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, and "$" will match before any newline. At the
114 cost of a little more overhead, you can do this by using the /m modifier
115 on the pattern match operator. (Older programs did this by setting C<$*>,
116 but this practice has been removed in perl 5.9.)
119 To simplify multi-line substitutions, the "." character never matches a
120 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
121 the string is a single line--even if it isn't.
126 The following standard quantifiers are recognized:
127 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
129 * Match 0 or more times
130 + Match 1 or more times
132 {n} Match exactly n times
133 {n,} Match at least n times
134 {n,m} Match at least n but not more than m times
136 (If a curly bracket occurs in any other context, it is treated
137 as a regular character. In particular, the lower bound
138 is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+"
139 modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited
140 to integral values less than a preset limit defined when perl is built.
141 This is usually 32766 on the most common platforms. The actual limit can
142 be seen in the error message generated by code such as this:
144 $_ **= $_ , / {$_} / for 2 .. 42;
146 By default, a quantified subpattern is "greedy", that is, it will match as
147 many times as possible (given a particular starting location) while still
148 allowing the rest of the pattern to match. If you want it to match the
149 minimum number of times possible, follow the quantifier with a "?". Note
150 that the meanings don't change, just the "greediness":
151 X<metacharacter> X<greedy> X<greedyness>
152 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
154 *? Match 0 or more times
155 +? Match 1 or more times
157 {n}? Match exactly n times
158 {n,}? Match at least n times
159 {n,m}? Match at least n but not more than m times
161 By default, when a quantified subpattern does not allow the rest of the
162 overall pattern to match, Perl will backtrack. However, this behaviour is
163 sometimes undesirable. Thus Perl provides the "possesive" quantifier form
166 *+ Match 0 or more times and give nothing back
167 ++ Match 1 or more times and give nothing back
168 ?+ Match 0 or 1 time and give nothing back
169 {n}+ Match exactly n times and give nothing back (redundant)
170 {n,}+ Match at least n times and give nothing back
171 {n,m}+ Match at least n but not more than m times and give nothing back
177 will never match, as the C<a++> will gobble up all the C<a>'s in the
178 string and won't leave any for the remaining part of the pattern. This
179 feature can be extremely useful to give perl hints about where it
180 shouldn't backtrack. For instance, the typical "match a double-quoted
181 string" problem can be most efficiently performed when written as:
183 /"(?:[^"\\]++|\\.)*+"/
185 as we know that if the final quote does not match, bactracking will not
186 help. See the independent subexpression C<< (?>...) >> for more details;
187 possessive quantifiers are just syntactic sugar for that construct. For
188 instance the above example could also be written as follows:
190 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
192 =head3 Escape sequences
194 Because patterns are processed as double quoted strings, the following
196 X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q>
197 X<\0> X<\c> X<\N> X<\x>
203 \a alarm (bell) (BEL)
204 \e escape (think troff) (ESC)
205 \033 octal char (think of a PDP-11)
207 \x{263a} wide hex char (Unicode SMILEY)
210 \l lowercase next char (think vi)
211 \u uppercase next char (think vi)
212 \L lowercase till \E (think vi)
213 \U uppercase till \E (think vi)
214 \E end case modification (think vi)
215 \Q quote (disable) pattern metacharacters till \E
217 If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u>
218 and C<\U> is taken from the current locale. See L<perllocale>. For
219 documentation of C<\N{name}>, see L<charnames>.
221 You cannot include a literal C<$> or C<@> within a C<\Q> sequence.
222 An unescaped C<$> or C<@> interpolates the corresponding variable,
223 while escaping will cause the literal string C<\$> to be matched.
224 You'll need to write something like C<m/\Quser\E\@\Qhost/>.
226 =head3 Character classes
228 In addition, Perl defines the following:
230 X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C>
231 X<word> X<whitespace>
233 \w Match a "word" character (alphanumeric plus "_")
234 \W Match a non-"word" character
235 \s Match a whitespace character
236 \S Match a non-whitespace character
237 \d Match a digit character
238 \D Match a non-digit character
239 \pP Match P, named property. Use \p{Prop} for longer names.
241 \X Match eXtended Unicode "combining character sequence",
242 equivalent to (?:\PM\pM*)
243 \C Match a single C char (octet) even under Unicode.
244 NOTE: breaks up characters into their UTF-8 bytes,
245 so you may end up with malformed pieces of UTF-8.
246 Unsupported in lookbehind.
247 \1 Backreference to a a specific group.
248 '1' may actually be any positive integer
249 \k<name> Named backreference
250 \N{name} Named unicode character, or unicode escape.
251 \x12 Hexadecimal escape sequence
252 \x{1234} Long hexadecimal escape sequence
254 A C<\w> matches a single alphanumeric character (an alphabetic
255 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
256 to match a string of Perl-identifier characters (which isn't the same
257 as matching an English word). If C<use locale> is in effect, the list
258 of alphabetic characters generated by C<\w> is taken from the current
259 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
260 C<\d>, and C<\D> within character classes, but if you try to use them
261 as endpoints of a range, that's not a range, the "-" is understood
262 literally. If Unicode is in effect, C<\s> matches also "\x{85}",
263 "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about
264 C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general.
265 You can define your own C<\p> and C<\P> properties, see L<perlunicode>.
268 The POSIX character class syntax
273 is also available. Note that the C<[> and C<]> braces are I<literal>;
274 they must always be used within a character class expression.
277 $string =~ /[[:alpha:]]/;
279 # this is not, and will generate a warning:
280 $string =~ /[:alpha:]/;
282 The available classes and their backslash equivalents (if available) are
285 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
286 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
307 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
311 Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
312 also the (very rare) "vertical tabulator", "\ck", chr(11).
316 A Perl extension, see above.
320 For example use C<[:upper:]> to match all the uppercase characters.
321 Note that the C<[]> are part of the C<[::]> construct, not part of the
322 whole character class. For example:
326 matches zero, one, any alphabetic character, and the percentage sign.
