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 (using blackslashes 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
85 The patterns used in Perl pattern matching derive from supplied in
86 the Version 8 regex routines. (The routines are derived
87 (distantly) from Henry Spencer's freely redistributable reimplementation
88 of the V8 routines.) See L<Version 8 Regular Expressions> for
91 In particular the following metacharacters have their standard I<egrep>-ish
94 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
97 \ Quote the next metacharacter
98 ^ Match the beginning of the line
99 . Match any character (except newline)
100 $ Match the end of the line (or before newline at the end)
105 By default, the "^" character is guaranteed to match only the
106 beginning of the string, the "$" character only the end (or before the
107 newline at the end), and Perl does certain optimizations with the
108 assumption that the string contains only one line. Embedded newlines
109 will not be matched by "^" or "$". You may, however, wish to treat a
110 string as a multi-line buffer, such that the "^" will match after any
111 newline within the string, and "$" will match before any newline. At the
112 cost of a little more overhead, you can do this by using the /m modifier
113 on the pattern match operator. (Older programs did this by setting C<$*>,
114 but this practice has been removed in perl 5.9.)
117 To simplify multi-line substitutions, the "." character never matches a
118 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
119 the string is a single line--even if it isn't.
122 The following standard quantifiers are recognized:
123 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
125 * Match 0 or more times
126 + Match 1 or more times
128 {n} Match exactly n times
129 {n,} Match at least n times
130 {n,m} Match at least n but not more than m times
132 (If a curly bracket occurs in any other context, it is treated
133 as a regular character. In particular, the lower bound
134 is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+"
135 modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited
136 to integral values less than a preset limit defined when perl is built.
137 This is usually 32766 on the most common platforms. The actual limit can
138 be seen in the error message generated by code such as this:
140 $_ **= $_ , / {$_} / for 2 .. 42;
142 By default, a quantified subpattern is "greedy", that is, it will match as
143 many times as possible (given a particular starting location) while still
144 allowing the rest of the pattern to match. If you want it to match the
145 minimum number of times possible, follow the quantifier with a "?". Note
146 that the meanings don't change, just the "greediness":
147 X<metacharacter> X<greedy> X<greedyness>
148 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
150 *? Match 0 or more times
151 +? Match 1 or more times
153 {n}? Match exactly n times
154 {n,}? Match at least n times
155 {n,m}? Match at least n but not more than m times
157 Because patterns are processed as double quoted strings, the following
159 X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q>
160 X<\0> X<\c> X<\N> X<\x>
166 \a alarm (bell) (BEL)
167 \e escape (think troff) (ESC)
168 \033 octal char (think of a PDP-11)
170 \x{263a} wide hex char (Unicode SMILEY)
173 \l lowercase next char (think vi)
174 \u uppercase next char (think vi)
175 \L lowercase till \E (think vi)
176 \U uppercase till \E (think vi)
177 \E end case modification (think vi)
178 \Q quote (disable) pattern metacharacters till \E
180 If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u>
181 and C<\U> is taken from the current locale. See L<perllocale>. For
182 documentation of C<\N{name}>, see L<charnames>.
184 You cannot include a literal C<$> or C<@> within a C<\Q> sequence.
185 An unescaped C<$> or C<@> interpolates the corresponding variable,
186 while escaping will cause the literal string C<\$> to be matched.
187 You'll need to write something like C<m/\Quser\E\@\Qhost/>.
189 In addition, Perl defines the following:
191 X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C>
192 X<word> X<whitespace>
194 \w Match a "word" character (alphanumeric plus "_")
195 \W Match a non-"word" character
196 \s Match a whitespace character
197 \S Match a non-whitespace character
198 \d Match a digit character
199 \D Match a non-digit character
200 \pP Match P, named property. Use \p{Prop} for longer names.
202 \X Match eXtended Unicode "combining character sequence",
203 equivalent to (?:\PM\pM*)
204 \C Match a single C char (octet) even under Unicode.
205 NOTE: breaks up characters into their UTF-8 bytes,
206 so you may end up with malformed pieces of UTF-8.
207 Unsupported in lookbehind.
209 A C<\w> matches a single alphanumeric character (an alphabetic
210 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
211 to match a string of Perl-identifier characters (which isn't the same
212 as matching an English word). If C<use locale> is in effect, the list
213 of alphabetic characters generated by C<\w> is taken from the current
214 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
215 C<\d>, and C<\D> within character classes, but if you try to use them
216 as endpoints of a range, that's not a range, the "-" is understood
217 literally. If Unicode is in effect, C<\s> matches also "\x{85}",
218 "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about
219 C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general.
220 You can define your own C<\p> and C<\P> properties, see L<perlunicode>.
223 The POSIX character class syntax
228 is also available. Note that the C<[> and C<]> braces are I<literal>;
229 they must always be used within a character class expression.
