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 a specific group.
249 '1' may actually be any positive integer
250 \k<name> Named backreference
251 \N{name} Named unicode character, or unicode escape.
252 \x12 Hexadecimal escape sequence
253 \x{1234} Long hexadecimal escape sequence
255 A C<\w> matches a single alphanumeric character (an alphabetic
256 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
257 to match a string of Perl-identifier characters (which isn't the same
258 as matching an English word). If C<use locale> is in effect, the list
259 of alphabetic characters generated by C<\w> is taken from the current
260 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
261 C<\d>, and C<\D> within character classes, but if you try to use them
262 as endpoints of a range, that's not a range, the "-" is understood
263 literally. If Unicode is in effect, C<\s> matches also "\x{85}",
264 "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about
265 C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general.
266 You can define your own C<\p> and C<\P> properties, see L<perlunicode>.
269 The POSIX character class syntax
274 is also available. Note that the C<[> and C<]> braces are I<literal>;
275 they must always be used within a character class expression.
278 $string =~ /[[:alpha:]]/;
280 # this is not, and will generate a warning:
281 $string =~ /[:alpha:]/;
283 The available classes and their backslash equivalents (if available) are
286 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
287 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
308 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
312 Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
313 also the (very rare) "vertical tabulator", "\ck", chr(11).
317 A Perl extension, see above.
321 For example use C<[:upper:]> to match all the uppercase characters.
322 Note that the C<[]> are part of the C<[::]> construct, not part of the
323 whole character class. For example:
327 matches zero, one, any alphabetic character, and the percentage sign.
329 The following equivalences to Unicode \p{} constructs and equivalent
330 backslash character classes (if available), will hold:
331 X<character class> X<\p> X<\p{}>
333 [[:...:]] \p{...} backslash
351 For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent.
353 If the C<utf8> pragma is not used but the C<locale> pragma is, the
354 classes correlate with the usual isalpha(3) interface (except for
357 The assumedly non-obviously named classes are:
364 Any control character. Usually characters that don't produce output as
365 such but instead control the terminal somehow: for example newline and
366 backspace are control characters. All characters with ord() less than
367 32 are most often classified as control characters (assuming ASCII,
368 the ISO Latin character sets, and Unicode), as is the character with
369 the ord() value of 127 (C<DEL>).
374 Any alphanumeric or punctuation (special) character.
379 Any alphanumeric or punctuation (special) character or the space character.
384 Any punctuation (special) character.
389 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
390 work just fine) it is included for completeness.
394 You can negate the [::] character classes by prefixing the class name
395 with a '^'. This is a Perl extension. For example:
396 X<character class, negation>
398 POSIX traditional Unicode
400 [[:^digit:]] \D \P{IsDigit}
401 [[:^space:]] \S \P{IsSpace}
402 [[:^word:]] \W \P{IsWord}
404 Perl respects the POSIX standard in that POSIX character classes are
405 only supported within a character class. The POSIX character classes
406 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
407 use them will cause an error.
411 Perl defines the following zero-width assertions:
412 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
413 X<regexp, zero-width assertion>
414 X<regular expression, zero-width assertion>
415 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
417 \b Match a word boundary
418 \B Match a non-(word boundary)
419 \A Match only at beginning of string
420 \Z Match only at end of string, or before newline at the end
421 \z Match only at end of string
422 \G Match only at pos() (e.g. at the end-of-match position
425 A word boundary (C<\b>) is a spot between two characters
426 that has a C<\w> on one side of it and a C<\W> on the other side
427 of it (in either order), counting the imaginary characters off the
428 beginning and end of the string as matching a C<\W>. (Within
429 character classes C<\b> represents backspace rather than a word
430 boundary, just as it normally does in any double-quoted string.)
431 The C<\A> and C<\Z> are just like "^" and "$", except that they
432 won't match multiple times when the C</m> modifier is used, while
433 "^" and "$" will match at every internal line boundary. To match
434 the actual end of the string and not ignore an optional trailing
436 X<\b> X<\A> X<\Z> X<\z> X</m>
438 The C<\G> assertion can be used to chain global matches (using
439 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
440 It is also useful when writing C<lex>-like scanners, when you have
441 several patterns that you want to match against consequent substrings
442 of your string, see the previous reference. The actual location
443 where C<\G> will match can also be influenced by using C<pos()> as
444 an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully
445 supported when anchored to the start of the pattern; while it
446 is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some
447 such uses (C</.\G/g>, for example) currently cause problems, and
448 it is recommended that you avoid such usage for now.
451 =head3 Capture buffers
453 The bracketing construct C<( ... )> creates capture buffers. To
454 refer to the digit'th buffer use \<digit> within the
455 match. Outside the match use "$" instead of "\". (The
456 \<digit> notation works in certain circumstances outside
457 the match. See the warning below about \1 vs $1 for details.)
458 Referring back to another part of the match is called a
460 X<regex, capture buffer> X<regexp, capture buffer>
461 X<regular expression, capture buffer> X<backreference>
463 There is no limit to the number of captured substrings that you may
464 use. However Perl also uses \10, \11, etc. as aliases for \010,
465 \011, etc. (Recall that 0 means octal, so \011 is the character at
466 number 9 in your coded character set; which would be the 10th character,
467 a horizontal tab under ASCII.) Perl resolves this
468 ambiguity by interpreting \10 as a backreference only if at least 10
469 left parentheses have opened before it. Likewise \11 is a
470 backreference only if at least 11 left parentheses have opened
471 before it. And so on. \1 through \9 are always interpreted as
474 Additionally, as of Perl 5.10 you may use named capture buffers and named
475 backreferences. The notation is C<< (?<name>...) >> and C<< \k<name> >>
476 (you may also use single quotes instead of angle brackets to quote the
477 name). The only difference with named capture buffers and unnamed ones is
478 that multiple buffers may have the same name and that the contents of
479 named capture buffers is available via the C<%+> hash. When multiple
480 groups share the same name C<$+{name}> and C<< \k<name> >> refer to the
481 leftmost defined group, thus it's possible to do things with named capture
482 buffers that would otherwise require C<(??{})> code to accomplish. Named
483 capture buffers are numbered just as normal capture buffers are and may be
484 referenced via the magic numeric variables or via numeric backreferences
489 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
491 /(.)\1/ # find first doubled char
492 and print "'$1' is the first doubled character\n";
494 /(?<char>.)\k<char>/ # ... a different way
495 and print "'$+{char}' is the first doubled character\n";
497 /(?<char>.)\1/ # ... mix and match
498 and print "'$1' is the first doubled character\n";
500 if (/Time: (..):(..):(..)/) { # parse out values
506 Several special variables also refer back to portions of the previous
507 match. C<$+> returns whatever the last bracket match matched.
