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
85 The patterns used in Perl pattern matching derive from supplied in
86 the Version 8 regex routines. (The routines are derived
87 (distantly) from Henry Spencer's freely redistributable reimplementation
88 of the V8 routines.) See L<Version 8 Regular Expressions> for
91 In particular the following metacharacters have their standard I<egrep>-ish
94 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
97 \ Quote the next metacharacter
98 ^ Match the beginning of the line
99 . Match any character (except newline)
100 $ Match the end of the line (or before newline at the end)
105 By default, the "^" character is guaranteed to match only the
106 beginning of the string, the "$" character only the end (or before the
107 newline at the end), and Perl does certain optimizations with the
108 assumption that the string contains only one line. Embedded newlines
109 will not be matched by "^" or "$". You may, however, wish to treat a
110 string as a multi-line buffer, such that the "^" will match after any
111 newline within the string, and "$" will match before any newline. At the
112 cost of a little more overhead, you can do this by using the /m modifier
113 on the pattern match operator. (Older programs did this by setting C<$*>,
114 but this practice has been removed in perl 5.9.)
117 To simplify multi-line substitutions, the "." character never matches a
118 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
119 the string is a single line--even if it isn't.
122 The following standard quantifiers are recognized:
123 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
125 * Match 0 or more times
126 + Match 1 or more times
128 {n} Match exactly n times
129 {n,} Match at least n times
130 {n,m} Match at least n but not more than m times
132 (If a curly bracket occurs in any other context, it is treated
133 as a regular character. In particular, the lower bound
134 is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+"
135 modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited
136 to integral values less than a preset limit defined when perl is built.
137 This is usually 32766 on the most common platforms. The actual limit can
138 be seen in the error message generated by code such as this:
140 $_ **= $_ , / {$_} / for 2 .. 42;
142 By default, a quantified subpattern is "greedy", that is, it will match as
143 many times as possible (given a particular starting location) while still
144 allowing the rest of the pattern to match. If you want it to match the
145 minimum number of times possible, follow the quantifier with a "?". Note
146 that the meanings don't change, just the "greediness":
147 X<metacharacter> X<greedy> X<greedyness>
148 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
150 *? Match 0 or more times
151 +? Match 1 or more times
153 {n}? Match exactly n times
154 {n,}? Match at least n times
155 {n,m}? Match at least n but not more than m times
157 By default, when a quantified subpattern does not allow the rest of the
158 overall pattern to match, Perl will backtrack. However, this behaviour is
159 sometimes undesirable. Thus Perl provides the "possesive" quantifier form
162 *+ Match 0 or more times and give nothing back
163 +? Match 1 or more times and give nothing back
164 ?+ Match 0 or 1 time and give nothing back
165 {n}+ Match exactly n times and give nothing back (redundant)
166 {n,}? Match at least n times and give nothing back
167 {n,m}? Match at least n but not more than m times and give nothing back
173 will never match, as the C<a++> will gobble up all the C<a>'s in the
174 string and won't leave any for the remaining part of the pattern. This
175 feature can be extremely useful to give perl hints about where it
176 shouldn't backtrack. For instance, the typical "match a double-quoted
177 string" problem can be most efficiently performed when written as:
179 /"(?:[^"\\]++|\\.)*+"/
181 as we know that if the final quote does not match, bactracking will not
182 help. See the independent subexpression C<< (?>...) >> for more details;
183 possessive quantifiers are just syntactic sugar for that construct. For
184 instance the above example could also be written as follows:
186 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
188 Because patterns are processed as double quoted strings, the following
190 X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q>
191 X<\0> X<\c> X<\N> X<\x>
197 \a alarm (bell) (BEL)
198 \e escape (think troff) (ESC)
199 \033 octal char (think of a PDP-11)
201 \x{263a} wide hex char (Unicode SMILEY)
204 \l lowercase next char (think vi)
205 \u uppercase next char (think vi)
206 \L lowercase till \E (think vi)
207 \U uppercase till \E (think vi)
208 \E end case modification (think vi)
209 \Q quote (disable) pattern metacharacters till \E
211 If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u>
212 and C<\U> is taken from the current locale. See L<perllocale>. For
213 documentation of C<\N{name}>, see L<charnames>.
215 You cannot include a literal C<$> or C<@> within a C<\Q> sequence.
216 An unescaped C<$> or C<@> interpolates the corresponding variable,
217 while escaping will cause the literal string C<\$> to be matched.
218 You'll need to write something like C<m/\Quser\E\@\Qhost/>.
220 In addition, Perl defines the following:
222 X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C>
223 X<word> X<whitespace>
225 \w Match a "word" character (alphanumeric plus "_")
226 \W Match a non-"word" character
227 \s Match a whitespace character
228 \S Match a non-whitespace character
229 \d Match a digit character
230 \D Match a non-digit character
231 \pP Match P, named property. Use \p{Prop} for longer names.
233 \X Match eXtended Unicode "combining character sequence",
234 equivalent to (?:\PM\pM*)
235 \C Match a single C char (octet) even under Unicode.
236 NOTE: breaks up characters into their UTF-8 bytes,
237 so you may end up with malformed pieces of UTF-8.
238 Unsupported in lookbehind.
239 \1 Backreference to a a specific group.
240 '1' may actually be any positive integer
241 \k<name> Named backreference
242 \N{name} Named unicode character, or unicode escape.
243 \x12 Hexadecimal escape sequence
244 \x{1234} Long hexadecimal escape sequence
246 A C<\w> matches a single alphanumeric character (an alphabetic
247 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
248 to match a string of Perl-identifier characters (which isn't the same
249 as matching an English word). If C<use locale> is in effect, the list
250 of alphabetic characters generated by C<\w> is taken from the current
251 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
252 C<\d>, and C<\D> within character classes, but if you try to use them
253 as endpoints of a range, that's not a range, the "-" is understood
254 literally. If Unicode is in effect, C<\s> matches also "\x{85}",
255 "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about
256 C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general.
257 You can define your own C<\p> and C<\P> properties, see L<perlunicode>.
260 The POSIX character class syntax
265 is also available. Note that the C<[> and C<]> braces are I<literal>;
266 they must always be used within a character class expression.