232 $string =~ /[[:alpha:]]/;
234 # this is not, and will generate a warning:
235 $string =~ /[:alpha:]/;
237 The available classes and their backslash equivalents (if available) are
240 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
241 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
262 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
266 Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
267 also the (very rare) "vertical tabulator", "\ck", chr(11).
271 A Perl extension, see above.
275 For example use C<[:upper:]> to match all the uppercase characters.
276 Note that the C<[]> are part of the C<[::]> construct, not part of the
277 whole character class. For example:
281 matches zero, one, any alphabetic character, and the percentage sign.
283 The following equivalences to Unicode \p{} constructs and equivalent
284 backslash character classes (if available), will hold:
285 X<character class> X<\p> X<\p{}>
287 [[:...:]] \p{...} backslash
305 For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent.
307 If the C<utf8> pragma is not used but the C<locale> pragma is, the
308 classes correlate with the usual isalpha(3) interface (except for
311 The assumedly non-obviously named classes are:
318 Any control character. Usually characters that don't produce output as
319 such but instead control the terminal somehow: for example newline and
320 backspace are control characters. All characters with ord() less than
321 32 are most often classified as control characters (assuming ASCII,
322 the ISO Latin character sets, and Unicode), as is the character with
323 the ord() value of 127 (C<DEL>).
328 Any alphanumeric or punctuation (special) character.
333 Any alphanumeric or punctuation (special) character or the space character.
338 Any punctuation (special) character.
343 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
344 work just fine) it is included for completeness.
348 You can negate the [::] character classes by prefixing the class name
349 with a '^'. This is a Perl extension. For example:
350 X<character class, negation>
352 POSIX traditional Unicode
354 [[:^digit:]] \D \P{IsDigit}
355 [[:^space:]] \S \P{IsSpace}
356 [[:^word:]] \W \P{IsWord}
358 Perl respects the POSIX standard in that POSIX character classes are
359 only supported within a character class. The POSIX character classes
360 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
361 use them will cause an error.
363 Perl defines the following zero-width assertions:
364 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
365 X<regexp, zero-width assertion>
366 X<regular expression, zero-width assertion>
367 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
369 \b Match a word boundary
370 \B Match a non-(word boundary)
371 \A Match only at beginning of string
372 \Z Match only at end of string, or before newline at the end
373 \z Match only at end of string
374 \G Match only at pos() (e.g. at the end-of-match position
377 A word boundary (C<\b>) is a spot between two characters
378 that has a C<\w> on one side of it and a C<\W> on the other side
379 of it (in either order), counting the imaginary characters off the
380 beginning and end of the string as matching a C<\W>. (Within
381 character classes C<\b> represents backspace rather than a word
382 boundary, just as it normally does in any double-quoted string.)
383 The C<\A> and C<\Z> are just like "^" and "$", except that they
384 won't match multiple times when the C</m> modifier is used, while
385 "^" and "$" will match at every internal line boundary. To match
386 the actual end of the string and not ignore an optional trailing
388 X<\b> X<\A> X<\Z> X<\z> X</m>
390 The C<\G> assertion can be used to chain global matches (using
391 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
392 It is also useful when writing C<lex>-like scanners, when you have
393 several patterns that you want to match against consequent substrings
394 of your string, see the previous reference. The actual location
395 where C<\G> will match can also be influenced by using C<pos()> as
396 an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully
397 supported when anchored to the start of the pattern; while it
398 is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some
399 such uses (C</.\G/g>, for example) currently cause problems, and
400 it is recommended that you avoid such usage for now.
403 The bracketing construct C<( ... )> creates capture buffers. To
404 refer to the digit'th buffer use \<digit> within the
405 match. Outside the match use "$" instead of "\". (The
406 \<digit> notation works in certain circumstances outside
407 the match. See the warning below about \1 vs $1 for details.)
408 Referring back to another part of the match is called a
410 X<regex, capture buffer> X<regexp, capture buffer>
411 X<regular expression, capture buffer> X<backreference>
413 There is no limit to the number of captured substrings that you may
414 use. However Perl also uses \10, \11, etc. as aliases for \010,
415 \011, etc. (Recall that 0 means octal, so \011 is the character at
416 number 9 in your coded character set; which would be the 10th character,
417 a horizontal tab under ASCII.) Perl resolves this
418 ambiguity by interpreting \10 as a backreference only if at least 10
419 left parentheses have opened before it. Likewise \11 is a
420 backreference only if at least 11 left parentheses have opened
421 before it. And so on. \1 through \9 are always interpreted as
426 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
428 if (/(.)\1/) { # find first doubled char
429 print "'$1' is the first doubled character\n";
432 if (/Time: (..):(..):(..)/) { # parse out values
438 Several special variables also refer back to portions of the previous
439 match. C<$+> returns whatever the last bracket match matched.