508 C<$&> returns the entire matched string. (At one point C<$0> did
509 also, but now it returns the name of the program.) C<$`> returns
510 everything before the matched string. C<$'> returns everything
511 after the matched string. And C<$^N> contains whatever was matched by
512 the most-recently closed group (submatch). C<$^N> can be used in
513 extended patterns (see below), for example to assign a submatch to a
515 X<$+> X<$^N> X<$&> X<$`> X<$'>
517 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
518 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
519 until the end of the enclosing block or until the next successful
520 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
521 X<$+> X<$^N> X<$&> X<$`> X<$'>
522 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
525 B<NOTE>: failed matches in Perl do not reset the match variables,
526 which makes it easier to write code that tests for a series of more
527 specific cases and remembers the best match.
529 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
530 C<$'> anywhere in the program, it has to provide them for every
531 pattern match. This may substantially slow your program. Perl
532 uses the same mechanism to produce $1, $2, etc, so you also pay a
533 price for each pattern that contains capturing parentheses. (To
534 avoid this cost while retaining the grouping behaviour, use the
535 extended regular expression C<(?: ... )> instead.) But if you never
536 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
537 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
538 if you can, but if you can't (and some algorithms really appreciate
539 them), once you've used them once, use them at will, because you've
540 already paid the price. As of 5.005, C<$&> is not so costly as the
544 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
545 C<\w>, C<\n>. Unlike some other regular expression languages, there
546 are no backslashed symbols that aren't alphanumeric. So anything
547 that looks like \\, \(, \), \<, \>, \{, or \} is always
548 interpreted as a literal character, not a metacharacter. This was
549 once used in a common idiom to disable or quote the special meanings
550 of regular expression metacharacters in a string that you want to
551 use for a pattern. Simply quote all non-"word" characters:
553 $pattern =~ s/(\W)/\\$1/g;
555 (If C<use locale> is set, then this depends on the current locale.)
556 Today it is more common to use the quotemeta() function or the C<\Q>
557 metaquoting escape sequence to disable all metacharacters' special
560 /$unquoted\Q$quoted\E$unquoted/
562 Beware that if you put literal backslashes (those not inside
563 interpolated variables) between C<\Q> and C<\E>, double-quotish
564 backslash interpolation may lead to confusing results. If you
565 I<need> to use literal backslashes within C<\Q...\E>,
566 consult L<perlop/"Gory details of parsing quoted constructs">.
568 =head2 Extended Patterns
570 Perl also defines a consistent extension syntax for features not
571 found in standard tools like B<awk> and B<lex>. The syntax is a
572 pair of parentheses with a question mark as the first thing within
573 the parentheses. The character after the question mark indicates
576 The stability of these extensions varies widely. Some have been
577 part of the core language for many years. Others are experimental
578 and may change without warning or be completely removed. Check
579 the documentation on an individual feature to verify its current
582 A question mark was chosen for this and for the minimal-matching
583 construct because 1) question marks are rare in older regular
584 expressions, and 2) whenever you see one, you should stop and
585 "question" exactly what is going on. That's psychology...
592 A comment. The text is ignored. If the C</x> modifier enables
593 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
594 the comment as soon as it sees a C<)>, so there is no way to put a literal
597 =item C<(?imsx-imsx)>
600 One or more embedded pattern-match modifiers, to be turned on (or
601 turned off, if preceded by C<->) for the remainder of the pattern or
602 the remainder of the enclosing pattern group (if any). This is
603 particularly useful for dynamic patterns, such as those read in from a
604 configuration file, read in as an argument, are specified in a table
605 somewhere, etc. Consider the case that some of which want to be case
606 sensitive and some do not. The case insensitive ones need to include
607 merely C<(?i)> at the front of the pattern. For example:
610 if ( /$pattern/i ) { }
614 $pattern = "(?i)foobar";
615 if ( /$pattern/ ) { }
617 These modifiers are restored at the end of the enclosing group. For example,
621 will match a repeated (I<including the case>!) word C<blah> in any
622 case, assuming C<x> modifier, and no C<i> modifier outside this
628 =item C<(?imsx-imsx:pattern)>
630 This is for clustering, not capturing; it groups subexpressions like
631 "()", but doesn't make backreferences as "()" does. So
633 @fields = split(/\b(?:a|b|c)\b/)
637 @fields = split(/\b(a|b|c)\b/)
639 but doesn't spit out extra fields. It's also cheaper not to capture
640 characters if you don't need to.
642 Any letters between C<?> and C<:> act as flags modifiers as with
643 C<(?imsx-imsx)>. For example,
645 /(?s-i:more.*than).*million/i
647 is equivalent to the more verbose
649 /(?:(?s-i)more.*than).*million/i
652 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
654 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
655 matches a word followed by a tab, without including the tab in C<$&>.
658 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
660 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
661 matches any occurrence of "foo" that isn't followed by "bar". Note
662 however that look-ahead and look-behind are NOT the same thing. You cannot
663 use this for look-behind.
665 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
666 will not do what you want. That's because the C<(?!foo)> is just saying that
667 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
668 match. You would have to do something like C</(?!foo)...bar/> for that. We
669 say "like" because there's the case of your "bar" not having three characters
670 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
671 Sometimes it's still easier just to say:
673 if (/bar/ && $` !~ /foo$/)
675 For look-behind see below.
677 =item C<(?<=pattern)>
678 X<(?<=)> X<look-behind, positive> X<lookbehind, positive>
680 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
681 matches a word that follows a tab, without including the tab in C<$&>.