269 $string =~ /[[:alpha:]]/;
271 # this is not, and will generate a warning:
272 $string =~ /[:alpha:]/;
274 The available classes and their backslash equivalents (if available) are
277 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
278 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
299 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
303 Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
304 also the (very rare) "vertical tabulator", "\ck", chr(11).
308 A Perl extension, see above.
312 For example use C<[:upper:]> to match all the uppercase characters.
313 Note that the C<[]> are part of the C<[::]> construct, not part of the
314 whole character class. For example:
318 matches zero, one, any alphabetic character, and the percentage sign.
320 The following equivalences to Unicode \p{} constructs and equivalent
321 backslash character classes (if available), will hold:
322 X<character class> X<\p> X<\p{}>
324 [[:...:]] \p{...} backslash
342 For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent.
344 If the C<utf8> pragma is not used but the C<locale> pragma is, the
345 classes correlate with the usual isalpha(3) interface (except for
348 The assumedly non-obviously named classes are:
355 Any control character. Usually characters that don't produce output as
356 such but instead control the terminal somehow: for example newline and
357 backspace are control characters. All characters with ord() less than
358 32 are most often classified as control characters (assuming ASCII,
359 the ISO Latin character sets, and Unicode), as is the character with
360 the ord() value of 127 (C<DEL>).
365 Any alphanumeric or punctuation (special) character.
370 Any alphanumeric or punctuation (special) character or the space character.
375 Any punctuation (special) character.
380 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
381 work just fine) it is included for completeness.
385 You can negate the [::] character classes by prefixing the class name
386 with a '^'. This is a Perl extension. For example:
387 X<character class, negation>
389 POSIX traditional Unicode
391 [[:^digit:]] \D \P{IsDigit}
392 [[:^space:]] \S \P{IsSpace}
393 [[:^word:]] \W \P{IsWord}
395 Perl respects the POSIX standard in that POSIX character classes are
396 only supported within a character class. The POSIX character classes
397 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
398 use them will cause an error.
400 Perl defines the following zero-width assertions:
401 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
402 X<regexp, zero-width assertion>
403 X<regular expression, zero-width assertion>
404 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
406 \b Match a word boundary
407 \B Match a non-(word boundary)
408 \A Match only at beginning of string
409 \Z Match only at end of string, or before newline at the end
410 \z Match only at end of string
411 \G Match only at pos() (e.g. at the end-of-match position
414 A word boundary (C<\b>) is a spot between two characters
415 that has a C<\w> on one side of it and a C<\W> on the other side
416 of it (in either order), counting the imaginary characters off the
417 beginning and end of the string as matching a C<\W>. (Within
418 character classes C<\b> represents backspace rather than a word
419 boundary, just as it normally does in any double-quoted string.)
420 The C<\A> and C<\Z> are just like "^" and "$", except that they
421 won't match multiple times when the C</m> modifier is used, while
422 "^" and "$" will match at every internal line boundary. To match
423 the actual end of the string and not ignore an optional trailing
425 X<\b> X<\A> X<\Z> X<\z> X</m>
427 The C<\G> assertion can be used to chain global matches (using
428 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
429 It is also useful when writing C<lex>-like scanners, when you have
430 several patterns that you want to match against consequent substrings
431 of your string, see the previous reference. The actual location
432 where C<\G> will match can also be influenced by using C<pos()> as
433 an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully
434 supported when anchored to the start of the pattern; while it
435 is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some
436 such uses (C</.\G/g>, for example) currently cause problems, and
437 it is recommended that you avoid such usage for now.
440 The bracketing construct C<( ... )> creates capture buffers. To
441 refer to the digit'th buffer use \<digit> within the
442 match. Outside the match use "$" instead of "\". (The
443 \<digit> notation works in certain circumstances outside
444 the match. See the warning below about \1 vs $1 for details.)
445 Referring back to another part of the match is called a
447 X<regex, capture buffer> X<regexp, capture buffer>
448 X<regular expression, capture buffer> X<backreference>
450 There is no limit to the number of captured substrings that you may
451 use. However Perl also uses \10, \11, etc. as aliases for \010,
452 \011, etc. (Recall that 0 means octal, so \011 is the character at
453 number 9 in your coded character set; which would be the 10th character,
454 a horizontal tab under ASCII.) Perl resolves this
455 ambiguity by interpreting \10 as a backreference only if at least 10
456 left parentheses have opened before it. Likewise \11 is a
457 backreference only if at least 11 left parentheses have opened
458 before it. And so on. \1 through \9 are always interpreted as
461 Additionally, as of Perl 5.10 you may use named capture buffers and named
462 backreferences. The notation is C<< (?<name>...) >> and C<< \k<name> >>
463 (you may also use single quotes instead of angle brackets to quote the
464 name). The only difference with named capture buffers and unnamed ones is
465 that multiple buffers may have the same name and that the contents of
466 named capture buffers is available via the C<%+> hash. When multiple
467 groups share the same name C<$+{name}> and C<< \k<name> >> refer to the
468 leftmost defined group, thus it's possible to do things with named capture
469 buffers that would otherwise require C<(??{})> code to accomplish. Named
470 capture buffers are numbered just as normal capture buffers are and may be
471 referenced via the magic numeric variables or via numeric backreferences
476 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
478 /(.)\1/ # find first doubled char
479 and print "'$1' is the first doubled character\n";
481 /(?<char>.)\k<char>/ # ... a different way
482 and print "'$+{char}' is the first doubled character\n";
484 /(?<char>.)\1/ # ... mix and match
485 and print "'$1' is the first doubled character\n";
487 if (/Time: (..):(..):(..)/) { # parse out values
493 Several special variables also refer back to portions of the previous
494 match. C<$+> returns whatever the last bracket match matched.
