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 Because patterns are processed as double quoted strings, the following
159 X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q>
160 X<\0> X<\c> X<\N> X<\x>
166 \a alarm (bell) (BEL)
167 \e escape (think troff) (ESC)
168 \033 octal char (think of a PDP-11)
170 \x{263a} wide hex char (Unicode SMILEY)
173 \l lowercase next char (think vi)
174 \u uppercase next char (think vi)
175 \L lowercase till \E (think vi)
176 \U uppercase till \E (think vi)
177 \E end case modification (think vi)
178 \Q quote (disable) pattern metacharacters till \E
180 If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u>
181 and C<\U> is taken from the current locale. See L<perllocale>. For
182 documentation of C<\N{name}>, see L<charnames>.
184 You cannot include a literal C<$> or C<@> within a C<\Q> sequence.
185 An unescaped C<$> or C<@> interpolates the corresponding variable,
186 while escaping will cause the literal string C<\$> to be matched.
187 You'll need to write something like C<m/\Quser\E\@\Qhost/>.
189 In addition, Perl defines the following:
191 X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C>
192 X<word> X<whitespace>
194 \w Match a "word" character (alphanumeric plus "_")
195 \W Match a non-"word" character
196 \s Match a whitespace character
197 \S Match a non-whitespace character
198 \d Match a digit character
199 \D Match a non-digit character
200 \pP Match P, named property. Use \p{Prop} for longer names.
202 \X Match eXtended Unicode "combining character sequence",
203 equivalent to (?:\PM\pM*)
204 \C Match a single C char (octet) even under Unicode.
205 NOTE: breaks up characters into their UTF-8 bytes,
206 so you may end up with malformed pieces of UTF-8.
207 Unsupported in lookbehind.
208 \1 Backreference to a a specific group.
209 '1' may actually be any positive integer
210 \k<name> Named backreference
211 \N{name} Named unicode character, or unicode escape.
212 \x12 Hexadecimal escape sequence
213 \x{1234} Long hexadecimal escape sequence
215 A C<\w> matches a single alphanumeric character (an alphabetic
216 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
217 to match a string of Perl-identifier characters (which isn't the same
218 as matching an English word). If C<use locale> is in effect, the list
219 of alphabetic characters generated by C<\w> is taken from the current
220 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
221 C<\d>, and C<\D> within character classes, but if you try to use them
222 as endpoints of a range, that's not a range, the "-" is understood
223 literally. If Unicode is in effect, C<\s> matches also "\x{85}",
224 "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about
225 C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general.
226 You can define your own C<\p> and C<\P> properties, see L<perlunicode>.
229 The POSIX character class syntax
234 is also available. Note that the C<[> and C<]> braces are I<literal>;
235 they must always be used within a character class expression.
238 $string =~ /[[:alpha:]]/;
240 # this is not, and will generate a warning:
241 $string =~ /[:alpha:]/;
243 The available classes and their backslash equivalents (if available) are
246 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
247 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
268 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
272 Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
273 also the (very rare) "vertical tabulator", "\ck", chr(11).
277 A Perl extension, see above.
281 For example use C<[:upper:]> to match all the uppercase characters.
282 Note that the C<[]> are part of the C<[::]> construct, not part of the
283 whole character class. For example:
287 matches zero, one, any alphabetic character, and the percentage sign.
289 The following equivalences to Unicode \p{} constructs and equivalent
290 backslash character classes (if available), will hold:
291 X<character class> X<\p> X<\p{}>
293 [[:...:]] \p{...} backslash
311 For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent.
313 If the C<utf8> pragma is not used but the C<locale> pragma is, the
314 classes correlate with the usual isalpha(3) interface (except for
317 The assumedly non-obviously named classes are:
324 Any control character. Usually characters that don't produce output as
325 such but instead control the terminal somehow: for example newline and
326 backspace are control characters. All characters with ord() less than
327 32 are most often classified as control characters (assuming ASCII,
328 the ISO Latin character sets, and Unicode), as is the character with
329 the ord() value of 127 (C<DEL>).
334 Any alphanumeric or punctuation (special) character.
339 Any alphanumeric or punctuation (special) character or the space character.
344 Any punctuation (special) character.
349 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
350 work just fine) it is included for completeness.
354 You can negate the [::] character classes by prefixing the class name
355 with a '^'. This is a Perl extension. For example:
356 X<character class, negation>
358 POSIX traditional Unicode
360 [[:^digit:]] \D \P{IsDigit}
361 [[:^space:]] \S \P{IsSpace}
362 [[:^word:]] \W \P{IsWord}
364 Perl respects the POSIX standard in that POSIX character classes are
365 only supported within a character class. The POSIX character classes
366 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
367 use them will cause an error.
369 Perl defines the following zero-width assertions:
370 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
371 X<regexp, zero-width assertion>
372 X<regular expression, zero-width assertion>
373 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
375 \b Match a word boundary
376 \B Match a non-(word boundary)
377 \A Match only at beginning of string
378 \Z Match only at end of string, or before newline at the end
379 \z Match only at end of string
380 \G Match only at pos() (e.g. at the end-of-match position
383 A word boundary (C<\b>) is a spot between two characters
384 that has a C<\w> on one side of it and a C<\W> on the other side
385 of it (in either order), counting the imaginary characters off the
386 beginning and end of the string as matching a C<\W>. (Within
387 character classes C<\b> represents backspace rather than a word
388 boundary, just as it normally does in any double-quoted string.)