328 The following equivalences to Unicode \p{} constructs and equivalent
329 backslash character classes (if available), will hold:
330 X<character class> X<\p> X<\p{}>
332 [[:...:]] \p{...} backslash
350 For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent.
352 If the C<utf8> pragma is not used but the C<locale> pragma is, the
353 classes correlate with the usual isalpha(3) interface (except for
356 The assumedly non-obviously named classes are:
363 Any control character. Usually characters that don't produce output as
364 such but instead control the terminal somehow: for example newline and
365 backspace are control characters. All characters with ord() less than
366 32 are most often classified as control characters (assuming ASCII,
367 the ISO Latin character sets, and Unicode), as is the character with
368 the ord() value of 127 (C<DEL>).
373 Any alphanumeric or punctuation (special) character.
378 Any alphanumeric or punctuation (special) character or the space character.
383 Any punctuation (special) character.
388 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
389 work just fine) it is included for completeness.
393 You can negate the [::] character classes by prefixing the class name
394 with a '^'. This is a Perl extension. For example:
395 X<character class, negation>
397 POSIX traditional Unicode
399 [[:^digit:]] \D \P{IsDigit}
400 [[:^space:]] \S \P{IsSpace}
401 [[:^word:]] \W \P{IsWord}
403 Perl respects the POSIX standard in that POSIX character classes are
404 only supported within a character class. The POSIX character classes
405 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
406 use them will cause an error.
410 Perl defines the following zero-width assertions:
411 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
412 X<regexp, zero-width assertion>
413 X<regular expression, zero-width assertion>
414 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
416 \b Match a word boundary
417 \B Match a non-(word boundary)
418 \A Match only at beginning of string
419 \Z Match only at end of string, or before newline at the end
420 \z Match only at end of string
421 \G Match only at pos() (e.g. at the end-of-match position
424 A word boundary (C<\b>) is a spot between two characters
425 that has a C<\w> on one side of it and a C<\W> on the other side
426 of it (in either order), counting the imaginary characters off the
427 beginning and end of the string as matching a C<\W>. (Within
428 character classes C<\b> represents backspace rather than a word
429 boundary, just as it normally does in any double-quoted string.)
430 The C<\A> and C<\Z> are just like "^" and "$", except that they
431 won't match multiple times when the C</m> modifier is used, while
432 "^" and "$" will match at every internal line boundary. To match
433 the actual end of the string and not ignore an optional trailing
435 X<\b> X<\A> X<\Z> X<\z> X</m>
437 The C<\G> assertion can be used to chain global matches (using
438 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
439 It is also useful when writing C<lex>-like scanners, when you have
440 several patterns that you want to match against consequent substrings
441 of your string, see the previous reference. The actual location
442 where C<\G> will match can also be influenced by using C<pos()> as
443 an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully
444 supported when anchored to the start of the pattern; while it
445 is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some
446 such uses (C</.\G/g>, for example) currently cause problems, and
447 it is recommended that you avoid such usage for now.
450 =head3 Capture buffers
452 The bracketing construct C<( ... )> creates capture buffers. To
453 refer to the digit'th buffer use \<digit> within the
454 match. Outside the match use "$" instead of "\". (The
455 \<digit> notation works in certain circumstances outside
456 the match. See the warning below about \1 vs $1 for details.)
457 Referring back to another part of the match is called a
459 X<regex, capture buffer> X<regexp, capture buffer>
460 X<regular expression, capture buffer> X<backreference>
462 There is no limit to the number of captured substrings that you may
463 use. However Perl also uses \10, \11, etc. as aliases for \010,
464 \011, etc. (Recall that 0 means octal, so \011 is the character at
465 number 9 in your coded character set; which would be the 10th character,
466 a horizontal tab under ASCII.) Perl resolves this
467 ambiguity by interpreting \10 as a backreference only if at least 10
468 left parentheses have opened before it. Likewise \11 is a
469 backreference only if at least 11 left parentheses have opened
470 before it. And so on. \1 through \9 are always interpreted as
473 Additionally, as of Perl 5.10 you may use named capture buffers and named
474 backreferences. The notation is C<< (?<name>...) >> and C<< \k<name> >>
475 (you may also use single quotes instead of angle brackets to quote the
476 name). The only difference with named capture buffers and unnamed ones is
477 that multiple buffers may have the same name and that the contents of
478 named capture buffers is available via the C<%+> hash. When multiple
479 groups share the same name C<$+{name}> and C<< \k<name> >> refer to the
480 leftmost defined group, thus it's possible to do things with named capture
481 buffers that would otherwise require C<(??{})> code to accomplish. Named
482 capture buffers are numbered just as normal capture buffers are and may be
483 referenced via the magic numeric variables or via numeric backreferences
488 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
490 /(.)\1/ # find first doubled char
491 and print "'$1' is the first doubled character\n";
493 /(?<char>.)\k<char>/ # ... a different way
494 and print "'$+{char}' is the first doubled character\n";
496 /(?<char>.)\1/ # ... mix and match
497 and print "'$1' is the first doubled character\n";
499 if (/Time: (..):(..):(..)/) { # parse out values
505 Several special variables also refer back to portions of the previous
506 match. C<$+> returns whatever the last bracket match matched.
507 C<$&> returns the entire matched string. (At one point C<$0> did
508 also, but now it returns the name of the program.) C<$`> returns
509 everything before the matched string. C<$'> returns everything
510 after the matched string. And C<$^N> contains whatever was matched by
511 the most-recently closed group (submatch). C<$^N> can be used in
512 extended patterns (see below), for example to assign a submatch to a
514 X<$+> X<$^N> X<$&> X<$`> X<$'>
516 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
517 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
518 until the end of the enclosing block or until the next successful
519 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
520 X<$+> X<$^N> X<$&> X<$`> X<$'>
521 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
524 B<NOTE>: failed matches in Perl do not reset the match variables,
525 which makes it easier to write code that tests for a series of more
526 specific cases and remembers the best match.