495 C<$&> returns the entire matched string. (At one point C<$0> did
496 also, but now it returns the name of the program.) C<$`> returns
497 everything before the matched string. C<$'> returns everything
498 after the matched string. And C<$^N> contains whatever was matched by
499 the most-recently closed group (submatch). C<$^N> can be used in
500 extended patterns (see below), for example to assign a submatch to a
502 X<$+> X<$^N> X<$&> X<$`> X<$'>
504 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
505 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
506 until the end of the enclosing block or until the next successful
507 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
508 X<$+> X<$^N> X<$&> X<$`> X<$'>
509 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
512 B<NOTE>: failed matches in Perl do not reset the match variables,
513 which makes it easier to write code that tests for a series of more
514 specific cases and remembers the best match.
516 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
517 C<$'> anywhere in the program, it has to provide them for every
518 pattern match. This may substantially slow your program. Perl
519 uses the same mechanism to produce $1, $2, etc, so you also pay a
520 price for each pattern that contains capturing parentheses. (To
521 avoid this cost while retaining the grouping behaviour, use the
522 extended regular expression C<(?: ... )> instead.) But if you never
523 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
524 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
525 if you can, but if you can't (and some algorithms really appreciate
526 them), once you've used them once, use them at will, because you've
527 already paid the price. As of 5.005, C<$&> is not so costly as the
531 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
532 C<\w>, C<\n>. Unlike some other regular expression languages, there
533 are no backslashed symbols that aren't alphanumeric. So anything
534 that looks like \\, \(, \), \<, \>, \{, or \} is always
535 interpreted as a literal character, not a metacharacter. This was
536 once used in a common idiom to disable or quote the special meanings
537 of regular expression metacharacters in a string that you want to
538 use for a pattern. Simply quote all non-"word" characters:
540 $pattern =~ s/(\W)/\\$1/g;
542 (If C<use locale> is set, then this depends on the current locale.)
543 Today it is more common to use the quotemeta() function or the C<\Q>
544 metaquoting escape sequence to disable all metacharacters' special
547 /$unquoted\Q$quoted\E$unquoted/
549 Beware that if you put literal backslashes (those not inside
550 interpolated variables) between C<\Q> and C<\E>, double-quotish
551 backslash interpolation may lead to confusing results. If you
552 I<need> to use literal backslashes within C<\Q...\E>,
553 consult L<perlop/"Gory details of parsing quoted constructs">.
555 =head2 Extended Patterns
557 Perl also defines a consistent extension syntax for features not
558 found in standard tools like B<awk> and B<lex>. The syntax is a
559 pair of parentheses with a question mark as the first thing within
560 the parentheses. The character after the question mark indicates
563 The stability of these extensions varies widely. Some have been
564 part of the core language for many years. Others are experimental
565 and may change without warning or be completely removed. Check
566 the documentation on an individual feature to verify its current
569 A question mark was chosen for this and for the minimal-matching
570 construct because 1) question marks are rare in older regular
571 expressions, and 2) whenever you see one, you should stop and
572 "question" exactly what is going on. That's psychology...
579 A comment. The text is ignored. If the C</x> modifier enables
580 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
581 the comment as soon as it sees a C<)>, so there is no way to put a literal
584 =item C<(?imsx-imsx)>
587 One or more embedded pattern-match modifiers, to be turned on (or
588 turned off, if preceded by C<->) for the remainder of the pattern or
589 the remainder of the enclosing pattern group (if any). This is
590 particularly useful for dynamic patterns, such as those read in from a
591 configuration file, read in as an argument, are specified in a table
592 somewhere, etc. Consider the case that some of which want to be case
593 sensitive and some do not. The case insensitive ones need to include
594 merely C<(?i)> at the front of the pattern. For example:
597 if ( /$pattern/i ) { }
601 $pattern = "(?i)foobar";
602 if ( /$pattern/ ) { }
604 These modifiers are restored at the end of the enclosing group. For example,
608 will match a repeated (I<including the case>!) word C<blah> in any
609 case, assuming C<x> modifier, and no C<i> modifier outside this
615 =item C<(?imsx-imsx:pattern)>
617 This is for clustering, not capturing; it groups subexpressions like
618 "()", but doesn't make backreferences as "()" does. So
620 @fields = split(/\b(?:a|b|c)\b/)
624 @fields = split(/\b(a|b|c)\b/)
626 but doesn't spit out extra fields. It's also cheaper not to capture
627 characters if you don't need to.
629 Any letters between C<?> and C<:> act as flags modifiers as with
630 C<(?imsx-imsx)>. For example,
632 /(?s-i:more.*than).*million/i
634 is equivalent to the more verbose
636 /(?:(?s-i)more.*than).*million/i
639 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
641 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
642 matches a word followed by a tab, without including the tab in C<$&>.
645 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
647 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
648 matches any occurrence of "foo" that isn't followed by "bar". Note
649 however that look-ahead and look-behind are NOT the same thing. You cannot
650 use this for look-behind.
652 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
653 will not do what you want. That's because the C<(?!foo)> is just saying that
654 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
655 match. You would have to do something like C</(?!foo)...bar/> for that. We
656 say "like" because there's the case of your "bar" not having three characters
657 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
658 Sometimes it's still easier just to say:
660 if (/bar/ && $` !~ /foo$/)
662 For look-behind see below.
664 =item C<(?<=pattern)>
665 X<(?<=)> X<look-behind, positive> X<lookbehind, positive>
667 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
668 matches a word that follows a tab, without including the tab in C<$&>.
669 Works only for fixed-width look-behind.
671 =item C<(?<!pattern)>
672 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
674 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
675 matches any occurrence of "foo" that does not follow "bar". Works
676 only for fixed-width look-behind.
678 =item C<(?'NAME'pattern)>
680 =item C<< (?<NAME>pattern) >>
681 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
683 A named capture buffer. Identical in every respect to normal capturing
684 parens C<()> but for the additional fact that C<%+> may be used after
685 a succesful match to refer to a named buffer. See C<perlvar> for more
686 details on the C<%+> hash.
