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
22 Matching operations can have various modifiers. Modifiers
23 that relate to the interpretation of the regular expression inside
24 are listed below. Modifiers that alter the way a regular expression
25 is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and
26 L<perlop/"Gory details of parsing quoted constructs">.
31 X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline>
33 Treat string as multiple lines. That is, change "^" and "$" from matching
34 the start or end of the string to matching the start or end of any
35 line anywhere within the string.
38 X</s> X<regex, single-line> X<regexp, single-line>
39 X<regular expression, single-line>
41 Treat string as single line. That is, change "." to match any character
42 whatsoever, even a newline, which normally it would not match.
44 Used together, as /ms, they let the "." match any character whatsoever,
45 while still allowing "^" and "$" to match, respectively, just after
46 and just before newlines within the string.
49 X</i> X<regex, case-insensitive> X<regexp, case-insensitive>
50 X<regular expression, case-insensitive>
52 Do case-insensitive pattern matching.
54 If C<use locale> is in effect, the case map is taken from the current
55 locale. See L<perllocale>.
60 Extend your pattern's legibility by permitting whitespace and comments.
63 X</p> X<regex, preserve> X<regexp, preserve>
65 Preserve the string matched such that ${^PREMATCH}, ${^MATCH}, and
66 ${^POSTMATCH} are available for use after matching.
71 Global matching, and keep the Current position after failed matching.
72 Unlike i, m, s and x, these two flags affect the way the regex is used
73 rather than the regex itself. See
74 L<perlretut/"Using regular expressions in Perl"> for further explanation
75 of the g and c modifiers.
79 These are usually written as "the C</x> modifier", even though the delimiter
80 in question might not really be a slash. Any of these
81 modifiers may also be embedded within the regular expression itself using
82 the C<(?...)> construct. See below.
84 The C</x> modifier itself needs a little more explanation. It tells
85 the regular expression parser to ignore most whitespace that is neither
86 backslashed nor within a character class. You can use this to break up
87 your regular expression into (slightly) more readable parts. The C<#>
88 character is also treated as a metacharacter introducing a comment,
89 just as in ordinary Perl code. This also means that if you want real
90 whitespace or C<#> characters in the pattern (outside a character
91 class, where they are unaffected by C</x>), then you'll either have to
92 escape them (using backslashes or C<\Q...\E>) or encode them using octal,
93 hex, or C<\N{}> escapes. Taken together, these features go a long way towards
94 making Perl's regular expressions more readable. Note that you have to
95 be careful not to include the pattern delimiter in the comment--perl has
96 no way of knowing you did not intend to close the pattern early. See
97 the C-comment deletion code in L<perlop>. Also note that anything inside
98 a C<\Q...\E> stays unaffected by C</x>. Also note that space is never allowed
99 within a L<quantifier|Quantifiers> such as C<{3}> or C<{5,}>, regardless of
100 C</x>, nor is space allowed before the C<{> or within the braces in C<\x{...}>
101 nor C<\N{U+...}>. Similarly space is not allowed before the C<{> in
102 C<\N{I<name>}> and is currently significant within the braces.
105 =head2 Regular Expressions
107 =head3 Metacharacters
109 The patterns used in Perl pattern matching evolved from those supplied in
110 the Version 8 regex routines. (The routines are derived
111 (distantly) from Henry Spencer's freely redistributable reimplementation
112 of the V8 routines.) See L<Version 8 Regular Expressions> for
115 In particular the following metacharacters have their standard I<egrep>-ish
118 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
121 \ Quote the next metacharacter
122 ^ Match the beginning of the line
123 . Match any character (except newline)
124 $ Match the end of the line (or before newline at the end)
129 By default, the "^" character is guaranteed to match only the
130 beginning of the string, the "$" character only the end (or before the
131 newline at the end), and Perl does certain optimizations with the
132 assumption that the string contains only one line. Embedded newlines
133 will not be matched by "^" or "$". You may, however, wish to treat a
134 string as a multi-line buffer, such that the "^" will match after any
135 newline within the string (except if the newline is the last character in
136 the string), and "$" will match before any newline. At the
137 cost of a little more overhead, you can do this by using the /m modifier
138 on the pattern match operator. (Older programs did this by setting C<$*>,
139 but this practice has been removed in perl 5.9.)
142 To simplify multi-line substitutions, the "." character never matches a
143 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
144 the string is a single line--even if it isn't.
149 The following standard quantifiers are recognized:
150 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
152 * Match 0 or more times
153 + Match 1 or more times
155 {n} Match exactly n times
156 {n,} Match at least n times
157 {n,m} Match at least n but not more than m times
159 (If a curly bracket occurs in any other context, it is treated
160 as a regular character. In particular, the lower bound
161 is not optional.) The "*" quantifier is equivalent to C<{0,}>, the "+"
162 quantifier to C<{1,}>, and the "?" quantifier to C<{0,1}>. n and m are limited
163 to integral values less than a preset limit defined when perl is built.
164 This is usually 32766 on the most common platforms. The actual limit can
165 be seen in the error message generated by code such as this:
167 $_ **= $_ , / {$_} / for 2 .. 42;
169 By default, a quantified subpattern is "greedy", that is, it will match as
170 many times as possible (given a particular starting location) while still
171 allowing the rest of the pattern to match. If you want it to match the
172 minimum number of times possible, follow the quantifier with a "?". Note
173 that the meanings don't change, just the "greediness":
174 X<metacharacter> X<greedy> X<greediness>
175 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
177 *? Match 0 or more times, not greedily
178 +? Match 1 or more times, not greedily
179 ?? Match 0 or 1 time, not greedily
180 {n}? Match exactly n times, not greedily
181 {n,}? Match at least n times, not greedily
182 {n,m}? Match at least n but not more than m times, not greedily
184 By default, when a quantified subpattern does not allow the rest of the
185 overall pattern to match, Perl will backtrack. However, this behaviour is
186 sometimes undesirable. Thus Perl provides the "possessive" quantifier form
189 *+ Match 0 or more times and give nothing back
190 ++ Match 1 or more times and give nothing back
191 ?+ Match 0 or 1 time and give nothing back
192 {n}+ Match exactly n times and give nothing back (redundant)
193 {n,}+ Match at least n times and give nothing back
194 {n,m}+ Match at least n but not more than m times and give nothing back
200 will never match, as the C<a++> will gobble up all the C<a>'s in the
201 string and won't leave any for the remaining part of the pattern. This
202 feature can be extremely useful to give perl hints about where it
203 shouldn't backtrack. For instance, the typical "match a double-quoted
204 string" problem can be most efficiently performed when written as:
206 /"(?:[^"\\]++|\\.)*+"/
208 as we know that if the final quote does not match, backtracking will not
209 help. See the independent subexpression C<< (?>...) >> for more details;
210 possessive quantifiers are just syntactic sugar for that construct. For
211 instance the above example could also be written as follows:
213 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
215 =head3 Escape sequences
217 Because patterns are processed as double quoted strings, the following
219 X<\t> X<\n> X<\r> X<\f> X<\e> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q>
220 X<\0> X<\c> X<\N> X<\x>
226 \a alarm (bell) (BEL)
227 \e escape (think troff) (ESC)
228 \033 octal char (example: ESC)
229 \x1B hex char (example: ESC)
230 \x{263a} long hex char (example: Unicode SMILEY)
231 \cK control char (example: VT)
232 \N{name} named Unicode character
233 \N{U+263D} Unicode character (example: FIRST QUARTER MOON)
234 \l lowercase next char (think vi)
235 \u uppercase next char (think vi)
236 \L lowercase till \E (think vi)
237 \U uppercase till \E (think vi)
238 \E end case modification (think vi)
239 \Q quote (disable) pattern metacharacters till \E
241 If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u>
242 and C<\U> is taken from the current locale. See L<perllocale>. For
243 documentation of C<\N{name}>, see L<charnames>.