389 The C<\A> and C<\Z> are just like "^" and "$", except that they
390 won't match multiple times when the C</m> modifier is used, while
391 "^" and "$" will match at every internal line boundary. To match
392 the actual end of the string and not ignore an optional trailing
394 X<\b> X<\A> X<\Z> X<\z> X</m>
396 The C<\G> assertion can be used to chain global matches (using
397 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
398 It is also useful when writing C<lex>-like scanners, when you have
399 several patterns that you want to match against consequent substrings
400 of your string, see the previous reference. The actual location
401 where C<\G> will match can also be influenced by using C<pos()> as
402 an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully
403 supported when anchored to the start of the pattern; while it
404 is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some
405 such uses (C</.\G/g>, for example) currently cause problems, and
406 it is recommended that you avoid such usage for now.
409 The bracketing construct C<( ... )> creates capture buffers. To
410 refer to the digit'th buffer use \<digit> within the
411 match. Outside the match use "$" instead of "\". (The
412 \<digit> notation works in certain circumstances outside
413 the match. See the warning below about \1 vs $1 for details.)
414 Referring back to another part of the match is called a
416 X<regex, capture buffer> X<regexp, capture buffer>
417 X<regular expression, capture buffer> X<backreference>
419 There is no limit to the number of captured substrings that you may
420 use. However Perl also uses \10, \11, etc. as aliases for \010,
421 \011, etc. (Recall that 0 means octal, so \011 is the character at
422 number 9 in your coded character set; which would be the 10th character,
423 a horizontal tab under ASCII.) Perl resolves this
424 ambiguity by interpreting \10 as a backreference only if at least 10
425 left parentheses have opened before it. Likewise \11 is a
426 backreference only if at least 11 left parentheses have opened
427 before it. And so on. \1 through \9 are always interpreted as
430 Additionally, as of Perl 5.10 you may use named capture buffers and named
431 backreferences. The notation is C<< (?<name>...) >> and C<< \k<name> >>
432 (you may also use single quotes instead of angle brackets to quote the
433 name). The only difference with named capture buffers and unnamed ones is
434 that multiple buffers may have the same name and that the contents of
435 named capture buffers is available via the C<%+> hash. When multiple
436 groups share the same name C<$+{name}> and C<< \k<name> >> refer to the
437 leftmost defined group, thus it's possible to do things with named capture
438 buffers that would otherwise require C<(??{})> code to accomplish. Named
439 capture buffers are numbered just as normal capture buffers are and may be
440 referenced via the magic numeric variables or via numeric backreferences
445 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
447 /(.)\1/ # find first doubled char
448 and print "'$1' is the first doubled character\n";
450 /(?<char>.)\k<char>/ # ... a different way
451 and print "'$+{char}' is the first doubled character\n";
453 /(?<char>.)\1/ # ... mix and match
454 and print "'$1' is the first doubled character\n";
456 if (/Time: (..):(..):(..)/) { # parse out values
462 Several special variables also refer back to portions of the previous
463 match. C<$+> returns whatever the last bracket match matched.
464 C<$&> returns the entire matched string. (At one point C<$0> did
465 also, but now it returns the name of the program.) C<$`> returns
466 everything before the matched string. C<$'> returns everything
467 after the matched string. And C<$^N> contains whatever was matched by
468 the most-recently closed group (submatch). C<$^N> can be used in
469 extended patterns (see below), for example to assign a submatch to a
471 X<$+> X<$^N> X<$&> X<$`> X<$'>
473 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
474 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
475 until the end of the enclosing block or until the next successful
476 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
477 X<$+> X<$^N> X<$&> X<$`> X<$'>
478 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
481 B<NOTE>: failed matches in Perl do not reset the match variables,
482 which makes it easier to write code that tests for a series of more
483 specific cases and remembers the best match.
485 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
486 C<$'> anywhere in the program, it has to provide them for every
487 pattern match. This may substantially slow your program. Perl
488 uses the same mechanism to produce $1, $2, etc, so you also pay a
489 price for each pattern that contains capturing parentheses. (To
490 avoid this cost while retaining the grouping behaviour, use the
491 extended regular expression C<(?: ... )> instead.) But if you never
492 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
493 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
494 if you can, but if you can't (and some algorithms really appreciate
495 them), once you've used them once, use them at will, because you've
496 already paid the price. As of 5.005, C<$&> is not so costly as the
500 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
501 C<\w>, C<\n>. Unlike some other regular expression languages, there
502 are no backslashed symbols that aren't alphanumeric. So anything
503 that looks like \\, \(, \), \<, \>, \{, or \} is always
504 interpreted as a literal character, not a metacharacter. This was
505 once used in a common idiom to disable or quote the special meanings
506 of regular expression metacharacters in a string that you want to
507 use for a pattern. Simply quote all non-"word" characters:
509 $pattern =~ s/(\W)/\\$1/g;
511 (If C<use locale> is set, then this depends on the current locale.)