688 If multiple distinct capture buffers have the same name then the
689 $+{NAME} will refer to the leftmost defined buffer in the match.
691 The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent.
693 B<NOTE:> While the notation of this construct is the same as the similar
694 function in .NET regexes, the behavior is not, in Perl the buffers are
695 numbered sequentially regardless of being named or not. Thus in the
700 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
701 the opposite which is what a .NET regex hacker might expect.
703 Currently NAME is restricted to word chars only. In other words, it
704 must match C</^\w+$/>.
706 =item C<< \k<name> >>
708 =item C<< \k'name' >>
710 Named backreference. Similar to numeric backreferences, except that
711 the group is designated by name and not number. If multiple groups
712 have the same name then it refers to the leftmost defined group in
715 It is an error to refer to a name not defined by a C<(?<NAME>)>
716 earlier in the pattern.
718 Both forms are equivalent.
721 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
723 B<WARNING>: This extended regular expression feature is considered
724 experimental, and may be changed without notice. Code executed that
725 has side effects may not perform identically from version to version
726 due to the effect of future optimisations in the regex engine.
728 This zero-width assertion evaluates any embedded Perl code. It
729 always succeeds, and its C<code> is not interpolated. Currently,
730 the rules to determine where the C<code> ends are somewhat convoluted.
732 This feature can be used together with the special variable C<$^N> to
733 capture the results of submatches in variables without having to keep
734 track of the number of nested parentheses. For example:
736 $_ = "The brown fox jumps over the lazy dog";
737 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
738 print "color = $color, animal = $animal\n";
740 Inside the C<(?{...})> block, C<$_> refers to the string the regular
741 expression is matching against. You can also use C<pos()> to know what is
742 the current position of matching within this string.
744 The C<code> is properly scoped in the following sense: If the assertion
745 is backtracked (compare L<"Backtracking">), all changes introduced after
746 C<local>ization are undone, so that
750 (?{ $cnt = 0 }) # Initialize $cnt.
754 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
758 (?{ $res = $cnt }) # On success copy to non-localized
762 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
763 introduced value, because the scopes that restrict C<local> operators
766 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
767 switch. If I<not> used in this way, the result of evaluation of
768 C<code> is put into the special variable C<$^R>. This happens
769 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
770 inside the same regular expression.
772 The assignment to C<$^R> above is properly localized, so the old
773 value of C<$^R> is restored if the assertion is backtracked; compare
776 Due to an unfortunate implementation issue, the Perl code contained in these
777 blocks is treated as a compile time closure that can have seemingly bizarre
778 consequences when used with lexically scoped variables inside of subroutines
779 or loops. There are various workarounds for this, including simply using
780 global variables instead. If you are using this construct and strange results
781 occur then check for the use of lexically scoped variables.
783 For reasons of security, this construct is forbidden if the regular
784 expression involves run-time interpolation of variables, unless the
785 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
786 variables contain results of C<qr//> operator (see
787 L<perlop/"qr/STRING/imosx">).
789 This restriction is because of the wide-spread and remarkably convenient
790 custom of using run-time determined strings as patterns. For example:
796 Before Perl knew how to execute interpolated code within a pattern,
797 this operation was completely safe from a security point of view,
798 although it could raise an exception from an illegal pattern. If
799 you turn on the C<use re 'eval'>, though, it is no longer secure,
800 so you should only do so if you are also using taint checking.
801 Better yet, use the carefully constrained evaluation within a Safe
802 compartment. See L<perlsec> for details about both these mechanisms.
804 Because perl's regex engine is not currently re-entrant, interpolated
805 code may not invoke the regex engine either directly with C<m//> or C<s///>),
806 or indirectly with functions such as C<split>.
808 =item C<(??{ code })>
810 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
812 B<WARNING>: This extended regular expression feature is considered
813 experimental, and may be changed without notice. Code executed that
814 has side effects may not perform identically from version to version
815 due to the effect of future optimisations in the regex engine.
817 This is a "postponed" regular subexpression. The C<code> is evaluated
818 at run time, at the moment this subexpression may match. The result
819 of evaluation is considered as a regular expression and matched as
820 if it were inserted instead of this construct. Note that this means
821 that the contents of capture buffers defined inside an eval'ed pattern
822 are not available outside of the pattern, and vice versa, there is no
823 way for the inner pattern to refer to a capture buffer defined outside.
826 ('a' x 100)=~/(??{'(.)' x 100})/
828 B<will> match, it will B<not> set $1.
830 The C<code> is not interpolated. As before, the rules to determine
831 where the C<code> ends are currently somewhat convoluted.
833 The following pattern matches a parenthesized group:
838 (?> [^()]+ ) # Non-parens without backtracking
840 (??{ $re }) # Group with matching parens
845 See also C<(?PARNO)> for a different, more efficient way to accomplish
848 Because perl's regex engine is not currently re-entrant, delayed
849 code may not invoke the regex engine either directly with C<m//> or C<s///>),
850 or indirectly with functions such as C<split>.
852 Recursing deeper than 50 times without consuming any input string will
853 result in a fatal error. The maximum depth is compiled into perl, so
854 changing it requires a custom build.
856 =item C<(?PARNO)> C<(?R)> C<(?0)>
857 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)>
858 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
860 Similar to C<(??{ code })> except it does not involve compiling any code,
861 instead it treats the contents of a capture buffer as an independent
862 pattern that must match at the current position. Capture buffers
863 contained by the pattern will have the value as determined by the
866 PARNO is a sequence of digits (not starting with 0) whose value reflects
867 the paren-number of the capture buffer to recurse to. C<(?R)> recurses to
868 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
871 The following pattern matches a function foo() which may contain
872 balanced parenthesis as the argument.