682 Works only for fixed-width look-behind.
684 =item C<(?<!pattern)>
685 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
687 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
688 matches any occurrence of "foo" that does not follow "bar". Works
689 only for fixed-width look-behind.
691 =item C<(?'NAME'pattern)>
693 =item C<< (?<NAME>pattern) >>
694 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
696 A named capture buffer. Identical in every respect to normal capturing
697 parens C<()> but for the additional fact that C<%+> may be used after
698 a succesful match to refer to a named buffer. See C<perlvar> for more
699 details on the C<%+> hash.
701 If multiple distinct capture buffers have the same name then the
702 $+{NAME} will refer to the leftmost defined buffer in the match.
704 The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent.
706 B<NOTE:> While the notation of this construct is the same as the similar
707 function in .NET regexes, the behavior is not, in Perl the buffers are
708 numbered sequentially regardless of being named or not. Thus in the
713 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
714 the opposite which is what a .NET regex hacker might expect.
716 Currently NAME is restricted to word chars only. In other words, it
717 must match C</^\w+$/>.
719 =item C<< \k<name> >>
721 =item C<< \k'name' >>
723 Named backreference. Similar to numeric backreferences, except that
724 the group is designated by name and not number. If multiple groups
725 have the same name then it refers to the leftmost defined group in
728 It is an error to refer to a name not defined by a C<(?<NAME>)>
729 earlier in the pattern.
731 Both forms are equivalent.
734 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
736 B<WARNING>: This extended regular expression feature is considered
737 experimental, and may be changed without notice. Code executed that
738 has side effects may not perform identically from version to version
739 due to the effect of future optimisations in the regex engine.
741 This zero-width assertion evaluates any embedded Perl code. It
742 always succeeds, and its C<code> is not interpolated. Currently,
743 the rules to determine where the C<code> ends are somewhat convoluted.
745 This feature can be used together with the special variable C<$^N> to
746 capture the results of submatches in variables without having to keep
747 track of the number of nested parentheses. For example:
749 $_ = "The brown fox jumps over the lazy dog";
750 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
751 print "color = $color, animal = $animal\n";
753 Inside the C<(?{...})> block, C<$_> refers to the string the regular
754 expression is matching against. You can also use C<pos()> to know what is
755 the current position of matching within this string.
757 The C<code> is properly scoped in the following sense: If the assertion
758 is backtracked (compare L<"Backtracking">), all changes introduced after
759 C<local>ization are undone, so that
763 (?{ $cnt = 0 }) # Initialize $cnt.
767 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
771 (?{ $res = $cnt }) # On success copy to non-localized
775 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
776 introduced value, because the scopes that restrict C<local> operators
779 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
780 switch. If I<not> used in this way, the result of evaluation of
781 C<code> is put into the special variable C<$^R>. This happens
782 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
783 inside the same regular expression.
785 The assignment to C<$^R> above is properly localized, so the old
786 value of C<$^R> is restored if the assertion is backtracked; compare
789 Due to an unfortunate implementation issue, the Perl code contained in these
790 blocks is treated as a compile time closure that can have seemingly bizarre
791 consequences when used with lexically scoped variables inside of subroutines
792 or loops. There are various workarounds for this, including simply using
793 global variables instead. If you are using this construct and strange results
794 occur then check for the use of lexically scoped variables.
796 For reasons of security, this construct is forbidden if the regular
797 expression involves run-time interpolation of variables, unless the
798 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
799 variables contain results of C<qr//> operator (see
800 L<perlop/"qr/STRING/imosx">).
802 This restriction is because of the wide-spread and remarkably convenient
803 custom of using run-time determined strings as patterns. For example:
809 Before Perl knew how to execute interpolated code within a pattern,
810 this operation was completely safe from a security point of view,
811 although it could raise an exception from an illegal pattern. If
812 you turn on the C<use re 'eval'>, though, it is no longer secure,
813 so you should only do so if you are also using taint checking.
814 Better yet, use the carefully constrained evaluation within a Safe
815 compartment. See L<perlsec> for details about both these mechanisms.
817 Because perl's regex engine is not currently re-entrant, interpolated
818 code may not invoke the regex engine either directly with C<m//> or C<s///>),
819 or indirectly with functions such as C<split>.
821 =item C<(??{ code })>
823 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
825 B<WARNING>: This extended regular expression feature is considered
826 experimental, and may be changed without notice. Code executed that
827 has side effects may not perform identically from version to version
828 due to the effect of future optimisations in the regex engine.
830 This is a "postponed" regular subexpression. The C<code> is evaluated
831 at run time, at the moment this subexpression may match. The result
832 of evaluation is considered as a regular expression and matched as
833 if it were inserted instead of this construct. Note that this means
834 that the contents of capture buffers defined inside an eval'ed pattern
835 are not available outside of the pattern, and vice versa, there is no
836 way for the inner pattern to refer to a capture buffer defined outside.
839 ('a' x 100)=~/(??{'(.)' x 100})/
841 B<will> match, it will B<not> set $1.
843 The C<code> is not interpolated. As before, the rules to determine
844 where the C<code> ends are currently somewhat convoluted.
846 The following pattern matches a parenthesized group:
851 (?> [^()]+ ) # Non-parens without backtracking
853 (??{ $re }) # Group with matching parens
858 See also C<(?PARNO)> for a different, more efficient way to accomplish
861 Because perl's regex engine is not currently re-entrant, delayed
862 code may not invoke the regex engine either directly with C<m//> or C<s///>),
863 or indirectly with functions such as C<split>.
865 Recursing deeper than 50 times without consuming any input string will
866 result in a fatal error. The maximum depth is compiled into perl, so
867 changing it requires a custom build.
