer use \<digit> within the
395 match. Outside the match use "$" instead of "\". (The
396 \<digit> notation works in certain circumstances outside
397 the match. See the warning below about \1 vs $1 for details.)
398 Referring back to another part of the match is called a
400 X<regex, capture buffer> X<regexp, capture buffer>
401 X<regular expression, capture buffer> X<backreference>
403 There is no limit to the number of captured substrings that you may
404 use. However Perl also uses \10, \11, etc. as aliases for \010,
405 \011, etc. (Recall that 0 means octal, so \011 is the character at
406 number 9 in your coded character set; which would be the 10th character,
407 a horizontal tab under ASCII.) Perl resolves this
408 ambiguity by interpreting \10 as a backreference only if at least 10
409 left parentheses have opened before it. Likewise \11 is a
410 backreference only if at least 11 left parentheses have opened
411 before it. And so on. \1 through \9 are always interpreted as
416 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
418 if (/(.)\1/) { # find first doubled char
419 print "'$1' is the first doubled character\n";
422 if (/Time: (..):(..):(..)/) { # parse out values
428 Several special variables also refer back to portions of the previous
429 match. C<$+> returns whatever the last bracket match matched.
430 C<$&> returns the entire matched string. (At one point C<$0> did
431 also, but now it returns the name of the program.) C<$`> returns
432 everything before the matched string. C<$'> returns everything
433 after the matched string. And C<$^N> contains whatever was matched by
434 the most-recently closed group (submatch). C<$^N> can be used in
435 extended patterns (see below), for example to assign a submatch to a
437 X<$+> X<$^N> X<$&> X<$`> X<$'>
439 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
440 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
441 until the end of the enclosing block or until the next successful
442 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
443 X<$+> X<$^N> X<$&> X<$`> X<$'>
444 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
447 B<NOTE>: failed matches in Perl do not reset the match variables,
448 which makes it easier to write code that tests for a series of more
449 specific cases and remembers the best match.
451 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
452 C<$'> anywhere in the program, it has to provide them for every
453 pattern match. This may substantially slow your program. Perl
454 uses the same mechanism to produce $1, $2, etc, so you also pay a
455 price for each pattern that contains capturing parentheses. (To
456 avoid this cost while retaining the grouping behaviour, use the
457 extended regular expression C<(?: ... )> instead.) But if you never
458 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
459 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
460 if you can, but if you can't (and some algorithms really appreciate
461 them), once you've used them once, use them at will, because you've
462 already paid the price. As of 5.005, C<$&> is not so costly as the
466 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
467 C<\w>, C<\n>. Unlike some other regular expression languages, there
468 are no backslashed symbols that aren't alphanumeric. So anything
469 that looks like \\, \(, \), \<, \>, \{, or \} is always
470 interpreted as a literal character, not a metacharacter. This was
471 once used in a common idiom to disable or quote the special meanings
472 of regular expression metacharacters in a string that you want to
473 use for a pattern. Simply quote all non-"word" characters:
475 $pattern =~ s/(\W)/\\$1/g;
477 (If C<use locale> is set, then this depends on the current locale.)
478 Today it is more common to use the quotemeta() function or the C<\Q>
479 metaquoting escape sequence to disable all metacharacters' special
482 /$unquoted\Q$quoted\E$unquoted/
484 Beware that if you put literal backslashes (those not inside
485 interpolated variables) between C<\Q> and C<\E>, double-quotish
486 backslash interpolation may lead to confusing results. If you
487 I<need> to use literal backslashes within C<\Q...\E>,
488 consult L<perlop/"Gory details of parsing quoted constructs">.
490 =head2 Extended Patterns
492 Perl also defines a consistent extension syntax for features not
493 found in standard tools like B<awk> and B<lex>. The syntax is a
494 pair of parentheses with a question mark as the first thing within
495 the parentheses. The character after the question mark indicates
498 The stability of these extensions varies widely. Some have been
499 part of the core language for many years. Others are experimental
500 and may change without warning or be completely removed. Check
501 the documentation on an individual feature to verify its current
504 A question mark was chosen for this and for the minimal-matching
505 construct because 1) question marks are rare in older regular
506 expressions, and 2) whenever you see one, you should stop and
507 "question" exactly what is going on. That's psychology...