874 $re = qr{ ( # paren group 1 (full function)
876 ( # paren group 2 (parens)
878 ( # paren group 3 (contents of parens)
880 (?> [^()]+ ) # Non-parens without backtracking
882 (?2) # Recurse to start of paren group 2
890 If the pattern was used as follows
892 'foo(bar(baz)+baz(bop))'=~/$re/
893 and print "\$1 = $1\n",
897 the output produced should be the following:
899 $1 = foo(bar(baz)+baz(bop))
900 $2 = (bar(baz)+baz(bop))
901 $3 = bar(baz)+baz(bop)
903 If there is no corresponding capture buffer defined, then it is a
904 fatal error. Recursing deeper than 50 times without consuming any input
905 string will also result in a fatal error. The maximum depth is compiled
906 into perl, so changing it requires a custom build.
908 B<Note> that this pattern does not behave the same way as the equivalent
909 PCRE or Python construct of the same form. In perl you can backtrack into
910 a recursed group, in PCRE and Python the recursed into group is treated
911 as atomic. Also, constructs like (?i:(?1)) or (?:(?i)(?1)) do not affect
912 the pattern being recursed into.
917 Recurse to a named subpattern. Identical to (?PARNO) except that the
918 parenthesis to recurse to is determined by name. If multiple parens have
919 the same name, then it recurses to the leftmost.
921 It is an error to refer to a name that is not declared somewhere in the
924 =item C<< (?>pattern) >>
925 X<backtrack> X<backtracking> X<atomic> X<possessive>
927 An "independent" subexpression, one which matches the substring
928 that a I<standalone> C<pattern> would match if anchored at the given
929 position, and it matches I<nothing other than this substring>. This
930 construct is useful for optimizations of what would otherwise be
931 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
932 It may also be useful in places where the "grab all you can, and do not
933 give anything back" semantic is desirable.
935 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
936 (anchored at the beginning of string, as above) will match I<all>
937 characters C<a> at the beginning of string, leaving no C<a> for
938 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
939 since the match of the subgroup C<a*> is influenced by the following
940 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
941 C<a*ab> will match fewer characters than a standalone C<a*>, since
942 this makes the tail match.
944 An effect similar to C<< (?>pattern) >> may be achieved by writing
945 C<(?=(pattern))\1>. This matches the same substring as a standalone
946 C<a+>, and the following C<\1> eats the matched string; it therefore
947 makes a zero-length assertion into an analogue of C<< (?>...) >>.
948 (The difference between these two constructs is that the second one
949 uses a capturing group, thus shifting ordinals of backreferences
950 in the rest of a regular expression.)
952 Consider this pattern:
963 That will efficiently match a nonempty group with matching parentheses
964 two levels deep or less. However, if there is no such group, it
965 will take virtually forever on a long string. That's because there
966 are so many different ways to split a long string into several
967 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
968 to a subpattern of the above pattern. Consider how the pattern
969 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
970 seconds, but that each extra letter doubles this time. This
971 exponential performance will make it appear that your program has
972 hung. However, a tiny change to this pattern
976 (?> [^()]+ ) # change x+ above to (?> x+ )
983 which uses C<< (?>...) >> matches exactly when the one above does (verifying
984 this yourself would be a productive exercise), but finishes in a fourth
985 the time when used on a similar string with 1000000 C<a>s. Be aware,
986 however, that this pattern currently triggers a warning message under
987 the C<use warnings> pragma or B<-w> switch saying it
988 C<"matches null string many times in regex">.
990 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
991 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
992 This was only 4 times slower on a string with 1000000 C<a>s.
994 The "grab all you can, and do not give anything back" semantic is desirable
995 in many situations where on the first sight a simple C<()*> looks like
996 the correct solution. Suppose we parse text with comments being delimited
997 by C<#> followed by some optional (horizontal) whitespace. Contrary to
998 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
999 the comment delimiter, because it may "give up" some whitespace if
1000 the remainder of the pattern can be made to match that way. The correct
1001 answer is either one of these:
1006 For example, to grab non-empty comments into $1, one should use either
1009 / (?> \# [ \t]* ) ( .+ ) /x;
1010 / \# [ \t]* ( [^ \t] .* ) /x;
1012 Which one you pick depends on which of these expressions better reflects
1013 the above specification of comments.
1015 In some literature this construct is called "atomic matching" or
1016 "possessive matching".
1018 Possessive quantifiers are equivalent to putting the item they are applied
1019 to inside of one of these constructs. The following equivalences apply:
1021 Quantifier Form Bracketing Form
1022 --------------- ---------------
1026 PAT{min,max}+ (?>PAT{min,max})
1028 =item C<(?(condition)yes-pattern|no-pattern)>
1031 =item C<(?(condition)yes-pattern)>
1033 Conditional expression. C<(condition)> should be either an integer in
1034 parentheses (which is valid if the corresponding pair of parentheses
1035 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
1036 name in angle brackets or single quotes (which is valid if a buffer
1037 with the given name matched), the special symbol (R) (true when
1038 evaluated inside of recursion or eval). Additionally the R may be
1039 followed by a number, (which will be true when evaluated when recursing
1040 inside of the appropriate group), or by C<&NAME> in which case it will
1041 be true only when evaluated during recursion in the named group.
1043 Here's a summary of the possible predicates:
1049 Checks if the numbered capturing buffer has matched something.
1051 =item (<NAME>) ('NAME')
1053 Checks if a buffer with the given name has matched something.
1057 Treats the code block as the condition
1061 Checks if the expression has been evaluated inside of recursion.
1065 Checks if the expression has been evaluated while executing directly
1066 inside of the n-th capture group. This check is the regex equivalent of
1068 if ((caller(0))[3] eq 'subname') { .. }
1070 In other words, it does not check the full recursion stack.
1074 Similar to C<(R1)>, this predicate checks to see if we're executing
1075 directly inside of the leftmost group with a given name (this is the same
1076 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1077 stack, but only the name of the innermost active recursion.
1081 In this case, the yes-pattern is never directly executed, and no
1082 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1083 See below for details.