869 =item C<(?PARNO)> C<(?R)> C<(?0)>
870 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)>
871 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
873 Similar to C<(??{ code })> except it does not involve compiling any code,
874 instead it treats the contents of a capture buffer as an independent
875 pattern that must match at the current position. Capture buffers
876 contained by the pattern will have the value as determined by the
879 PARNO is a sequence of digits (not starting with 0) whose value reflects
880 the paren-number of the capture buffer to recurse to. C<(?R)> recurses to
881 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
884 The following pattern matches a function foo() which may contain
885 balanced parentheses as the argument.
887 $re = qr{ ( # paren group 1 (full function)
889 ( # paren group 2 (parens)
891 ( # paren group 3 (contents of parens)
893 (?> [^()]+ ) # Non-parens without backtracking
895 (?2) # Recurse to start of paren group 2
903 If the pattern was used as follows
905 'foo(bar(baz)+baz(bop))'=~/$re/
906 and print "\$1 = $1\n",
910 the output produced should be the following:
912 $1 = foo(bar(baz)+baz(bop))
913 $2 = (bar(baz)+baz(bop))
914 $3 = bar(baz)+baz(bop)
916 If there is no corresponding capture buffer defined, then it is a
917 fatal error. Recursing deeper than 50 times without consuming any input
918 string will also result in a fatal error. The maximum depth is compiled
919 into perl, so changing it requires a custom build.
921 B<Note> that this pattern does not behave the same way as the equivalent
922 PCRE or Python construct of the same form. In perl you can backtrack into
923 a recursed group, in PCRE and Python the recursed into group is treated
924 as atomic. Also, constructs like (?i:(?1)) or (?:(?i)(?1)) do not affect
925 the pattern being recursed into.
930 Recurse to a named subpattern. Identical to (?PARNO) except that the
931 parenthesis to recurse to is determined by name. If multiple parens have
932 the same name, then it recurses to the leftmost.
934 It is an error to refer to a name that is not declared somewhere in the
937 =item C<(?(condition)yes-pattern|no-pattern)>
940 =item C<(?(condition)yes-pattern)>
942 Conditional expression. C<(condition)> should be either an integer in
943 parentheses (which is valid if the corresponding pair of parentheses
944 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
945 name in angle brackets or single quotes (which is valid if a buffer
946 with the given name matched), or the special symbol (R) (true when
947 evaluated inside of recursion or eval). Additionally the R may be
948 followed by a number, (which will be true when evaluated when recursing
949 inside of the appropriate group), or by C<&NAME>, in which case it will
950 be true only when evaluated during recursion in the named group.
952 Here's a summary of the possible predicates:
958 Checks if the numbered capturing buffer has matched something.
960 =item (<NAME>) ('NAME')
962 Checks if a buffer with the given name has matched something.
966 Treats the code block as the condition.
970 Checks if the expression has been evaluated inside of recursion.
974 Checks if the expression has been evaluated while executing directly
975 inside of the n-th capture group. This check is the regex equivalent of
977 if ((caller(0))[3] eq 'subname') { ... }
979 In other words, it does not check the full recursion stack.
983 Similar to C<(R1)>, this predicate checks to see if we're executing
984 directly inside of the leftmost group with a given name (this is the same
985 logic used by C<(?&NAME)> to disambiguate). It does not check the full
986 stack, but only the name of the innermost active recursion.
990 In this case, the yes-pattern is never directly executed, and no
991 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
992 See below for details.
1003 matches a chunk of non-parentheses, possibly included in parentheses
1006 A special form is the C<(DEFINE)> predicate, which never executes directly
1007 its yes-pattern, and does not allow a no-pattern. This allows to define
1008 subpatterns which will be executed only by using the recursion mechanism.
1009 This way, you can define a set of regular expression rules that can be
1010 bundled into any pattern you choose.
1012 It is recommended that for this usage you put the DEFINE block at the
1013 end of the pattern, and that you name any subpatterns defined within it.
1015 Also, it's worth noting that patterns defined this way probably will
1016 not be as efficient, as the optimiser is not very clever about
1019 An example of how this might be used is as follows:
1021 /(?<NAME>(&NAME_PAT))(?<ADDR>(&ADDRESS_PAT))
1027 Note that capture buffers matched inside of recursion are not accessible
1028 after the recursion returns, so the extra layer of capturing buffers are
1029 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1030 C<$+{NAME}> would be.
1032 =item C<< (?>pattern) >>
1033 X<backtrack> X<backtracking> X<atomic> X<possessive>
1035 An "independent" subexpression, one which matches the substring
1036 that a I<standalone> C<pattern> would match if anchored at the given
1037 position, and it matches I<nothing other than this substring>. This
1038 construct is useful for optimizations of what would otherwise be
1039 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1040 It may also be useful in places where the "grab all you can, and do not
1041 give anything back" semantic is desirable.
1043 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1044 (anchored at the beginning of string, as above) will match I<all>
1045 characters C<a> at the beginning of string, leaving no C<a> for
1046 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1047 since the match of the subgroup C<a*> is influenced by the following
1048 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1049 C<a*ab> will match fewer characters than a standalone C<a*>, since
1050 this makes the tail match.
1052 An effect similar to C<< (?>pattern) >> may be achieved by writing
1053 C<(?=(pattern))\1>. This matches the same substring as a standalone
1054 C<a+>, and the following C<\1> eats the matched string; it therefore
1055 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1056 (The difference between these two constructs is that the second one
1057 uses a capturing group, thus shifting ordinals of backreferences
1058 in the rest of a regular expression.)
1060 Consider this pattern:
1071 That will efficiently match a nonempty group with matching parentheses
1072 two levels deep or less. However, if there is no such group, it
1073 will take virtually forever on a long string. That's because there
1074 are so many different ways to split a long string into several
1075 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1076 to a subpattern of the above pattern. Consider how the pattern
1077 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1078 seconds, but that each extra letter doubles this time. This
1079 exponential performance will make it appear that your program has
1080 hung. However, a tiny change to this pattern
1084 (?> [^()]+ ) # change x+ above to (?> x+ )
1091 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1092 this yourself would be a productive exercise), but finishes in a fourth
1093 the time when used on a similar string with 1000000 C<a>s. Be aware,
1094 however, that this pattern currently triggers a warning message under
1095 the C<use warnings> pragma or B<-w> switch saying it
1096 C<"matches null string many times in regex">.
1098 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1099 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1100 This was only 4 times slower on a string with 1000000 C<a>s.
