2 X<regular expression> X<regex> X<regexp>
4 perlre - Perl regular expressions
8 This page describes the syntax of regular expressions in Perl.
10 If you haven't used regular expressions before, a quick-start
11 introduction is available in L<perlrequick>, and a longer tutorial
12 introduction is available in L<perlretut>.
14 For reference on how regular expressions are used in matching
15 operations, plus various examples of the same, see discussions of
16 C<m//>, C<s///>, C<qr//> and C<??> in L<perlop/"Regexp Quote-Like
19 Matching operations can have various modifiers. Modifiers
20 that relate to the interpretation of the regular expression inside
21 are listed below. Modifiers that alter the way a regular expression
22 is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and
23 L<perlop/"Gory details of parsing quoted constructs">.
28 X</i> X<regex, case-insensitive> X<regexp, case-insensitive>
29 X<regular expression, case-insensitive>
31 Do case-insensitive pattern matching.
33 If C<use locale> is in effect, the case map is taken from the current
34 locale. See L<perllocale>.
37 X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline>
39 Treat string as multiple lines. That is, change "^" and "$" from matching
40 the start or end of the string to matching the start or end of any
41 line anywhere within the string.
44 X</s> X<regex, single-line> X<regexp, single-line>
45 X<regular expression, single-line>
47 Treat string as single line. That is, change "." to match any character
48 whatsoever, even a newline, which normally it would not match.
50 Used together, as /ms, they let the "." match any character whatsoever,
51 while still allowing "^" and "$" to match, respectively, just after
52 and just before newlines within the string.
57 Extend your pattern's legibility by permitting whitespace and comments.
61 These are usually written as "the C</x> modifier", even though the delimiter
62 in question might not really be a slash. Any of these
63 modifiers may also be embedded within the regular expression itself using
64 the C<(?...)> construct. See below.
66 The C</x> modifier itself needs a little more explanation. It tells
67 the regular expression parser to ignore whitespace that is neither
68 backslashed nor within a character class. You can use this to break up
69 your regular expression into (slightly) more readable parts. The C<#>
70 character is also treated as a metacharacter introducing a comment,
71 just as in ordinary Perl code. This also means that if you want real
72 whitespace or C<#> characters in the pattern (outside a character
73 class, where they are unaffected by C</x>), then you'll either have to
74 escape them (using backslashes or C<\Q...\E>) or encode them using octal
75 or hex escapes. Taken together, these features go a long way towards
76 making Perl's regular expressions more readable. Note that you have to
77 be careful not to include the pattern delimiter in the comment--perl has
78 no way of knowing you did not intend to close the pattern early. See
79 the C-comment deletion code in L<perlop>. Also note that anything inside
80 a C<\Q...\E> stays unaffected by C</x>.
83 =head2 Regular Expressions
87 The patterns used in Perl pattern matching derive from supplied in
88 the Version 8 regex routines. (The routines are derived
89 (distantly) from Henry Spencer's freely redistributable reimplementation
90 of the V8 routines.) See L<Version 8 Regular Expressions> for
93 In particular the following metacharacters have their standard I<egrep>-ish
96 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
99 \ Quote the next metacharacter
100 ^ Match the beginning of the line
101 . Match any character (except newline)
102 $ Match the end of the line (or before newline at the end)
107 By default, the "^" character is guaranteed to match only the
108 beginning of the string, the "$" character only the end (or before the
109 newline at the end), and Perl does certain optimizations with the
110 assumption that the string contains only one line. Embedded newlines
111 will not be matched by "^" or "$". You may, however, wish to treat a
112 string as a multi-line buffer, such that the "^" will match after any
113 newline within the string (except if the newline is the last character in
114 the string), and "$" will match before any newline. At the
115 cost of a little more overhead, you can do this by using the /m modifier
116 on the pattern match operator. (Older programs did this by setting C<$*>,
117 but this practice has been removed in perl 5.9.)
120 To simplify multi-line substitutions, the "." character never matches a
121 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
122 the string is a single line--even if it isn't.
127 The following standard quantifiers are recognized:
128 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
130 * Match 0 or more times
131 + Match 1 or more times
133 {n} Match exactly n times
134 {n,} Match at least n times
135 {n,m} Match at least n but not more than m times
137 (If a curly bracket occurs in any other context, it is treated
138 as a regular character. In particular, the lower bound
139 is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+"
140 modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited
141 to integral values less than a preset limit defined when perl is built.
142 This is usually 32766 on the most common platforms. The actual limit can
143 be seen in the error message generated by code such as this:
145 $_ **= $_ , / {$_} / for 2 .. 42;
147 By default, a quantified subpattern is "greedy", that is, it will match as
148 many times as possible (given a particular starting location) while still
149 allowing the rest of the pattern to match. If you want it to match the
150 minimum number of times possible, follow the quantifier with a "?". Note
151 that the meanings don't change, just the "greediness":
152 X<metacharacter> X<greedy> X<greedyness>
153 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
155 *? Match 0 or more times
156 +? Match 1 or more times
158 {n}? Match exactly n times
159 {n,}? Match at least n times
160 {n,m}? Match at least n but not more than m times
162 By default, when a quantified subpattern does not allow the rest of the
163 overall pattern to match, Perl will backtrack. However, this behaviour is
164 sometimes undesirable. Thus Perl provides the "possesive" quantifier form
167 *+ Match 0 or more times and give nothing back
168 ++ Match 1 or more times and give nothing back
169 ?+ Match 0 or 1 time and give nothing back
170 {n}+ Match exactly n times and give nothing back (redundant)
171 {n,}+ Match at least n times and give nothing back
172 {n,m}+ Match at least n but not more than m times and give nothing back
178 will never match, as the C<a++> will gobble up all the C<a>'s in the
179 string and won't leave any for the remaining part of the pattern. This
180 feature can be extremely useful to give perl hints about where it
181 shouldn't backtrack. For instance, the typical "match a double-quoted
182 string" problem can be most efficiently performed when written as:
184 /"(?:[^"\\]++|\\.)*+"/
186 as we know that if the final quote does not match, bactracking will not
187 help. See the independent subexpression C<< (?>...) >> for more details;
188 possessive quantifiers are just syntactic sugar for that construct. For
189 instance the above example could also be written as follows:
191 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
193 =head3 Escape sequences
195 Because patterns are processed as double quoted strings, the following
197 X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q>
198 X<\0> X<\c> X<\N> X<\x>
204 \a alarm (bell) (BEL)
205 \e escape (think troff) (ESC)
206 \033 octal char (think of a PDP-11)
208 \x{263a} wide hex char (Unicode SMILEY)
211 \l lowercase next char (think vi)
212 \u uppercase next char (think vi)
213 \L lowercase till \E (think vi)
214 \U uppercase till \E (think vi)
215 \E end case modification (think vi)
216 \Q quote (disable) pattern metacharacters till \E
218 If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u>
219 and C<\U> is taken from the current locale. See L<perllocale>. For
220 documentation of C<\N{name}>, see L<charnames>.
222 You cannot include a literal C<$> or C<@> within a C<\Q> sequence.
223 An unescaped C<$> or C<@> interpolates the corresponding variable,
224 while escaping will cause the literal string C<\$> to be matched.