528 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
529 C<$'> anywhere in the program, it has to provide them for every
530 pattern match. This may substantially slow your program. Perl
531 uses the same mechanism to produce $1, $2, etc, so you also pay a
532 price for each pattern that contains capturing parentheses. (To
533 avoid this cost while retaining the grouping behaviour, use the
534 extended regular expression C<(?: ... )> instead.) But if you never
535 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
536 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
537 if you can, but if you can't (and some algorithms really appreciate
538 them), once you've used them once, use them at will, because you've
539 already paid the price. As of 5.005, C<$&> is not so costly as the
543 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
544 C<\w>, C<\n>. Unlike some other regular expression languages, there
545 are no backslashed symbols that aren't alphanumeric. So anything
546 that looks like \\, \(, \), \<, \>, \{, or \} is always
547 interpreted as a literal character, not a metacharacter. This was
548 once used in a common idiom to disable or quote the special meanings
549 of regular expression metacharacters in a string that you want to
550 use for a pattern. Simply quote all non-"word" characters:
552 $pattern =~ s/(\W)/\\$1/g;
554 (If C<use locale> is set, then this depends on the current locale.)
555 Today it is more common to use the quotemeta() function or the C<\Q>
556 metaquoting escape sequence to disable all metacharacters' special
559 /$unquoted\Q$quoted\E$unquoted/
561 Beware that if you put literal backslashes (those not inside
562 interpolated variables) between C<\Q> and C<\E>, double-quotish
563 backslash interpolation may lead to confusing results. If you
564 I<need> to use literal backslashes within C<\Q...\E>,
565 consult L<perlop/"Gory details of parsing quoted constructs">.
567 =head2 Extended Patterns
569 Perl also defines a consistent extension syntax for features not
570 found in standard tools like B<awk> and B<lex>. The syntax is a
571 pair of parentheses with a question mark as the first thing within
572 the parentheses. The character after the question mark indicates
575 The stability of these extensions varies widely. Some have been
576 part of the core language for many years. Others are experimental
577 and may change without warning or be completely removed. Check
578 the documentation on an individual feature to verify its current
581 A question mark was chosen for this and for the minimal-matching
582 construct because 1) question marks are rare in older regular
583 expressions, and 2) whenever you see one, you should stop and
584 "question" exactly what is going on. That's psychology...
591 A comment. The text is ignored. If the C</x> modifier enables
592 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
593 the comment as soon as it sees a C<)>, so there is no way to put a literal
596 =item C<(?imsx-imsx)>
599 One or more embedded pattern-match modifiers, to be turned on (or
600 turned off, if preceded by C<->) for the remainder of the pattern or
601 the remainder of the enclosing pattern group (if any). This is
602 particularly useful for dynamic patterns, such as those read in from a
603 configuration file, read in as an argument, are specified in a table
604 somewhere, etc. Consider the case that some of which want to be case
605 sensitive and some do not. The case insensitive ones need to include
606 merely C<(?i)> at the front of the pattern. For example:
609 if ( /$pattern/i ) { }
613 $pattern = "(?i)foobar";
614 if ( /$pattern/ ) { }
616 These modifiers are restored at the end of the enclosing group. For example,
620 will match a repeated (I<including the case>!) word C<blah> in any
621 case, assuming C<x> modifier, and no C<i> modifier outside this
627 =item C<(?imsx-imsx:pattern)>
629 This is for clustering, not capturing; it groups subexpressions like
630 "()", but doesn't make backreferences as "()" does. So
632 @fields = split(/\b(?:a|b|c)\b/)
636 @fields = split(/\b(a|b|c)\b/)
638 but doesn't spit out extra fields. It's also cheaper not to capture
639 characters if you don't need to.
641 Any letters between C<?> and C<:> act as flags modifiers as with
642 C<(?imsx-imsx)>. For example,
644 /(?s-i:more.*than).*million/i
646 is equivalent to the more verbose
648 /(?:(?s-i)more.*than).*million/i
651 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
653 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
654 matches a word followed by a tab, without including the tab in C<$&>.
657 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
659 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
660 matches any occurrence of "foo" that isn't followed by "bar". Note
661 however that look-ahead and look-behind are NOT the same thing. You cannot
662 use this for look-behind.
664 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
665 will not do what you want. That's because the C<(?!foo)> is just saying that
666 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
667 match. You would have to do something like C</(?!foo)...bar/> for that. We
668 say "like" because there's the case of your "bar" not having three characters
669 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
670 Sometimes it's still easier just to say:
672 if (/bar/ && $` !~ /foo$/)
674 For look-behind see below.
676 =item C<(?<=pattern)>
677 X<(?<=)> X<look-behind, positive> X<lookbehind, positive>
679 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
680 matches a word that follows a tab, without including the tab in C<$&>.
681 Works only for fixed-width look-behind.
683 =item C<(?<!pattern)>
684 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
686 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
687 matches any occurrence of "foo" that does not follow "bar". Works
688 only for fixed-width look-behind.
690 =item C<(?'NAME'pattern)>
692 =item C<< (?<NAME>pattern) >>
693 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
695 A named capture buffer. Identical in every respect to normal capturing
696 parens C<()> but for the additional fact that C<%+> may be used after
697 a succesful match to refer to a named buffer. See C<perlvar> for more
698 details on the C<%+> hash.
700 If multiple distinct capture buffers have the same name then the
701 $+{NAME} will refer to the leftmost defined buffer in the match.
703 The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent.
705 B<NOTE:> While the notation of this construct is the same as the similar
706 function in .NET regexes, the behavior is not, in Perl the buffers are
707 numbered sequentially regardless of being named or not. Thus in the
712 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
713 the opposite which is what a .NET regex hacker might expect.