1102 The "grab all you can, and do not give anything back" semantic is desirable
1103 in many situations where on the first sight a simple C<()*> looks like
1104 the correct solution. Suppose we parse text with comments being delimited
1105 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1106 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1107 the comment delimiter, because it may "give up" some whitespace if
1108 the remainder of the pattern can be made to match that way. The correct
1109 answer is either one of these:
1114 For example, to grab non-empty comments into $1, one should use either
1117 / (?> \# [ \t]* ) ( .+ ) /x;
1118 / \# [ \t]* ( [^ \t] .* ) /x;
1120 Which one you pick depends on which of these expressions better reflects
1121 the above specification of comments.
1123 In some literature this construct is called "atomic matching" or
1124 "possessive matching".
1126 Possessive quantifiers are equivalent to putting the item they are applied
1127 to inside of one of these constructs. The following equivalences apply:
1129 Quantifier Form Bracketing Form
1130 --------------- ---------------
1134 PAT{min,max}+ (?>PAT{min,max})
1138 =head2 Special Backtracking Control Verbs
1140 B<WARNING:> These patterns are experimental and subject to change or
1141 removal in a future version of perl. Their usage in production code should
1142 be noted to avoid problems during upgrades.
1144 These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1145 otherwise stated the ARG argument is optional; in some cases, it is
1148 Any pattern containing a special backtracking verb that allows an argument
1149 has the special behaviour that when executed it sets the current packages'
1150 C<$REGERROR> variable. In this case, the following rules apply:
1152 On failure, this variable will be set to the ARG value of the verb
1153 pattern, if the verb was involved in the failure of the match. If the ARG
1154 part of the pattern was omitted, then C<$REGERROR> will be set to TRUE.
1156 On a successful match this variable will be set to FALSE.
1158 B<NOTE:> C<$REGERROR> is not a magic variable in the same sense than
1159 C<$1> and most other regex related variables. It is not local to a
1160 scope, nor readonly but instead a volatile package variable similar to
1161 C<$AUTOLOAD>. Use C<local> to localize changes to it to a specific scope
1164 If a pattern does not contain a special backtracking verb that allows an
1165 argument, then C<$REGERROR> is not touched at all.
1169 =item Verbs that take an argument
1173 =item C<(*NOMATCH)> C<(*NOMATCH:NAME)>
1174 X<(*NOMATCH)> X<(*NOMATCH:NAME)>
1176 This zero-width pattern commits the match at the current point, preventing
1177 the engine from backtracking on failure to the left of the this point.
1178 Consider the pattern C<A (*NOMATCH) B>, where A and B are complex patterns.
1179 Until the C<(*NOMATCH)> is reached, A may backtrack as necessary to match.
1180 Once it is reached, matching continues in B, which may also backtrack as
1181 necessary; however, should B not match, then no further backtracking will
1182 take place, and the pattern will fail outright at that starting position.
1184 The following example counts all the possible matching strings in a
1185 pattern (without actually matching any of them).
1187 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1188 print "Count=$count\n";
1203 If we add a C<(*NOMATCH)> before the count like the following
1205 'aaab' =~ /a+b?(*NOMATCH)(?{print "$&\n"; $count++})(*FAIL)/;
1206 print "Count=$count\n";
1208 we prevent backtracking and find the count of the longest matching
1209 at each matching startpoint like so:
1216 Any number of C<(*NOMATCH)> assertions may be used in a pattern.
1218 See also C<< (?>pattern) >> and possessive quantifiers for other
1219 ways to control backtracking.
1221 =item C<(*MARK)> C<(*MARK:NAME)>
1224 This zero-width pattern can be used to mark the point in a string
1225 reached when a certain part of the pattern has been successfully
1226 matched. This mark may be given a name. A later C<(*CUT)> pattern
1227 will then cut at that point if backtracked into on failure. Any
1228 number of (*MARK) patterns are allowed, and the NAME portion is
1229 optional and may be duplicated.
1231 See C<*CUT> for more detail.
1233 =item C<(*CUT)> C<(*CUT:NAME)>
1236 This zero-width pattern is similar to C<(*NOMATCH)>, except that on
1237 failure it also signifies that whatever text that was matched leading up
1238 to the C<(*CUT)> pattern being executed cannot be part of a match, I<even
1239 if started from a later point>. This effectively means that the regex
1240 engine moves forward to this position on failure and tries to match
1241 again, (assuming that there is sufficient room to match).
1243 The name of the C<(*CUT:NAME)> pattern has special significance. If a
1244 C<(*MARK:NAME)> was encountered while matching, then it is the position
1245 where that pattern was executed that is used for the "cut point" in the
1246 string. If no mark of that name was encountered, then the cut is done at
1247 the point where the C<(*CUT)> was. Similarly if no NAME is specified in
1248 the C<(*CUT)>, and if a C<(*MARK)> with any name (or none) is encountered,
1249 then that C<(*MARK)>'s cursor point will be used. If the C<(*CUT)> is not
1250 preceded by a C<(*MARK)>, then the cut point is where the string was when
1251 the C<(*CUT)> was encountered.
1253 Compare the following to the examples in C<(*NOMATCH)>, note the string
1256 'aaabaaab' =~ /a+b?(*CUT)(?{print "$&\n"; $count++})(*FAIL)/;
1257 print "Count=$count\n";
1265 Once the 'aaab' at the start of the string has matched, and the C<(*CUT)>
1266 executed, the next startpoint will be where the cursor was when the
1267 C<(*CUT)> was executed.
1272 This zero-width pattern is similar to C<(*CUT)> except that it causes
1273 the match to fail outright. No attempts to match will occur again.