1094 matches a chunk of non-parentheses, possibly included in parentheses
1097 A special form is the C<(DEFINE)> predicate, which never executes directly
1098 its yes-pattern, and does not allow a no-pattern. This allows to define
1099 subpatterns which will be executed only by using the recursion mechanism.
1100 This way, you can define a set of regular expression rules that can be
1101 bundled into any pattern you choose.
1103 It is recommended that for this usage you put the DEFINE block at the
1104 end of the pattern, and that you name any subpatterns defined within it.
1106 Also, it's worth noting that patterns defined this way probably will
1107 not be as efficient, as the optimiser is not very clever about
1108 handling them. YMMV.
1110 An example of how this might be used is as follows:
1112 /(?<NAME>(&NAME_PAT))(?<ADDR>(&ADDRESS_PAT))
1118 Note that capture buffers matched inside of recursion are not accessible
1119 after the recursion returns, so the extra layer of capturing buffers are
1120 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1121 C<$+{NAME}> would be.
1126 X<backtrack> X<backtracking>
1128 NOTE: This section presents an abstract approximation of regular
1129 expression behavior. For a more rigorous (and complicated) view of
1130 the rules involved in selecting a match among possible alternatives,
1131 see L<Combining pieces together>.
1133 A fundamental feature of regular expression matching involves the
1134 notion called I<backtracking>, which is currently used (when needed)
1135 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
1136 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1137 internally, but the general principle outlined here is valid.
1139 For a regular expression to match, the I<entire> regular expression must
1140 match, not just part of it. So if the beginning of a pattern containing a
1141 quantifier succeeds in a way that causes later parts in the pattern to
1142 fail, the matching engine backs up and recalculates the beginning
1143 part--that's why it's called backtracking.
1145 Here is an example of backtracking: Let's say you want to find the
1146 word following "foo" in the string "Food is on the foo table.":
1148 $_ = "Food is on the foo table.";
1149 if ( /\b(foo)\s+(\w+)/i ) {
1150 print "$2 follows $1.\n";
1153 When the match runs, the first part of the regular expression (C<\b(foo)>)
1154 finds a possible match right at the beginning of the string, and loads up
1155 $1 with "Foo". However, as soon as the matching engine sees that there's
1156 no whitespace following the "Foo" that it had saved in $1, it realizes its
1157 mistake and starts over again one character after where it had the
1158 tentative match. This time it goes all the way until the next occurrence
1159 of "foo". The complete regular expression matches this time, and you get
1160 the expected output of "table follows foo."
1162 Sometimes minimal matching can help a lot. Imagine you'd like to match
1163 everything between "foo" and "bar". Initially, you write something
1166 $_ = "The food is under the bar in the barn.";
1167 if ( /foo(.*)bar/ ) {
1171 Which perhaps unexpectedly yields:
1173 got <d is under the bar in the >
1175 That's because C<.*> was greedy, so you get everything between the
1176 I<first> "foo" and the I<last> "bar". Here it's more effective
1177 to use minimal matching to make sure you get the text between a "foo"
1178 and the first "bar" thereafter.
1180 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1181 got <d is under the >
1183 Here's another example: let's say you'd like to match a number at the end
1184 of a string, and you also want to keep the preceding part of the match.
1187 $_ = "I have 2 numbers: 53147";
1188 if ( /(.*)(\d*)/ ) { # Wrong!
1189 print "Beginning is <$1>, number is <$2>.\n";
1192 That won't work at all, because C<.*> was greedy and gobbled up the
1193 whole string. As C<\d*> can match on an empty string the complete
1194 regular expression matched successfully.
1196 Beginning is <I have 2 numbers: 53147>, number is <>.
1198 Here are some variants, most of which don't work:
1200 $_ = "I have 2 numbers: 53147";
1213 printf "%-12s ", $pat;
1215 print "<$1> <$2>\n";
1221 That will print out:
1223 (.*)(\d*) <I have 2 numbers: 53147> <>
1224 (.*)(\d+) <I have 2 numbers: 5314> <7>
1226 (.*?)(\d+) <I have > <2>
1227 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1228 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1229 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1230 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1232 As you see, this can be a bit tricky. It's important to realize that a
1233 regular expression is merely a set of assertions that gives a definition
1234 of success. There may be 0, 1, or several different ways that the
1235 definition might succeed against a particular string. And if there are
1236 multiple ways it might succeed, you need to understand backtracking to
1237 know which variety of success you will achieve.
1239 When using look-ahead assertions and negations, this can all get even
1240 trickier. Imagine you'd like to find a sequence of non-digits not
1241 followed by "123". You might try to write that as
1244 if ( /^\D*(?!123)/ ) { # Wrong!
1245 print "Yup, no 123 in $_\n";
1248 But that isn't going to match; at least, not the way you're hoping. It
1249 claims that there is no 123 in the string. Here's a clearer picture of
1250 why that pattern matches, contrary to popular expectations:
1255 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1256 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1258 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1259 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1267 You might have expected test 3 to fail because it seems to a more
1268 general purpose version of test 1. The important difference between
1269 them is that test 3 contains a quantifier (C<\D*>) and so can use
1270 backtracking, whereas test 1 will not. What's happening is
1271 that you've asked "Is it true that at the start of $x, following 0 or more
1272 non-digits, you have something that's not 123?" If the pattern matcher had
1273 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1276 The search engine will initially match C<\D*> with "ABC". Then it will
1277 try to match C<(?!123> with "123", which fails. But because
1278 a quantifier (C<\D*>) has been used in the regular expression, the
1279 search engine can backtrack and retry the match differently
1280 in the hope of matching the complete regular expression.
1282 The pattern really, I<really> wants to succeed, so it uses the
1283 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1284 time. Now there's indeed something following "AB" that is not
1285 "123". It's "C123", which suffices.