245 You cannot include a literal C<$> or C<@> within a C<\Q> sequence.
246 An unescaped C<$> or C<@> interpolates the corresponding variable,
247 while escaping will cause the literal string C<\$> to be matched.
248 You'll need to write something like C<m/\Quser\E\@\Qhost/>.
250 =head3 Character Classes and other Special Escapes
252 In addition, Perl defines the following:
253 X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C>
254 X<\g> X<\k> X<\N> X<\K> X<\v> X<\V> X<\h> X<\H>
255 X<word> X<whitespace> X<character class> X<backreference>
257 \w Match a "word" character (alphanumeric plus "_")
258 \W Match a non-"word" character
259 \s Match a whitespace character
260 \S Match a non-whitespace character
261 \d Match a digit character
262 \D Match a non-digit character
263 \pP Match P, named property. Use \p{Prop} for longer names.
265 \X Match Unicode "eXtended grapheme cluster"
266 \C Match a single C char (octet) even under Unicode.
267 NOTE: breaks up characters into their UTF-8 bytes,
268 so you may end up with malformed pieces of UTF-8.
269 Unsupported in lookbehind.
270 \1 Backreference to a specific group.
271 '1' may actually be any positive integer.
272 \g1 Backreference to a specific or previous group,
273 \g{-1} number may be negative indicating a previous buffer and may
274 optionally be wrapped in curly brackets for safer parsing.
275 \g{name} Named backreference
276 \k<name> Named backreference
277 \K Keep the stuff left of the \K, don't include it in $&
278 \N Any character but \n (experimental)
279 \v Vertical whitespace
280 \V Not vertical whitespace
281 \h Horizontal whitespace
282 \H Not horizontal whitespace
285 A C<\w> matches a single alphanumeric character (an alphabetic
286 character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
287 to match a string of Perl-identifier characters (which isn't the same
288 as matching an English word). If C<use locale> is in effect, the list
289 of alphabetic characters generated by C<\w> is taken from the current
290 locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
291 C<\d>, and C<\D> within character classes, but they aren't usable
292 as either end of a range. If any of them precedes or follows a "-",
293 the "-" is understood literally. If Unicode is in effect, C<\s> matches
294 also "\x{85}", "\x{2028}", and "\x{2029}". See L<perlunicode> for more
295 details about C<\pP>, C<\PP>, C<\X> and the possibility of defining
296 your own C<\p> and C<\P> properties, and L<perluniintro> about Unicode
300 C<\R> will atomically match a linebreak, including the network line-ending
301 "\x0D\x0A". Specifically, X<\R> is exactly equivalent to
303 (?>\x0D\x0A?|[\x0A-\x0C\x85\x{2028}\x{2029}])
305 B<Note:> C<\R> has no special meaning inside of a character class;
306 use C<\v> instead (vertical whitespace).
309 Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the
310 character whose name is C<NAME>; and similarly when of the form
311 C<\N{U+I<wide hex char>}>, it matches the character whose Unicode ordinal is
312 I<wide hex char>. Otherwise it matches any character but C<\n>.
314 The POSIX character class syntax
319 is also available. Note that the C<[> and C<]> brackets are I<literal>;
320 they must always be used within a character class expression.
323 $string =~ /[[:alpha:]]/;
325 # this is not, and will generate a warning:
326 $string =~ /[:alpha:]/;
328 The following table shows the mapping of POSIX character class
329 names, common escapes, literal escape sequences and their equivalent
330 Unicode style property names.
331 X<character class> X<\p> X<\p{}>
332 X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
333 X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
335 B<Note:> up to Perl 5.10 the property names used were shared with
336 standard Unicode properties, this was changed in Perl 5.11, see
337 L<perl5110delta> for details.
339 POSIX Esc Class Property Note
340 --------------------------------------------------------
341 alnum [0-9A-Za-z] IsPosixAlnum
342 alpha [A-Za-z] IsPosixAlpha
343 ascii [\000-\177] IsASCII
344 blank [\011 ] IsPosixBlank [1]
345 cntrl [\0-\37\177] IsPosixCntrl
346 digit \d [0-9] IsPosixDigit
347 graph [!-~] IsPosixGraph
348 lower [a-z] IsPosixLower
349 print [ -~] IsPosixPrint
350 punct [!-/:-@[-`{-~] IsPosixPunct
351 space [\11-\15 ] IsPosixSpace [2]
352 \s [\11\12\14\15 ] IsPerlSpace [2]
353 upper [A-Z] IsPosixUpper
354 word \w [0-9A-Z_a-z] IsPerlWord [3]
355 xdigit [0-9A-Fa-f] IsXDigit
361 A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
365 Note that C<\s> and C<[[:space:]]> are B<not> equivalent as C<[[:space:]]>
366 includes also the (very rare) "vertical tabulator", "\cK" or chr(11) in
371 A Perl extension, see above.
375 For example use C<[:upper:]> to match all the uppercase characters.
376 Note that the C<[]> are part of the C<[::]> construct, not part of the
377 whole character class. For example:
381 matches zero, one, any alphabetic character, and the percent sign.
383 The other named classes are:
390 Any control character. Usually characters that don't produce output as
391 such but instead control the terminal somehow: for example newline and
392 backspace are control characters. All characters with ord() less than
393 32 are usually classified as control characters (assuming ASCII,
394 the ISO Latin character sets, and Unicode), as is the character with
395 the ord() value of 127 (C<DEL>).
400 Any alphanumeric or punctuation (special) character.
405 Any alphanumeric or punctuation (special) character or the space character.
410 Any punctuation (special) character.
415 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
416 work just fine) it is included for completeness.
420 You can negate the [::] character classes by prefixing the class name
421 with a '^'. This is a Perl extension. For example:
422 X<character class, negation>
424 POSIX traditional Unicode
426 [[:^digit:]] \D \P{IsPosixDigit}
427 [[:^space:]] \S \P{IsPosixSpace}
428 [[:^word:]] \W \P{IsPerlWord}
430 Perl respects the POSIX standard in that POSIX character classes are
431 only supported within a character class. The POSIX character classes
432 [.cc.] and [=cc=] are recognized but B<not> supported and trying to
433 use them will cause an error.
437 Perl defines the following zero-width assertions:
438 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
439 X<regexp, zero-width assertion>
440 X<regular expression, zero-width assertion>
441 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
443 \b Match a word boundary
444 \B Match except at a word boundary
445 \A Match only at beginning of string
446 \Z Match only at end of string, or before newline at the end
447 \z Match only at end of string
448 \G Match only at pos() (e.g. at the end-of-match position
451 A word boundary (C<\b>) is a spot between two characters
452 that has a C<\w> on one side of it and a C<\W> on the other side
453 of it (in either order), counting the imaginary characters off the
454 beginning and end of the string as matching a C<\W>. (Within
455 character classes C<\b> represents backspace rather than a word
456 boundary, just as it normally does in any double-quoted string.)
457 The C<\A> and C<\Z> are just like "^" and "$", except that they
458 won't match multiple times when the C</m> modifier is used, while
459 "^" and "$" will match at every internal line boundary. To match
460 the actual end of the string and not ignore an optional trailing
462 X<\b> X<\A> X<\Z> X<\z> X</m>
464 The C<\G> assertion can be used to chain global matches (using
465 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
466 It is also useful when writing C<lex>-like scanners, when you have
467 several patterns that you want to match against consequent substrings
468 of your string, see the previous reference. The actual location
469 where C<\G> will match can also be influenced by using C<pos()> as
470 an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
471 matches is modified somewhat, in that contents to the left of C<\G> is
472 not counted when determining the length of the match. Thus the following
473 will not match forever:
482 It will print 'A' and then terminate, as it considers the match to
483 be zero-width, and thus will not match at the same position twice in a
486 It is worth noting that C<\G> improperly used can result in an infinite
487 loop. Take care when using patterns that include C<\G> in an alternation.