225 You'll need to write something like C<m/\Quser\E\@\Qhost/>.
227 =head3 Character classes
229 In addition, Perl defines the following:
231 X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C>
232 X<word> X<whitespace>
234 \w Match a "word" character (alphanumeric plus "_")
235 \W Match a non-"word" character
236 \s Match a whitespace character
237 \S Match a non-whitespace character
238 \d Match a digit character
239 \D Match a non-digit character
240 \pP Match P, named property. Use \p{Prop} for longer names.
242 \X Match eXtended Unicode "combining character sequence",
243 equivalent to (?:\PM\pM*)
244 \C Match a single C char (octet) even under Unicode.
245 NOTE: breaks up characters into their UTF-8 bytes,
246 so you may end up with malformed pieces of UTF-8.
247 Unsupported in lookbehind.
248 \1 Backreference to a specific group.
249 '1' may actually be any positive integer.
250 \R1 Relative backreference to a preceding closed group.
251 '1' may actually be any positive integer.
252 \k<name> Named backreference
253 \N{name} Named unicode character, or unicode escape
254 \x12 Hexadecimal escape sequence
255 \x{1234} Long hexadecimal escape sequence
257 A C<\w> matches a single alphanumeric character (an alphabetic
258 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
259 to match a string of Perl-identifier characters (which isn't the same
260 as matching an English word). If C<use locale> is in effect, the list
261 of alphabetic characters generated by C<\w> is taken from the current
262 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
263 C<\d>, and C<\D> within character classes, but if you try to use them
264 as endpoints of a range, that's not a range, the "-" is understood
265 literally. If Unicode is in effect, C<\s> matches also "\x{85}",
266 "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about
267 C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general.
268 You can define your own C<\p> and C<\P> properties, see L<perlunicode>.
271 The POSIX character class syntax
276 is also available. Note that the C<[> and C<]> braces are I<literal>;
277 they must always be used within a character class expression.
280 $string =~ /[[:alpha:]]/;
282 # this is not, and will generate a warning:
283 $string =~ /[:alpha:]/;
285 The available classes and their backslash equivalents (if available) are
288 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
289 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
310 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
314 Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
315 also the (very rare) "vertical tabulator", "\ck", chr(11).
319 A Perl extension, see above.
323 For example use C<[:upper:]> to match all the uppercase characters.
324 Note that the C<[]> are part of the C<[::]> construct, not part of the
325 whole character class. For example:
329 matches zero, one, any alphabetic character, and the percentage sign.
331 The following equivalences to Unicode \p{} constructs and equivalent
332 backslash character classes (if available), will hold:
333 X<character class> X<\p> X<\p{}>
335 [[:...:]] \p{...} backslash
353 For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent.
355 If the C<utf8> pragma is not used but the C<locale> pragma is, the
356 classes correlate with the usual isalpha(3) interface (except for
359 The assumedly non-obviously named classes are:
366 Any control character. Usually characters that don't produce output as
367 such but instead control the terminal somehow: for example newline and
368 backspace are control characters. All characters with ord() less than
369 32 are most often classified as control characters (assuming ASCII,
370 the ISO Latin character sets, and Unicode), as is the character with
371 the ord() value of 127 (C<DEL>).
376 Any alphanumeric or punctuation (special) character.
381 Any alphanumeric or punctuation (special) character or the space character.
386 Any punctuation (special) character.
391 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
392 work just fine) it is included for completeness.
396 You can negate the [::] character classes by prefixing the class name
397 with a '^'. This is a Perl extension. For example:
398 X<character class, negation>
400 POSIX traditional Unicode
402 [[:^digit:]] \D \P{IsDigit}
403 [[:^space:]] \S \P{IsSpace}
404 [[:^word:]] \W \P{IsWord}
406 Perl respects the POSIX standard in that POSIX character classes are
407 only supported within a character class. The POSIX character classes
408 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
409 use them will cause an error.
413 Perl defines the following zero-width assertions:
414 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
415 X<regexp, zero-width assertion>
416 X<regular expression, zero-width assertion>
417 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
419 \b Match a word boundary
420 \B Match a non-(word boundary)
421 \A Match only at beginning of string
422 \Z Match only at end of string, or before newline at the end
423 \z Match only at end of string
424 \G Match only at pos() (e.g. at the end-of-match position
427 A word boundary (C<\b>) is a spot between two characters
428 that has a C<\w> on one side of it and a C<\W> on the other side
429 of it (in either order), counting the imaginary characters off the
430 beginning and end of the string as matching a C<\W>. (Within
431 character classes C<\b> represents backspace rather than a word
432 boundary, just as it normally does in any double-quoted string.)
433 The C<\A> and C<\Z> are just like "^" and "$", except that they
434 won't match multiple times when the C</m> modifier is used, while
435 "^" and "$" will match at every internal line boundary. To match
436 the actual end of the string and not ignore an optional trailing
438 X<\b> X<\A> X<\Z> X<\z> X</m>
440 The C<\G> assertion can be used to chain global matches (using
441 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
442 It is also useful when writing C<lex>-like scanners, when you have
443 several patterns that you want to match against consequent substrings
444 of your string, see the previous reference. The actual location
445 where C<\G> will match can also be influenced by using C<pos()> as
446 an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
447 matches is modified somewhat, in that contents to the left of C<\G> is
448 not counted when determining the length of the match. Thus the following
449 will not match forever:
458 It will print 'A' and then terminate, as it considers the match to
459 be zero-width, and thus will not match at the same position twice in a
462 It is worth noting that C<\G> improperly used can result in an infinite
463 loop. Take care when using patterns that include C<\G> in an alternation.
465 =head3 Capture buffers
467 The bracketing construct C<( ... )> creates capture buffers. To
468 refer to the digit'th buffer use \<digit> within the
469 match. Outside the match use "$" instead of "\". (The
470 \<digit> notation works in certain circumstances outside
471 the match. See the warning below about \1 vs $1 for details.)
472 Referring back to another part of the match is called a
474 X<regex, capture buffer> X<regexp, capture buffer>
475 X<regular expression, capture buffer> X<backreference>
477 There is no limit to the number of captured substrings that you may
478 use. However Perl also uses \10, \11, etc. as aliases for \010,
479 \011, etc. (Recall that 0 means octal, so \011 is the character at
480 number 9 in your coded character set; which would be the 10th character,
481 a horizontal tab under ASCII.) Perl resolves this
482 ambiguity by interpreting \10 as a backreference only if at least 10
483 left parentheses have opened before it. Likewise \11 is a
484 backreference only if at least 11 left parentheses have opened
485 before it. And so on. \1 through \9 are always interpreted as
488 X<relative backreference>
489 In Perl 5.10 it is possible to relatively address a capture buffer by
490 using the C<\RNNN> notation, where C<NNN> is negative offset to a
491 preceding capture buffer. Thus C<\R1> refers to the last buffer,
492 C<\R2> refers to the buffer before that. For example:
498 \R1 # backref to buffer 3
499 \R3 # backref to buffer 1
503 and would match the same as C</(Y) ( (X) $3 $1 )/x>.