715 Currently NAME is restricted to word chars only. In other words, it
716 must match C</^\w+$/>.
718 =item C<< \k<name> >>
720 =item C<< \k'name' >>
722 Named backreference. Similar to numeric backreferences, except that
723 the group is designated by name and not number. If multiple groups
724 have the same name then it refers to the leftmost defined group in
727 It is an error to refer to a name not defined by a C<(?<NAME>)>
728 earlier in the pattern.
730 Both forms are equivalent.
733 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
735 B<WARNING>: This extended regular expression feature is considered
736 experimental, and may be changed without notice. Code executed that
737 has side effects may not perform identically from version to version
738 due to the effect of future optimisations in the regex engine.
740 This zero-width assertion evaluates any embedded Perl code. It
741 always succeeds, and its C<code> is not interpolated. Currently,
742 the rules to determine where the C<code> ends are somewhat convoluted.
744 This feature can be used together with the special variable C<$^N> to
745 capture the results of submatches in variables without having to keep
746 track of the number of nested parentheses. For example:
748 $_ = "The brown fox jumps over the lazy dog";
749 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
750 print "color = $color, animal = $animal\n";
752 Inside the C<(?{...})> block, C<$_> refers to the string the regular
753 expression is matching against. You can also use C<pos()> to know what is
754 the current position of matching within this string.
756 The C<code> is properly scoped in the following sense: If the assertion
757 is backtracked (compare L<"Backtracking">), all changes introduced after
758 C<local>ization are undone, so that
762 (?{ $cnt = 0 }) # Initialize $cnt.
766 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
770 (?{ $res = $cnt }) # On success copy to non-localized
774 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
775 introduced value, because the scopes that restrict C<local> operators
778 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
779 switch. If I<not> used in this way, the result of evaluation of
780 C<code> is put into the special variable C<$^R>. This happens
781 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
782 inside the same regular expression.
784 The assignment to C<$^R> above is properly localized, so the old
785 value of C<$^R> is restored if the assertion is backtracked; compare
788 Due to an unfortunate implementation issue, the Perl code contained in these
789 blocks is treated as a compile time closure that can have seemingly bizarre
790 consequences when used with lexically scoped variables inside of subroutines
791 or loops. There are various workarounds for this, including simply using
792 global variables instead. If you are using this construct and strange results
793 occur then check for the use of lexically scoped variables.
795 For reasons of security, this construct is forbidden if the regular
796 expression involves run-time interpolation of variables, unless the
797 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
798 variables contain results of C<qr//> operator (see
799 L<perlop/"qr/STRING/imosx">).
801 This restriction is because of the wide-spread and remarkably convenient
802 custom of using run-time determined strings as patterns. For example:
808 Before Perl knew how to execute interpolated code within a pattern,
809 this operation was completely safe from a security point of view,
810 although it could raise an exception from an illegal pattern. If
811 you turn on the C<use re 'eval'>, though, it is no longer secure,
812 so you should only do so if you are also using taint checking.
813 Better yet, use the carefully constrained evaluation within a Safe
814 compartment. See L<perlsec> for details about both these mechanisms.
816 Because perl's regex engine is not currently re-entrant, interpolated
817 code may not invoke the regex engine either directly with C<m//> or C<s///>),
818 or indirectly with functions such as C<split>.
820 =item C<(??{ code })>
822 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
824 B<WARNING>: This extended regular expression feature is considered
825 experimental, and may be changed without notice. Code executed that
826 has side effects may not perform identically from version to version
827 due to the effect of future optimisations in the regex engine.
829 This is a "postponed" regular subexpression. The C<code> is evaluated
830 at run time, at the moment this subexpression may match. The result
831 of evaluation is considered as a regular expression and matched as
832 if it were inserted instead of this construct. Note that this means
833 that the contents of capture buffers defined inside an eval'ed pattern
834 are not available outside of the pattern, and vice versa, there is no
835 way for the inner pattern to refer to a capture buffer defined outside.
838 ('a' x 100)=~/(??{'(.)' x 100})/
840 B<will> match, it will B<not> set $1.
842 The C<code> is not interpolated. As before, the rules to determine
843 where the C<code> ends are currently somewhat convoluted.
845 The following pattern matches a parenthesized group:
850 (?> [^()]+ ) # Non-parens without backtracking
852 (??{ $re }) # Group with matching parens
857 See also C<(?PARNO)> for a different, more efficient way to accomplish
860 Because perl's regex engine is not currently re-entrant, delayed
861 code may not invoke the regex engine either directly with C<m//> or C<s///>),
862 or indirectly with functions such as C<split>.
864 Recursing deeper than 50 times without consuming any input string will
865 result in a fatal error. The maximum depth is compiled into perl, so
866 changing it requires a custom build.
868 =item C<(?PARNO)> C<(?R)> C<(?0)>
869 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)>
870 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
872 Similar to C<(??{ code })> except it does not involve compiling any code,
873 instead it treats the contents of a capture buffer as an independent
874 pattern that must match at the current position. Capture buffers
875 contained by the pattern will have the value as determined by the
878 PARNO is a sequence of digits (not starting with 0) whose value reflects
879 the paren-number of the capture buffer to recurse to. C<(?R)> recurses to
880 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
883 The following pattern matches a function foo() which may contain
884 balanced parenthesis as the argument.
886 $re = qr{ ( # paren group 1 (full function)
888 ( # paren group 2 (parens)
890 ( # paren group 3 (contents of parens)
892 (?> [^()]+ ) # Non-parens without backtracking
894 (?2) # Recurse to start of paren group 2
902 If the pattern was used as follows
904 'foo(bar(baz)+baz(bop))'=~/$re/
905 and print "\$1 = $1\n",
909 the output produced should be the following:
911 $1 = foo(bar(baz)+baz(bop))
912 $2 = (bar(baz)+baz(bop))
913 $3 = bar(baz)+baz(bop)
915 If there is no corresponding capture buffer defined, then it is a
916 fatal error. Recursing deeper than 50 times without consuming any input
917 string will also result in a fatal error. The maximum depth is compiled
918 into perl, so changing it requires a custom build.