489 =head3 Capture buffers
491 The bracketing construct C<( ... )> creates capture buffers. To refer
492 to the current contents of a buffer later on, within the same pattern,
493 use \1 for the first, \2 for the second, and so on.
494 Outside the match use "$" instead of "\". (The
495 \<digit> notation works in certain circumstances outside
496 the match. See the warning below about \1 vs $1 for details.)
497 Referring back to another part of the match is called a
499 X<regex, capture buffer> X<regexp, capture buffer>
500 X<regular expression, capture buffer> X<backreference>
502 There is no limit to the number of captured substrings that you may
503 use. However Perl also uses \10, \11, etc. as aliases for \010,
504 \011, etc. (Recall that 0 means octal, so \011 is the character at
505 number 9 in your coded character set; which would be the 10th character,
506 a horizontal tab under ASCII.) Perl resolves this
507 ambiguity by interpreting \10 as a backreference only if at least 10
508 left parentheses have opened before it. Likewise \11 is a
509 backreference only if at least 11 left parentheses have opened
510 before it. And so on. \1 through \9 are always interpreted as
512 If the bracketing group did not match, the associated backreference won't
513 match either. (This can happen if the bracketing group is optional, or
514 in a different branch of an alternation.)
516 X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
517 In order to provide a safer and easier way to construct patterns using
518 backreferences, Perl provides the C<\g{N}> notation (starting with perl
519 5.10.0). The curly brackets are optional, however omitting them is less
520 safe as the meaning of the pattern can be changed by text (such as digits)
521 following it. When N is a positive integer the C<\g{N}> notation is
522 exactly equivalent to using normal backreferences. When N is a negative
523 integer then it is a relative backreference referring to the previous N'th
524 capturing group. When the bracket form is used and N is not an integer, it
525 is treated as a reference to a named buffer.
527 Thus C<\g{-1}> refers to the last buffer, C<\g{-2}> refers to the
528 buffer before that. For example:
534 \g{-1} # backref to buffer 3
535 \g{-3} # backref to buffer 1
539 and would match the same as C</(Y) ( (X) \3 \1 )/x>.
541 Additionally, as of Perl 5.10.0 you may use named capture buffers and named
542 backreferences. The notation is C<< (?<name>...) >> to declare and C<< \k<name> >>
543 to reference. You may also use apostrophes instead of angle brackets to delimit the
544 name; and you may use the bracketed C<< \g{name} >> backreference syntax.
545 It's possible to refer to a named capture buffer by absolute and relative number as well.
546 Outside the pattern, a named capture buffer is available via the C<%+> hash.
547 When different buffers within the same pattern have the same name, C<$+{name}>
548 and C<< \k<name> >> refer to the leftmost defined group. (Thus it's possible
549 to do things with named capture buffers that would otherwise require C<(??{})>
551 X<named capture buffer> X<regular expression, named capture buffer>
552 X<%+> X<$+{name}> X<< \k<name> >>
556 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
558 /(.)\1/ # find first doubled char
559 and print "'$1' is the first doubled character\n";
561 /(?<char>.)\k<char>/ # ... a different way
562 and print "'$+{char}' is the first doubled character\n";
564 /(?'char'.)\1/ # ... mix and match
565 and print "'$1' is the first doubled character\n";
567 if (/Time: (..):(..):(..)/) { # parse out values
573 Several special variables also refer back to portions of the previous
574 match. C<$+> returns whatever the last bracket match matched.
575 C<$&> returns the entire matched string. (At one point C<$0> did
576 also, but now it returns the name of the program.) C<$`> returns
577 everything before the matched string. C<$'> returns everything
578 after the matched string. And C<$^N> contains whatever was matched by
579 the most-recently closed group (submatch). C<$^N> can be used in
580 extended patterns (see below), for example to assign a submatch to a
582 X<$+> X<$^N> X<$&> X<$`> X<$'>
584 The numbered match variables ($1, $2, $3, etc.) and the related punctuation
585 set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
586 until the end of the enclosing block or until the next successful
587 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
588 X<$+> X<$^N> X<$&> X<$`> X<$'>
589 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
592 B<NOTE>: Failed matches in Perl do not reset the match variables,
593 which makes it easier to write code that tests for a series of more
594 specific cases and remembers the best match.
596 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
597 C<$'> anywhere in the program, it has to provide them for every
598 pattern match. This may substantially slow your program. Perl
599 uses the same mechanism to produce $1, $2, etc, so you also pay a
600 price for each pattern that contains capturing parentheses. (To
601 avoid this cost while retaining the grouping behaviour, use the
602 extended regular expression C<(?: ... )> instead.) But if you never
603 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
604 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
605 if you can, but if you can't (and some algorithms really appreciate
606 them), once you've used them once, use them at will, because you've
607 already paid the price. As of 5.005, C<$&> is not so costly as the
611 As a workaround for this problem, Perl 5.10.0 introduces C<${^PREMATCH}>,
612 C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&>
613 and C<$'>, B<except> that they are only guaranteed to be defined after a
614 successful match that was executed with the C</p> (preserve) modifier.
615 The use of these variables incurs no global performance penalty, unlike
616 their punctuation char equivalents, however at the trade-off that you
617 have to tell perl when you want to use them.
620 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
621 C<\w>, C<\n>. Unlike some other regular expression languages, there
622 are no backslashed symbols that aren't alphanumeric. So anything
623 that looks like \\, \(, \), \<, \>, \{, or \} is always
624 interpreted as a literal character, not a metacharacter. This was
625 once used in a common idiom to disable or quote the special meanings
626 of regular expression metacharacters in a string that you want to
627 use for a pattern. Simply quote all non-"word" characters:
629 $pattern =~ s/(\W)/\\$1/g;
631 (If C<use locale> is set, then this depends on the current locale.)
632 Today it is more common to use the quotemeta() function or the C<\Q>
633 metaquoting escape sequence to disable all metacharacters' special
636 /$unquoted\Q$quoted\E$unquoted/
638 Beware that if you put literal backslashes (those not inside
639 interpolated variables) between C<\Q> and C<\E>, double-quotish
640 backslash interpolation may lead to confusing results. If you
641 I<need> to use literal backslashes within C<\Q...\E>,
642 consult L<perlop/"Gory details of parsing quoted constructs">.
644 =head2 Extended Patterns
646 Perl also defines a consistent extension syntax for features not
647 found in standard tools like B<awk> and B<lex>. The syntax is a
648 pair of parentheses with a question mark as the first thing within
649 the parentheses. The character after the question mark indicates
652 The stability of these extensions varies widely. Some have been
653 part of the core language for many years. Others are experimental
654 and may change without warning or be completely removed. Check
655 the documentation on an individual feature to verify its current
658 A question mark was chosen for this and for the minimal-matching
659 construct because 1) question marks are rare in older regular
660 expressions, and 2) whenever you see one, you should stop and
661 "question" exactly what is going on. That's psychology...
668 A comment. The text is ignored. If the C</x> modifier enables
669 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
670 the comment as soon as it sees a C<)>, so there is no way to put a literal
673 =item C<(?pimsx-imsx)>
676 One or more embedded pattern-match modifiers, to be turned on (or
677 turned off, if preceded by C<->) for the remainder of the pattern or
678 the remainder of the enclosing pattern group (if any). This is
679 particularly useful for dynamic patterns, such as those read in from a
680 configuration file, taken from an argument, or specified in a table
681 somewhere. Consider the case where some patterns want to be case
682 sensitive and some do not: The case insensitive ones merely need to
683 include C<(?i)> at the front of the pattern. For example:
686 if ( /$pattern/i ) { }
690 $pattern = "(?i)foobar";
691 if ( /$pattern/ ) { }
693 These modifiers are restored at the end of the enclosing group. For example,
697 will match C<blah> in any case, some spaces, and an exact (I<including the case>!)