505 Additionally, as of Perl 5.10 you may use named capture buffers and named
506 backreferences. The notation is C<< (?<name>...) >> and C<< \k<name> >>
507 (you may also use single quotes instead of angle brackets to quote the
508 name). The only difference with named capture buffers and unnamed ones is
509 that multiple buffers may have the same name and that the contents of
510 named capture buffers is available via the C<%+> hash. When multiple
511 groups share the same name C<$+{name}> and C<< \k<name> >> refer to the
512 leftmost defined group, thus it's possible to do things with named capture
513 buffers that would otherwise require C<(??{})> code to accomplish. Named
514 capture buffers are numbered just as normal capture buffers are and may be
515 referenced via the magic numeric variables or via numeric backreferences
520 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
522 /(.)\1/ # find first doubled char
523 and print "'$1' is the first doubled character\n";
525 /(?<char>.)\k<char>/ # ... a different way
526 and print "'$+{char}' is the first doubled character\n";
528 /(?<char>.)\1/ # ... mix and match
529 and print "'$1' is the first doubled character\n";
531 if (/Time: (..):(..):(..)/) { # parse out values
537 Several special variables also refer back to portions of the previous
538 match. C<$+> returns whatever the last bracket match matched.
539 C<$&> returns the entire matched string. (At one point C<$0> did
540 also, but now it returns the name of the program.) C<$`> returns
541 everything before the matched string. C<$'> returns everything
542 after the matched string. And C<$^N> contains whatever was matched by
543 the most-recently closed group (submatch). C<$^N> can be used in
544 extended patterns (see below), for example to assign a submatch to a
546 X<$+> X<$^N> X<$&> X<$`> X<$'>
548 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
549 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
550 until the end of the enclosing block or until the next successful
551 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
552 X<$+> X<$^N> X<$&> X<$`> X<$'>
553 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
556 B<NOTE>: failed matches in Perl do not reset the match variables,
557 which makes it easier to write code that tests for a series of more
558 specific cases and remembers the best match.
560 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
561 C<$'> anywhere in the program, it has to provide them for every
562 pattern match. This may substantially slow your program. Perl
563 uses the same mechanism to produce $1, $2, etc, so you also pay a
564 price for each pattern that contains capturing parentheses. (To
565 avoid this cost while retaining the grouping behaviour, use the
566 extended regular expression C<(?: ... )> instead.) But if you never
567 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
568 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
569 if you can, but if you can't (and some algorithms really appreciate
570 them), once you've used them once, use them at will, because you've
571 already paid the price. As of 5.005, C<$&> is not so costly as the
575 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
576 C<\w>, C<\n>. Unlike some other regular expression languages, there
577 are no backslashed symbols that aren't alphanumeric. So anything
578 that looks like \\, \(, \), \<, \>, \{, or \} is always
579 interpreted as a literal character, not a metacharacter. This was
580 once used in a common idiom to disable or quote the special meanings
581 of regular expression metacharacters in a string that you want to
582 use for a pattern. Simply quote all non-"word" characters:
584 $pattern =~ s/(\W)/\\$1/g;
586 (If C<use locale> is set, then this depends on the current locale.)
587 Today it is more common to use the quotemeta() function or the C<\Q>
588 metaquoting escape sequence to disable all metacharacters' special
591 /$unquoted\Q$quoted\E$unquoted/
593 Beware that if you put literal backslashes (those not inside
594 interpolated variables) between C<\Q> and C<\E>, double-quotish
595 backslash interpolation may lead to confusing results. If you
596 I<need> to use literal backslashes within C<\Q...\E>,
597 consult L<perlop/"Gory details of parsing quoted constructs">.
599 =head2 Extended Patterns
601 Perl also defines a consistent extension syntax for features not
602 found in standard tools like B<awk> and B<lex>. The syntax is a
603 pair of parentheses with a question mark as the first thing within
604 the parentheses. The character after the question mark indicates
607 The stability of these extensions varies widely. Some have been
608 part of the core language for many years. Others are experimental
609 and may change without warning or be completely removed. Check
610 the documentation on an individual feature to verify its current
613 A question mark was chosen for this and for the minimal-matching
614 construct because 1) question marks are rare in older regular
615 expressions, and 2) whenever you see one, you should stop and
616 "question" exactly what is going on. That's psychology...
623 A comment. The text is ignored. If the C</x> modifier enables
624 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
625 the comment as soon as it sees a C<)>, so there is no way to put a literal
628 =item C<(?imsx-imsx)>
631 One or more embedded pattern-match modifiers, to be turned on (or
632 turned off, if preceded by C<->) for the remainder of the pattern or
633 the remainder of the enclosing pattern group (if any). This is
634 particularly useful for dynamic patterns, such as those read in from a
635 configuration file, read in as an argument, are specified in a table
636 somewhere, etc. Consider the case that some of which want to be case
637 sensitive and some do not. The case insensitive ones need to include
638 merely C<(?i)> at the front of the pattern. For example:
641 if ( /$pattern/i ) { }
645 $pattern = "(?i)foobar";
646 if ( /$pattern/ ) { }
648 These modifiers are restored at the end of the enclosing group. For example,
652 will match a repeated (I<including the case>!) word C<blah> in any
653 case, assuming C<x> modifier, and no C<i> modifier outside this
659 =item C<(?imsx-imsx:pattern)>
661 This is for clustering, not capturing; it groups subexpressions like
662 "()", but doesn't make backreferences as "()" does. So
664 @fields = split(/\b(?:a|b|c)\b/)
668 @fields = split(/\b(a|b|c)\b/)
670 but doesn't spit out extra fields. It's also cheaper not to capture
671 characters if you don't need to.
673 Any letters between C<?> and C<:> act as flags modifiers as with
674 C<(?imsx-imsx)>. For example,
676 /(?s-i:more.*than).*million/i
678 is equivalent to the more verbose
680 /(?:(?s-i)more.*than).*million/i
683 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
685 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
686 matches a word followed by a tab, without including the tab in C<$&>.
689 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
691 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
692 matches any occurrence of "foo" that isn't followed by "bar". Note
693 however that look-ahead and look-behind are NOT the same thing. You cannot
694 use this for look-behind.
696 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
697 will not do what you want. That's because the C<(?!foo)> is just saying that
698 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
699 match. You would have to do something like C</(?!foo)...bar/> for that. We
700 say "like" because there's the case of your "bar" not having three characters
701 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
702 Sometimes it's still easier just to say:
704 if (/bar/ && $` !~ /foo$/)
706 For look-behind see below.
708 =item C<(?<=pattern)>
709 X<(?<=)> X<look-behind, positive> X<lookbehind, positive>
711 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
712 matches a word that follows a tab, without including the tab in C<$&>.
713 Works only for fixed-width look-behind.
715 =item C<(?<!pattern)>
716 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
718 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
719 matches any occurrence of "foo" that does not follow "bar". Works
720 only for fixed-width look-behind.
722 =item C<(?'NAME'pattern)>
724 =item C<< (?<NAME>pattern) >>
725 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
727 A named capture buffer. Identical in every respect to normal capturing
728 parens C<()> but for the additional fact that C<%+> may be used after
729 a succesful match to refer to a named buffer. See C<perlvar> for more
730 details on the C<%+> hash.
732 If multiple distinct capture buffers have the same name then the
733 $+{NAME} will refer to the leftmost defined buffer in the match.