1275 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
1276 print "Count=$count\n";
1283 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1284 does not match, the regex engine will not try any further matching on the
1289 =item Verbs without an argument
1293 =item C<(*FAIL)> C<(*F)>
1296 This pattern matches nothing and always fails. It can be used to force the
1297 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1298 fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
1300 It is probably useful only when combined with C<(?{})> or C<(??{})>.
1305 B<WARNING:> This feature is highly experimental. It is not recommended
1306 for production code.
1308 This pattern matches nothing and causes the end of successful matching at
1309 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1310 whether there is actually more to match in the string. When inside of a
1311 nested pattern, such as recursion or a dynamically generated subbpattern
1312 via C<(??{})>, only the innermost pattern is ended immediately.
1314 If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are
1315 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1318 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
1320 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
1321 be set. If another branch in the inner parens were matched, such as in the
1322 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
1329 X<backtrack> X<backtracking>
1331 NOTE: This section presents an abstract approximation of regular
1332 expression behavior. For a more rigorous (and complicated) view of
1333 the rules involved in selecting a match among possible alternatives,
1334 see L<Combining pieces together>.
1336 A fundamental feature of regular expression matching involves the
1337 notion called I<backtracking>, which is currently used (when needed)
1338 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
1339 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1340 internally, but the general principle outlined here is valid.
1342 For a regular expression to match, the I<entire> regular expression must
1343 match, not just part of it. So if the beginning of a pattern containing a
1344 quantifier succeeds in a way that causes later parts in the pattern to
1345 fail, the matching engine backs up and recalculates the beginning
1346 part--that's why it's called backtracking.
1348 Here is an example of backtracking: Let's say you want to find the
1349 word following "foo" in the string "Food is on the foo table.":
1351 $_ = "Food is on the foo table.";
1352 if ( /\b(foo)\s+(\w+)/i ) {
1353 print "$2 follows $1.\n";
1356 When the match runs, the first part of the regular expression (C<\b(foo)>)
1357 finds a possible match right at the beginning of the string, and loads up
1358 $1 with "Foo". However, as soon as the matching engine sees that there's
1359 no whitespace following the "Foo" that it had saved in $1, it realizes its
1360 mistake and starts over again one character after where it had the
1361 tentative match. This time it goes all the way until the next occurrence
1362 of "foo". The complete regular expression matches this time, and you get
1363 the expected output of "table follows foo."
1365 Sometimes minimal matching can help a lot. Imagine you'd like to match
1366 everything between "foo" and "bar". Initially, you write something
1369 $_ = "The food is under the bar in the barn.";
1370 if ( /foo(.*)bar/ ) {
1374 Which perhaps unexpectedly yields:
1376 got <d is under the bar in the >
1378 That's because C<.*> was greedy, so you get everything between the
1379 I<first> "foo" and the I<last> "bar". Here it's more effective
1380 to use minimal matching to make sure you get the text between a "foo"
1381 and the first "bar" thereafter.
1383 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1384 got <d is under the >
1386 Here's another example: let's say you'd like to match a number at the end
1387 of a string, and you also want to keep the preceding part of the match.
1390 $_ = "I have 2 numbers: 53147";
1391 if ( /(.*)(\d*)/ ) { # Wrong!
1392 print "Beginning is <$1>, number is <$2>.\n";
1395 That won't work at all, because C<.*> was greedy and gobbled up the
1396 whole string. As C<\d*> can match on an empty string the complete
1397 regular expression matched successfully.
1399 Beginning is <I have 2 numbers: 53147>, number is <>.
1401 Here are some variants, most of which don't work:
1403 $_ = "I have 2 numbers: 53147";
1416 printf "%-12s ", $pat;
1418 print "<$1> <$2>\n";
1424 That will print out:
1426 (.*)(\d*) <I have 2 numbers: 53147> <>
1427 (.*)(\d+) <I have 2 numbers: 5314> <7>
1429 (.*?)(\d+) <I have > <2>
1430 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1431 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1432 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1433 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1435 As you see, this can be a bit tricky. It's important to realize that a
1436 regular expression is merely a set of assertions that gives a definition
1437 of success. There may be 0, 1, or several different ways that the
1438 definition might succeed against a particular string. And if there are
1439 multiple ways it might succeed, you need to understand backtracking to
1440 know which variety of success you will achieve.
1442 When using look-ahead assertions and negations, this can all get even
1443 trickier. Imagine you'd like to find a sequence of non-digits not
1444 followed by "123". You might try to write that as
1447 if ( /^\D*(?!123)/ ) { # Wrong!
1448 print "Yup, no 123 in $_\n";
1451 But that isn't going to match; at least, not the way you're hoping. It
1452 claims that there is no 123 in the string. Here's a clearer picture of
1453 why that pattern matches, contrary to popular expectations:
1458 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1459 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1461 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1462 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1470 You might have expected test 3 to fail because it seems to a more
1471 general purpose version of test 1. The important difference between
1472 them is that test 3 contains a quantifier (C<\D*>) and so can use
1473 backtracking, whereas test 1 will not. What's happening is
1474 that you've asked "Is it true that at the start of $x, following 0 or more
1475 non-digits, you have something that's not 123?" If the pattern matcher had
1476 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1479 The search engine will initially match C<\D*> with "ABC". Then it will
1480 try to match C<(?!123> with "123", which fails. But because
1481 a quantifier (C<\D*>) has been used in the regular expression, the
1482 search engine can backtrack and retry the match differently
1483 in the hope of matching the complete regular expression.
1485 The pattern really, I<really> wants to succeed, so it uses the
1486 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1487 time. Now there's indeed something following "AB" that is not
1488 "123". It's "C123", which suffices.