698 repetition of the previous word, assuming the C</x> modifier, and no C</i>
699 modifier outside this group.
701 These modifiers do not carry over into named subpatterns called in the
702 enclosing group. In other words, a pattern such as C<((?i)(&NAME))> does not
703 change the case-sensitivity of the "NAME" pattern.
705 Note that the C<p> modifier is special in that it can only be enabled,
706 not disabled, and that its presence anywhere in a pattern has a global
707 effect. Thus C<(?-p)> and C<(?-p:...)> are meaningless and will warn
708 when executed under C<use warnings>.
713 =item C<(?imsx-imsx:pattern)>
715 This is for clustering, not capturing; it groups subexpressions like
716 "()", but doesn't make backreferences as "()" does. So
718 @fields = split(/\b(?:a|b|c)\b/)
722 @fields = split(/\b(a|b|c)\b/)
724 but doesn't spit out extra fields. It's also cheaper not to capture
725 characters if you don't need to.
727 Any letters between C<?> and C<:> act as flags modifiers as with
728 C<(?imsx-imsx)>. For example,
730 /(?s-i:more.*than).*million/i
732 is equivalent to the more verbose
734 /(?:(?s-i)more.*than).*million/i
737 X<(?|)> X<Branch reset>
739 This is the "branch reset" pattern, which has the special property
740 that the capture buffers are numbered from the same starting point
741 in each alternation branch. It is available starting from perl 5.10.0.
743 Capture buffers are numbered from left to right, but inside this
744 construct the numbering is restarted for each branch.
746 The numbering within each branch will be as normal, and any buffers
747 following this construct will be numbered as though the construct
748 contained only one branch, that being the one with the most capture
751 This construct will be useful when you want to capture one of a
752 number of alternative matches.
754 Consider the following pattern. The numbers underneath show in
755 which buffer the captured content will be stored.
758 # before ---------------branch-reset----------- after
759 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
762 Be careful when using the branch reset pattern in combination with
763 named captures. Named captures are implemented as being aliases to
764 numbered buffers holding the captures, and that interferes with the
765 implementation of the branch reset pattern. If you are using named
766 captures in a branch reset pattern, it's best to use the same names,
767 in the same order, in each of the alternations:
769 /(?| (?<a> x ) (?<b> y )
770 | (?<a> z ) (?<b> w )) /x
772 Not doing so may lead to surprises:
774 "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x;
775 say $+ {a}; # Prints '12'
776 say $+ {b}; # *Also* prints '12'.
778 The problem here is that both the buffer named C<< a >> and the buffer
779 named C<< b >> are aliases for the buffer belonging to C<< $1 >>.
781 =item Look-Around Assertions
782 X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
784 Look-around assertions are zero width patterns which match a specific
785 pattern without including it in C<$&>. Positive assertions match when
786 their subpattern matches, negative assertions match when their subpattern
787 fails. Look-behind matches text up to the current match position,
788 look-ahead matches text following the current match position.
793 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
795 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
796 matches a word followed by a tab, without including the tab in C<$&>.
799 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
801 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
802 matches any occurrence of "foo" that isn't followed by "bar". Note
803 however that look-ahead and look-behind are NOT the same thing. You cannot
804 use this for look-behind.
806 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
807 will not do what you want. That's because the C<(?!foo)> is just saying that
808 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
809 match. You would have to do something like C</(?!foo)...bar/> for that. We
810 say "like" because there's the case of your "bar" not having three characters
811 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
812 Sometimes it's still easier just to say:
814 if (/bar/ && $` !~ /foo$/)
816 For look-behind see below.
818 =item C<(?<=pattern)> C<\K>
819 X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K>
821 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
822 matches a word that follows a tab, without including the tab in C<$&>.
823 Works only for fixed-width look-behind.
825 There is a special form of this construct, called C<\K>, which causes the
826 regex engine to "keep" everything it had matched prior to the C<\K> and
827 not include it in C<$&>. This effectively provides variable length
828 look-behind. The use of C<\K> inside of another look-around assertion
829 is allowed, but the behaviour is currently not well defined.
831 For various reasons C<\K> may be significantly more efficient than the
832 equivalent C<< (?<=...) >> construct, and it is especially useful in
833 situations where you want to efficiently remove something following
834 something else in a string. For instance
838 can be rewritten as the much more efficient
842 =item C<(?<!pattern)>
843 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
845 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
846 matches any occurrence of "foo" that does not follow "bar". Works
847 only for fixed-width look-behind.
851 =item C<(?'NAME'pattern)>
853 =item C<< (?<NAME>pattern) >>
854 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
856 A named capture buffer. Identical in every respect to normal capturing
857 parentheses C<()> but for the additional fact that C<%+> or C<%-> may be
858 used after a successful match to refer to a named buffer. See C<perlvar>
859 for more details on the C<%+> and C<%-> hashes.
861 If multiple distinct capture buffers have the same name then the
862 $+{NAME} will refer to the leftmost defined buffer in the match.
864 The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent.
866 B<NOTE:> While the notation of this construct is the same as the similar
867 function in .NET regexes, the behavior is not. In Perl the buffers are
868 numbered sequentially regardless of being named or not. Thus in the
873 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
874 the opposite which is what a .NET regex hacker might expect.
876 Currently NAME is restricted to simple identifiers only.
877 In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
878 its Unicode extension (see L<utf8>),
879 though it isn't extended by the locale (see L<perllocale>).
881 B<NOTE:> In order to make things easier for programmers with experience
882 with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >>
883 may be used instead of C<< (?<NAME>pattern) >>; however this form does not
884 support the use of single quotes as a delimiter for the name.
886 =item C<< \k<NAME> >>
888 =item C<< \k'NAME' >>
890 Named backreference. Similar to numeric backreferences, except that
891 the group is designated by name and not number. If multiple groups
892 have the same name then it refers to the leftmost defined group in
895 It is an error to refer to a name not defined by a C<< (?<NAME>) >>
896 earlier in the pattern.
898 Both forms are equivalent.
900 B<NOTE:> In order to make things easier for programmers with experience
901 with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >>
902 may be used instead of C<< \k<NAME> >>.
905 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
907 B<WARNING>: This extended regular expression feature is considered
908 experimental, and may be changed without notice. Code executed that
909 has side effects may not perform identically from version to version
910 due to the effect of future optimisations in the regex engine.
912 This zero-width assertion evaluates any embedded Perl code. It
913 always succeeds, and its C<code> is not interpolated. Currently,
914 the rules to determine where the C<code> ends are somewhat convoluted.
916 This feature can be used together with the special variable C<$^N> to
917 capture the results of submatches in variables without having to keep
918 track of the number of nested parentheses. For example:
920 $_ = "The brown fox jumps over the lazy dog";
921 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
922 print "color = $color, animal = $animal\n";
924 Inside the C<(?{...})> block, C<$_> refers to the string the regular
925 expression is matching against. You can also use C<pos()> to know what is
926 the current position of matching within this string.
928 The C<code> is properly scoped in the following sense: If the assertion
929 is backtracked (compare L<"Backtracking">), all changes introduced after
930 C<local>ization are undone, so that
934 (?{ $cnt = 0 }) # Initialize $cnt.