440 C<$&> returns the entire matched string. (At one point C<$0> did
441 also, but now it returns the name of the program.) C<$`> returns
442 everything before the matched string. C<$'> returns everything
443 after the matched string. And C<$^N> contains whatever was matched by
444 the most-recently closed group (submatch). C<$^N> can be used in
445 extended patterns (see below), for example to assign a submatch to a
447 X<$+> X<$^N> X<$&> X<$`> X<$'>
449 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
450 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
451 until the end of the enclosing block or until the next successful
452 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
453 X<$+> X<$^N> X<$&> X<$`> X<$'>
454 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
457 B<NOTE>: failed matches in Perl do not reset the match variables,
458 which makes it easier to write code that tests for a series of more
459 specific cases and remembers the best match.
461 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
462 C<$'> anywhere in the program, it has to provide them for every
463 pattern match. This may substantially slow your program. Perl
464 uses the same mechanism to produce $1, $2, etc, so you also pay a
465 price for each pattern that contains capturing parentheses. (To
466 avoid this cost while retaining the grouping behaviour, use the
467 extended regular expression C<(?: ... )> instead.) But if you never
468 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
469 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
470 if you can, but if you can't (and some algorithms really appreciate
471 them), once you've used them once, use them at will, because you've
472 already paid the price. As of 5.005, C<$&> is not so costly as the
476 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
477 C<\w>, C<\n>. Unlike some other regular expression languages, there
478 are no backslashed symbols that aren't alphanumeric. So anything
479 that looks like \\, \(, \), \<, \>, \{, or \} is always
480 interpreted as a literal character, not a metacharacter. This was
481 once used in a common idiom to disable or quote the special meanings
482 of regular expression metacharacters in a string that you want to
483 use for a pattern. Simply quote all non-"word" characters:
485 $pattern =~ s/(\W)/\\$1/g;
487 (If C<use locale> is set, then this depends on the current locale.)
488 Today it is more common to use the quotemeta() function or the C<\Q>
489 metaquoting escape sequence to disable all metacharacters' special
492 /$unquoted\Q$quoted\E$unquoted/
494 Beware that if you put literal backslashes (those not inside
495 interpolated variables) between C<\Q> and C<\E>, double-quotish
496 backslash interpolation may lead to confusing results. If you
497 I<need> to use literal backslashes within C<\Q...\E>,
498 consult L<perlop/"Gory details of parsing quoted constructs">.
500 =head2 Extended Patterns
502 Perl also defines a consistent extension syntax for features not
503 found in standard tools like B<awk> and B<lex>. The syntax is a
504 pair of parentheses with a question mark as the first thing within
505 the parentheses. The character after the question mark indicates
508 The stability of these extensions varies widely. Some have been
509 part of the core language for many years. Others are experimental
510 and may change without warning or be completely removed. Check
511 the documentation on an individual feature to verify its current
514 A question mark was chosen for this and for the minimal-matching
515 construct because 1) question marks are rare in older regular
516 expressions, and 2) whenever you see one, you should stop and
517 "question" exactly what is going on. That's psychology...
524 A comment. The text is ignored. If the C</x> modifier enables
525 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
526 the comment as soon as it sees a C<)>, so there is no way to put a literal
529 =item C<(?imsx-imsx)>
532 One or more embedded pattern-match modifiers, to be turned on (or
533 turned off, if preceded by C<->) for the remainder of the pattern or
534 the remainder of the enclosing pattern group (if any). This is
535 particularly useful for dynamic patterns, such as those read in from a
536 configuration file, read in as an argument, are specified in a table
537 somewhere, etc. Consider the case that some of which want to be case
538 sensitive and some do not. The case insensitive ones need to include
539 merely C<(?i)> at the front of the pattern. For example:
542 if ( /$pattern/i ) { }
546 $pattern = "(?i)foobar";
547 if ( /$pattern/ ) { }
549 These modifiers are restored at the end of the enclosing group. For example,
553 will match a repeated (I<including the case>!) word C<blah> in any
554 case, assuming C<x> modifier, and no C<i> modifier outside this
560 =item C<(?imsx-imsx:pattern)>
562 This is for clustering, not capturing; it groups subexpressions like
563 "()", but doesn't make backreferences as "()" does. So
565 @fields = split(/\b(?:a|b|c)\b/)
569 @fields = split(/\b(a|b|c)\b/)
571 but doesn't spit out extra fields. It's also cheaper not to capture
572 characters if you don't need to.
574 Any letters between C<?> and C<:> act as flags modifiers as with
575 C<(?imsx-imsx)>. For example,
577 /(?s-i:more.*than).*million/i
579 is equivalent to the more verbose
581 /(?:(?s-i)more.*than).*million/i
584 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
586 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
587 matches a word followed by a tab, without including the tab in C<$&>.
590 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
592 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
593 matches any occurrence of "foo" that isn't followed by "bar". Note
594 however that look-ahead and look-behind are NOT the same thing. You cannot
595 use this for look-behind.