1490 We can deal with this by using both an assertion and a negation.
1491 We'll say that the first part in $1 must be followed both by a digit
1492 and by something that's not "123". Remember that the look-aheads
1493 are zero-width expressions--they only look, but don't consume any
1494 of the string in their match. So rewriting this way produces what
1495 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1497 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1498 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1502 In other words, the two zero-width assertions next to each other work as though
1503 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1504 matches only if you're at the beginning of the line AND the end of the
1505 line simultaneously. The deeper underlying truth is that juxtaposition in
1506 regular expressions always means AND, except when you write an explicit OR
1507 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1508 although the attempted matches are made at different positions because "a"
1509 is not a zero-width assertion, but a one-width assertion.
1511 B<WARNING>: particularly complicated regular expressions can take
1512 exponential time to solve because of the immense number of possible
1513 ways they can use backtracking to try match. For example, without
1514 internal optimizations done by the regular expression engine, this will
1515 take a painfully long time to run:
1517 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1519 And if you used C<*>'s in the internal groups instead of limiting them
1520 to 0 through 5 matches, then it would take forever--or until you ran
1521 out of stack space. Moreover, these internal optimizations are not
1522 always applicable. For example, if you put C<{0,5}> instead of C<*>
1523 on the external group, no current optimization is applicable, and the
1524 match takes a long time to finish.
1526 A powerful tool for optimizing such beasts is what is known as an
1527 "independent group",
1528 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1529 zero-length look-ahead/look-behind assertions will not backtrack to make
1530 the tail match, since they are in "logical" context: only
1531 whether they match is considered relevant. For an example
1532 where side-effects of look-ahead I<might> have influenced the
1533 following match, see L<C<< (?>pattern) >>>.
1535 =head2 Version 8 Regular Expressions
1536 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1538 In case you're not familiar with the "regular" Version 8 regex
1539 routines, here are the pattern-matching rules not described above.
1541 Any single character matches itself, unless it is a I<metacharacter>
1542 with a special meaning described here or above. You can cause
1543 characters that normally function as metacharacters to be interpreted
1544 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1545 character; "\\" matches a "\"). A series of characters matches that
1546 series of characters in the target string, so the pattern C<blurfl>
1547 would match "blurfl" in the target string.
1549 You can specify a character class, by enclosing a list of characters
1550 in C<[]>, which will match any one character from the list. If the
1551 first character after the "[" is "^", the class matches any character not
1552 in the list. Within a list, the "-" character specifies a
1553 range, so that C<a-z> represents all characters between "a" and "z",
1554 inclusive. If you want either "-" or "]" itself to be a member of a
1555 class, put it at the start of the list (possibly after a "^"), or
1556 escape it with a backslash. "-" is also taken literally when it is
1557 at the end of the list, just before the closing "]". (The
1558 following all specify the same class of three characters: C<[-az]>,
1559 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1560 specifies a class containing twenty-six characters, even on EBCDIC
1561 based coded character sets.) Also, if you try to use the character
1562 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1563 a range, that's not a range, the "-" is understood literally.
1565 Note also that the whole range idea is rather unportable between
1566 character sets--and even within character sets they may cause results
1567 you probably didn't expect. A sound principle is to use only ranges
1568 that begin from and end at either alphabets of equal case ([a-e],
1569 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1570 spell out the character sets in full.
1572 Characters may be specified using a metacharacter syntax much like that
1573 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1574 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1575 of octal digits, matches the character whose coded character set value
1576 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1577 matches the character whose numeric value is I<nn>. The expression \cI<x>
1578 matches the character control-I<x>. Finally, the "." metacharacter
1579 matches any character except "\n" (unless you use C</s>).
1581 You can specify a series of alternatives for a pattern using "|" to
1582 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1583 or "foe" in the target string (as would C<f(e|i|o)e>). The
1584 first alternative includes everything from the last pattern delimiter
1585 ("(", "[", or the beginning of the pattern) up to the first "|", and
1586 the last alternative contains everything from the last "|" to the next
1587 pattern delimiter. That's why it's common practice to include
1588 alternatives in parentheses: to minimize confusion about where they
1591 Alternatives are tried from left to right, so the first
1592 alternative found for which the entire expression matches, is the one that
1593 is chosen. This means that alternatives are not necessarily greedy. For
1594 example: when matching C<foo|foot> against "barefoot", only the "foo"
1595 part will match, as that is the first alternative tried, and it successfully
1596 matches the target string. (This might not seem important, but it is
1597 important when you are capturing matched text using parentheses.)
1599 Also remember that "|" is interpreted as a literal within square brackets,
1600 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1602 Within a pattern, you may designate subpatterns for later reference
1603 by enclosing them in parentheses, and you may refer back to the
1604 I<n>th subpattern later in the pattern using the metacharacter
1605 \I<n>. Subpatterns are numbered based on the left to right order
1606 of their opening parenthesis. A backreference matches whatever
1607 actually matched the subpattern in the string being examined, not
1608 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1609 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1610 1 matched "0x", even though the rule C<0|0x> could potentially match
1611 the leading 0 in the second number.
1613 =head2 Warning on \1 vs $1
1615 Some people get too used to writing things like:
1617 $pattern =~ s/(\W)/\\\1/g;
1619 This is grandfathered for the RHS of a substitute to avoid shocking the
1620 B<sed> addicts, but it's a dirty habit to get into. That's because in
1621 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1622 the usual double-quoted string means a control-A. The customary Unix
1623 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1624 of doing that, you get yourself into trouble if you then add an C</e>
1627 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1633 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1634 C<${1}000>. The operation of interpolation should not be confused
1635 with the operation of matching a backreference. Certainly they mean two
1636 different things on the I<left> side of the C<s///>.
1638 =head2 Repeated patterns matching zero-length substring
1640 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1642 Regular expressions provide a terse and powerful programming language. As
1643 with most other power tools, power comes together with the ability
1646 A common abuse of this power stems from the ability to make infinite
1647 loops using regular expressions, with something as innocuous as:
1649 'foo' =~ m{ ( o? )* }x;
1651 The C<o?> can match at the beginning of C<'foo'>, and since the position
1652 in the string is not moved by the match, C<o?> would match again and again
1653 because of the C<*> modifier. Another common way to create a similar cycle
1654 is with the looping modifier C<//g>:
1656 @matches = ( 'foo' =~ m{ o? }xg );
1660 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1662 or the loop implied by split().