938 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
942 (?{ $res = $cnt }) # On success copy to non-localized
946 will set C<$res = 4>. Note that after the match, C<$cnt> returns to the globally
947 introduced value, because the scopes that restrict C<local> operators
950 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
951 switch. If I<not> used in this way, the result of evaluation of
952 C<code> is put into the special variable C<$^R>. This happens
953 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
954 inside the same regular expression.
956 The assignment to C<$^R> above is properly localized, so the old
957 value of C<$^R> is restored if the assertion is backtracked; compare
960 For reasons of security, this construct is forbidden if the regular
961 expression involves run-time interpolation of variables, unless the
962 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
963 variables contain results of C<qr//> operator (see
964 L<perlop/"qr/STRING/imosx">).
966 This restriction is due to the wide-spread and remarkably convenient
967 custom of using run-time determined strings as patterns. For example:
973 Before Perl knew how to execute interpolated code within a pattern,
974 this operation was completely safe from a security point of view,
975 although it could raise an exception from an illegal pattern. If
976 you turn on the C<use re 'eval'>, though, it is no longer secure,
977 so you should only do so if you are also using taint checking.
978 Better yet, use the carefully constrained evaluation within a Safe
979 compartment. See L<perlsec> for details about both these mechanisms.
981 B<WARNING>: Use of lexical (C<my>) variables in these blocks is
982 broken. The result is unpredictable and will make perl unstable. The
983 workaround is to use global (C<our>) variables.
985 B<WARNING>: Because Perl's regex engine is currently not re-entrant,
986 interpolated code may not invoke the regex engine either directly with
987 C<m//> or C<s///>), or indirectly with functions such as
988 C<split>. Invoking the regex engine in these blocks will make perl
991 =item C<(??{ code })>
993 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
995 B<WARNING>: This extended regular expression feature is considered
996 experimental, and may be changed without notice. Code executed that
997 has side effects may not perform identically from version to version
998 due to the effect of future optimisations in the regex engine.
1000 This is a "postponed" regular subexpression. The C<code> is evaluated
1001 at run time, at the moment this subexpression may match. The result
1002 of evaluation is considered as a regular expression and matched as
1003 if it were inserted instead of this construct. Note that this means
1004 that the contents of capture buffers defined inside an eval'ed pattern
1005 are not available outside of the pattern, and vice versa, there is no
1006 way for the inner pattern to refer to a capture buffer defined outside.
1009 ('a' x 100)=~/(??{'(.)' x 100})/
1011 B<will> match, it will B<not> set $1.
1013 The C<code> is not interpolated. As before, the rules to determine
1014 where the C<code> ends are currently somewhat convoluted.
1016 The following pattern matches a parenthesized group:
1021 (?> [^()]+ ) # Non-parens without backtracking
1023 (??{ $re }) # Group with matching parens
1028 See also C<(?PARNO)> for a different, more efficient way to accomplish
1031 For reasons of security, this construct is forbidden if the regular
1032 expression involves run-time interpolation of variables, unless the
1033 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
1034 variables contain results of C<qr//> operator (see
1035 L<perlop/"qr/STRING/imosx">).
1037 Because perl's regex engine is not currently re-entrant, delayed
1038 code may not invoke the regex engine either directly with C<m//> or C<s///>),
1039 or indirectly with functions such as C<split>.
1041 Recursing deeper than 50 times without consuming any input string will
1042 result in a fatal error. The maximum depth is compiled into perl, so
1043 changing it requires a custom build.
1045 =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)>
1046 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
1047 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
1048 X<regex, relative recursion>
1050 Similar to C<(??{ code })> except it does not involve compiling any code,
1051 instead it treats the contents of a capture buffer as an independent
1052 pattern that must match at the current position. Capture buffers
1053 contained by the pattern will have the value as determined by the
1054 outermost recursion.
1056 PARNO is a sequence of digits (not starting with 0) whose value reflects
1057 the paren-number of the capture buffer to recurse to. C<(?R)> recurses to
1058 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
1059 C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed
1060 to be relative, with negative numbers indicating preceding capture buffers
1061 and positive ones following. Thus C<(?-1)> refers to the most recently
1062 declared buffer, and C<(?+1)> indicates the next buffer to be declared.
1063 Note that the counting for relative recursion differs from that of
1064 relative backreferences, in that with recursion unclosed buffers B<are>
1067 The following pattern matches a function foo() which may contain
1068 balanced parentheses as the argument.
1070 $re = qr{ ( # paren group 1 (full function)
1072 ( # paren group 2 (parens)
1074 ( # paren group 3 (contents of parens)
1076 (?> [^()]+ ) # Non-parens without backtracking
1078 (?2) # Recurse to start of paren group 2
1086 If the pattern was used as follows
1088 'foo(bar(baz)+baz(bop))'=~/$re/
1089 and print "\$1 = $1\n",
1093 the output produced should be the following:
1095 $1 = foo(bar(baz)+baz(bop))
1096 $2 = (bar(baz)+baz(bop))
1097 $3 = bar(baz)+baz(bop)
1099 If there is no corresponding capture buffer defined, then it is a
1100 fatal error. Recursing deeper than 50 times without consuming any input
1101 string will also result in a fatal error. The maximum depth is compiled
1102 into perl, so changing it requires a custom build.
1104 The following shows how using negative indexing can make it
1105 easier to embed recursive patterns inside of a C<qr//> construct
1108 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
1109 if (/foo $parens \s+ + \s+ bar $parens/x) {
1110 # do something here...
1113 B<Note> that this pattern does not behave the same way as the equivalent
1114 PCRE or Python construct of the same form. In Perl you can backtrack into
1115 a recursed group, in PCRE and Python the recursed into group is treated
1116 as atomic. Also, modifiers are resolved at compile time, so constructs
1117 like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
1123 Recurse to a named subpattern. Identical to C<(?PARNO)> except that the
1124 parenthesis to recurse to is determined by name. If multiple parentheses have
1125 the same name, then it recurses to the leftmost.
1127 It is an error to refer to a name that is not declared somewhere in the
1130 B<NOTE:> In order to make things easier for programmers with experience
1131 with the Python or PCRE regex engines the pattern C<< (?P>NAME) >>
1132 may be used instead of C<< (?&NAME) >>.
1134 =item C<(?(condition)yes-pattern|no-pattern)>
1137 =item C<(?(condition)yes-pattern)>
1139 Conditional expression. C<(condition)> should be either an integer in
1140 parentheses (which is valid if the corresponding pair of parentheses
1141 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
1142 name in angle brackets or single quotes (which is valid if a buffer
1143 with the given name matched), or the special symbol (R) (true when
1144 evaluated inside of recursion or eval). Additionally the R may be
1145 followed by a number, (which will be true when evaluated when recursing
1146 inside of the appropriate group), or by C<&NAME>, in which case it will
1147 be true only when evaluated during recursion in the named group.
1149 Here's a summary of the possible predicates:
1155 Checks if the numbered capturing buffer has matched something.
1157 =item (<NAME>) ('NAME')
1159 Checks if a buffer with the given name has matched something.
1163 Treats the code block as the condition.
1167 Checks if the expression has been evaluated inside of recursion.
1171 Checks if the expression has been evaluated while executing directly
1172 inside of the n-th capture group. This check is the regex equivalent of
1174 if ((caller(0))[3] eq 'subname') { ... }
1176 In other words, it does not check the full recursion stack.
1180 Similar to C<(R1)>, this predicate checks to see if we're executing
1181 directly inside of the leftmost group with a given name (this is the same
1182 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1183 stack, but only the name of the innermost active recursion.
1187 In this case, the yes-pattern is never directly executed, and no
1188 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1189 See below for details.
1200 matches a chunk of non-parentheses, possibly included in parentheses
1203 A special form is the C<(DEFINE)> predicate, which never executes directly
1204 its yes-pattern, and does not allow a no-pattern. This allows to define
1205 subpatterns which will be executed only by using the recursion mechanism.