514 A comment. The text is ignored. If the C</x> modifier enables
515 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
516 the comment as soon as it sees a C<)>, so there is no way to put a literal
519 =item C<(?imsx-imsx)>
522 One or more embedded pattern-match modifiers, to be turned on (or
523 turned off, if preceded by C<->) for the remainder of the pattern or
524 the remainder of the enclosing pattern group (if any). This is
525 particularly useful for dynamic patterns, such as those read in from a
526 configuration file, read in as an argument, are specified in a table
527 somewhere, etc. Consider the case that some of which want to be case
528 sensitive and some do not. The case insensitive ones need to include
529 merely C<(?i)> at the front of the pattern. For example:
532 if ( /$pattern/i ) { }
536 $pattern = "(?i)foobar";
537 if ( /$pattern/ ) { }
539 These modifiers are restored at the end of the enclosing group. For example,
543 will match a repeated (I<including the case>!) word C<blah> in any
544 case, assuming C<x> modifier, and no C<i> modifier outside this
550 =item C<(?imsx-imsx:pattern)>
552 This is for clustering, not capturing; it groups subexpressions like
553 "()", but doesn't make backreferences as "()" does. So
555 @fields = split(/\b(?:a|b|c)\b/)
559 @fields = split(/\b(a|b|c)\b/)
561 but doesn't spit out extra fields. It's also cheaper not to capture
562 characters if you don't need to.
564 Any letters between C<?> and C<:> act as flags modifiers as with
565 C<(?imsx-imsx)>. For example,
567 /(?s-i:more.*than).*million/i
569 is equivalent to the more verbose
571 /(?:(?s-i)more.*than).*million/i
574 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
576 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
577 matches a word followed by a tab, without including the tab in C<$&>.
580 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
582 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
583 matches any occurrence of "foo" that isn't followed by "bar". Note
584 however that look-ahead and look-behind are NOT the same thing. You cannot
585 use this for look-behind.
587 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
588 will not do what you want. That's because the C<(?!foo)> is just saying that
589 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
590 match. You would have to do something like C</(?!foo)...bar/> for that. We
591 say "like" because there's the case of your "bar" not having three characters
592 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
593 Sometimes it's still easier just to say:
595 if (/bar/ && $` !~ /foo$/)
597 For look-behind see below.
599 =item C<(?<=pattern)>
600 X<(?<=)> X<look-behind, positive> X<lookbehind, positive>
602 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
603 matches a word that follows a tab, without including the tab in C<$&>.
604 Works only for fixed-width look-behind.
606 =item C<(?<!pattern)>
607 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
609 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
610 matches any occurrence of "foo" that does not follow "bar". Works
611 only for fixed-width look-behind.
614 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
616 B<WARNING>: This extended regular expression feature is considered
617 highly experimental, and may be changed or deleted without notice.
619 This zero-width assertion evaluates any embedded Perl code. It
620 always succeeds, and its C<code> is not interpolated. Currently,
621 the rules to determine where the C<code> ends are somewhat convoluted.
623 This feature can be used together with the special variable C<$^N> to
624 capture the results of submatches in variables without having to keep
625 track of the number of nested parentheses. For example:
627 $_ = "The brown fox jumps over the lazy dog";
628 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
629 print "color = $color, animal = $animal\n";
631 Inside the C<(?{...})> block, C<$_> refers to the string the regular
632 expression is matching against. You can also use C<pos()> to know what is
633 the current position of matching within this string.
635 The C<code> is properly scoped in the following sense: If the assertion
636 is backtracked (compare L<"Backtracking">), all changes introduced after
637 C<local>ization are undone, so that
641 (?{ $cnt = 0 }) # Initialize $cnt.
645 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
649 (?{ $res = $cnt }) # On success copy to non-localized
653 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
654 introduced value, because the scopes that restrict C<local> operators
657 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
658 switch. If I<not> used in this way, the result of evaluation of
659 C<code> is put into the special variable C<$^R>. This happens
660 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
661 inside the same regular expression.
663 The assignment to C<$^R> above is properly localized, so the old
664 value of C<$^R> is restored if the assertion is backtracked; compare
667 For reasons of security, this construct is forbidden if the regular
668 expression involves run-time interpolation of variables, unless the
669 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
670 variables contain results of C<qr//> operator (see
671 L<perlop/"qr/STRING/imosx">).
673 This restriction is because of the wide-spread and remarkably convenient
674 custom of using run-time determined strings as patterns. For example:
680 Before Perl knew how to execute interpolated code within a pattern,
681 this operation was completely safe from a security point of view,
682 although it could raise an exception from an illegal pattern. If
683 you turn on the C<use re 'eval'>, though, it is no longer secure,
684 so you should only do so if you are also using taint checking.