1287 We can deal with this by using both an assertion and a negation.
1288 We'll say that the first part in $1 must be followed both by a digit
1289 and by something that's not "123". Remember that the look-aheads
1290 are zero-width expressions--they only look, but don't consume any
1291 of the string in their match. So rewriting this way produces what
1292 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1294 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1295 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1299 In other words, the two zero-width assertions next to each other work as though
1300 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1301 matches only if you're at the beginning of the line AND the end of the
1302 line simultaneously. The deeper underlying truth is that juxtaposition in
1303 regular expressions always means AND, except when you write an explicit OR
1304 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1305 although the attempted matches are made at different positions because "a"
1306 is not a zero-width assertion, but a one-width assertion.
1308 B<WARNING>: particularly complicated regular expressions can take
1309 exponential time to solve because of the immense number of possible
1310 ways they can use backtracking to try match. For example, without
1311 internal optimizations done by the regular expression engine, this will
1312 take a painfully long time to run:
1314 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1316 And if you used C<*>'s in the internal groups instead of limiting them
1317 to 0 through 5 matches, then it would take forever--or until you ran
1318 out of stack space. Moreover, these internal optimizations are not
1319 always applicable. For example, if you put C<{0,5}> instead of C<*>
1320 on the external group, no current optimization is applicable, and the
1321 match takes a long time to finish.
1323 A powerful tool for optimizing such beasts is what is known as an
1324 "independent group",
1325 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1326 zero-length look-ahead/look-behind assertions will not backtrack to make
1327 the tail match, since they are in "logical" context: only
1328 whether they match is considered relevant. For an example
1329 where side-effects of look-ahead I<might> have influenced the
1330 following match, see L<C<< (?>pattern) >>>.
1332 =head2 Version 8 Regular Expressions
1333 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1335 In case you're not familiar with the "regular" Version 8 regex
1336 routines, here are the pattern-matching rules not described above.
1338 Any single character matches itself, unless it is a I<metacharacter>
1339 with a special meaning described here or above. You can cause
1340 characters that normally function as metacharacters to be interpreted
1341 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1342 character; "\\" matches a "\"). A series of characters matches that
1343 series of characters in the target string, so the pattern C<blurfl>
1344 would match "blurfl" in the target string.
1346 You can specify a character class, by enclosing a list of characters
1347 in C<[]>, which will match any one character from the list. If the
1348 first character after the "[" is "^", the class matches any character not
1349 in the list. Within a list, the "-" character specifies a
1350 range, so that C<a-z> represents all characters between "a" and "z",
1351 inclusive. If you want either "-" or "]" itself to be a member of a
1352 class, put it at the start of the list (possibly after a "^"), or
1353 escape it with a backslash. "-" is also taken literally when it is
1354 at the end of the list, just before the closing "]". (The
1355 following all specify the same class of three characters: C<[-az]>,
1356 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1357 specifies a class containing twenty-six characters, even on EBCDIC
1358 based coded character sets.) Also, if you try to use the character
1359 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1360 a range, that's not a range, the "-" is understood literally.
1362 Note also that the whole range idea is rather unportable between
1363 character sets--and even within character sets they may cause results
1364 you probably didn't expect. A sound principle is to use only ranges
1365 that begin from and end at either alphabets of equal case ([a-e],
1366 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1367 spell out the character sets in full.
1369 Characters may be specified using a metacharacter syntax much like that
1370 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1371 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1372 of octal digits, matches the character whose coded character set value
1373 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1374 matches the character whose numeric value is I<nn>. The expression \cI<x>
1375 matches the character control-I<x>. Finally, the "." metacharacter
1376 matches any character except "\n" (unless you use C</s>).
1378 You can specify a series of alternatives for a pattern using "|" to
1379 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1380 or "foe" in the target string (as would C<f(e|i|o)e>). The
1381 first alternative includes everything from the last pattern delimiter
1382 ("(", "[", or the beginning of the pattern) up to the first "|", and
1383 the last alternative contains everything from the last "|" to the next
1384 pattern delimiter. That's why it's common practice to include
1385 alternatives in parentheses: to minimize confusion about where they
1388 Alternatives are tried from left to right, so the first
1389 alternative found for which the entire expression matches, is the one that
1390 is chosen. This means that alternatives are not necessarily greedy. For
1391 example: when matching C<foo|foot> against "barefoot", only the "foo"
1392 part will match, as that is the first alternative tried, and it successfully
1393 matches the target string. (This might not seem important, but it is
1394 important when you are capturing matched text using parentheses.)
1396 Also remember that "|" is interpreted as a literal within square brackets,
1397 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1399 Within a pattern, you may designate subpatterns for later reference
1400 by enclosing them in parentheses, and you may refer back to the
1401 I<n>th subpattern later in the pattern using the metacharacter
1402 \I<n>. Subpatterns are numbered based on the left to right order
1403 of their opening parenthesis. A backreference matches whatever
1404 actually matched the subpattern in the string being examined, not
1405 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1406 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1407 1 matched "0x", even though the rule C<0|0x> could potentially match
1408 the leading 0 in the second number.
1410 =head2 Warning on \1 vs $1
1412 Some people get too used to writing things like:
1414 $pattern =~ s/(\W)/\\\1/g;
1416 This is grandfathered for the RHS of a substitute to avoid shocking the
1417 B<sed> addicts, but it's a dirty habit to get into. That's because in
1418 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1419 the usual double-quoted string means a control-A. The customary Unix
1420 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1421 of doing that, you get yourself into trouble if you then add an C</e>
1424 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1430 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1431 C<${1}000>. The operation of interpolation should not be confused
1432 with the operation of matching a backreference. Certainly they mean two
1433 different things on the I<left> side of the C<s///>.
1435 =head2 Repeated patterns matching zero-length substring
1437 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1439 Regular expressions provide a terse and powerful programming language. As
1440 with most other power tools, power comes together with the ability
1443 A common abuse of this power stems from the ability to make infinite
1444 loops using regular expressions, with something as innocuous as:
1446 'foo' =~ m{ ( o? )* }x;
1448 The C<o?> can match at the beginning of C<'foo'>, and since the position
1449 in the string is not moved by the match, C<o?> would match again and again
1450 because of the C<*> modifier. Another common way to create a similar cycle
1451 is with the looping modifier C<//g>:
1453 @matches = ( 'foo' =~ m{ o? }xg );
1457 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1459 or the loop implied by split().