512 Today it is more common to use the quotemeta() function or the C<\Q>
513 metaquoting escape sequence to disable all metacharacters' special
516 /$unquoted\Q$quoted\E$unquoted/
518 Beware that if you put literal backslashes (those not inside
519 interpolated variables) between C<\Q> and C<\E>, double-quotish
520 backslash interpolation may lead to confusing results. If you
521 I<need> to use literal backslashes within C<\Q...\E>,
522 consult L<perlop/"Gory details of parsing quoted constructs">.
524 =head2 Extended Patterns
526 Perl also defines a consistent extension syntax for features not
527 found in standard tools like B<awk> and B<lex>. The syntax is a
528 pair of parentheses with a question mark as the first thing within
529 the parentheses. The character after the question mark indicates
532 The stability of these extensions varies widely. Some have been
533 part of the core language for many years. Others are experimental
534 and may change without warning or be completely removed. Check
535 the documentation on an individual feature to verify its current
538 A question mark was chosen for this and for the minimal-matching
539 construct because 1) question marks are rare in older regular
540 expressions, and 2) whenever you see one, you should stop and
541 "question" exactly what is going on. That's psychology...
548 A comment. The text is ignored. If the C</x> modifier enables
549 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
550 the comment as soon as it sees a C<)>, so there is no way to put a literal
553 =item C<(?imsx-imsx)>
556 One or more embedded pattern-match modifiers, to be turned on (or
557 turned off, if preceded by C<->) for the remainder of the pattern or
558 the remainder of the enclosing pattern group (if any). This is
559 particularly useful for dynamic patterns, such as those read in from a
560 configuration file, read in as an argument, are specified in a table
561 somewhere, etc. Consider the case that some of which want to be case
562 sensitive and some do not. The case insensitive ones need to include
563 merely C<(?i)> at the front of the pattern. For example:
566 if ( /$pattern/i ) { }
570 $pattern = "(?i)foobar";
571 if ( /$pattern/ ) { }
573 These modifiers are restored at the end of the enclosing group. For example,
577 will match a repeated (I<including the case>!) word C<blah> in any
578 case, assuming C<x> modifier, and no C<i> modifier outside this
584 =item C<(?imsx-imsx:pattern)>
586 This is for clustering, not capturing; it groups subexpressions like
587 "()", but doesn't make backreferences as "()" does. So
589 @fields = split(/\b(?:a|b|c)\b/)
593 @fields = split(/\b(a|b|c)\b/)
595 but doesn't spit out extra fields. It's also cheaper not to capture
596 characters if you don't need to.
598 Any letters between C<?> and C<:> act as flags modifiers as with
599 C<(?imsx-imsx)>. For example,
601 /(?s-i:more.*than).*million/i
603 is equivalent to the more verbose
605 /(?:(?s-i)more.*than).*million/i
608 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
610 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
611 matches a word followed by a tab, without including the tab in C<$&>.
614 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
616 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
617 matches any occurrence of "foo" that isn't followed by "bar". Note
618 however that look-ahead and look-behind are NOT the same thing. You cannot
619 use this for look-behind.
621 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
622 will not do what you want. That's because the C<(?!foo)> is just saying that
623 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
624 match. You would have to do something like C</(?!foo)...bar/> for that. We
625 say "like" because there's the case of your "bar" not having three characters
626 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
627 Sometimes it's still easier just to say:
629 if (/bar/ && $` !~ /foo$/)
631 For look-behind see below.
633 =item C<(?<=pattern)>
634 X<(?<=)> X<look-behind, positive> X<lookbehind, positive>
636 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
637 matches a word that follows a tab, without including the tab in C<$&>.
638 Works only for fixed-width look-behind.
640 =item C<(?<!pattern)>
641 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
643 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
644 matches any occurrence of "foo" that does not follow "bar". Works
645 only for fixed-width look-behind.
647 =item C<(?'NAME'pattern)>
649 =item C<< (?<NAME>pattern) >>
650 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
652 A named capture buffer. Identical in every respect to normal capturing
653 parens C<()> but for the additional fact that C<%+> may be used after
654 a succesful match to refer to a named buffer. See C<perlvar> for more
655 details on the C<%+> hash.
657 If multiple distinct capture buffers have the same name then the
658 $+{NAME} will refer to the leftmost defined buffer in the match.
660 The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent.
662 B<NOTE:> While the notation of this construct is the same as the similar
663 function in .NET regexes, the behavior is not, in Perl the buffers are
664 numbered sequentially regardless of being named or not. Thus in the
669 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
670 the opposite which is what a .NET regex hacker might expect.
672 Currently NAME is restricted to word chars only. In other words, it
673 must match C</^\w+$/>.
675 =item C<< \k<name> >>
677 =item C<< \k'name' >>
679 Named backreference. Similar to numeric backreferences, except that
680 the group is designated by name and not number. If multiple groups
681 have the same name then it refers to the leftmost defined group in
684 It is an error to refer to a name not defined by a C<(?<NAME>)>
685 earlier in the pattern.
687 Both forms are equivalent.
690 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
692 B<WARNING>: This extended regular expression feature is considered
693 highly experimental, and may be changed or deleted without notice.
695 This zero-width assertion evaluates any embedded Perl code. It
696 always succeeds, and its C<code> is not interpolated. Currently,
697 the rules to determine where the C<code> ends are somewhat convoluted.