597 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
598 will not do what you want. That's because the C<(?!foo)> is just saying that
599 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
600 match. You would have to do something like C</(?!foo)...bar/> for that. We
601 say "like" because there's the case of your "bar" not having three characters
602 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
603 Sometimes it's still easier just to say:
605 if (/bar/ && $` !~ /foo$/)
607 For look-behind see below.
609 =item C<(?<=pattern)>
610 X<(?<=)> X<look-behind, positive> X<lookbehind, positive>
612 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
613 matches a word that follows a tab, without including the tab in C<$&>.
614 Works only for fixed-width look-behind.
616 =item C<(?<!pattern)>
617 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
619 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
620 matches any occurrence of "foo" that does not follow "bar". Works
621 only for fixed-width look-behind.
624 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
626 B<WARNING>: This extended regular expression feature is considered
627 highly experimental, and may be changed or deleted without notice.
629 This zero-width assertion evaluates any embedded Perl code. It
630 always succeeds, and its C<code> is not interpolated. Currently,
631 the rules to determine where the C<code> ends are somewhat convoluted.
633 This feature can be used together with the special variable C<$^N> to
634 capture the results of submatches in variables without having to keep
635 track of the number of nested parentheses. For example:
637 $_ = "The brown fox jumps over the lazy dog";
638 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
639 print "color = $color, animal = $animal\n";
641 Inside the C<(?{...})> block, C<$_> refers to the string the regular
642 expression is matching against. You can also use C<pos()> to know what is
643 the current position of matching within this string.
645 The C<code> is properly scoped in the following sense: If the assertion
646 is backtracked (compare L<"Backtracking">), all changes introduced after
647 C<local>ization are undone, so that
651 (?{ $cnt = 0 }) # Initialize $cnt.
655 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
659 (?{ $res = $cnt }) # On success copy to non-localized
663 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
664 introduced value, because the scopes that restrict C<local> operators
667 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
668 switch. If I<not> used in this way, the result of evaluation of
669 C<code> is put into the special variable C<$^R>. This happens
670 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
671 inside the same regular expression.
673 The assignment to C<$^R> above is properly localized, so the old
674 value of C<$^R> is restored if the assertion is backtracked; compare
677 For reasons of security, this construct is forbidden if the regular
678 expression involves run-time interpolation of variables, unless the
679 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
680 variables contain results of C<qr//> operator (see
681 L<perlop/"qr/STRING/imosx">).
683 This restriction is because of the wide-spread and remarkably convenient
684 custom of using run-time determined strings as patterns. For example:
690 Before Perl knew how to execute interpolated code within a pattern,
691 this operation was completely safe from a security point of view,
692 although it could raise an exception from an illegal pattern. If
693 you turn on the C<use re 'eval'>, though, it is no longer secure,
694 so you should only do so if you are also using taint checking.
695 Better yet, use the carefully constrained evaluation within a Safe
696 compartment. See L<perlsec> for details about both these mechanisms.
698 Because perl's regex engine is not currently re-entrant, interpolated
699 code may not invoke the regex engine either directly with C<m//> or C<s///>),
700 or indirectly with functions such as C<split>.
702 =item C<(??{ code })>
704 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
705 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
707 B<WARNING>: This extended regular expression feature is considered
708 highly experimental, and may be changed or deleted without notice.
709 A simplified version of the syntax may be introduced for commonly
712 This is a "postponed" regular subexpression. The C<code> is evaluated
713 at run time, at the moment this subexpression may match. The result
714 of evaluation is considered as a regular expression and matched as
715 if it were inserted instead of this construct.
717 The C<code> is not interpolated. As before, the rules to determine
718 where the C<code> ends are currently somewhat convoluted.
720 The following pattern matches a parenthesized group:
725 (?> [^()]+ ) # Non-parens without backtracking
727 (??{ $re }) # Group with matching parens
732 Because perl's regex engine is not currently re-entrant, delayed
733 code may not invoke the regex engine either directly with C<m//> or C<s///>),
734 or indirectly with functions such as C<split>.
736 =item C<< (?>pattern) >>
737 X<backtrack> X<backtracking>
739 B<WARNING>: This extended regular expression feature is considered
740 highly experimental, and may be changed or deleted without notice.
742 An "independent" subexpression, one which matches the substring
743 that a I<standalone> C<pattern> would match if anchored at the given
744 position, and it matches I<nothing other than this substring>. This
745 construct is useful for optimizations of what would otherwise be
746 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
747 It may also be useful in places where the "grab all you can, and do not
748 give anything back" semantic is desirable.
750 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
751 (anchored at the beginning of string, as above) will match I<all>
752 characters C<a> at the beginning of string, leaving no C<a> for
753 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
754 since the match of the subgroup C<a*> is influenced by the following
755 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
756 C<a*ab> will match fewer characters than a standalone C<a*>, since
757 this makes the tail match.