920 B<Note> that this pattern does not behave the same way as the equivalent
921 PCRE or Python construct of the same form. In perl you can backtrack into
922 a recursed group, in PCRE and Python the recursed into group is treated
923 as atomic. Also, constructs like (?i:(?1)) or (?:(?i)(?1)) do not affect
924 the pattern being recursed into.
929 Recurse to a named subpattern. Identical to (?PARNO) except that the
930 parenthesis to recurse to is determined by name. If multiple parens have
931 the same name, then it recurses to the leftmost.
933 It is an error to refer to a name that is not declared somewhere in the
936 =item C<< (?>pattern) >>
937 X<backtrack> X<backtracking> X<atomic> X<possessive>
939 An "independent" subexpression, one which matches the substring
940 that a I<standalone> C<pattern> would match if anchored at the given
941 position, and it matches I<nothing other than this substring>. This
942 construct is useful for optimizations of what would otherwise be
943 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
944 It may also be useful in places where the "grab all you can, and do not
945 give anything back" semantic is desirable.
947 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
948 (anchored at the beginning of string, as above) will match I<all>
949 characters C<a> at the beginning of string, leaving no C<a> for
950 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
951 since the match of the subgroup C<a*> is influenced by the following
952 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
953 C<a*ab> will match fewer characters than a standalone C<a*>, since
954 this makes the tail match.
956 An effect similar to C<< (?>pattern) >> may be achieved by writing
957 C<(?=(pattern))\1>. This matches the same substring as a standalone
958 C<a+>, and the following C<\1> eats the matched string; it therefore
959 makes a zero-length assertion into an analogue of C<< (?>...) >>.
960 (The difference between these two constructs is that the second one
961 uses a capturing group, thus shifting ordinals of backreferences
962 in the rest of a regular expression.)
964 Consider this pattern:
975 That will efficiently match a nonempty group with matching parentheses
976 two levels deep or less. However, if there is no such group, it
977 will take virtually forever on a long string. That's because there
978 are so many different ways to split a long string into several
979 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
980 to a subpattern of the above pattern. Consider how the pattern
981 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
982 seconds, but that each extra letter doubles this time. This
983 exponential performance will make it appear that your program has
984 hung. However, a tiny change to this pattern
988 (?> [^()]+ ) # change x+ above to (?> x+ )
995 which uses C<< (?>...) >> matches exactly when the one above does (verifying
996 this yourself would be a productive exercise), but finishes in a fourth
997 the time when used on a similar string with 1000000 C<a>s. Be aware,
998 however, that this pattern currently triggers a warning message under
999 the C<use warnings> pragma or B<-w> switch saying it
1000 C<"matches null string many times in regex">.
1002 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1003 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1004 This was only 4 times slower on a string with 1000000 C<a>s.
1006 The "grab all you can, and do not give anything back" semantic is desirable
1007 in many situations where on the first sight a simple C<()*> looks like
1008 the correct solution. Suppose we parse text with comments being delimited
1009 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1010 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1011 the comment delimiter, because it may "give up" some whitespace if
1012 the remainder of the pattern can be made to match that way. The correct
1013 answer is either one of these:
1018 For example, to grab non-empty comments into $1, one should use either
1021 / (?> \# [ \t]* ) ( .+ ) /x;
1022 / \# [ \t]* ( [^ \t] .* ) /x;
1024 Which one you pick depends on which of these expressions better reflects
1025 the above specification of comments.
1027 In some literature this construct is called "atomic matching" or
1028 "possessive matching".
1030 Possessive quantifiers are equivalent to putting the item they are applied
1031 to inside of one of these constructs. The following equivalences apply:
1033 Quantifier Form Bracketing Form
1034 --------------- ---------------
1038 PAT{min,max}+ (?>PAT{min,max})
1040 =item C<(?(condition)yes-pattern|no-pattern)>
1043 =item C<(?(condition)yes-pattern)>
1045 Conditional expression. C<(condition)> should be either an integer in
1046 parentheses (which is valid if the corresponding pair of parentheses
1047 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
1048 name in angle brackets or single quotes (which is valid if a buffer
1049 with the given name matched), the special symbol (R) (true when
1050 evaluated inside of recursion or eval). Additionally the R may be
1051 followed by a number, (which will be true when evaluated when recursing
1052 inside of the appropriate group), or by C<&NAME> in which case it will
1053 be true only when evaluated during recursion in the named group.
1055 Here's a summary of the possible predicates:
1061 Checks if the numbered capturing buffer has matched something.
1063 =item (<NAME>) ('NAME')
1065 Checks if a buffer with the given name has matched something.
1069 Treats the code block as the condition
1073 Checks if the expression has been evaluated inside of recursion.
1077 Checks if the expression has been evaluated while executing directly
1078 inside of the n-th capture group. This check is the regex equivalent of
1080 if ((caller(0))[3] eq 'subname') { .. }
1082 In other words, it does not check the full recursion stack.
1086 Similar to C<(R1)>, this predicate checks to see if we're executing
1087 directly inside of the leftmost group with a given name (this is the same
1088 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1089 stack, but only the name of the innermost active recursion.
1093 In this case, the yes-pattern is never directly executed, and no
1094 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1095 See below for details.
1106 matches a chunk of non-parentheses, possibly included in parentheses
1109 A special form is the C<(DEFINE)> predicate, which never executes directly
1110 its yes-pattern, and does not allow a no-pattern. This allows to define
1111 subpatterns which will be executed only by using the recursion mechanism.
1112 This way, you can define a set of regular expression rules that can be
1113 bundled into any pattern you choose.