699 This feature can be used together with the special variable C<$^N> to
700 capture the results of submatches in variables without having to keep
701 track of the number of nested parentheses. For example:
703 $_ = "The brown fox jumps over the lazy dog";
704 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
705 print "color = $color, animal = $animal\n";
707 Inside the C<(?{...})> block, C<$_> refers to the string the regular
708 expression is matching against. You can also use C<pos()> to know what is
709 the current position of matching within this string.
711 The C<code> is properly scoped in the following sense: If the assertion
712 is backtracked (compare L<"Backtracking">), all changes introduced after
713 C<local>ization are undone, so that
717 (?{ $cnt = 0 }) # Initialize $cnt.
721 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
725 (?{ $res = $cnt }) # On success copy to non-localized
729 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
730 introduced value, because the scopes that restrict C<local> operators
733 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
734 switch. If I<not> used in this way, the result of evaluation of
735 C<code> is put into the special variable C<$^R>. This happens
736 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
737 inside the same regular expression.
739 The assignment to C<$^R> above is properly localized, so the old
740 value of C<$^R> is restored if the assertion is backtracked; compare
743 Due to an unfortunate implementation issue, the Perl code contained in these
744 blocks is treated as a compile time closure that can have seemingly bizarre
745 consequences when used with lexically scoped variables inside of subroutines
746 or loops. There are various workarounds for this, including simply using
747 global variables instead. If you are using this construct and strange results
748 occur then check for the use of lexically scoped variables.
750 For reasons of security, this construct is forbidden if the regular
751 expression involves run-time interpolation of variables, unless the
752 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
753 variables contain results of C<qr//> operator (see
754 L<perlop/"qr/STRING/imosx">).
756 This restriction is because of the wide-spread and remarkably convenient
757 custom of using run-time determined strings as patterns. For example:
763 Before Perl knew how to execute interpolated code within a pattern,
764 this operation was completely safe from a security point of view,
765 although it could raise an exception from an illegal pattern. If
766 you turn on the C<use re 'eval'>, though, it is no longer secure,
767 so you should only do so if you are also using taint checking.
768 Better yet, use the carefully constrained evaluation within a Safe
769 compartment. See L<perlsec> for details about both these mechanisms.
771 Because perl's regex engine is not currently re-entrant, interpolated
772 code may not invoke the regex engine either directly with C<m//> or C<s///>),
773 or indirectly with functions such as C<split>.
775 =item C<(??{ code })>
777 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
779 B<WARNING>: This extended regular expression feature is considered
780 highly experimental, and may be changed or deleted without notice.
781 A simplified version of the syntax may be introduced for commonly
784 This is a "postponed" regular subexpression. The C<code> is evaluated
785 at run time, at the moment this subexpression may match. The result
786 of evaluation is considered as a regular expression and matched as
787 if it were inserted instead of this construct. Note that this means
788 that the contents of capture buffers defined inside an eval'ed pattern
789 are not available outside of the pattern, and vice versa, there is no
790 way for the inner pattern to refer to a capture buffer defined outside.
793 ('a' x 100)=~/(??{'(.)' x 100})/
795 B<will> match, it will B<not> set $1.
797 The C<code> is not interpolated. As before, the rules to determine
798 where the C<code> ends are currently somewhat convoluted.
800 The following pattern matches a parenthesized group:
805 (?> [^()]+ ) # Non-parens without backtracking
807 (??{ $re }) # Group with matching parens
812 See also C<(?PARNO)> for a different, more efficient way to accomplish
815 Because perl's regex engine is not currently re-entrant, delayed
816 code may not invoke the regex engine either directly with C<m//> or C<s///>),
817 or indirectly with functions such as C<split>.
819 Recursing deeper than 50 times without consuming any input string will
820 result in a fatal error. The maximum depth is compiled into perl, so
821 changing it requires a custom build.
823 =item C<(?PARNO)> C<(?R)>
826 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
828 B<WARNING>: This extended regular expression feature is considered
829 highly experimental, and may be changed or deleted without notice.
831 Similar to C<(??{ code })> except it does not involve compiling any code,
832 instead it treats the contents of a capture buffer as an independent
833 pattern that must match at the current position. Capture buffers
834 contained by the pattern will have the value as determined by the
837 PARNO is a sequence of digits not starting with 0 whose value
838 reflects the paren-number of the capture buffer to recurse to.
839 C<(?R)> curses to the beginning of the pattern.
841 The following pattern matches a function foo() which may contain
842 balanced parenthesis as the argument.
844 $re = qr{ ( # paren group 1 (full function)
846 ( # paren group 2 (parens)
848 ( # paren group 3 (contents of parens)
850 (?> [^()]+ ) # Non-parens without backtracking
852 (?2) # Recurse to start of paren group 2
860 If the pattern was used as follows
862 'foo(bar(baz)+baz(bop))'=~/$re/
863 and print "\$1 = $1\n",
867 the output produced should be the following:
869 $1 = foo(bar(baz)+baz(bop))
870 $2 = (bar(baz)+baz(bop))
871 $3 = bar(baz)+baz(bop)
873 If there is no corresponding capture buffer defined, then it is a
874 fatal error. Recursing deeper than 50 times without consuming any input
875 string will also result in a fatal error. The maximum depth is compiled
876 into perl, so changing it requires a custom build.