759 An effect similar to C<< (?>pattern) >> may be achieved by writing
760 C<(?=(pattern))\1>. This matches the same substring as a standalone
761 C<a+>, and the following C<\1> eats the matched string; it therefore
762 makes a zero-length assertion into an analogue of C<< (?>...) >>.
763 (The difference between these two constructs is that the second one
764 uses a capturing group, thus shifting ordinals of backreferences
765 in the rest of a regular expression.)
767 Consider this pattern:
778 That will efficiently match a nonempty group with matching parentheses
779 two levels deep or less. However, if there is no such group, it
780 will take virtually forever on a long string. That's because there
781 are so many different ways to split a long string into several
782 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
783 to a subpattern of the above pattern. Consider how the pattern
784 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
785 seconds, but that each extra letter doubles this time. This
786 exponential performance will make it appear that your program has
787 hung. However, a tiny change to this pattern
791 (?> [^()]+ ) # change x+ above to (?> x+ )
798 which uses C<< (?>...) >> matches exactly when the one above does (verifying
799 this yourself would be a productive exercise), but finishes in a fourth
800 the time when used on a similar string with 1000000 C<a>s. Be aware,
801 however, that this pattern currently triggers a warning message under
802 the C<use warnings> pragma or B<-w> switch saying it
803 C<"matches null string many times in regex">.
805 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
806 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
807 This was only 4 times slower on a string with 1000000 C<a>s.
809 The "grab all you can, and do not give anything back" semantic is desirable
810 in many situations where on the first sight a simple C<()*> looks like
811 the correct solution. Suppose we parse text with comments being delimited
812 by C<#> followed by some optional (horizontal) whitespace. Contrary to
813 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
814 the comment delimiter, because it may "give up" some whitespace if
815 the remainder of the pattern can be made to match that way. The correct
816 answer is either one of these:
821 For example, to grab non-empty comments into $1, one should use either
824 / (?> \# [ \t]* ) ( .+ ) /x;
825 / \# [ \t]* ( [^ \t] .* ) /x;
827 Which one you pick depends on which of these expressions better reflects
828 the above specification of comments.
830 =item C<(?(condition)yes-pattern|no-pattern)>
833 =item C<(?(condition)yes-pattern)>
835 B<WARNING>: This extended regular expression feature is considered
836 highly experimental, and may be changed or deleted without notice.
838 Conditional expression. C<(condition)> should be either an integer in
839 parentheses (which is valid if the corresponding pair of parentheses
840 matched), or look-ahead/look-behind/evaluate zero-width assertion.
849 matches a chunk of non-parentheses, possibly included in parentheses
855 X<backtrack> X<backtracking>
857 NOTE: This section presents an abstract approximation of regular
858 expression behavior. For a more rigorous (and complicated) view of
859 the rules involved in selecting a match among possible alternatives,
860 see L<Combining pieces together>.
862 A fundamental feature of regular expression matching involves the
863 notion called I<backtracking>, which is currently used (when needed)
864 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
865 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
866 internally, but the general principle outlined here is valid.
868 For a regular expression to match, the I<entire> regular expression must
869 match, not just part of it. So if the beginning of a pattern containing a
870 quantifier succeeds in a way that causes later parts in the pattern to
871 fail, the matching engine backs up and recalculates the beginning
872 part--that's why it's called backtracking.
874 Here is an example of backtracking: Let's say you want to find the
875 word following "foo" in the string "Food is on the foo table.":
877 $_ = "Food is on the foo table.";
878 if ( /\b(foo)\s+(\w+)/i ) {
879 print "$2 follows $1.\n";
882 When the match runs, the first part of the regular expression (C<\b(foo)>)
883 finds a possible match right at the beginning of the string, and loads up
884 $1 with "Foo". However, as soon as the matching engine sees that there's
885 no whitespace following the "Foo" that it had saved in $1, it realizes its
886 mistake and starts over again one character after where it had the
887 tentative match. This time it goes all the way until the next occurrence
888 of "foo". The complete regular expression matches this time, and you get
889 the expected output of "table follows foo."
891 Sometimes minimal matching can help a lot. Imagine you'd like to match
892 everything between "foo" and "bar". Initially, you write something
895 $_ = "The food is under the bar in the barn.";
896 if ( /foo(.*)bar/ ) {
900 Which perhaps unexpectedly yields:
902 got <d is under the bar in the >
904 That's because C<.*> was greedy, so you get everything between the
905 I<first> "foo" and the I<last> "bar". Here it's more effective
906 to use minimal matching to make sure you get the text between a "foo"
907 and the first "bar" thereafter.
909 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
910 got <d is under the >
912 Here's another example: let's say you'd like to match a number at the end
913 of a string, and you also want to keep the preceding part of the match.
916 $_ = "I have 2 numbers: 53147";
917 if ( /(.*)(\d*)/ ) { # Wrong!