1664 However, long experience has shown that many programming tasks may
1665 be significantly simplified by using repeated subexpressions that
1666 may match zero-length substrings. Here's a simple example being:
1668 @chars = split //, $string; # // is not magic in split
1669 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1671 Thus Perl allows such constructs, by I<forcefully breaking
1672 the infinite loop>. The rules for this are different for lower-level
1673 loops given by the greedy modifiers C<*+{}>, and for higher-level
1674 ones like the C</g> modifier or split() operator.
1676 The lower-level loops are I<interrupted> (that is, the loop is
1677 broken) when Perl detects that a repeated expression matched a
1678 zero-length substring. Thus
1680 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1682 is made equivalent to
1684 m{ (?: NON_ZERO_LENGTH )*
1689 The higher level-loops preserve an additional state between iterations:
1690 whether the last match was zero-length. To break the loop, the following
1691 match after a zero-length match is prohibited to have a length of zero.
1692 This prohibition interacts with backtracking (see L<"Backtracking">),
1693 and so the I<second best> match is chosen if the I<best> match is of
1701 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1702 match given by non-greedy C<??> is the zero-length match, and the I<second
1703 best> match is what is matched by C<\w>. Thus zero-length matches
1704 alternate with one-character-long matches.
1706 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1707 position one notch further in the string.
1709 The additional state of being I<matched with zero-length> is associated with
1710 the matched string, and is reset by each assignment to pos().
1711 Zero-length matches at the end of the previous match are ignored
1714 =head2 Combining pieces together
1716 Each of the elementary pieces of regular expressions which were described
1717 before (such as C<ab> or C<\Z>) could match at most one substring
1718 at the given position of the input string. However, in a typical regular
1719 expression these elementary pieces are combined into more complicated
1720 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1721 (in these examples C<S> and C<T> are regular subexpressions).
1723 Such combinations can include alternatives, leading to a problem of choice:
1724 if we match a regular expression C<a|ab> against C<"abc">, will it match
1725 substring C<"a"> or C<"ab">? One way to describe which substring is
1726 actually matched is the concept of backtracking (see L<"Backtracking">).
1727 However, this description is too low-level and makes you think
1728 in terms of a particular implementation.
1730 Another description starts with notions of "better"/"worse". All the
1731 substrings which may be matched by the given regular expression can be
1732 sorted from the "best" match to the "worst" match, and it is the "best"
1733 match which is chosen. This substitutes the question of "what is chosen?"
1734 by the question of "which matches are better, and which are worse?".
1736 Again, for elementary pieces there is no such question, since at most
1737 one match at a given position is possible. This section describes the
1738 notion of better/worse for combining operators. In the description
1739 below C<S> and C<T> are regular subexpressions.
1745 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1746 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1747 which can be matched by C<T>.
1749 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1752 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1753 C<B> is better match for C<T> than C<B'>.
1757 When C<S> can match, it is a better match than when only C<T> can match.
1759 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1760 two matches for C<T>.
1762 =item C<S{REPEAT_COUNT}>
1764 Matches as C<SSS...S> (repeated as many times as necessary).
1768 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1770 =item C<S{min,max}?>
1772 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1774 =item C<S?>, C<S*>, C<S+>
1776 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1778 =item C<S??>, C<S*?>, C<S+?>
1780 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1784 Matches the best match for C<S> and only that.
1786 =item C<(?=S)>, C<(?<=S)>
1788 Only the best match for C<S> is considered. (This is important only if
1789 C<S> has capturing parentheses, and backreferences are used somewhere
1790 else in the whole regular expression.)
1792 =item C<(?!S)>, C<(?<!S)>
1794 For this grouping operator there is no need to describe the ordering, since
1795 only whether or not C<S> can match is important.
1797 =item C<(??{ EXPR })>, C<(?PARNO)>
1799 The ordering is the same as for the regular expression which is
1800 the result of EXPR, or the pattern contained by capture buffer PARNO.
1802 =item C<(?(condition)yes-pattern|no-pattern)>
1804 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1805 already determined. The ordering of the matches is the same as for the
1806 chosen subexpression.
1810 The above recipes describe the ordering of matches I<at a given position>.
1811 One more rule is needed to understand how a match is determined for the
1812 whole regular expression: a match at an earlier position is always better
1813 than a match at a later position.
1815 =head2 Creating custom RE engines
1817 Overloaded constants (see L<overload>) provide a simple way to extend
1818 the functionality of the RE engine.
1820 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
1821 matches at boundary between whitespace characters and non-whitespace
1822 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1823 at these positions, so we want to have each C<\Y|> in the place of the
1824 more complicated version. We can create a module C<customre> to do
1832 die "No argument to customre::import allowed" if @_;
1833 overload::constant 'qr' => \&convert;
1836 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1838 # We must also take care of not escaping the legitimate \\Y|
1839 # sequence, hence the presence of '\\' in the conversion rules.
1840 my %rules = ( '\\' => '\\\\',
1841 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
1847 { $rules{$1} or invalid($re,$1) }sgex;
1851 Now C<use customre> enables the new escape in constant regular
1852 expressions, i.e., those without any runtime variable interpolations.
1853 As documented in L<overload>, this conversion will work only over
1854 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
1855 part of this regular expression needs to be converted explicitly
1856 (but only if the special meaning of C<\Y|> should be enabled inside $re):
1861 $re = customre::convert $re;
1866 This document varies from difficult to understand to completely
1867 and utterly opaque. The wandering prose riddled with jargon is
1868 hard to fathom in several places.
1870 This document needs a rewrite that separates the tutorial content
1871 from the reference content.
1879 L<perlop/"Regexp Quote-Like Operators">.
1881 L<perlop/"Gory details of parsing quoted constructs">.
1891 I<Mastering Regular Expressions> by Jeffrey Friedl, published
1892 by O'Reilly and Associates.