878 B<Note> that this pattern does not behave the same way as the equivalent
879 PCRE or Python construct of the same form. In perl you can backtrack into
880 a recursed group, in PCRE and Python the recursed into group is treated
881 as atomic. Also, constructs like (?i:(?1)) or (?:(?i)(?1)) do not affect
882 the pattern being recursed into.
884 =item C<< (?>pattern) >>
885 X<backtrack> X<backtracking> X<atomic> X<possessive>
887 B<WARNING>: This extended regular expression feature is considered
888 highly experimental, and may be changed or deleted without notice.
890 An "independent" subexpression, one which matches the substring
891 that a I<standalone> C<pattern> would match if anchored at the given
892 position, and it matches I<nothing other than this substring>. This
893 construct is useful for optimizations of what would otherwise be
894 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
895 It may also be useful in places where the "grab all you can, and do not
896 give anything back" semantic is desirable.
898 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
899 (anchored at the beginning of string, as above) will match I<all>
900 characters C<a> at the beginning of string, leaving no C<a> for
901 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
902 since the match of the subgroup C<a*> is influenced by the following
903 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
904 C<a*ab> will match fewer characters than a standalone C<a*>, since
905 this makes the tail match.
907 An effect similar to C<< (?>pattern) >> may be achieved by writing
908 C<(?=(pattern))\1>. This matches the same substring as a standalone
909 C<a+>, and the following C<\1> eats the matched string; it therefore
910 makes a zero-length assertion into an analogue of C<< (?>...) >>.
911 (The difference between these two constructs is that the second one
912 uses a capturing group, thus shifting ordinals of backreferences
913 in the rest of a regular expression.)
915 Consider this pattern:
926 That will efficiently match a nonempty group with matching parentheses
927 two levels deep or less. However, if there is no such group, it
928 will take virtually forever on a long string. That's because there
929 are so many different ways to split a long string into several
930 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
931 to a subpattern of the above pattern. Consider how the pattern
932 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
933 seconds, but that each extra letter doubles this time. This
934 exponential performance will make it appear that your program has
935 hung. However, a tiny change to this pattern
939 (?> [^()]+ ) # change x+ above to (?> x+ )
946 which uses C<< (?>...) >> matches exactly when the one above does (verifying
947 this yourself would be a productive exercise), but finishes in a fourth
948 the time when used on a similar string with 1000000 C<a>s. Be aware,
949 however, that this pattern currently triggers a warning message under
950 the C<use warnings> pragma or B<-w> switch saying it
951 C<"matches null string many times in regex">.
953 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
954 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
955 This was only 4 times slower on a string with 1000000 C<a>s.
957 The "grab all you can, and do not give anything back" semantic is desirable
958 in many situations where on the first sight a simple C<()*> looks like
959 the correct solution. Suppose we parse text with comments being delimited
960 by C<#> followed by some optional (horizontal) whitespace. Contrary to
961 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
962 the comment delimiter, because it may "give up" some whitespace if
963 the remainder of the pattern can be made to match that way. The correct
964 answer is either one of these:
969 For example, to grab non-empty comments into $1, one should use either
972 / (?> \# [ \t]* ) ( .+ ) /x;
973 / \# [ \t]* ( [^ \t] .* ) /x;
975 Which one you pick depends on which of these expressions better reflects
976 the above specification of comments.
978 In some literature this construct is called "atomic matching" or
979 "possessive matching".
981 =item C<(?(condition)yes-pattern|no-pattern)>
984 =item C<(?(condition)yes-pattern)>
986 B<WARNING>: This extended regular expression feature is considered
987 highly experimental, and may be changed or deleted without notice.
989 Conditional expression. C<(condition)> should be either an integer in
990 parentheses (which is valid if the corresponding pair of parentheses
991 matched), or look-ahead/look-behind/evaluate zero-width assertion.
1000 matches a chunk of non-parentheses, possibly included in parentheses
1006 X<backtrack> X<backtracking>
1008 NOTE: This section presents an abstract approximation of regular
1009 expression behavior. For a more rigorous (and complicated) view of
1010 the rules involved in selecting a match among possible alternatives,
1011 see L<Combining pieces together>.
1013 A fundamental feature of regular expression matching involves the
1014 notion called I<backtracking>, which is currently used (when needed)
1015 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
1016 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1017 internally, but the general principle outlined here is valid.
1019 For a regular expression to match, the I<entire> regular expression must
1020 match, not just part of it. So if the beginning of a pattern containing a
1021 quantifier succeeds in a way that causes later parts in the pattern to
1022 fail, the matching engine backs up and recalculates the beginning
1023 part--that's why it's called backtracking.
1025 Here is an example of backtracking: Let's say you want to find the
1026 word following "foo" in the string "Food is on the foo table.":
1028 $_ = "Food is on the foo table.";
1029 if ( /\b(foo)\s+(\w+)/i ) {
1030 print "$2 follows $1.\n";
1033 When the match runs, the first part of the regular expression (C<\b(foo)>)
1034 finds a possible match right at the beginning of the string, and loads up
1035 $1 with "Foo". However, as soon as the matching engine sees that there's
1036 no whitespace following the "Foo" that it had saved in $1, it realizes its
1037 mistake and starts over again one character after where it had the
1038 tentative match. This time it goes all the way until the next occurrence
1039 of "foo". The complete regular expression matches this time, and you get
1040 the expected output of "table follows foo."