918 print "Beginning is <$1>, number is <$2>.\n";
921 That won't work at all, because C<.*> was greedy and gobbled up the
922 whole string. As C<\d*> can match on an empty string the complete
923 regular expression matched successfully.
925 Beginning is <I have 2 numbers: 53147>, number is <>.
927 Here are some variants, most of which don't work:
929 $_ = "I have 2 numbers: 53147";
942 printf "%-12s ", $pat;
952 (.*)(\d*) <I have 2 numbers: 53147> <>
953 (.*)(\d+) <I have 2 numbers: 5314> <7>
955 (.*?)(\d+) <I have > <2>
956 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
957 (.*?)(\d+)$ <I have 2 numbers: > <53147>
958 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
959 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
961 As you see, this can be a bit tricky. It's important to realize that a
962 regular expression is merely a set of assertions that gives a definition
963 of success. There may be 0, 1, or several different ways that the
964 definition might succeed against a particular string. And if there are
965 multiple ways it might succeed, you need to understand backtracking to
966 know which variety of success you will achieve.
968 When using look-ahead assertions and negations, this can all get even
969 trickier. Imagine you'd like to find a sequence of non-digits not
970 followed by "123". You might try to write that as
973 if ( /^\D*(?!123)/ ) { # Wrong!
974 print "Yup, no 123 in $_\n";
977 But that isn't going to match; at least, not the way you're hoping. It
978 claims that there is no 123 in the string. Here's a clearer picture of
979 why that pattern matches, contrary to popular expectations:
984 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
985 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
987 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
988 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
996 You might have expected test 3 to fail because it seems to a more
997 general purpose version of test 1. The important difference between
998 them is that test 3 contains a quantifier (C<\D*>) and so can use
999 backtracking, whereas test 1 will not. What's happening is
1000 that you've asked "Is it true that at the start of $x, following 0 or more
1001 non-digits, you have something that's not 123?" If the pattern matcher had
1002 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1005 The search engine will initially match C<\D*> with "ABC". Then it will
1006 try to match C<(?!123> with "123", which fails. But because
1007 a quantifier (C<\D*>) has been used in the regular expression, the
1008 search engine can backtrack and retry the match differently
1009 in the hope of matching the complete regular expression.
1011 The pattern really, I<really> wants to succeed, so it uses the
1012 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1013 time. Now there's indeed something following "AB" that is not
1014 "123". It's "C123", which suffices.
1016 We can deal with this by using both an assertion and a negation.
1017 We'll say that the first part in $1 must be followed both by a digit
1018 and by something that's not "123". Remember that the look-aheads
1019 are zero-width expressions--they only look, but don't consume any
1020 of the string in their match. So rewriting this way produces what
1021 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1023 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1024 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1028 In other words, the two zero-width assertions next to each other work as though
1029 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1030 matches only if you're at the beginning of the line AND the end of the
1031 line simultaneously. The deeper underlying truth is that juxtaposition in
1032 regular expressions always means AND, except when you write an explicit OR
1033 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1034 although the attempted matches are made at different positions because "a"
1035 is not a zero-width assertion, but a one-width assertion.
1037 B<WARNING>: particularly complicated regular expressions can take
1038 exponential time to solve because of the immense number of possible
1039 ways they can use backtracking to try match. For example, without
1040 internal optimizations done by the regular expression engine, this will
1041 take a painfully long time to run:
1043 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1045 And if you used C<*>'s in the internal groups instead of limiting them
1046 to 0 through 5 matches, then it would take forever--or until you ran
1047 out of stack space. Moreover, these internal optimizations are not
1048 always applicable. For example, if you put C<{0,5}> instead of C<*>
1049 on the external group, no current optimization is applicable, and the
1050 match takes a long time to finish.
1052 A powerful tool for optimizing such beasts is what is known as an
1053 "independent group",
1054 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1055 zero-length look-ahead/look-behind assertions will not backtrack to make
1056 the tail match, since they are in "logical" context: only
1057 whether they match is considered relevant. For an example
1058 where side-effects of look-ahead I<might> have influenced the
1059 following match, see L<C<< (?>pattern) >>>.
1061 =head2 Version 8 Regular Expressions
1062 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1064 In case you're not familiar with the "regular" Version 8 regex
1065 routines, here are the pattern-matching rules not described above.
1067 Any single character matches itself, unless it is a I<metacharacter>
1068 with a special meaning described here or above. You can cause
1069 characters that normally function as metacharacters to be interpreted
1070 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1071 character; "\\" matches a "\"). A series of characters matches that
1072 series of characters in the target string, so the pattern C<blurfl>
1073 would match "blurfl" in the target string.