735 The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent.
737 B<NOTE:> While the notation of this construct is the same as the similar
738 function in .NET regexes, the behavior is not, in Perl the buffers are
739 numbered sequentially regardless of being named or not. Thus in the
744 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
745 the opposite which is what a .NET regex hacker might expect.
747 Currently NAME is restricted to word chars only. In other words, it
748 must match C</^\w+$/>.
750 =item C<< \k<name> >>
752 =item C<< \k'name' >>
754 Named backreference. Similar to numeric backreferences, except that
755 the group is designated by name and not number. If multiple groups
756 have the same name then it refers to the leftmost defined group in
759 It is an error to refer to a name not defined by a C<(?<NAME>)>
760 earlier in the pattern.
762 Both forms are equivalent.
765 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
767 B<WARNING>: This extended regular expression feature is considered
768 experimental, and may be changed without notice. Code executed that
769 has side effects may not perform identically from version to version
770 due to the effect of future optimisations in the regex engine.
772 This zero-width assertion evaluates any embedded Perl code. It
773 always succeeds, and its C<code> is not interpolated. Currently,
774 the rules to determine where the C<code> ends are somewhat convoluted.
776 This feature can be used together with the special variable C<$^N> to
777 capture the results of submatches in variables without having to keep
778 track of the number of nested parentheses. For example:
780 $_ = "The brown fox jumps over the lazy dog";
781 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
782 print "color = $color, animal = $animal\n";
784 Inside the C<(?{...})> block, C<$_> refers to the string the regular
785 expression is matching against. You can also use C<pos()> to know what is
786 the current position of matching within this string.
788 The C<code> is properly scoped in the following sense: If the assertion
789 is backtracked (compare L<"Backtracking">), all changes introduced after
790 C<local>ization are undone, so that
794 (?{ $cnt = 0 }) # Initialize $cnt.
798 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
802 (?{ $res = $cnt }) # On success copy to non-localized
806 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
807 introduced value, because the scopes that restrict C<local> operators
810 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
811 switch. If I<not> used in this way, the result of evaluation of
812 C<code> is put into the special variable C<$^R>. This happens
813 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
814 inside the same regular expression.
816 The assignment to C<$^R> above is properly localized, so the old
817 value of C<$^R> is restored if the assertion is backtracked; compare
820 Due to an unfortunate implementation issue, the Perl code contained in these
821 blocks is treated as a compile time closure that can have seemingly bizarre
822 consequences when used with lexically scoped variables inside of subroutines
823 or loops. There are various workarounds for this, including simply using
824 global variables instead. If you are using this construct and strange results
825 occur then check for the use of lexically scoped variables.
827 For reasons of security, this construct is forbidden if the regular
828 expression involves run-time interpolation of variables, unless the
829 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
830 variables contain results of C<qr//> operator (see
831 L<perlop/"qr/STRING/imosx">).
833 This restriction is because of the wide-spread and remarkably convenient
834 custom of using run-time determined strings as patterns. For example:
840 Before Perl knew how to execute interpolated code within a pattern,
841 this operation was completely safe from a security point of view,
842 although it could raise an exception from an illegal pattern. If
843 you turn on the C<use re 'eval'>, though, it is no longer secure,
844 so you should only do so if you are also using taint checking.
845 Better yet, use the carefully constrained evaluation within a Safe
846 compartment. See L<perlsec> for details about both these mechanisms.
848 Because perl's regex engine is not currently re-entrant, interpolated
849 code may not invoke the regex engine either directly with C<m//> or C<s///>),
850 or indirectly with functions such as C<split>.
852 =item C<(??{ code })>
854 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
856 B<WARNING>: This extended regular expression feature is considered
857 experimental, and may be changed without notice. Code executed that
858 has side effects may not perform identically from version to version
859 due to the effect of future optimisations in the regex engine.
861 This is a "postponed" regular subexpression. The C<code> is evaluated
862 at run time, at the moment this subexpression may match. The result
863 of evaluation is considered as a regular expression and matched as
864 if it were inserted instead of this construct. Note that this means
865 that the contents of capture buffers defined inside an eval'ed pattern
866 are not available outside of the pattern, and vice versa, there is no
867 way for the inner pattern to refer to a capture buffer defined outside.
870 ('a' x 100)=~/(??{'(.)' x 100})/
872 B<will> match, it will B<not> set $1.
874 The C<code> is not interpolated. As before, the rules to determine
875 where the C<code> ends are currently somewhat convoluted.
877 The following pattern matches a parenthesized group:
882 (?> [^()]+ ) # Non-parens without backtracking
884 (??{ $re }) # Group with matching parens
889 See also C<(?PARNO)> for a different, more efficient way to accomplish
892 Because perl's regex engine is not currently re-entrant, delayed
893 code may not invoke the regex engine either directly with C<m//> or C<s///>),
894 or indirectly with functions such as C<split>.
896 Recursing deeper than 50 times without consuming any input string will
897 result in a fatal error. The maximum depth is compiled into perl, so
898 changing it requires a custom build.
900 =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)>
901 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
902 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
903 X<regex, relative recursion>
905 Similar to C<(??{ code })> except it does not involve compiling any code,
906 instead it treats the contents of a capture buffer as an independent
907 pattern that must match at the current position. Capture buffers
908 contained by the pattern will have the value as determined by the
911 PARNO is a sequence of digits (not starting with 0) whose value reflects
912 the paren-number of the capture buffer to recurse to. C<(?R)> recurses to
913 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
914 C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed
915 to be relative, with negative numbers indicating preceding capture buffers
916 and positive ones following. Thus C<(?-1)> refers to the most recently
917 declared buffer, and C<(?+1)> indicates the next buffer to be declared.
918 Note that the counting for relative recursion differs from that of
919 relative backreferences, in that with recursion unclosed buffers B<are>
922 The following pattern matches a function foo() which may contain
923 balanced parentheses as the argument.
925 $re = qr{ ( # paren group 1 (full function)
927 ( # paren group 2 (parens)
929 ( # paren group 3 (contents of parens)
931 (?> [^()]+ ) # Non-parens without backtracking
933 (?2) # Recurse to start of paren group 2
941 If the pattern was used as follows
943 'foo(bar(baz)+baz(bop))'=~/$re/
944 and print "\$1 = $1\n",
948 the output produced should be the following:
950 $1 = foo(bar(baz)+baz(bop))
951 $2 = (bar(baz)+baz(bop))
952 $3 = bar(baz)+baz(bop)
954 If there is no corresponding capture buffer defined, then it is a
955 fatal error. Recursing deeper than 50 times without consuming any input
956 string will also result in a fatal error. The maximum depth is compiled
957 into perl, so changing it requires a custom build.
959 The following shows how using negative indexing can make it
960 easier to embed recursive patterns inside of a C<qr//> construct
963 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
964 if (/foo $parens \s+ + \s+ bar $parens/x) {
965 # do something here...
968 B<Note> that this pattern does not behave the same way as the equivalent
969 PCRE or Python construct of the same form. In perl you can backtrack into
970 a recursed group, in PCRE and Python the recursed into group is treated
971 as atomic. Also, modifiers are resolved at compile time, so constructs
972 like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
978 Recurse to a named subpattern. Identical to (?PARNO) except that the
979 parenthesis to recurse to is determined by name. If multiple parens have
980 the same name, then it recurses to the leftmost.