685 Better yet, use the carefully constrained evaluation within a Safe
686 compartment. See L<perlsec> for details about both these mechanisms.
688 =item C<(??{ code })>
690 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
691 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
693 B<WARNING>: This extended regular expression feature is considered
694 highly experimental, and may be changed or deleted without notice.
695 A simplified version of the syntax may be introduced for commonly
698 This is a "postponed" regular subexpression. The C<code> is evaluated
699 at run time, at the moment this subexpression may match. The result
700 of evaluation is considered as a regular expression and matched as
701 if it were inserted instead of this construct.
703 The C<code> is not interpolated. As before, the rules to determine
704 where the C<code> ends are currently somewhat convoluted.
706 The following pattern matches a parenthesized group:
711 (?> [^()]+ ) # Non-parens without backtracking
713 (??{ $re }) # Group with matching parens
718 =item C<< (?>pattern) >>
719 X<backtrack> X<backtracking>
721 B<WARNING>: This extended regular expression feature is considered
722 highly experimental, and may be changed or deleted without notice.
724 An "independent" subexpression, one which matches the substring
725 that a I<standalone> C<pattern> would match if anchored at the given
726 position, and it matches I<nothing other than this substring>. This
727 construct is useful for optimizations of what would otherwise be
728 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
729 It may also be useful in places where the "grab all you can, and do not
730 give anything back" semantic is desirable.
732 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
733 (anchored at the beginning of string, as above) will match I<all>
734 characters C<a> at the beginning of string, leaving no C<a> for
735 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
736 since the match of the subgroup C<a*> is influenced by the following
737 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
738 C<a*ab> will match fewer characters than a standalone C<a*>, since
739 this makes the tail match.
741 An effect similar to C<< (?>pattern) >> may be achieved by writing
742 C<(?=(pattern))\1>. This matches the same substring as a standalone
743 C<a+>, and the following C<\1> eats the matched string; it therefore
744 makes a zero-length assertion into an analogue of C<< (?>...) >>.
745 (The difference between these two constructs is that the second one
746 uses a capturing group, thus shifting ordinals of backreferences
747 in the rest of a regular expression.)
749 Consider this pattern:
760 That will efficiently match a nonempty group with matching parentheses
761 two levels deep or less. However, if there is no such group, it
762 will take virtually forever on a long string. That's because there
763 are so many different ways to split a long string into several
764 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
765 to a subpattern of the above pattern. Consider how the pattern
766 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
767 seconds, but that each extra letter doubles this time. This
768 exponential performance will make it appear that your program has
769 hung. However, a tiny change to this pattern
773 (?> [^()]+ ) # change x+ above to (?> x+ )
780 which uses C<< (?>...) >> matches exactly when the one above does (verifying
781 this yourself would be a productive exercise), but finishes in a fourth
782 the time when used on a similar string with 1000000 C<a>s. Be aware,
783 however, that this pattern currently triggers a warning message under
784 the C<use warnings> pragma or B<-w> switch saying it
785 C<"matches null string many times in regex">.
787 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
788 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
789 This was only 4 times slower on a string with 1000000 C<a>s.
791 The "grab all you can, and do not give anything back" semantic is desirable
792 in many situations where on the first sight a simple C<()*> looks like
793 the correct solution. Suppose we parse text with comments being delimited
794 by C<#> followed by some optional (horizontal) whitespace. Contrary to
795 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
796 the comment delimiter, because it may "give up" some whitespace if
797 the remainder of the pattern can be made to match that way. The correct
798 answer is either one of these:
803 For example, to grab non-empty comments into $1, one should use either
806 / (?> \# [ \t]* ) ( .+ ) /x;
807 / \# [ \t]* ( [^ \t] .* ) /x;
809 Which one you pick depends on which of these expressions better reflects
810 the above specification of comments.
812 =item C<(?(condition)yes-pattern|no-pattern)>
815 =item C<(?(condition)yes-pattern)>
817 B<WARNING>: This extended regular expression feature is considered
818 highly experimental, and may be changed or deleted without notice.