1075 You can specify a character class, by enclosing a list of characters
1076 in C<[]>, which will match any one character from the list. If the
1077 first character after the "[" is "^", the class matches any character not
1078 in the list. Within a list, the "-" character specifies a
1079 range, so that C<a-z> represents all characters between "a" and "z",
1080 inclusive. If you want either "-" or "]" itself to be a member of a
1081 class, put it at the start of the list (possibly after a "^"), or
1082 escape it with a backslash. "-" is also taken literally when it is
1083 at the end of the list, just before the closing "]". (The
1084 following all specify the same class of three characters: C<[-az]>,
1085 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1086 specifies a class containing twenty-six characters, even on EBCDIC
1087 based coded character sets.) Also, if you try to use the character
1088 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1089 a range, that's not a range, the "-" is understood literally.
1091 Note also that the whole range idea is rather unportable between
1092 character sets--and even within character sets they may cause results
1093 you probably didn't expect. A sound principle is to use only ranges
1094 that begin from and end at either alphabets of equal case ([a-e],
1095 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1096 spell out the character sets in full.
1098 Characters may be specified using a metacharacter syntax much like that
1099 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1100 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1101 of octal digits, matches the character whose coded character set value
1102 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1103 matches the character whose numeric value is I<nn>. The expression \cI<x>
1104 matches the character control-I<x>. Finally, the "." metacharacter
1105 matches any character except "\n" (unless you use C</s>).
1107 You can specify a series of alternatives for a pattern using "|" to
1108 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1109 or "foe" in the target string (as would C<f(e|i|o)e>). The
1110 first alternative includes everything from the last pattern delimiter
1111 ("(", "[", or the beginning of the pattern) up to the first "|", and
1112 the last alternative contains everything from the last "|" to the next
1113 pattern delimiter. That's why it's common practice to include
1114 alternatives in parentheses: to minimize confusion about where they
1117 Alternatives are tried from left to right, so the first
1118 alternative found for which the entire expression matches, is the one that
1119 is chosen. This means that alternatives are not necessarily greedy. For
1120 example: when matching C<foo|foot> against "barefoot", only the "foo"
1121 part will match, as that is the first alternative tried, and it successfully
1122 matches the target string. (This might not seem important, but it is
1123 important when you are capturing matched text using parentheses.)
1125 Also remember that "|" is interpreted as a literal within square brackets,
1126 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1128 Within a pattern, you may designate subpatterns for later reference
1129 by enclosing them in parentheses, and you may refer back to the
1130 I<n>th subpattern later in the pattern using the metacharacter
1131 \I<n>. Subpatterns are numbered based on the left to right order
1132 of their opening parenthesis. A backreference matches whatever
1133 actually matched the subpattern in the string being examined, not
1134 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1135 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1136 1 matched "0x", even though the rule C<0|0x> could potentially match
1137 the leading 0 in the second number.
1139 =head2 Warning on \1 vs $1
1141 Some people get too used to writing things like:
1143 $pattern =~ s/(\W)/\\\1/g;
1145 This is grandfathered for the RHS of a substitute to avoid shocking the
1146 B<sed> addicts, but it's a dirty habit to get into. That's because in
1147 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1148 the usual double-quoted string means a control-A. The customary Unix
1149 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1150 of doing that, you get yourself into trouble if you then add an C</e>
1153 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1159 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1160 C<${1}000>. The operation of interpolation should not be confused
1161 with the operation of matching a backreference. Certainly they mean two
1162 different things on the I<left> side of the C<s///>.
1164 =head2 Repeated patterns matching zero-length substring
1166 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1168 Regular expressions provide a terse and powerful programming language. As
1169 with most other power tools, power comes together with the ability
1172 A common abuse of this power stems from the ability to make infinite
1173 loops using regular expressions, with something as innocuous as:
1175 'foo' =~ m{ ( o? )* }x;
1177 The C<o?> can match at the beginning of C<'foo'>, and since the position
1178 in the string is not moved by the match, C<o?> would match again and again
1179 because of the C<*> modifier. Another common way to create a similar cycle
1180 is with the looping modifier C<//g>:
1182 @matches = ( 'foo' =~ m{ o? }xg );
1186 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1188 or the loop implied by split().
1190 However, long experience has shown that many programming tasks may
1191 be significantly simplified by using repeated subexpressions that
1192 may match zero-length substrings. Here's a simple example being:
1194 @chars = split //, $string; # // is not magic in split
1195 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1197 Thus Perl allows such constructs, by I<forcefully breaking
1198 the infinite loop>. The rules for this are different for lower-level
1199 loops given by the greedy modifiers C<*+{}>, and for higher-level
1200 ones like the C</g> modifier or split() operator.
1202 The lower-level loops are I<interrupted> (that is, the loop is
1203 broken) when Perl detects that a repeated expression matched a
1204 zero-length substring. Thus
1206 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1208 is made equivalent to
1210 m{ (?: NON_ZERO_LENGTH )*
1215 The higher level-loops preserve an additional state between iterations:
1216 whether the last match was zero-length. To break the loop, the following
1217 match after a zero-length match is prohibited to have a length of zero.