1115 It is recommended that for this usage you put the DEFINE block at the
1116 end of the pattern, and that you name any subpatterns defined within it.
1118 Also, it's worth noting that patterns defined this way probably will
1119 not be as efficient, as the optimiser is not very clever about
1120 handling them. YMMV.
1122 An example of how this might be used is as follows:
1124 /(?<NAME>(&NAME_PAT))(?<ADDR>(&ADDRESS_PAT))
1130 Note that capture buffers matched inside of recursion are not accessible
1131 after the recursion returns, so the extra layer of capturing buffers are
1132 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1133 C<$+{NAME}> would be.
1138 X<backtrack> X<backtracking>
1140 NOTE: This section presents an abstract approximation of regular
1141 expression behavior. For a more rigorous (and complicated) view of
1142 the rules involved in selecting a match among possible alternatives,
1143 see L<Combining pieces together>.
1145 A fundamental feature of regular expression matching involves the
1146 notion called I<backtracking>, which is currently used (when needed)
1147 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
1148 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1149 internally, but the general principle outlined here is valid.
1151 For a regular expression to match, the I<entire> regular expression must
1152 match, not just part of it. So if the beginning of a pattern containing a
1153 quantifier succeeds in a way that causes later parts in the pattern to
1154 fail, the matching engine backs up and recalculates the beginning
1155 part--that's why it's called backtracking.
1157 Here is an example of backtracking: Let's say you want to find the
1158 word following "foo" in the string "Food is on the foo table.":
1160 $_ = "Food is on the foo table.";
1161 if ( /\b(foo)\s+(\w+)/i ) {
1162 print "$2 follows $1.\n";
1165 When the match runs, the first part of the regular expression (C<\b(foo)>)
1166 finds a possible match right at the beginning of the string, and loads up
1167 $1 with "Foo". However, as soon as the matching engine sees that there's
1168 no whitespace following the "Foo" that it had saved in $1, it realizes its
1169 mistake and starts over again one character after where it had the
1170 tentative match. This time it goes all the way until the next occurrence
1171 of "foo". The complete regular expression matches this time, and you get
1172 the expected output of "table follows foo."
1174 Sometimes minimal matching can help a lot. Imagine you'd like to match
1175 everything between "foo" and "bar". Initially, you write something
1178 $_ = "The food is under the bar in the barn.";
1179 if ( /foo(.*)bar/ ) {
1183 Which perhaps unexpectedly yields:
1185 got <d is under the bar in the >
1187 That's because C<.*> was greedy, so you get everything between the
1188 I<first> "foo" and the I<last> "bar". Here it's more effective
1189 to use minimal matching to make sure you get the text between a "foo"
1190 and the first "bar" thereafter.
1192 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1193 got <d is under the >
1195 Here's another example: let's say you'd like to match a number at the end
1196 of a string, and you also want to keep the preceding part of the match.
1199 $_ = "I have 2 numbers: 53147";
1200 if ( /(.*)(\d*)/ ) { # Wrong!
1201 print "Beginning is <$1>, number is <$2>.\n";
1204 That won't work at all, because C<.*> was greedy and gobbled up the
1205 whole string. As C<\d*> can match on an empty string the complete
1206 regular expression matched successfully.
1208 Beginning is <I have 2 numbers: 53147>, number is <>.
1210 Here are some variants, most of which don't work:
1212 $_ = "I have 2 numbers: 53147";
1225 printf "%-12s ", $pat;
1227 print "<$1> <$2>\n";
1233 That will print out:
1235 (.*)(\d*) <I have 2 numbers: 53147> <>
1236 (.*)(\d+) <I have 2 numbers: 5314> <7>
1238 (.*?)(\d+) <I have > <2>
1239 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1240 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1241 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1242 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1244 As you see, this can be a bit tricky. It's important to realize that a
1245 regular expression is merely a set of assertions that gives a definition
1246 of success. There may be 0, 1, or several different ways that the
1247 definition might succeed against a particular string. And if there are
1248 multiple ways it might succeed, you need to understand backtracking to
1249 know which variety of success you will achieve.
1251 When using look-ahead assertions and negations, this can all get even
1252 trickier. Imagine you'd like to find a sequence of non-digits not
1253 followed by "123". You might try to write that as
1256 if ( /^\D*(?!123)/ ) { # Wrong!
1257 print "Yup, no 123 in $_\n";
1260 But that isn't going to match; at least, not the way you're hoping. It
1261 claims that there is no 123 in the string. Here's a clearer picture of
1262 why that pattern matches, contrary to popular expectations:
1267 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1268 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1270 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1271 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1279 You might have expected test 3 to fail because it seems to a more
1280 general purpose version of test 1. The important difference between
1281 them is that test 3 contains a quantifier (C<\D*>) and so can use
1282 backtracking, whereas test 1 will not. What's happening is
1283 that you've asked "Is it true that at the start of $x, following 0 or more
1284 non-digits, you have something that's not 123?" If the pattern matcher had
1285 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1288 The search engine will initially match C<\D*> with "ABC". Then it will
1289 try to match C<(?!123> with "123", which fails. But because
1290 a quantifier (C<\D*>) has been used in the regular expression, the
1291 search engine can backtrack and retry the match differently
1292 in the hope of matching the complete regular expression.
1294 The pattern really, I<really> wants to succeed, so it uses the
1295 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1296 time. Now there's indeed something following "AB" that is not
1297 "123". It's "C123", which suffices.
1299 We can deal with this by using both an assertion and a negation.
1300 We'll say that the first part in $1 must be followed both by a digit
1301 and by something that's not "123". Remember that the look-aheads
1302 are zero-width expressions--they only look, but don't consume any
1303 of the string in their match. So rewriting this way produces what
1304 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1306 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1307 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1311 In other words, the two zero-width assertions next to each other work as though
1312 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1313 matches only if you're at the beginning of the line AND the end of the
1314 line simultaneously. The deeper underlying truth is that juxtaposition in
1315 regular expressions always means AND, except when you write an explicit OR
1316 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1317 although the attempted matches are made at different positions because "a"
1318 is not a zero-width assertion, but a one-width assertion.