1206 This way, you can define a set of regular expression rules that can be
1207 bundled into any pattern you choose.
1209 It is recommended that for this usage you put the DEFINE block at the
1210 end of the pattern, and that you name any subpatterns defined within it.
1212 Also, it's worth noting that patterns defined this way probably will
1213 not be as efficient, as the optimiser is not very clever about
1216 An example of how this might be used is as follows:
1218 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
1224 Note that capture buffers matched inside of recursion are not accessible
1225 after the recursion returns, so the extra layer of capturing buffers is
1226 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1227 C<$+{NAME}> would be.
1229 =item C<< (?>pattern) >>
1230 X<backtrack> X<backtracking> X<atomic> X<possessive>
1232 An "independent" subexpression, one which matches the substring
1233 that a I<standalone> C<pattern> would match if anchored at the given
1234 position, and it matches I<nothing other than this substring>. This
1235 construct is useful for optimizations of what would otherwise be
1236 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1237 It may also be useful in places where the "grab all you can, and do not
1238 give anything back" semantic is desirable.
1240 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1241 (anchored at the beginning of string, as above) will match I<all>
1242 characters C<a> at the beginning of string, leaving no C<a> for
1243 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1244 since the match of the subgroup C<a*> is influenced by the following
1245 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1246 C<a*ab> will match fewer characters than a standalone C<a*>, since
1247 this makes the tail match.
1249 An effect similar to C<< (?>pattern) >> may be achieved by writing
1250 C<(?=(pattern))\1>. This matches the same substring as a standalone
1251 C<a+>, and the following C<\1> eats the matched string; it therefore
1252 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1253 (The difference between these two constructs is that the second one
1254 uses a capturing group, thus shifting ordinals of backreferences
1255 in the rest of a regular expression.)
1257 Consider this pattern:
1268 That will efficiently match a nonempty group with matching parentheses
1269 two levels deep or less. However, if there is no such group, it
1270 will take virtually forever on a long string. That's because there
1271 are so many different ways to split a long string into several
1272 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1273 to a subpattern of the above pattern. Consider how the pattern
1274 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1275 seconds, but that each extra letter doubles this time. This
1276 exponential performance will make it appear that your program has
1277 hung. However, a tiny change to this pattern
1281 (?> [^()]+ ) # change x+ above to (?> x+ )
1288 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1289 this yourself would be a productive exercise), but finishes in a fourth
1290 the time when used on a similar string with 1000000 C<a>s. Be aware,
1291 however, that this pattern currently triggers a warning message under
1292 the C<use warnings> pragma or B<-w> switch saying it
1293 C<"matches null string many times in regex">.
1295 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1296 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1297 This was only 4 times slower on a string with 1000000 C<a>s.
1299 The "grab all you can, and do not give anything back" semantic is desirable
1300 in many situations where on the first sight a simple C<()*> looks like
1301 the correct solution. Suppose we parse text with comments being delimited
1302 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1303 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1304 the comment delimiter, because it may "give up" some whitespace if
1305 the remainder of the pattern can be made to match that way. The correct
1306 answer is either one of these:
1311 For example, to grab non-empty comments into $1, one should use either
1314 / (?> \# [ \t]* ) ( .+ ) /x;
1315 / \# [ \t]* ( [^ \t] .* ) /x;
1317 Which one you pick depends on which of these expressions better reflects
1318 the above specification of comments.
1320 In some literature this construct is called "atomic matching" or
1321 "possessive matching".
1323 Possessive quantifiers are equivalent to putting the item they are applied
1324 to inside of one of these constructs. The following equivalences apply:
1326 Quantifier Form Bracketing Form
1327 --------------- ---------------
1331 PAT{min,max}+ (?>PAT{min,max})
1335 =head2 Special Backtracking Control Verbs
1337 B<WARNING:> These patterns are experimental and subject to change or
1338 removal in a future version of Perl. Their usage in production code should
1339 be noted to avoid problems during upgrades.
1341 These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1342 otherwise stated the ARG argument is optional; in some cases, it is
1345 Any pattern containing a special backtracking verb that allows an argument
1346 has the special behaviour that when executed it sets the current package's
1347 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1350 On failure, the C<$REGERROR> variable will be set to the ARG value of the
1351 verb pattern, if the verb was involved in the failure of the match. If the
1352 ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1353 name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1354 none. Also, the C<$REGMARK> variable will be set to FALSE.
1356 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1357 the C<$REGMARK> variable will be set to the name of the last
1358 C<(*MARK:NAME)> pattern executed. See the explanation for the
1359 C<(*MARK:NAME)> verb below for more details.
1361 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1362 and most other regex related variables. They are not local to a scope, nor
1363 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1364 Use C<local> to localize changes to them to a specific scope if necessary.
1366 If a pattern does not contain a special backtracking verb that allows an
1367 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
1371 =item Verbs that take an argument
1375 =item C<(*PRUNE)> C<(*PRUNE:NAME)>
1376 X<(*PRUNE)> X<(*PRUNE:NAME)>
1378 This zero-width pattern prunes the backtracking tree at the current point
1379 when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1380 where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1381 A may backtrack as necessary to match. Once it is reached, matching
1382 continues in B, which may also backtrack as necessary; however, should B
1383 not match, then no further backtracking will take place, and the pattern
1384 will fail outright at the current starting position.
1386 The following example counts all the possible matching strings in a
1387 pattern (without actually matching any of them).
1389 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1390 print "Count=$count\n";
1405 If we add a C<(*PRUNE)> before the count like the following
1407 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
1408 print "Count=$count\n";
1410 we prevent backtracking and find the count of the longest matching
1411 at each matching starting point like so:
1418 Any number of C<(*PRUNE)> assertions may be used in a pattern.
1420 See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1421 control backtracking. In some cases, the use of C<(*PRUNE)> can be
1422 replaced with a C<< (?>pattern) >> with no functional difference; however,
1423 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1424 C<< (?>pattern) >> alone.
1427 =item C<(*SKIP)> C<(*SKIP:NAME)>
1430 This zero-width pattern is similar to C<(*PRUNE)>, except that on
1431 failure it also signifies that whatever text that was matched leading up
1432 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1433 of this pattern. This effectively means that the regex engine "skips" forward
1434 to this position on failure and tries to match again, (assuming that
1435 there is sufficient room to match).
1437 The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1438 C<(*MARK:NAME)> was encountered while matching, then it is that position
1439 which is used as the "skip point". If no C<(*MARK)> of that name was
1440 encountered, then the C<(*SKIP)> operator has no effect. When used
1441 without a name the "skip point" is where the match point was when
1442 executing the (*SKIP) pattern.
1444 Compare the following to the examples in C<(*PRUNE)>, note the string
1447 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
1448 print "Count=$count\n";
1456 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
1457 executed, the next starting point will be where the cursor was when the
1458 C<(*SKIP)> was executed.
1460 =item C<(*MARK:NAME)> C<(*:NAME)>
1461 X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)>
1463 This zero-width pattern can be used to mark the point reached in a string
1464 when a certain part of the pattern has been successfully matched. This
1465 mark may be given a name. A later C<(*SKIP)> pattern will then skip
1466 forward to that point if backtracked into on failure. Any number of
1467 C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated.
1469 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1470 can be used to "label" a pattern branch, so that after matching, the
1471 program can determine which branches of the pattern were involved in the
1474 When a match is successful, the C<$REGMARK> variable will be set to the
1475 name of the most recently executed C<(*MARK:NAME)> that was involved
1478 This can be used to determine which branch of a pattern was matched
1479 without using a separate capture buffer for each branch, which in turn
1480 can result in a performance improvement, as perl cannot optimize
1481 C</(?:(x)|(y)|(z))/> as efficiently as something like
1482 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1484 When a match has failed, and unless another verb has been involved in
1485 failing the match and has provided its own name to use, the C<$REGERROR>
1486 variable will be set to the name of the most recently executed
1489 See C<(*SKIP)> for more details.
1491 As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
1493 =item C<(*THEN)> C<(*THEN:NAME)>
1495 This is similar to the "cut group" operator C<::> from Perl 6. Like
1496 C<(*PRUNE)>, this verb always matches, and when backtracked into on
1497 failure, it causes the regex engine to try the next alternation in the
1498 innermost enclosing group (capturing or otherwise).