982 It is an error to refer to a name that is not declared somewhere in the
985 =item C<(?(condition)yes-pattern|no-pattern)>
988 =item C<(?(condition)yes-pattern)>
990 Conditional expression. C<(condition)> should be either an integer in
991 parentheses (which is valid if the corresponding pair of parentheses
992 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
993 name in angle brackets or single quotes (which is valid if a buffer
994 with the given name matched), or the special symbol (R) (true when
995 evaluated inside of recursion or eval). Additionally the R may be
996 followed by a number, (which will be true when evaluated when recursing
997 inside of the appropriate group), or by C<&NAME>, in which case it will
998 be true only when evaluated during recursion in the named group.
1000 Here's a summary of the possible predicates:
1006 Checks if the numbered capturing buffer has matched something.
1008 =item (<NAME>) ('NAME')
1010 Checks if a buffer with the given name has matched something.
1014 Treats the code block as the condition.
1018 Checks if the expression has been evaluated inside of recursion.
1022 Checks if the expression has been evaluated while executing directly
1023 inside of the n-th capture group. This check is the regex equivalent of
1025 if ((caller(0))[3] eq 'subname') { ... }
1027 In other words, it does not check the full recursion stack.
1031 Similar to C<(R1)>, this predicate checks to see if we're executing
1032 directly inside of the leftmost group with a given name (this is the same
1033 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1034 stack, but only the name of the innermost active recursion.
1038 In this case, the yes-pattern is never directly executed, and no
1039 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1040 See below for details.
1051 matches a chunk of non-parentheses, possibly included in parentheses
1054 A special form is the C<(DEFINE)> predicate, which never executes directly
1055 its yes-pattern, and does not allow a no-pattern. This allows to define
1056 subpatterns which will be executed only by using the recursion mechanism.
1057 This way, you can define a set of regular expression rules that can be
1058 bundled into any pattern you choose.
1060 It is recommended that for this usage you put the DEFINE block at the
1061 end of the pattern, and that you name any subpatterns defined within it.
1063 Also, it's worth noting that patterns defined this way probably will
1064 not be as efficient, as the optimiser is not very clever about
1067 An example of how this might be used is as follows:
1069 /(?<NAME>(&NAME_PAT))(?<ADDR>(&ADDRESS_PAT))
1075 Note that capture buffers matched inside of recursion are not accessible
1076 after the recursion returns, so the extra layer of capturing buffers are
1077 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1078 C<$+{NAME}> would be.
1080 =item C<< (?>pattern) >>
1081 X<backtrack> X<backtracking> X<atomic> X<possessive>
1083 An "independent" subexpression, one which matches the substring
1084 that a I<standalone> C<pattern> would match if anchored at the given
1085 position, and it matches I<nothing other than this substring>. This
1086 construct is useful for optimizations of what would otherwise be
1087 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1088 It may also be useful in places where the "grab all you can, and do not
1089 give anything back" semantic is desirable.
1091 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1092 (anchored at the beginning of string, as above) will match I<all>
1093 characters C<a> at the beginning of string, leaving no C<a> for
1094 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1095 since the match of the subgroup C<a*> is influenced by the following
1096 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1097 C<a*ab> will match fewer characters than a standalone C<a*>, since
1098 this makes the tail match.
1100 An effect similar to C<< (?>pattern) >> may be achieved by writing
1101 C<(?=(pattern))\1>. This matches the same substring as a standalone
1102 C<a+>, and the following C<\1> eats the matched string; it therefore
1103 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1104 (The difference between these two constructs is that the second one
1105 uses a capturing group, thus shifting ordinals of backreferences
1106 in the rest of a regular expression.)
1108 Consider this pattern:
1119 That will efficiently match a nonempty group with matching parentheses
1120 two levels deep or less. However, if there is no such group, it
1121 will take virtually forever on a long string. That's because there
1122 are so many different ways to split a long string into several
1123 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1124 to a subpattern of the above pattern. Consider how the pattern
1125 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1126 seconds, but that each extra letter doubles this time. This
1127 exponential performance will make it appear that your program has
1128 hung. However, a tiny change to this pattern
1132 (?> [^()]+ ) # change x+ above to (?> x+ )
1139 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1140 this yourself would be a productive exercise), but finishes in a fourth
1141 the time when used on a similar string with 1000000 C<a>s. Be aware,
1142 however, that this pattern currently triggers a warning message under
1143 the C<use warnings> pragma or B<-w> switch saying it
1144 C<"matches null string many times in regex">.
1146 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1147 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1148 This was only 4 times slower on a string with 1000000 C<a>s.
1150 The "grab all you can, and do not give anything back" semantic is desirable
1151 in many situations where on the first sight a simple C<()*> looks like
1152 the correct solution. Suppose we parse text with comments being delimited
1153 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1154 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1155 the comment delimiter, because it may "give up" some whitespace if
1156 the remainder of the pattern can be made to match that way. The correct
1157 answer is either one of these:
1162 For example, to grab non-empty comments into $1, one should use either
1165 / (?> \# [ \t]* ) ( .+ ) /x;
1166 / \# [ \t]* ( [^ \t] .* ) /x;
1168 Which one you pick depends on which of these expressions better reflects
1169 the above specification of comments.
1171 In some literature this construct is called "atomic matching" or
1172 "possessive matching".
1174 Possessive quantifiers are equivalent to putting the item they are applied
1175 to inside of one of these constructs. The following equivalences apply:
1177 Quantifier Form Bracketing Form
1178 --------------- ---------------
1182 PAT{min,max}+ (?>PAT{min,max})
1186 =head2 Special Backtracking Control Verbs
1188 B<WARNING:> These patterns are experimental and subject to change or
1189 removal in a future version of perl. Their usage in production code should
1190 be noted to avoid problems during upgrades.
1192 These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1193 otherwise stated the ARG argument is optional; in some cases, it is
1196 Any pattern containing a special backtracking verb that allows an argument
1197 has the special behaviour that when executed it sets the current packages'
1198 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1201 On failure, the C<$REGERROR> variable will be set to the ARG value of the
1202 verb pattern, if the verb was involved in the failure of the match. If the
1203 ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1204 name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1205 none. Also, the C<$REGMARK> variable will be set to FALSE.
1207 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1208 the C<$REGMARK> variable will be set to the name of the last
1209 C<(*MARK:NAME)> pattern executed. See the explanation for the
1210 C<(*MARK:NAME)> verb below for more details.
1212 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1213 and most other regex related variables. They are not local to a scope, nor
1214 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1215 Use C<local> to localize changes to them to a specific scope if necessary.
1217 If a pattern does not contain a special backtracking verb that allows an
1218 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
1222 =item Verbs that take an argument
1226 =item C<(*PRUNE)> C<(*PRUNE:NAME)>
1227 X<(*PRUNE)> X<(*PRUNE:NAME)>
1229 This zero-width pattern prunes the backtracking tree at the current point
1230 when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1231 where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1232 A may backtrack as necessary to match. Once it is reached, matching
1233 continues in B, which may also backtrack as necessary; however, should B
1234 not match, then no further backtracking will take place, and the pattern
1235 will fail outright at the current starting position.