1320 B<WARNING>: particularly complicated regular expressions can take
1321 exponential time to solve because of the immense number of possible
1322 ways they can use backtracking to try match. For example, without
1323 internal optimizations done by the regular expression engine, this will
1324 take a painfully long time to run:
1326 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1328 And if you used C<*>'s in the internal groups instead of limiting them
1329 to 0 through 5 matches, then it would take forever--or until you ran
1330 out of stack space. Moreover, these internal optimizations are not
1331 always applicable. For example, if you put C<{0,5}> instead of C<*>
1332 on the external group, no current optimization is applicable, and the
1333 match takes a long time to finish.
1335 A powerful tool for optimizing such beasts is what is known as an
1336 "independent group",
1337 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1338 zero-length look-ahead/look-behind assertions will not backtrack to make
1339 the tail match, since they are in "logical" context: only
1340 whether they match is considered relevant. For an example
1341 where side-effects of look-ahead I<might> have influenced the
1342 following match, see L<C<< (?>pattern) >>>.
1344 =head2 Version 8 Regular Expressions
1345 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1347 In case you're not familiar with the "regular" Version 8 regex
1348 routines, here are the pattern-matching rules not described above.
1350 Any single character matches itself, unless it is a I<metacharacter>
1351 with a special meaning described here or above. You can cause
1352 characters that normally function as metacharacters to be interpreted
1353 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1354 character; "\\" matches a "\"). A series of characters matches that
1355 series of characters in the target string, so the pattern C<blurfl>
1356 would match "blurfl" in the target string.
1358 You can specify a character class, by enclosing a list of characters
1359 in C<[]>, which will match any one character from the list. If the
1360 first character after the "[" is "^", the class matches any character not
1361 in the list. Within a list, the "-" character specifies a
1362 range, so that C<a-z> represents all characters between "a" and "z",
1363 inclusive. If you want either "-" or "]" itself to be a member of a
1364 class, put it at the start of the list (possibly after a "^"), or
1365 escape it with a backslash. "-" is also taken literally when it is
1366 at the end of the list, just before the closing "]". (The
1367 following all specify the same class of three characters: C<[-az]>,
1368 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1369 specifies a class containing twenty-six characters, even on EBCDIC
1370 based coded character sets.) Also, if you try to use the character
1371 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1372 a range, that's not a range, the "-" is understood literally.
1374 Note also that the whole range idea is rather unportable between
1375 character sets--and even within character sets they may cause results
1376 you probably didn't expect. A sound principle is to use only ranges
1377 that begin from and end at either alphabets of equal case ([a-e],
1378 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1379 spell out the character sets in full.
1381 Characters may be specified using a metacharacter syntax much like that
1382 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1383 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1384 of octal digits, matches the character whose coded character set value
1385 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1386 matches the character whose numeric value is I<nn>. The expression \cI<x>
1387 matches the character control-I<x>. Finally, the "." metacharacter
1388 matches any character except "\n" (unless you use C</s>).
1390 You can specify a series of alternatives for a pattern using "|" to
1391 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1392 or "foe" in the target string (as would C<f(e|i|o)e>). The
1393 first alternative includes everything from the last pattern delimiter
1394 ("(", "[", or the beginning of the pattern) up to the first "|", and
1395 the last alternative contains everything from the last "|" to the next
1396 pattern delimiter. That's why it's common practice to include
1397 alternatives in parentheses: to minimize confusion about where they
1400 Alternatives are tried from left to right, so the first
1401 alternative found for which the entire expression matches, is the one that
1402 is chosen. This means that alternatives are not necessarily greedy. For
1403 example: when matching C<foo|foot> against "barefoot", only the "foo"
1404 part will match, as that is the first alternative tried, and it successfully
1405 matches the target string. (This might not seem important, but it is
1406 important when you are capturing matched text using parentheses.)
1408 Also remember that "|" is interpreted as a literal within square brackets,
1409 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1411 Within a pattern, you may designate subpatterns for later reference
1412 by enclosing them in parentheses, and you may refer back to the
1413 I<n>th subpattern later in the pattern using the metacharacter
1414 \I<n>. Subpatterns are numbered based on the left to right order
1415 of their opening parenthesis. A backreference matches whatever
1416 actually matched the subpattern in the string being examined, not
1417 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1418 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1419 1 matched "0x", even though the rule C<0|0x> could potentially match
1420 the leading 0 in the second number.
1422 =head2 Warning on \1 vs $1
1424 Some people get too used to writing things like:
1426 $pattern =~ s/(\W)/\\\1/g;
1428 This is grandfathered for the RHS of a substitute to avoid shocking the
1429 B<sed> addicts, but it's a dirty habit to get into. That's because in
1430 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1431 the usual double-quoted string means a control-A. The customary Unix
1432 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1433 of doing that, you get yourself into trouble if you then add an C</e>
1436 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1442 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1443 C<${1}000>. The operation of interpolation should not be confused
1444 with the operation of matching a backreference. Certainly they mean two
1445 different things on the I<left> side of the C<s///>.
1447 =head2 Repeated patterns matching zero-length substring
1449 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1451 Regular expressions provide a terse and powerful programming language. As
1452 with most other power tools, power comes together with the ability
1455 A common abuse of this power stems from the ability to make infinite
1456 loops using regular expressions, with something as innocuous as:
1458 'foo' =~ m{ ( o? )* }x;
1460 The C<o?> can match at the beginning of C<'foo'>, and since the position
1461 in the string is not moved by the match, C<o?> would match again and again
1462 because of the C<*> modifier. Another common way to create a similar cycle
1463 is with the looping modifier C<//g>:
1465 @matches = ( 'foo' =~ m{ o? }xg );
1469 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1471 or the loop implied by split().