1042 Sometimes minimal matching can help a lot. Imagine you'd like to match
1043 everything between "foo" and "bar". Initially, you write something
1046 $_ = "The food is under the bar in the barn.";
1047 if ( /foo(.*)bar/ ) {
1051 Which perhaps unexpectedly yields:
1053 got <d is under the bar in the >
1055 That's because C<.*> was greedy, so you get everything between the
1056 I<first> "foo" and the I<last> "bar". Here it's more effective
1057 to use minimal matching to make sure you get the text between a "foo"
1058 and the first "bar" thereafter.
1060 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1061 got <d is under the >
1063 Here's another example: let's say you'd like to match a number at the end
1064 of a string, and you also want to keep the preceding part of the match.
1067 $_ = "I have 2 numbers: 53147";
1068 if ( /(.*)(\d*)/ ) { # Wrong!
1069 print "Beginning is <$1>, number is <$2>.\n";
1072 That won't work at all, because C<.*> was greedy and gobbled up the
1073 whole string. As C<\d*> can match on an empty string the complete
1074 regular expression matched successfully.
1076 Beginning is <I have 2 numbers: 53147>, number is <>.
1078 Here are some variants, most of which don't work:
1080 $_ = "I have 2 numbers: 53147";
1093 printf "%-12s ", $pat;
1095 print "<$1> <$2>\n";
1101 That will print out:
1103 (.*)(\d*) <I have 2 numbers: 53147> <>
1104 (.*)(\d+) <I have 2 numbers: 5314> <7>
1106 (.*?)(\d+) <I have > <2>
1107 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1108 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1109 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1110 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1112 As you see, this can be a bit tricky. It's important to realize that a
1113 regular expression is merely a set of assertions that gives a definition
1114 of success. There may be 0, 1, or several different ways that the
1115 definition might succeed against a particular string. And if there are
1116 multiple ways it might succeed, you need to understand backtracking to
1117 know which variety of success you will achieve.
1119 When using look-ahead assertions and negations, this can all get even
1120 trickier. Imagine you'd like to find a sequence of non-digits not
1121 followed by "123". You might try to write that as
1124 if ( /^\D*(?!123)/ ) { # Wrong!
1125 print "Yup, no 123 in $_\n";
1128 But that isn't going to match; at least, not the way you're hoping. It
1129 claims that there is no 123 in the string. Here's a clearer picture of
1130 why that pattern matches, contrary to popular expectations:
1135 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1136 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1138 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1139 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1147 You might have expected test 3 to fail because it seems to a more
1148 general purpose version of test 1. The important difference between
1149 them is that test 3 contains a quantifier (C<\D*>) and so can use
1150 backtracking, whereas test 1 will not. What's happening is
1151 that you've asked "Is it true that at the start of $x, following 0 or more
1152 non-digits, you have something that's not 123?" If the pattern matcher had
1153 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1156 The search engine will initially match C<\D*> with "ABC". Then it will
1157 try to match C<(?!123> with "123", which fails. But because
1158 a quantifier (C<\D*>) has been used in the regular expression, the
1159 search engine can backtrack and retry the match differently
1160 in the hope of matching the complete regular expression.
1162 The pattern really, I<really> wants to succeed, so it uses the
1163 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1164 time. Now there's indeed something following "AB" that is not
1165 "123". It's "C123", which suffices.
1167 We can deal with this by using both an assertion and a negation.
1168 We'll say that the first part in $1 must be followed both by a digit
1169 and by something that's not "123". Remember that the look-aheads
1170 are zero-width expressions--they only look, but don't consume any
1171 of the string in their match. So rewriting this way produces what
1172 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1174 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1175 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1179 In other words, the two zero-width assertions next to each other work as though
1180 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1181 matches only if you're at the beginning of the line AND the end of the
1182 line simultaneously. The deeper underlying truth is that juxtaposition in
1183 regular expressions always means AND, except when you write an explicit OR
1184 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1185 although the attempted matches are made at different positions because "a"
1186 is not a zero-width assertion, but a one-width assertion.
1188 B<WARNING>: particularly complicated regular expressions can take
1189 exponential time to solve because of the immense number of possible
1190 ways they can use backtracking to try match. For example, without
1191 internal optimizations done by the regular expression engine, this will
1192 take a painfully long time to run:
1194 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1196 And if you used C<*>'s in the internal groups instead of limiting them
1197 to 0 through 5 matches, then it would take forever--or until you ran
1198 out of stack space. Moreover, these internal optimizations are not
1199 always applicable. For example, if you put C<{0,5}> instead of C<*>
1200 on the external group, no current optimization is applicable, and the
1201 match takes a long time to finish.
1203 A powerful tool for optimizing such beasts is what is known as an
1204 "independent group",
1205 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1206 zero-length look-ahead/look-behind assertions will not backtrack to make
1207 the tail match, since they are in "logical" context: only
1208 whether they match is considered relevant. For an example
1209 where side-effects of look-ahead I<might> have influenced the
1210 following match, see L<C<< (?>pattern) >>>.