1237 The following example counts all the possible matching strings in a
1238 pattern (without actually matching any of them).
1240 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1241 print "Count=$count\n";
1256 If we add a C<(*PRUNE)> before the count like the following
1258 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
1259 print "Count=$count\n";
1261 we prevent backtracking and find the count of the longest matching
1262 at each matching startpoint like so:
1269 Any number of C<(*PRUNE)> assertions may be used in a pattern.
1271 See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1272 control backtracking. In some cases, the use of C<(*PRUNE)> can be
1273 replaced with a C<< (?>pattern) >> with no functional difference; however,
1274 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1275 C<< (?>pattern) >> alone.
1278 =item C<(*SKIP)> C<(*SKIP:NAME)>
1281 This zero-width pattern is similar to C<(*PRUNE)>, except that on
1282 failure it also signifies that whatever text that was matched leading up
1283 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1284 of this pattern. This effectively means that the regex engine "skips" forward
1285 to this position on failure and tries to match again, (assuming that
1286 there is sufficient room to match).
1288 The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1289 C<(*MARK:NAME)> was encountered while matching, then it is that position
1290 which is used as the "skip point". If no C<(*MARK)> of that name was
1291 encountered, then the C<(*SKIP)> operator has no effect. When used
1292 without a name the "skip point" is where the match point was when
1293 executing the (*SKIP) pattern.
1295 Compare the following to the examples in C<(*PRUNE)>, note the string
1298 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
1299 print "Count=$count\n";
1307 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
1308 executed, the next startpoint will be where the cursor was when the
1309 C<(*SKIP)> was executed.
1311 As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
1313 =item C<(*MARK:NAME)> C<(*:NAME)>
1314 X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)>
1316 This zero-width pattern can be used to mark the point reached in a string
1317 when a certain part of the pattern has been successfully matched. This
1318 mark may be given a name. A later C<(*SKIP)> pattern will then skip
1319 forward to that point if backtracked into on failure. Any number of
1320 C<(*MARK)> patterns are allowed, and the NAME portion is optional and may
1323 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1324 can be used to "label" a pattern branch, so that after matching, the
1325 program can determine which branches of the pattern were involved in the
1328 When a match is successful, the C<$REGMARK> variable will be set to the
1329 name of the most recently executed C<(*MARK:NAME)> that was involved
1332 This can be used to determine which branch of a pattern was matched
1333 without using a seperate capture buffer for each branch, which in turn
1334 can result in a performance improvement, as perl cannot optimize
1335 C</(?:(x)|(y)|(z))/> as efficiently as something like
1336 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1338 When a match has failed, and unless another verb has been involved in
1339 failing the match and has provided its own name to use, the C<$REGERROR>
1340 variable will be set to the name of the most recently executed
1343 See C<(*SKIP)> for more details.
1345 =item C<(*THEN)> C<(*THEN:NAME)>
1347 This is similar to the "cut group" operator C<::> from Perl6. Like
1348 C<(*PRUNE)>, this verb always matches, and when backtracked into on
1349 failure, it causes the regex engine to try the next alternation in the
1350 innermost enclosing group (capturing or otherwise).
1352 Its name comes from the observation that this operation combined with the
1353 alternation operator (C<|>) can be used to create what is essentially a
1354 pattern-based if/then/else block:
1356 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
1358 Note that if this operator is used and NOT inside of an alternation then
1359 it acts exactly like the C<(*PRUNE)> operator.
1369 / ( A (*THEN) B | C (*THEN) D ) /
1373 / ( A (*PRUNE) B | C (*PRUNE) D ) /
1375 as after matching the A but failing on the B the C<(*THEN)> verb will
1376 backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
1381 This is the Perl6 "commit pattern" C<< <commit> >> or C<:::>. It's a
1382 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
1383 into on failure it causes the match to fail outright. No further attempts
1384 to find a valid match by advancing the start pointer will occur again.
1387 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
1388 print "Count=$count\n";
1395 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1396 does not match, the regex engine will not try any further matching on the
1401 =item Verbs without an argument
1405 =item C<(*FAIL)> C<(*F)>
1408 This pattern matches nothing and always fails. It can be used to force the
1409 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1410 fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
1412 It is probably useful only when combined with C<(?{})> or C<(??{})>.
1417 B<WARNING:> This feature is highly experimental. It is not recommended
1418 for production code.
1420 This pattern matches nothing and causes the end of successful matching at
1421 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1422 whether there is actually more to match in the string. When inside of a
1423 nested pattern, such as recursion or a dynamically generated subbpattern
1424 via C<(??{})>, only the innermost pattern is ended immediately.
1426 If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are
1427 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1430 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
1432 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
1433 be set. If another branch in the inner parens were matched, such as in the
1434 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
1441 X<backtrack> X<backtracking>
1443 NOTE: This section presents an abstract approximation of regular
1444 expression behavior. For a more rigorous (and complicated) view of
1445 the rules involved in selecting a match among possible alternatives,
1446 see L<Combining pieces together>.
1448 A fundamental feature of regular expression matching involves the
1449 notion called I<backtracking>, which is currently used (when needed)
1450 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
1451 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1452 internally, but the general principle outlined here is valid.
1454 For a regular expression to match, the I<entire> regular expression must
1455 match, not just part of it. So if the beginning of a pattern containing a
1456 quantifier succeeds in a way that causes later parts in the pattern to
1457 fail, the matching engine backs up and recalculates the beginning
1458 part--that's why it's called backtracking.
1460 Here is an example of backtracking: Let's say you want to find the
1461 word following "foo" in the string "Food is on the foo table.":
1463 $_ = "Food is on the foo table.";
1464 if ( /\b(foo)\s+(\w+)/i ) {
1465 print "$2 follows $1.\n";
1468 When the match runs, the first part of the regular expression (C<\b(foo)>)
1469 finds a possible match right at the beginning of the string, and loads up
1470 $1 with "Foo". However, as soon as the matching engine sees that there's
1471 no whitespace following the "Foo" that it had saved in $1, it realizes its
1472 mistake and starts over again one character after where it had the
1473 tentative match. This time it goes all the way until the next occurrence
1474 of "foo". The complete regular expression matches this time, and you get
1475 the expected output of "table follows foo."
1477 Sometimes minimal matching can help a lot. Imagine you'd like to match
1478 everything between "foo" and "bar". Initially, you write something
1481 $_ = "The food is under the bar in the barn.";
1482 if ( /foo(.*)bar/ ) {
1486 Which perhaps unexpectedly yields:
1488 got <d is under the bar in the >
1490 That's because C<.*> was greedy, so you get everything between the
1491 I<first> "foo" and the I<last> "bar". Here it's more effective
1492 to use minimal matching to make sure you get the text between a "foo"
1493 and the first "bar" thereafter.
1495 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1496 got <d is under the >
1498 Here's another example: let's say you'd like to match a number at the end
1499 of a string, and you also want to keep the preceding part of the match.
1502 $_ = "I have 2 numbers: 53147";
1503 if ( /(.*)(\d*)/ ) { # Wrong!
1504 print "Beginning is <$1>, number is <$2>.\n";
1507 That won't work at all, because C<.*> was greedy and gobbled up the
1508 whole string. As C<\d*> can match on an empty string the complete
1509 regular expression matched successfully.