820 Conditional expression. C<(condition)> should be either an integer in
821 parentheses (which is valid if the corresponding pair of parentheses
822 matched), or look-ahead/look-behind/evaluate zero-width assertion.
831 matches a chunk of non-parentheses, possibly included in parentheses
837 X<backtrack> X<backtracking>
839 NOTE: This section presents an abstract approximation of regular
840 expression behavior. For a more rigorous (and complicated) view of
841 the rules involved in selecting a match among possible alternatives,
842 see L<Combining pieces together>.
844 A fundamental feature of regular expression matching involves the
845 notion called I<backtracking>, which is currently used (when needed)
846 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
847 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
848 internally, but the general principle outlined here is valid.
850 For a regular expression to match, the I<entire> regular expression must
851 match, not just part of it. So if the beginning of a pattern containing a
852 quantifier succeeds in a way that causes later parts in the pattern to
853 fail, the matching engine backs up and recalculates the beginning
854 part--that's why it's called backtracking.
856 Here is an example of backtracking: Let's say you want to find the
857 word following "foo" in the string "Food is on the foo table.":
859 $_ = "Food is on the foo table.";
860 if ( /\b(foo)\s+(\w+)/i ) {
861 print "$2 follows $1.\n";
864 When the match runs, the first part of the regular expression (C<\b(foo)>)
865 finds a possible match right at the beginning of the string, and loads up
866 $1 with "Foo". However, as soon as the matching engine sees that there's
867 no whitespace following the "Foo" that it had saved in $1, it realizes its
868 mistake and starts over again one character after where it had the
869 tentative match. This time it goes all the way until the next occurrence
870 of "foo". The complete regular expression matches this time, and you get
871 the expected output of "table follows foo."
873 Sometimes minimal matching can help a lot. Imagine you'd like to match
874 everything between "foo" and "bar". Initially, you write something
877 $_ = "The food is under the bar in the barn.";
878 if ( /foo(.*)bar/ ) {
882 Which perhaps unexpectedly yields:
884 got <d is under the bar in the >
886 That's because C<.*> was greedy, so you get everything between the
887 I<first> "foo" and the I<last> "bar". Here it's more effective
888 to use minimal matching to make sure you get the text between a "foo"
889 and the first "bar" thereafter.
891 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
892 got <d is under the >
894 Here's another example: let's say you'd like to match a number at the end
895 of a string, and you also want to keep the preceding part of the match.
898 $_ = "I have 2 numbers: 53147";
899 if ( /(.*)(\d*)/ ) { # Wrong!
900 print "Beginning is <$1>, number is <$2>.\n";
903 That won't work at all, because C<.*> was greedy and gobbled up the
904 whole string. As C<\d*> can match on an empty string the complete
905 regular expression matched successfully.
907 Beginning is <I have 2 numbers: 53147>, number is <>.
909 Here are some variants, most of which don't work:
911 $_ = "I have 2 numbers: 53147";
924 printf "%-12s ", $pat;
934 (.*)(\d*) <I have 2 numbers: 53147> <>
935 (.*)(\d+) <I have 2 numbers: 5314> <7>
937 (.*?)(\d+) <I have > <2>
938 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
939 (.*?)(\d+)$ <I have 2 numbers: > <53147>
940 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
941 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
943 As you see, this can be a bit tricky. It's important to realize that a
944 regular expression is merely a set of assertions that gives a definition
945 of success. There may be 0, 1, or several different ways that the
946 definition might succeed against a particular string. And if there are
947 multiple ways it might succeed, you need to understand backtracking to
948 know which variety of success you will achieve.
950 When using look-ahead assertions and negations, this can all get even
951 trickier. Imagine you'd like to find a sequence of non-digits not
952 followed by "123". You might try to write that as
955 if ( /^\D*(?!123)/ ) { # Wrong!