1212 =head2 Version 8 Regular Expressions
1213 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1215 In case you're not familiar with the "regular" Version 8 regex
1216 routines, here are the pattern-matching rules not described above.
1218 Any single character matches itself, unless it is a I<metacharacter>
1219 with a special meaning described here or above. You can cause
1220 characters that normally function as metacharacters to be interpreted
1221 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1222 character; "\\" matches a "\"). A series of characters matches that
1223 series of characters in the target string, so the pattern C<blurfl>
1224 would match "blurfl" in the target string.
1226 You can specify a character class, by enclosing a list of characters
1227 in C<[]>, which will match any one character from the list. If the
1228 first character after the "[" is "^", the class matches any character not
1229 in the list. Within a list, the "-" character specifies a
1230 range, so that C<a-z> represents all characters between "a" and "z",
1231 inclusive. If you want either "-" or "]" itself to be a member of a
1232 class, put it at the start of the list (possibly after a "^"), or
1233 escape it with a backslash. "-" is also taken literally when it is
1234 at the end of the list, just before the closing "]". (The
1235 following all specify the same class of three characters: C<[-az]>,
1236 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1237 specifies a class containing twenty-six characters, even on EBCDIC
1238 based coded character sets.) Also, if you try to use the character
1239 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1240 a range, that's not a range, the "-" is understood literally.
1242 Note also that the whole range idea is rather unportable between
1243 character sets--and even within character sets they may cause results
1244 you probably didn't expect. A sound principle is to use only ranges
1245 that begin from and end at either alphabets of equal case ([a-e],
1246 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1247 spell out the character sets in full.
1249 Characters may be specified using a metacharacter syntax much like that
1250 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1251 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1252 of octal digits, matches the character whose coded character set value
1253 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1254 matches the character whose numeric value is I<nn>. The expression \cI<x>
1255 matches the character control-I<x>. Finally, the "." metacharacter
1256 matches any character except "\n" (unless you use C</s>).
1258 You can specify a series of alternatives for a pattern using "|" to
1259 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1260 or "foe" in the target string (as would C<f(e|i|o)e>). The
1261 first alternative includes everything from the last pattern delimiter
1262 ("(", "[", or the beginning of the pattern) up to the first "|", and
1263 the last alternative contains everything from the last "|" to the next
1264 pattern delimiter. That's why it's common practice to include
1265 alternatives in parentheses: to minimize confusion about where they
1268 Alternatives are tried from left to right, so the first
1269 alternative found for which the entire expression matches, is the one that
1270 is chosen. This means that alternatives are not necessarily greedy. For
1271 example: when matching C<foo|foot> against "barefoot", only the "foo"
1272 part will match, as that is the first alternative tried, and it successfully
1273 matches the target string. (This might not seem important, but it is
1274 important when you are capturing matched text using parentheses.)
1276 Also remember that "|" is interpreted as a literal within square brackets,
1277 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1279 Within a pattern, you may designate subpatterns for later reference
1280 by enclosing them in parentheses, and you may refer back to the
1281 I<n>th subpattern later in the pattern using the metacharacter
1282 \I<n>. Subpatterns are numbered based on the left to right order
1283 of their opening parenthesis. A backreference matches whatever
1284 actually matched the subpattern in the string being examined, not
1285 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1286 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1287 1 matched "0x", even though the rule C<0|0x> could potentially match
1288 the leading 0 in the second number.
1290 =head2 Warning on \1 vs $1
1292 Some people get too used to writing things like:
1294 $pattern =~ s/(\W)/\\\1/g;
1296 This is grandfathered for the RHS of a substitute to avoid shocking the
1297 B<sed> addicts, but it's a dirty habit to get into. That's because in
1298 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1299 the usual double-quoted string means a control-A. The customary Unix
1300 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1301 of doing that, you get yourself into trouble if you then add an C</e>
1304 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1310 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1311 C<${1}000>. The operation of interpolation should not be confused
1312 with the operation of matching a backreference. Certainly they mean two
1313 different things on the I<left> side of the C<s///>.
1315 =head2 Repeated patterns matching zero-length substring
1317 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1319 Regular expressions provide a terse and powerful programming language. As
1320 with most other power tools, power comes together with the ability
1323 A common abuse of this power stems from the ability to make infinite
1324 loops using regular expressions, with something as innocuous as:
1326 'foo' =~ m{ ( o? )* }x;
1328 The C<o?> can match at the beginning of C<'foo'>, and since the position
1329 in the string is not moved by the match, C<o?> would match again and again
1330 because of the C<*> modifier. Another common way to create a similar cycle
1331 is with the looping modifier C<//g>:
1333 @matches = ( 'foo' =~ m{ o? }xg );
1337 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1339 or the loop implied by split().
1341 However, long experience has shown that many programming tasks may
1342 be significantly simplified by using repeated subexpressions that
1343 may match zero-length substrings. Here's a simple example being:
1345 @chars = split //, $string; # // is not magic in split
1346 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1348 Thus Perl allows such constructs, by I<forcefully breaking
1349 the infinite loop>. The rules for this are different for lower-level
1350 loops given by the greedy modifiers C<*+{}>, and for higher-level
1351 ones like the C</g> modifier or split() operator.