1511 Beginning is <I have 2 numbers: 53147>, number is <>.
1513 Here are some variants, most of which don't work:
1515 $_ = "I have 2 numbers: 53147";
1528 printf "%-12s ", $pat;
1530 print "<$1> <$2>\n";
1536 That will print out:
1538 (.*)(\d*) <I have 2 numbers: 53147> <>
1539 (.*)(\d+) <I have 2 numbers: 5314> <7>
1541 (.*?)(\d+) <I have > <2>
1542 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1543 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1544 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1545 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1547 As you see, this can be a bit tricky. It's important to realize that a
1548 regular expression is merely a set of assertions that gives a definition
1549 of success. There may be 0, 1, or several different ways that the
1550 definition might succeed against a particular string. And if there are
1551 multiple ways it might succeed, you need to understand backtracking to
1552 know which variety of success you will achieve.
1554 When using look-ahead assertions and negations, this can all get even
1555 trickier. Imagine you'd like to find a sequence of non-digits not
1556 followed by "123". You might try to write that as
1559 if ( /^\D*(?!123)/ ) { # Wrong!
1560 print "Yup, no 123 in $_\n";
1563 But that isn't going to match; at least, not the way you're hoping. It
1564 claims that there is no 123 in the string. Here's a clearer picture of
1565 why that pattern matches, contrary to popular expectations:
1570 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1571 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1573 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1574 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1582 You might have expected test 3 to fail because it seems to a more
1583 general purpose version of test 1. The important difference between
1584 them is that test 3 contains a quantifier (C<\D*>) and so can use
1585 backtracking, whereas test 1 will not. What's happening is
1586 that you've asked "Is it true that at the start of $x, following 0 or more
1587 non-digits, you have something that's not 123?" If the pattern matcher had
1588 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1591 The search engine will initially match C<\D*> with "ABC". Then it will
1592 try to match C<(?!123> with "123", which fails. But because
1593 a quantifier (C<\D*>) has been used in the regular expression, the
1594 search engine can backtrack and retry the match differently
1595 in the hope of matching the complete regular expression.
1597 The pattern really, I<really> wants to succeed, so it uses the
1598 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1599 time. Now there's indeed something following "AB" that is not
1600 "123". It's "C123", which suffices.
1602 We can deal with this by using both an assertion and a negation.
1603 We'll say that the first part in $1 must be followed both by a digit
1604 and by something that's not "123". Remember that the look-aheads
1605 are zero-width expressions--they only look, but don't consume any
1606 of the string in their match. So rewriting this way produces what
1607 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1609 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1610 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1614 In other words, the two zero-width assertions next to each other work as though
1615 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1616 matches only if you're at the beginning of the line AND the end of the
1617 line simultaneously. The deeper underlying truth is that juxtaposition in
1618 regular expressions always means AND, except when you write an explicit OR
1619 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1620 although the attempted matches are made at different positions because "a"
1621 is not a zero-width assertion, but a one-width assertion.
1623 B<WARNING>: particularly complicated regular expressions can take
1624 exponential time to solve because of the immense number of possible
1625 ways they can use backtracking to try match. For example, without
1626 internal optimizations done by the regular expression engine, this will
1627 take a painfully long time to run:
1629 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1631 And if you used C<*>'s in the internal groups instead of limiting them
1632 to 0 through 5 matches, then it would take forever--or until you ran
1633 out of stack space. Moreover, these internal optimizations are not
1634 always applicable. For example, if you put C<{0,5}> instead of C<*>
1635 on the external group, no current optimization is applicable, and the
1636 match takes a long time to finish.
1638 A powerful tool for optimizing such beasts is what is known as an
1639 "independent group",
1640 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1641 zero-length look-ahead/look-behind assertions will not backtrack to make
1642 the tail match, since they are in "logical" context: only
1643 whether they match is considered relevant. For an example
1644 where side-effects of look-ahead I<might> have influenced the
1645 following match, see L<C<< (?>pattern) >>>.
1647 =head2 Version 8 Regular Expressions
1648 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1650 In case you're not familiar with the "regular" Version 8 regex
1651 routines, here are the pattern-matching rules not described above.
1653 Any single character matches itself, unless it is a I<metacharacter>
1654 with a special meaning described here or above. You can cause
1655 characters that normally function as metacharacters to be interpreted
1656 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1657 character; "\\" matches a "\"). A series of characters matches that
1658 series of characters in the target string, so the pattern C<blurfl>
1659 would match "blurfl" in the target string.
1661 You can specify a character class, by enclosing a list of characters
1662 in C<[]>, which will match any character from the list. If the
1663 first character after the "[" is "^", the class matches any character not
1664 in the list. Within a list, the "-" character specifies a
1665 range, so that C<a-z> represents all characters between "a" and "z",
1666 inclusive. If you want either "-" or "]" itself to be a member of a
1667 class, put it at the start of the list (possibly after a "^"), or
1668 escape it with a backslash. "-" is also taken literally when it is
1669 at the end of the list, just before the closing "]". (The
1670 following all specify the same class of three characters: C<[-az]>,
1671 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1672 specifies a class containing twenty-six characters, even on EBCDIC-based
1673 character sets.) Also, if you try to use the character
1674 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1675 a range, the "-" is understood literally.
1677 Note also that the whole range idea is rather unportable between
1678 character sets--and even within character sets they may cause results
1679 you probably didn't expect. A sound principle is to use only ranges
1680 that begin from and end at either alphabets of equal case ([a-e],
1681 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1682 spell out the character sets in full.
1684 Characters may be specified using a metacharacter syntax much like that
1685 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1686 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1687 of octal digits, matches the character whose coded character set value
1688 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1689 matches the character whose numeric value is I<nn>. The expression \cI<x>
1690 matches the character control-I<x>. Finally, the "." metacharacter
1691 matches any character except "\n" (unless you use C</s>).
1693 You can specify a series of alternatives for a pattern using "|" to
1694 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1695 or "foe" in the target string (as would C<f(e|i|o)e>). The
1696 first alternative includes everything from the last pattern delimiter
1697 ("(", "[", or the beginning of the pattern) up to the first "|", and
1698 the last alternative contains everything from the last "|" to the next
1699 pattern delimiter. That's why it's common practice to include
1700 alternatives in parentheses: to minimize confusion about where they
1703 Alternatives are tried from left to right, so the first
1704 alternative found for which the entire expression matches, is the one that
1705 is chosen. This means that alternatives are not necessarily greedy. For
1706 example: when matching C<foo|foot> against "barefoot", only the "foo"
1707 part will match, as that is the first alternative tried, and it successfully
1708 matches the target string. (This might not seem important, but it is
1709 important when you are capturing matched text using parentheses.)
1711 Also remember that "|" is interpreted as a literal within square brackets,
1712 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1714 Within a pattern, you may designate subpatterns for later reference
1715 by enclosing them in parentheses, and you may refer back to the
1716 I<n>th subpattern later in the pattern using the metacharacter
1717 \I<n>. Subpatterns are numbered based on the left to right order
1718 of their opening parenthesis. A backreference matches whatever
1719 actually matched the subpattern in the string being examined, not
1720 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1721 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1722 1 matched "0x", even though the rule C<0|0x> could potentially match
1723 the leading 0 in the second number.