1500 Its name comes from the observation that this operation combined with the
1501 alternation operator (C<|>) can be used to create what is essentially a
1502 pattern-based if/then/else block:
1504 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
1506 Note that if this operator is used and NOT inside of an alternation then
1507 it acts exactly like the C<(*PRUNE)> operator.
1517 / ( A (*THEN) B | C (*THEN) D ) /
1521 / ( A (*PRUNE) B | C (*PRUNE) D ) /
1523 as after matching the A but failing on the B the C<(*THEN)> verb will
1524 backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
1529 This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a
1530 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
1531 into on failure it causes the match to fail outright. No further attempts
1532 to find a valid match by advancing the start pointer will occur again.
1535 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
1536 print "Count=$count\n";
1543 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1544 does not match, the regex engine will not try any further matching on the
1549 =item Verbs without an argument
1553 =item C<(*FAIL)> C<(*F)>
1556 This pattern matches nothing and always fails. It can be used to force the
1557 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1558 fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
1560 It is probably useful only when combined with C<(?{})> or C<(??{})>.
1565 B<WARNING:> This feature is highly experimental. It is not recommended
1566 for production code.
1568 This pattern matches nothing and causes the end of successful matching at
1569 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1570 whether there is actually more to match in the string. When inside of a
1571 nested pattern, such as recursion, or in a subpattern dynamically generated
1572 via C<(??{})>, only the innermost pattern is ended immediately.
1574 If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are
1575 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1578 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
1580 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
1581 be set. If another branch in the inner parentheses were matched, such as in the
1582 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
1589 X<backtrack> X<backtracking>
1591 NOTE: This section presents an abstract approximation of regular
1592 expression behavior. For a more rigorous (and complicated) view of
1593 the rules involved in selecting a match among possible alternatives,
1594 see L<Combining RE Pieces>.
1596 A fundamental feature of regular expression matching involves the
1597 notion called I<backtracking>, which is currently used (when needed)
1598 by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>,
1599 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1600 internally, but the general principle outlined here is valid.
1602 For a regular expression to match, the I<entire> regular expression must
1603 match, not just part of it. So if the beginning of a pattern containing a
1604 quantifier succeeds in a way that causes later parts in the pattern to
1605 fail, the matching engine backs up and recalculates the beginning
1606 part--that's why it's called backtracking.
1608 Here is an example of backtracking: Let's say you want to find the
1609 word following "foo" in the string "Food is on the foo table.":
1611 $_ = "Food is on the foo table.";
1612 if ( /\b(foo)\s+(\w+)/i ) {
1613 print "$2 follows $1.\n";
1616 When the match runs, the first part of the regular expression (C<\b(foo)>)
1617 finds a possible match right at the beginning of the string, and loads up
1618 $1 with "Foo". However, as soon as the matching engine sees that there's
1619 no whitespace following the "Foo" that it had saved in $1, it realizes its
1620 mistake and starts over again one character after where it had the
1621 tentative match. This time it goes all the way until the next occurrence
1622 of "foo". The complete regular expression matches this time, and you get
1623 the expected output of "table follows foo."
1625 Sometimes minimal matching can help a lot. Imagine you'd like to match
1626 everything between "foo" and "bar". Initially, you write something
1629 $_ = "The food is under the bar in the barn.";
1630 if ( /foo(.*)bar/ ) {
1634 Which perhaps unexpectedly yields:
1636 got <d is under the bar in the >
1638 That's because C<.*> was greedy, so you get everything between the
1639 I<first> "foo" and the I<last> "bar". Here it's more effective
1640 to use minimal matching to make sure you get the text between a "foo"
1641 and the first "bar" thereafter.
1643 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1644 got <d is under the >
1646 Here's another example. Let's say you'd like to match a number at the end
1647 of a string, and you also want to keep the preceding part of the match.
1650 $_ = "I have 2 numbers: 53147";
1651 if ( /(.*)(\d*)/ ) { # Wrong!
1652 print "Beginning is <$1>, number is <$2>.\n";
1655 That won't work at all, because C<.*> was greedy and gobbled up the
1656 whole string. As C<\d*> can match on an empty string the complete
1657 regular expression matched successfully.
1659 Beginning is <I have 2 numbers: 53147>, number is <>.
1661 Here are some variants, most of which don't work:
1663 $_ = "I have 2 numbers: 53147";
1676 printf "%-12s ", $pat;
1678 print "<$1> <$2>\n";
1684 That will print out:
1686 (.*)(\d*) <I have 2 numbers: 53147> <>
1687 (.*)(\d+) <I have 2 numbers: 5314> <7>
1689 (.*?)(\d+) <I have > <2>
1690 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1691 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1692 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1693 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1695 As you see, this can be a bit tricky. It's important to realize that a
1696 regular expression is merely a set of assertions that gives a definition
1697 of success. There may be 0, 1, or several different ways that the
1698 definition might succeed against a particular string. And if there are
1699 multiple ways it might succeed, you need to understand backtracking to
1700 know which variety of success you will achieve.
1702 When using look-ahead assertions and negations, this can all get even
1703 trickier. Imagine you'd like to find a sequence of non-digits not
1704 followed by "123". You might try to write that as
1707 if ( /^\D*(?!123)/ ) { # Wrong!
1708 print "Yup, no 123 in $_\n";
1711 But that isn't going to match; at least, not the way you're hoping. It
1712 claims that there is no 123 in the string. Here's a clearer picture of
1713 why that pattern matches, contrary to popular expectations:
1718 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1719 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1721 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1722 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1730 You might have expected test 3 to fail because it seems to a more
1731 general purpose version of test 1. The important difference between
1732 them is that test 3 contains a quantifier (C<\D*>) and so can use
1733 backtracking, whereas test 1 will not. What's happening is
1734 that you've asked "Is it true that at the start of $x, following 0 or more
1735 non-digits, you have something that's not 123?" If the pattern matcher had
1736 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1739 The search engine will initially match C<\D*> with "ABC". Then it will
1740 try to match C<(?!123> with "123", which fails. But because
1741 a quantifier (C<\D*>) has been used in the regular expression, the
1742 search engine can backtrack and retry the match differently
1743 in the hope of matching the complete regular expression.
1745 The pattern really, I<really> wants to succeed, so it uses the
1746 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1747 time. Now there's indeed something following "AB" that is not
1748 "123". It's "C123", which suffices.
1750 We can deal with this by using both an assertion and a negation.
1751 We'll say that the first part in $1 must be followed both by a digit
1752 and by something that's not "123". Remember that the look-aheads
1753 are zero-width expressions--they only look, but don't consume any
1754 of the string in their match. So rewriting this way produces what
1755 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1757 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1758 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1762 In other words, the two zero-width assertions next to each other work as though
1763 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1764 matches only if you're at the beginning of the line AND the end of the
1765 line simultaneously. The deeper underlying truth is that juxtaposition in
1766 regular expressions always means AND, except when you write an explicit OR
1767 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1768 although the attempted matches are made at different positions because "a"
1769 is not a zero-width assertion, but a one-width assertion.