956 print "Yup, no 123 in $_\n";
959 But that isn't going to match; at least, not the way you're hoping. It
960 claims that there is no 123 in the string. Here's a clearer picture of
961 why that pattern matches, contrary to popular expectations:
966 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
967 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
969 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
970 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
978 You might have expected test 3 to fail because it seems to a more
979 general purpose version of test 1. The important difference between
980 them is that test 3 contains a quantifier (C<\D*>) and so can use
981 backtracking, whereas test 1 will not. What's happening is
982 that you've asked "Is it true that at the start of $x, following 0 or more
983 non-digits, you have something that's not 123?" If the pattern matcher had
984 let C<\D*> expand to "ABC", this would have caused the whole pattern to
987 The search engine will initially match C<\D*> with "ABC". Then it will
988 try to match C<(?!123> with "123", which fails. But because
989 a quantifier (C<\D*>) has been used in the regular expression, the
990 search engine can backtrack and retry the match differently
991 in the hope of matching the complete regular expression.
993 The pattern really, I<really> wants to succeed, so it uses the
994 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
995 time. Now there's indeed something following "AB" that is not
996 "123". It's "C123", which suffices.
998 We can deal with this by using both an assertion and a negation.
999 We'll say that the first part in $1 must be followed both by a digit
1000 and by something that's not "123". Remember that the look-aheads
1001 are zero-width expressions--they only look, but don't consume any
1002 of the string in their match. So rewriting this way produces what
1003 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1005 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1006 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1010 In other words, the two zero-width assertions next to each other work as though
1011 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1012 matches only if you're at the beginning of the line AND the end of the
1013 line simultaneously. The deeper underlying truth is that juxtaposition in
1014 regular expressions always means AND, except when you write an explicit OR
1015 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1016 although the attempted matches are made at different positions because "a"
1017 is not a zero-width assertion, but a one-width assertion.
1019 B<WARNING>: particularly complicated regular expressions can take
1020 exponential time to solve because of the immense number of possible
1021 ways they can use backtracking to try match. For example, without
1022 internal optimizations done by the regular expression engine, this will
1023 take a painfully long time to run:
1025 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1027 And if you used C<*>'s in the internal groups instead of limiting them
1028 to 0 through 5 matches, then it would take forever--or until you ran
1029 out of stack space. Moreover, these internal optimizations are not
1030 always applicable. For example, if you put C<{0,5}> instead of C<*>
1031 on the external group, no current optimization is applicable, and the
1032 match takes a long time to finish.
1034 A powerful tool for optimizing such beasts is what is known as an
1035 "independent group",
1036 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1037 zero-length look-ahead/look-behind assertions will not backtrack to make
1038 the tail match, since they are in "logical" context: only
1039 whether they match is considered relevant. For an example
1040 where side-effects of look-ahead I<might> have influenced the
1041 following match, see L<C<< (?>pattern) >>>.
1043 =head2 Version 8 Regular Expressions
1044 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1046 In case you're not familiar with the "regular" Version 8 regex
1047 routines, here are the pattern-matching rules not described above.
1049 Any single character matches itself, unless it is a I<metacharacter>
1050 with a special meaning described here or above. You can cause
1051 characters that normally function as metacharacters to be interpreted
1052 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1053 character; "\\" matches a "\"). A series of characters matches that
1054 series of characters in the target string, so the pattern C<blurfl>
1055 would match "blurfl" in the target string.
1057 You can specify a character class, by enclosing a list of characters
1058 in C<[]>, which will match any one character from the list. If the
1059 first character after the "[" is "^", the class matches any character not
1060 in the list. Within a list, the "-" character specifies a
1061 range, so that C<a-z> represents all characters between "a" and "z",
1062 inclusive. If you want either "-" or "]" itself to be a member of a
1063 class, put it at the start of the list (possibly after a "^"), or
1064 escape it with a backslash. "-" is also taken literally when it is
1065 at the end of the list, just before the closing "]". (The
1066 following all specify the same class of three characters: C<[-az]>,
1067 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1068 specifies a class containing twenty-six characters, even on EBCDIC
1069 based coded character sets.) Also, if you try to use the character
1070 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1071 a range, that's not a range, the "-" is understood literally.
1073 Note also that the whole range idea is rather unportable between
1074 character sets--and even within character sets they may cause results
1075 you probably didn't expect. A sound principle is to use only ranges
1076 that begin from and end at either alphabets of equal case ([a-e],
1077 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1078 spell out the character sets in full.
1080 Characters may be specified using a metacharacter syntax much like that
1081 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1082 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1083 of octal digits, matches the character whose coded character set value
1084 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1085 matches the character whose numeric value is I<nn>. The expression \cI<x>
1086 matches the character control-I<x>. Finally, the "." metacharacter
1087 matches any character except "\n" (unless you use C</s>).