1218 This prohibition interacts with backtracking (see L<"Backtracking">),
1219 and so the I<second best> match is chosen if the I<best> match is of
1227 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1228 match given by non-greedy C<??> is the zero-length match, and the I<second
1229 best> match is what is matched by C<\w>. Thus zero-length matches
1230 alternate with one-character-long matches.
1232 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1233 position one notch further in the string.
1235 The additional state of being I<matched with zero-length> is associated with
1236 the matched string, and is reset by each assignment to pos().
1237 Zero-length matches at the end of the previous match are ignored
1240 =head2 Combining pieces together
1242 Each of the elementary pieces of regular expressions which were described
1243 before (such as C<ab> or C<\Z>) could match at most one substring
1244 at the given position of the input string. However, in a typical regular
1245 expression these elementary pieces are combined into more complicated
1246 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1247 (in these examples C<S> and C<T> are regular subexpressions).
1249 Such combinations can include alternatives, leading to a problem of choice:
1250 if we match a regular expression C<a|ab> against C<"abc">, will it match
1251 substring C<"a"> or C<"ab">? One way to describe which substring is
1252 actually matched is the concept of backtracking (see L<"Backtracking">).
1253 However, this description is too low-level and makes you think
1254 in terms of a particular implementation.
1256 Another description starts with notions of "better"/"worse". All the
1257 substrings which may be matched by the given regular expression can be
1258 sorted from the "best" match to the "worst" match, and it is the "best"
1259 match which is chosen. This substitutes the question of "what is chosen?"
1260 by the question of "which matches are better, and which are worse?".
1262 Again, for elementary pieces there is no such question, since at most
1263 one match at a given position is possible. This section describes the
1264 notion of better/worse for combining operators. In the description
1265 below C<S> and C<T> are regular subexpressions.
1271 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1272 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1273 which can be matched by C<T>.
1275 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1278 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1279 C<B> is better match for C<T> than C<B'>.
1283 When C<S> can match, it is a better match than when only C<T> can match.
1285 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1286 two matches for C<T>.
1288 =item C<S{REPEAT_COUNT}>
1290 Matches as C<SSS...S> (repeated as many times as necessary).
1294 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1296 =item C<S{min,max}?>
1298 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1300 =item C<S?>, C<S*>, C<S+>
1302 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1304 =item C<S??>, C<S*?>, C<S+?>
1306 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1310 Matches the best match for C<S> and only that.
1312 =item C<(?=S)>, C<(?<=S)>
1314 Only the best match for C<S> is considered. (This is important only if
1315 C<S> has capturing parentheses, and backreferences are used somewhere
1316 else in the whole regular expression.)
1318 =item C<(?!S)>, C<(?<!S)>
1320 For this grouping operator there is no need to describe the ordering, since
1321 only whether or not C<S> can match is important.
1323 =item C<(??{ EXPR })>
1325 The ordering is the same as for the regular expression which is
1328 =item C<(?(condition)yes-pattern|no-pattern)>
1330 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1331 already determined. The ordering of the matches is the same as for the
1332 chosen subexpression.
1336 The above recipes describe the ordering of matches I<at a given position>.
1337 One more rule is needed to understand how a match is determined for the
1338 whole regular expression: a match at an earlier position is always better
1339 than a match at a later position.
1341 =head2 Creating custom RE engines
1343 Overloaded constants (see L<overload>) provide a simple way to extend
1344 the functionality of the RE engine.
1346 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
1347 matches at boundary between whitespace characters and non-whitespace
1348 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1349 at these positions, so we want to have each C<\Y|> in the place of the
1350 more complicated version. We can create a module C<customre> to do
1358 die "No argument to customre::import allowed" if @_;
1359 overload::constant 'qr' => \&convert;
1362 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1364 # We must also take care of not escaping the legitimate \\Y|
1365 # sequence, hence the presence of '\\' in the conversion rules.
1366 my %rules = ( '\\' => '\\\\',
1367 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
1373 { $rules{$1} or invalid($re,$1) }sgex;
1377 Now C<use customre> enables the new escape in constant regular
1378 expressions, i.e., those without any runtime variable interpolations.
1379 As documented in L<overload>, this conversion will work only over
1380 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
1381 part of this regular expression needs to be converted explicitly
1382 (but only if the special meaning of C<\Y|> should be enabled inside $re):
1387 $re = customre::convert $re;
1392 This document varies from difficult to understand to completely
1393 and utterly opaque. The wandering prose riddled with jargon is
1394 hard to fathom in several places.
1396 This document needs a rewrite that separates the tutorial content
1397 from the reference content.
1405 L<perlop/"Regexp Quote-Like Operators">.
1407 L<perlop/"Gory details of parsing quoted constructs">.
1417 I<Mastering Regular Expressions> by Jeffrey Friedl, published
1418 by O'Reilly and Associates.