1461 However, long experience has shown that many programming tasks may
1462 be significantly simplified by using repeated subexpressions that
1463 may match zero-length substrings. Here's a simple example being:
1465 @chars = split //, $string; # // is not magic in split
1466 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1468 Thus Perl allows such constructs, by I<forcefully breaking
1469 the infinite loop>. The rules for this are different for lower-level
1470 loops given by the greedy modifiers C<*+{}>, and for higher-level
1471 ones like the C</g> modifier or split() operator.
1473 The lower-level loops are I<interrupted> (that is, the loop is
1474 broken) when Perl detects that a repeated expression matched a
1475 zero-length substring. Thus
1477 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1479 is made equivalent to
1481 m{ (?: NON_ZERO_LENGTH )*
1486 The higher level-loops preserve an additional state between iterations:
1487 whether the last match was zero-length. To break the loop, the following
1488 match after a zero-length match is prohibited to have a length of zero.
1489 This prohibition interacts with backtracking (see L<"Backtracking">),
1490 and so the I<second best> match is chosen if the I<best> match is of
1498 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1499 match given by non-greedy C<??> is the zero-length match, and the I<second
1500 best> match is what is matched by C<\w>. Thus zero-length matches
1501 alternate with one-character-long matches.
1503 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1504 position one notch further in the string.
1506 The additional state of being I<matched with zero-length> is associated with
1507 the matched string, and is reset by each assignment to pos().
1508 Zero-length matches at the end of the previous match are ignored
1511 =head2 Combining pieces together
1513 Each of the elementary pieces of regular expressions which were described
1514 before (such as C<ab> or C<\Z>) could match at most one substring
1515 at the given position of the input string. However, in a typical regular
1516 expression these elementary pieces are combined into more complicated
1517 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1518 (in these examples C<S> and C<T> are regular subexpressions).
1520 Such combinations can include alternatives, leading to a problem of choice:
1521 if we match a regular expression C<a|ab> against C<"abc">, will it match
1522 substring C<"a"> or C<"ab">? One way to describe which substring is
1523 actually matched is the concept of backtracking (see L<"Backtracking">).
1524 However, this description is too low-level and makes you think
1525 in terms of a particular implementation.
1527 Another description starts with notions of "better"/"worse". All the
1528 substrings which may be matched by the given regular expression can be
1529 sorted from the "best" match to the "worst" match, and it is the "best"
1530 match which is chosen. This substitutes the question of "what is chosen?"
1531 by the question of "which matches are better, and which are worse?".
1533 Again, for elementary pieces there is no such question, since at most
1534 one match at a given position is possible. This section describes the
1535 notion of better/worse for combining operators. In the description
1536 below C<S> and C<T> are regular subexpressions.
1542 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1543 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1544 which can be matched by C<T>.
1546 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1549 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1550 C<B> is better match for C<T> than C<B'>.
1554 When C<S> can match, it is a better match than when only C<T> can match.
1556 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1557 two matches for C<T>.
1559 =item C<S{REPEAT_COUNT}>
1561 Matches as C<SSS...S> (repeated as many times as necessary).
1565 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1567 =item C<S{min,max}?>
1569 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1571 =item C<S?>, C<S*>, C<S+>
1573 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1575 =item C<S??>, C<S*?>, C<S+?>
1577 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1581 Matches the best match for C<S> and only that.
1583 =item C<(?=S)>, C<(?<=S)>
1585 Only the best match for C<S> is considered. (This is important only if
1586 C<S> has capturing parentheses, and backreferences are used somewhere
1587 else in the whole regular expression.)
1589 =item C<(?!S)>, C<(?<!S)>
1591 For this grouping operator there is no need to describe the ordering, since
1592 only whether or not C<S> can match is important.
1594 =item C<(??{ EXPR })>, C<(?PARNO)>
1596 The ordering is the same as for the regular expression which is
1597 the result of EXPR, or the pattern contained by capture buffer PARNO.
1599 =item C<(?(condition)yes-pattern|no-pattern)>
1601 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1602 already determined. The ordering of the matches is the same as for the
1603 chosen subexpression.
1607 The above recipes describe the ordering of matches I<at a given position>.
1608 One more rule is needed to understand how a match is determined for the
1609 whole regular expression: a match at an earlier position is always better
1610 than a match at a later position.
1612 =head2 Creating custom RE engines
1614 Overloaded constants (see L<overload>) provide a simple way to extend
1615 the functionality of the RE engine.
1617 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
1618 matches at boundary between whitespace characters and non-whitespace
1619 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1620 at these positions, so we want to have each C<\Y|> in the place of the
1621 more complicated version. We can create a module C<customre> to do
1629 die "No argument to customre::import allowed" if @_;
1630 overload::constant 'qr' => \&convert;
1633 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1635 # We must also take care of not escaping the legitimate \\Y|
1636 # sequence, hence the presence of '\\' in the conversion rules.
1637 my %rules = ( '\\' => '\\\\',
1638 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
1644 { $rules{$1} or invalid($re,$1) }sgex;
1648 Now C<use customre> enables the new escape in constant regular
1649 expressions, i.e., those without any runtime variable interpolations.
1650 As documented in L<overload>, this conversion will work only over
1651 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
1652 part of this regular expression needs to be converted explicitly
1653 (but only if the special meaning of C<\Y|> should be enabled inside $re):
1658 $re = customre::convert $re;
1663 This document varies from difficult to understand to completely
1664 and utterly opaque. The wandering prose riddled with jargon is
1665 hard to fathom in several places.
1667 This document needs a rewrite that separates the tutorial content
1668 from the reference content.
1676 L<perlop/"Regexp Quote-Like Operators">.
1678 L<perlop/"Gory details of parsing quoted constructs">.
1688 I<Mastering Regular Expressions> by Jeffrey Friedl, published
1689 by O'Reilly and Associates.