1473 However, long experience has shown that many programming tasks may
1474 be significantly simplified by using repeated subexpressions that
1475 may match zero-length substrings. Here's a simple example being:
1477 @chars = split //, $string; # // is not magic in split
1478 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1480 Thus Perl allows such constructs, by I<forcefully breaking
1481 the infinite loop>. The rules for this are different for lower-level
1482 loops given by the greedy modifiers C<*+{}>, and for higher-level
1483 ones like the C</g> modifier or split() operator.
1485 The lower-level loops are I<interrupted> (that is, the loop is
1486 broken) when Perl detects that a repeated expression matched a
1487 zero-length substring. Thus
1489 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1491 is made equivalent to
1493 m{ (?: NON_ZERO_LENGTH )*
1498 The higher level-loops preserve an additional state between iterations:
1499 whether the last match was zero-length. To break the loop, the following
1500 match after a zero-length match is prohibited to have a length of zero.
1501 This prohibition interacts with backtracking (see L<"Backtracking">),
1502 and so the I<second best> match is chosen if the I<best> match is of
1510 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1511 match given by non-greedy C<??> is the zero-length match, and the I<second
1512 best> match is what is matched by C<\w>. Thus zero-length matches
1513 alternate with one-character-long matches.
1515 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1516 position one notch further in the string.
1518 The additional state of being I<matched with zero-length> is associated with
1519 the matched string, and is reset by each assignment to pos().
1520 Zero-length matches at the end of the previous match are ignored
1523 =head2 Combining pieces together
1525 Each of the elementary pieces of regular expressions which were described
1526 before (such as C<ab> or C<\Z>) could match at most one substring
1527 at the given position of the input string. However, in a typical regular
1528 expression these elementary pieces are combined into more complicated
1529 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1530 (in these examples C<S> and C<T> are regular subexpressions).
1532 Such combinations can include alternatives, leading to a problem of choice:
1533 if we match a regular expression C<a|ab> against C<"abc">, will it match
1534 substring C<"a"> or C<"ab">? One way to describe which substring is
1535 actually matched is the concept of backtracking (see L<"Backtracking">).
1536 However, this description is too low-level and makes you think
1537 in terms of a particular implementation.
1539 Another description starts with notions of "better"/"worse". All the
1540 substrings which may be matched by the given regular expression can be
1541 sorted from the "best" match to the "worst" match, and it is the "best"
1542 match which is chosen. This substitutes the question of "what is chosen?"
1543 by the question of "which matches are better, and which are worse?".
1545 Again, for elementary pieces there is no such question, since at most
1546 one match at a given position is possible. This section describes the
1547 notion of better/worse for combining operators. In the description
1548 below C<S> and C<T> are regular subexpressions.
1554 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1555 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1556 which can be matched by C<T>.
1558 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1561 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1562 C<B> is better match for C<T> than C<B'>.
1566 When C<S> can match, it is a better match than when only C<T> can match.
1568 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1569 two matches for C<T>.
1571 =item C<S{REPEAT_COUNT}>
1573 Matches as C<SSS...S> (repeated as many times as necessary).
1577 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1579 =item C<S{min,max}?>
1581 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1583 =item C<S?>, C<S*>, C<S+>
1585 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1587 =item C<S??>, C<S*?>, C<S+?>
1589 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1593 Matches the best match for C<S> and only that.
1595 =item C<(?=S)>, C<(?<=S)>
1597 Only the best match for C<S> is considered. (This is important only if
1598 C<S> has capturing parentheses, and backreferences are used somewhere
1599 else in the whole regular expression.)
1601 =item C<(?!S)>, C<(?<!S)>
1603 For this grouping operator there is no need to describe the ordering, since
1604 only whether or not C<S> can match is important.
1606 =item C<(??{ EXPR })>, C<(?PARNO)>
1608 The ordering is the same as for the regular expression which is
1609 the result of EXPR, or the pattern contained by capture buffer PARNO.
1611 =item C<(?(condition)yes-pattern|no-pattern)>
1613 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1614 already determined. The ordering of the matches is the same as for the
1615 chosen subexpression.
1619 The above recipes describe the ordering of matches I<at a given position>.
1620 One more rule is needed to understand how a match is determined for the
1621 whole regular expression: a match at an earlier position is always better
1622 than a match at a later position.
1624 =head2 Creating custom RE engines
1626 Overloaded constants (see L<overload>) provide a simple way to extend
1627 the functionality of the RE engine.
1629 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
1630 matches at boundary between whitespace characters and non-whitespace
1631 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1632 at these positions, so we want to have each C<\Y|> in the place of the
1633 more complicated version. We can create a module C<customre> to do
1641 die "No argument to customre::import allowed" if @_;
1642 overload::constant 'qr' => \&convert;
1645 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1647 # We must also take care of not escaping the legitimate \\Y|
1648 # sequence, hence the presence of '\\' in the conversion rules.
1649 my %rules = ( '\\' => '\\\\',
1650 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
1656 { $rules{$1} or invalid($re,$1) }sgex;
1660 Now C<use customre> enables the new escape in constant regular
1661 expressions, i.e., those without any runtime variable interpolations.
1662 As documented in L<overload>, this conversion will work only over
1663 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
1664 part of this regular expression needs to be converted explicitly
1665 (but only if the special meaning of C<\Y|> should be enabled inside $re):
1670 $re = customre::convert $re;
1675 This document varies from difficult to understand to completely
1676 and utterly opaque. The wandering prose riddled with jargon is
1677 hard to fathom in several places.
1679 This document needs a rewrite that separates the tutorial content
1680 from the reference content.
1688 L<perlop/"Regexp Quote-Like Operators">.
1690 L<perlop/"Gory details of parsing quoted constructs">.
1700 I<Mastering Regular Expressions> by Jeffrey Friedl, published
1701 by O'Reilly and Associates.