1725 =head2 Warning on \1 vs $1
1727 Some people get too used to writing things like:
1729 $pattern =~ s/(\W)/\\\1/g;
1731 This is grandfathered for the RHS of a substitute to avoid shocking the
1732 B<sed> addicts, but it's a dirty habit to get into. That's because in
1733 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1734 the usual double-quoted string means a control-A. The customary Unix
1735 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1736 of doing that, you get yourself into trouble if you then add an C</e>
1739 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1745 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1746 C<${1}000>. The operation of interpolation should not be confused
1747 with the operation of matching a backreference. Certainly they mean two
1748 different things on the I<left> side of the C<s///>.
1750 =head2 Repeated patterns matching zero-length substring
1752 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1754 Regular expressions provide a terse and powerful programming language. As
1755 with most other power tools, power comes together with the ability
1758 A common abuse of this power stems from the ability to make infinite
1759 loops using regular expressions, with something as innocuous as:
1761 'foo' =~ m{ ( o? )* }x;
1763 The C<o?> can match at the beginning of C<'foo'>, and since the position
1764 in the string is not moved by the match, C<o?> would match again and again
1765 because of the C<*> modifier. Another common way to create a similar cycle
1766 is with the looping modifier C<//g>:
1768 @matches = ( 'foo' =~ m{ o? }xg );
1772 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1774 or the loop implied by split().
1776 However, long experience has shown that many programming tasks may
1777 be significantly simplified by using repeated subexpressions that
1778 may match zero-length substrings. Here's a simple example being:
1780 @chars = split //, $string; # // is not magic in split
1781 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1783 Thus Perl allows such constructs, by I<forcefully breaking
1784 the infinite loop>. The rules for this are different for lower-level
1785 loops given by the greedy modifiers C<*+{}>, and for higher-level
1786 ones like the C</g> modifier or split() operator.
1788 The lower-level loops are I<interrupted> (that is, the loop is
1789 broken) when Perl detects that a repeated expression matched a
1790 zero-length substring. Thus
1792 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1794 is made equivalent to
1796 m{ (?: NON_ZERO_LENGTH )*
1801 The higher level-loops preserve an additional state between iterations:
1802 whether the last match was zero-length. To break the loop, the following
1803 match after a zero-length match is prohibited to have a length of zero.
1804 This prohibition interacts with backtracking (see L<"Backtracking">),
1805 and so the I<second best> match is chosen if the I<best> match is of
1813 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1814 match given by non-greedy C<??> is the zero-length match, and the I<second
1815 best> match is what is matched by C<\w>. Thus zero-length matches
1816 alternate with one-character-long matches.
1818 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1819 position one notch further in the string.
1821 The additional state of being I<matched with zero-length> is associated with
1822 the matched string, and is reset by each assignment to pos().
1823 Zero-length matches at the end of the previous match are ignored
1826 =head2 Combining pieces together
1828 Each of the elementary pieces of regular expressions which were described
1829 before (such as C<ab> or C<\Z>) could match at most one substring
1830 at the given position of the input string. However, in a typical regular
1831 expression these elementary pieces are combined into more complicated
1832 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1833 (in these examples C<S> and C<T> are regular subexpressions).
1835 Such combinations can include alternatives, leading to a problem of choice:
1836 if we match a regular expression C<a|ab> against C<"abc">, will it match
1837 substring C<"a"> or C<"ab">? One way to describe which substring is
1838 actually matched is the concept of backtracking (see L<"Backtracking">).
1839 However, this description is too low-level and makes you think
1840 in terms of a particular implementation.
1842 Another description starts with notions of "better"/"worse". All the
1843 substrings which may be matched by the given regular expression can be
1844 sorted from the "best" match to the "worst" match, and it is the "best"
1845 match which is chosen. This substitutes the question of "what is chosen?"
1846 by the question of "which matches are better, and which are worse?".
1848 Again, for elementary pieces there is no such question, since at most
1849 one match at a given position is possible. This section describes the
1850 notion of better/worse for combining operators. In the description
1851 below C<S> and C<T> are regular subexpressions.
1857 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1858 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1859 which can be matched by C<T>.
1861 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1864 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1865 C<B> is better match for C<T> than C<B'>.
1869 When C<S> can match, it is a better match than when only C<T> can match.
1871 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1872 two matches for C<T>.
1874 =item C<S{REPEAT_COUNT}>
1876 Matches as C<SSS...S> (repeated as many times as necessary).
1880 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1882 =item C<S{min,max}?>
1884 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1886 =item C<S?>, C<S*>, C<S+>
1888 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1890 =item C<S??>, C<S*?>, C<S+?>
1892 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1896 Matches the best match for C<S> and only that.
1898 =item C<(?=S)>, C<(?<=S)>
1900 Only the best match for C<S> is considered. (This is important only if
1901 C<S> has capturing parentheses, and backreferences are used somewhere
1902 else in the whole regular expression.)
1904 =item C<(?!S)>, C<(?<!S)>
1906 For this grouping operator there is no need to describe the ordering, since
1907 only whether or not C<S> can match is important.
1909 =item C<(??{ EXPR })>, C<(?PARNO)>
1911 The ordering is the same as for the regular expression which is
1912 the result of EXPR, or the pattern contained by capture buffer PARNO.
1914 =item C<(?(condition)yes-pattern|no-pattern)>
1916 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1917 already determined. The ordering of the matches is the same as for the
1918 chosen subexpression.
1922 The above recipes describe the ordering of matches I<at a given position>.
1923 One more rule is needed to understand how a match is determined for the
1924 whole regular expression: a match at an earlier position is always better
1925 than a match at a later position.
1927 =head2 Creating custom RE engines
1929 Overloaded constants (see L<overload>) provide a simple way to extend
1930 the functionality of the RE engine.
1932 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
1933 matches at boundary between whitespace characters and non-whitespace
1934 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1935 at these positions, so we want to have each C<\Y|> in the place of the
1936 more complicated version. We can create a module C<customre> to do
1944 die "No argument to customre::import allowed" if @_;
1945 overload::constant 'qr' => \&convert;
1948 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1950 # We must also take care of not escaping the legitimate \\Y|
1951 # sequence, hence the presence of '\\' in the conversion rules.
1952 my %rules = ( '\\' => '\\\\',
1953 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
1959 { $rules{$1} or invalid($re,$1) }sgex;
1963 Now C<use customre> enables the new escape in constant regular
1964 expressions, i.e., those without any runtime variable interpolations.
1965 As documented in L<overload>, this conversion will work only over
1966 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
1967 part of this regular expression needs to be converted explicitly
1968 (but only if the special meaning of C<\Y|> should be enabled inside $re):
1973 $re = customre::convert $re;
1978 This document varies from difficult to understand to completely
1979 and utterly opaque. The wandering prose riddled with jargon is
1980 hard to fathom in several places.
1982 This document needs a rewrite that separates the tutorial content
1983 from the reference content.
1991 L<perlop/"Regexp Quote-Like Operators">.
1993 L<perlop/"Gory details of parsing quoted constructs">.
2003 I<Mastering Regular Expressions> by Jeffrey Friedl, published
2004 by O'Reilly and Associates.