1089 You can specify a series of alternatives for a pattern using "|" to
1090 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1091 or "foe" in the target string (as would C<f(e|i|o)e>). The
1092 first alternative includes everything from the last pattern delimiter
1093 ("(", "[", or the beginning of the pattern) up to the first "|", and
1094 the last alternative contains everything from the last "|" to the next
1095 pattern delimiter. That's why it's common practice to include
1096 alternatives in parentheses: to minimize confusion about where they
1099 Alternatives are tried from left to right, so the first
1100 alternative found for which the entire expression matches, is the one that
1101 is chosen. This means that alternatives are not necessarily greedy. For
1102 example: when matching C<foo|foot> against "barefoot", only the "foo"
1103 part will match, as that is the first alternative tried, and it successfully
1104 matches the target string. (This might not seem important, but it is
1105 important when you are capturing matched text using parentheses.)
1107 Also remember that "|" is interpreted as a literal within square brackets,
1108 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1110 Within a pattern, you may designate subpatterns for later reference
1111 by enclosing them in parentheses, and you may refer back to the
1112 I<n>th subpattern later in the pattern using the metacharacter
1113 \I<n>. Subpatterns are numbered based on the left to right order
1114 of their opening parenthesis. A backreference matches whatever
1115 actually matched the subpattern in the string being examined, not
1116 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1117 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1118 1 matched "0x", even though the rule C<0|0x> could potentially match
1119 the leading 0 in the second number.
1121 =head2 Warning on \1 vs $1
1123 Some people get too used to writing things like:
1125 $pattern =~ s/(\W)/\\\1/g;
1127 This is grandfathered for the RHS of a substitute to avoid shocking the
1128 B<sed> addicts, but it's a dirty habit to get into. That's because in
1129 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1130 the usual double-quoted string means a control-A. The customary Unix
1131 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1132 of doing that, you get yourself into trouble if you then add an C</e>
1135 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1141 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1142 C<${1}000>. The operation of interpolation should not be confused
1143 with the operation of matching a backreference. Certainly they mean two
1144 different things on the I<left> side of the C<s///>.
1146 =head2 Repeated patterns matching zero-length substring
1148 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1150 Regular expressions provide a terse and powerful programming language. As
1151 with most other power tools, power comes together with the ability
1154 A common abuse of this power stems from the ability to make infinite
1155 loops using regular expressions, with something as innocuous as:
1157 'foo' =~ m{ ( o? )* }x;
1159 The C<o?> can match at the beginning of C<'foo'>, and since the position
1160 in the string is not moved by the match, C<o?> would match again and again
1161 because of the C<*> modifier. Another common way to create a similar cycle
1162 is with the looping modifier C<//g>:
1164 @matches = ( 'foo' =~ m{ o? }xg );
1168 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1170 or the loop implied by split().
1172 However, long experience has shown that many programming tasks may
1173 be significantly simplified by using repeated subexpressions that
1174 may match zero-length substrings. Here's a simple example being:
1176 @chars = split //, $string; # // is not magic in split
1177 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1179 Thus Perl allows such constructs, by I<forcefully breaking
1180 the infinite loop>. The rules for this are different for lower-level
1181 loops given by the greedy modifiers C<*+{}>, and for higher-level
1182 ones like the C</g> modifier or split() operator.
1184 The lower-level loops are I<interrupted> (that is, the loop is
1185 broken) when Perl detects that a repeated expression matched a
1186 zero-length substring. Thus
1188 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1190 is made equivalent to
1192 m{ (?: NON_ZERO_LENGTH )*
1197 The higher level-loops preserve an additional state between iterations:
1198 whether the last match was zero-length. To break the loop, the following
1199 match after a zero-length match is prohibited to have a length of zero.
1200 This prohibition interacts with backtracking (see L<"Backtracking">),
1201 and so the I<second best> match is chosen if the I<best> match is of
1209 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1210 match given by non-greedy C<??> is the zero-length match, and the I<second
1211 best> match is what is matched by C<\w>. Thus zero-length matches
1212 alternate with one-character-long matches.
1214 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1215 position one notch further in the string.
1217 The additional state of being I<matched with zero-length> is associated with
1218 the matched string, and is reset by each assignment to pos().