1353 The lower-level loops are I<interrupted> (that is, the loop is
1354 broken) when Perl detects that a repeated expression matched a
1355 zero-length substring. Thus
1357 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1359 is made equivalent to
1361 m{ (?: NON_ZERO_LENGTH )*
1366 The higher level-loops preserve an additional state between iterations:
1367 whether the last match was zero-length. To break the loop, the following
1368 match after a zero-length match is prohibited to have a length of zero.
1369 This prohibition interacts with backtracking (see L<"Backtracking">),
1370 and so the I<second best> match is chosen if the I<best> match is of
1378 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1379 match given by non-greedy C<??> is the zero-length match, and the I<second
1380 best> match is what is matched by C<\w>. Thus zero-length matches
1381 alternate with one-character-long matches.
1383 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1384 position one notch further in the string.
1386 The additional state of being I<matched with zero-length> is associated with
1387 the matched string, and is reset by each assignment to pos().
1388 Zero-length matches at the end of the previous match are ignored
1391 =head2 Combining pieces together
1393 Each of the elementary pieces of regular expressions which were described
1394 before (such as C<ab> or C<\Z>) could match at most one substring
1395 at the given position of the input string. However, in a typical regular
1396 expression these elementary pieces are combined into more complicated
1397 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1398 (in these examples C<S> and C<T> are regular subexpressions).
1400 Such combinations can include alternatives, leading to a problem of choice:
1401 if we match a regular expression C<a|ab> against C<"abc">, will it match
1402 substring C<"a"> or C<"ab">? One way to describe which substring is
1403 actually matched is the concept of backtracking (see L<"Backtracking">).
1404 However, this description is too low-level and makes you think
1405 in terms of a particular implementation.
1407 Another description starts with notions of "better"/"worse". All the
1408 substrings which may be matched by the given regular expression can be
1409 sorted from the "best" match to the "worst" match, and it is the "best"
1410 match which is chosen. This substitutes the question of "what is chosen?"
1411 by the question of "which matches are better, and which are worse?".
1413 Again, for elementary pieces there is no such question, since at most
1414 one match at a given position is possible. This section describes the
1415 notion of better/worse for combining operators. In the description
1416 below C<S> and C<T> are regular subexpressions.
1422 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1423 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1424 which can be matched by C<T>.
1426 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1429 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1430 C<B> is better match for C<T> than C<B'>.
1434 When C<S> can match, it is a better match than when only C<T> can match.
1436 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1437 two matches for C<T>.
1439 =item C<S{REPEAT_COUNT}>
1441 Matches as C<SSS...S> (repeated as many times as necessary).
1445 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1447 =item C<S{min,max}?>
1449 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1451 =item C<S?>, C<S*>, C<S+>
1453 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1455 =item C<S??>, C<S*?>, C<S+?>
1457 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1461 Matches the best match for C<S> and only that.
1463 =item C<(?=S)>, C<(?<=S)>
1465 Only the best match for C<S> is considered. (This is important only if
1466 C<S> has capturing parentheses, and backreferences are used somewhere
1467 else in the whole regular expression.)
1469 =item C<(?!S)>, C<(?<!S)>
1471 For this grouping operator there is no need to describe the ordering, since
1472 only whether or not C<S> can match is important.
1474 =item C<(??{ EXPR })>, C<(?PARNO)>
1476 The ordering is the same as for the regular expression which is
1477 the result of EXPR, or the pattern contained by capture buffer PARNO.
1479 =item C<(?(condition)yes-pattern|no-pattern)>
1481 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1482 already determined. The ordering of the matches is the same as for the
1483 chosen subexpression.
1487 The above recipes describe the ordering of matches I<at a given position>.
1488 One more rule is needed to understand how a match is determined for the
1489 whole regular expression: a match at an earlier position is always better
1490 than a match at a later position.
1492 =head2 Creating custom RE engines
1494 Overloaded constants (see L<overload>) provide a simple way to extend
1495 the functionality of the RE engine.
1497 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
1498 matches at boundary between whitespace characters and non-whitespace
1499 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1500 at these positions, so we want to have each C<\Y|> in the place of the
1501 more complicated version. We can create a module C<customre> to do
1509 die "No argument to customre::import allowed" if @_;
1510 overload::constant 'qr' => \&convert;
1513 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1515 # We must also take care of not escaping the legitimate \\Y|
1516 # sequence, hence the presence of '\\' in the conversion rules.
1517 my %rules = ( '\\' => '\\\\',
1518 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
1524 { $rules{$1} or invalid($re,$1) }sgex;
1528 Now C<use customre> enables the new escape in constant regular
1529 expressions, i.e., those without any runtime variable interpolations.
1530 As documented in L<overload>, this conversion will work only over
1531 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
1532 part of this regular expression needs to be converted explicitly
1533 (but only if the special meaning of C<\Y|> should be enabled inside $re):
1538 $re = customre::convert $re;
1543 This document varies from difficult to understand to completely
1544 and utterly opaque. The wandering prose riddled with jargon is
1545 hard to fathom in several places.
1547 This document needs a rewrite that separates the tutorial content
1548 from the reference content.
1556 L<perlop/"Regexp Quote-Like Operators">.
1558 L<perlop/"Gory details of parsing quoted constructs">.
1568 I<Mastering Regular Expressions> by Jeffrey Friedl, published
1569 by O'Reilly and Associates.