1219 Zero-length matches at the end of the previous match are ignored
1222 =head2 Combining pieces together
1224 Each of the elementary pieces of regular expressions which were described
1225 before (such as C<ab> or C<\Z>) could match at most one substring
1226 at the given position of the input string. However, in a typical regular
1227 expression these elementary pieces are combined into more complicated
1228 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1229 (in these examples C<S> and C<T> are regular subexpressions).
1231 Such combinations can include alternatives, leading to a problem of choice:
1232 if we match a regular expression C<a|ab> against C<"abc">, will it match
1233 substring C<"a"> or C<"ab">? One way to describe which substring is
1234 actually matched is the concept of backtracking (see L<"Backtracking">).
1235 However, this description is too low-level and makes you think
1236 in terms of a particular implementation.
1238 Another description starts with notions of "better"/"worse". All the
1239 substrings which may be matched by the given regular expression can be
1240 sorted from the "best" match to the "worst" match, and it is the "best"
1241 match which is chosen. This substitutes the question of "what is chosen?"
1242 by the question of "which matches are better, and which are worse?".
1244 Again, for elementary pieces there is no such question, since at most
1245 one match at a given position is possible. This section describes the
1246 notion of better/worse for combining operators. In the description
1247 below C<S> and C<T> are regular subexpressions.
1253 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1254 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1255 which can be matched by C<T>.
1257 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1260 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1261 C<B> is better match for C<T> than C<B'>.
1265 When C<S> can match, it is a better match than when only C<T> can match.
1267 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1268 two matches for C<T>.
1270 =item C<S{REPEAT_COUNT}>
1272 Matches as C<SSS...S> (repeated as many times as necessary).
1276 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1278 =item C<S{min,max}?>
1280 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1282 =item C<S?>, C<S*>, C<S+>
1284 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1286 =item C<S??>, C<S*?>, C<S+?>
1288 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1292 Matches the best match for C<S> and only that.
1294 =item C<(?=S)>, C<(?<=S)>
1296 Only the best match for C<S> is considered. (This is important only if
1297 C<S> has capturing parentheses, and backreferences are used somewhere
1298 else in the whole regular expression.)
1300 =item C<(?!S)>, C<(?<!S)>
1302 For this grouping operator there is no need to describe the ordering, since
1303 only whether or not C<S> can match is important.
1305 =item C<(??{ EXPR })>
1307 The ordering is the same as for the regular expression which is
1310 =item C<(?(condition)yes-pattern|no-pattern)>
1312 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1313 already determined. The ordering of the matches is the same as for the
1314 chosen subexpression.
1318 The above recipes describe the ordering of matches I<at a given position>.
1319 One more rule is needed to understand how a match is determined for the
1320 whole regular expression: a match at an earlier position is always better
1321 than a match at a later position.
1323 =head2 Creating custom RE engines
1325 Overloaded constants (see L<overload>) provide a simple way to extend
1326 the functionality of the RE engine.
1328 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
1329 matches at boundary between whitespace characters and non-whitespace
1330 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1331 at these positions, so we want to have each C<\Y|> in the place of the
1332 more complicated version. We can create a module C<customre> to do
1340 die "No argument to customre::import allowed" if @_;
1341 overload::constant 'qr' => \&convert;
1344 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1346 # We must also take care of not escaping the legitimate \\Y|
1347 # sequence, hence the presence of '\\' in the conversion rules.
1348 my %rules = ( '\\' => '\\\\',
1349 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
1355 { $rules{$1} or invalid($re,$1) }sgex;
1359 Now C<use customre> enables the new escape in constant regular
1360 expressions, i.e., those without any runtime variable interpolations.
1361 As documented in L<overload>, this conversion will work only over
1362 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
1363 part of this regular expression needs to be converted explicitly
1364 (but only if the special meaning of C<\Y|> should be enabled inside $re):
1369 $re = customre::convert $re;
1374 This document varies from difficult to understand to completely
1375 and utterly opaque. The wandering prose riddled with jargon is
1376 hard to fathom in several places.
1378 This document needs a rewrite that separates the tutorial content
1379 from the reference content.
1387 L<perlop/"Regexp Quote-Like Operators">.
1389 L<perlop/"Gory details of parsing quoted constructs">.
1399 I<Mastering Regular Expressions> by Jeffrey Friedl, published
1400 by O'Reilly and Associates.