1771 B<WARNING>: Particularly complicated regular expressions can take
1772 exponential time to solve because of the immense number of possible
1773 ways they can use backtracking to try for a match. For example, without
1774 internal optimizations done by the regular expression engine, this will
1775 take a painfully long time to run:
1777 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1779 And if you used C<*>'s in the internal groups instead of limiting them
1780 to 0 through 5 matches, then it would take forever--or until you ran
1781 out of stack space. Moreover, these internal optimizations are not
1782 always applicable. For example, if you put C<{0,5}> instead of C<*>
1783 on the external group, no current optimization is applicable, and the
1784 match takes a long time to finish.
1786 A powerful tool for optimizing such beasts is what is known as an
1787 "independent group",
1788 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1789 zero-length look-ahead/look-behind assertions will not backtrack to make
1790 the tail match, since they are in "logical" context: only
1791 whether they match is considered relevant. For an example
1792 where side-effects of look-ahead I<might> have influenced the
1793 following match, see L<C<< (?>pattern) >>>.
1795 =head2 Version 8 Regular Expressions
1796 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1798 In case you're not familiar with the "regular" Version 8 regex
1799 routines, here are the pattern-matching rules not described above.
1801 Any single character matches itself, unless it is a I<metacharacter>
1802 with a special meaning described here or above. You can cause
1803 characters that normally function as metacharacters to be interpreted
1804 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1805 character; "\\" matches a "\"). This escape mechanism is also required
1806 for the character used as the pattern delimiter.
1808 A series of characters matches that series of characters in the target
1809 string, so the pattern C<blurfl> would match "blurfl" in the target
1812 You can specify a character class, by enclosing a list of characters
1813 in C<[]>, which will match any character from the list. If the
1814 first character after the "[" is "^", the class matches any character not
1815 in the list. Within a list, the "-" character specifies a
1816 range, so that C<a-z> represents all characters between "a" and "z",
1817 inclusive. If you want either "-" or "]" itself to be a member of a
1818 class, put it at the start of the list (possibly after a "^"), or
1819 escape it with a backslash. "-" is also taken literally when it is
1820 at the end of the list, just before the closing "]". (The
1821 following all specify the same class of three characters: C<[-az]>,
1822 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1823 specifies a class containing twenty-six characters, even on EBCDIC-based
1824 character sets.) Also, if you try to use the character
1825 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1826 a range, the "-" is understood literally.
1828 Note also that the whole range idea is rather unportable between
1829 character sets--and even within character sets they may cause results
1830 you probably didn't expect. A sound principle is to use only ranges
1831 that begin from and end at either alphabetics of equal case ([a-e],
1832 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1833 spell out the character sets in full.
1835 Characters may be specified using a metacharacter syntax much like that
1836 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1837 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1838 of octal digits, matches the character whose coded character set value
1839 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1840 matches the character whose numeric value is I<nn>. The expression \cI<x>
1841 matches the character control-I<x>. Finally, the "." metacharacter
1842 matches any character except "\n" (unless you use C</s>).
1844 You can specify a series of alternatives for a pattern using "|" to
1845 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1846 or "foe" in the target string (as would C<f(e|i|o)e>). The
1847 first alternative includes everything from the last pattern delimiter
1848 ("(", "[", or the beginning of the pattern) up to the first "|", and
1849 the last alternative contains everything from the last "|" to the next
1850 pattern delimiter. That's why it's common practice to include
1851 alternatives in parentheses: to minimize confusion about where they
1854 Alternatives are tried from left to right, so the first
1855 alternative found for which the entire expression matches, is the one that
1856 is chosen. This means that alternatives are not necessarily greedy. For
1857 example: when matching C<foo|foot> against "barefoot", only the "foo"
1858 part will match, as that is the first alternative tried, and it successfully
1859 matches the target string. (This might not seem important, but it is
1860 important when you are capturing matched text using parentheses.)
1862 Also remember that "|" is interpreted as a literal within square brackets,
1863 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1865 Within a pattern, you may designate subpatterns for later reference
1866 by enclosing them in parentheses, and you may refer back to the
1867 I<n>th subpattern later in the pattern using the metacharacter
1868 \I<n>. Subpatterns are numbered based on the left to right order
1869 of their opening parenthesis. A backreference matches whatever
1870 actually matched the subpattern in the string being examined, not
1871 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1872 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1873 1 matched "0x", even though the rule C<0|0x> could potentially match
1874 the leading 0 in the second number.
1876 =head2 Warning on \1 Instead of $1
1878 Some people get too used to writing things like:
1880 $pattern =~ s/(\W)/\\\1/g;
1882 This is grandfathered for the RHS of a substitute to avoid shocking the
1883 B<sed> addicts, but it's a dirty habit to get into. That's because in
1884 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1885 the usual double-quoted string means a control-A. The customary Unix
1886 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1887 of doing that, you get yourself into trouble if you then add an C</e>
1890 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1896 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1897 C<${1}000>. The operation of interpolation should not be confused
1898 with the operation of matching a backreference. Certainly they mean two
1899 different things on the I<left> side of the C<s///>.
1901 =head2 Repeated Patterns Matching a Zero-length Substring
1903 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1905 Regular expressions provide a terse and powerful programming language. As
1906 with most other power tools, power comes together with the ability
1909 A common abuse of this power stems from the ability to make infinite
1910 loops using regular expressions, with something as innocuous as:
1912 'foo' =~ m{ ( o? )* }x;
1914 The C<o?> matches at the beginning of C<'foo'>, and since the position
1915 in the string is not moved by the match, C<o?> would match again and again
1916 because of the C<*> quantifier. Another common way to create a similar cycle
1917 is with the looping modifier C<//g>:
1919 @matches = ( 'foo' =~ m{ o? }xg );
1923 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1925 or the loop implied by split().
1927 However, long experience has shown that many programming tasks may
1928 be significantly simplified by using repeated subexpressions that
1929 may match zero-length substrings. Here's a simple example being:
1931 @chars = split //, $string; # // is not magic in split
1932 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1934 Thus Perl allows such constructs, by I<forcefully breaking
1935 the infinite loop>. The rules for this are different for lower-level
1936 loops given by the greedy quantifiers C<*+{}>, and for higher-level
1937 ones like the C</g> modifier or split() operator.
1939 The lower-level loops are I<interrupted> (that is, the loop is
1940 broken) when Perl detects that a repeated expression matched a
1941 zero-length substring. Thus
1943 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1945 is made equivalent to
1947 m{ (?: NON_ZERO_LENGTH )*
1952 The higher level-loops preserve an additional state between iterations:
1953 whether the last match was zero-length. To break the loop, the following
1954 match after a zero-length match is prohibited to have a length of zero.
1955 This prohibition interacts with backtracking (see L<"Backtracking">),
1956 and so the I<second best> match is chosen if the I<best> match is of
1964 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1965 match given by non-greedy C<??> is the zero-length match, and the I<second
1966 best> match is what is matched by C<\w>. Thus zero-length matches
1967 alternate with one-character-long matches.
1969 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1970 position one notch further in the string.
1972 The additional state of being I<matched with zero-length> is associated with
1973 the matched string, and is reset by each assignment to pos().
1974 Zero-length matches at the end of the previous match are ignored
1977 =head2 Combining RE Pieces
1979 Each of the elementary pieces of regular expressions which were described
1980 before (such as C<ab> or C<\Z>) could match at most one substring
1981 at the given position of the input string. However, in a typical regular
1982 expression these elementary pieces are combined into more complicated
1983 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1984 (in these examples C<S> and C<T> are regular subexpressions).
1986 Such combinations can include alternatives, leading to a problem of choice:
1987 if we match a regular expression C<a|ab> against C<"abc">, will it match
1988 substring C<"a"> or C<"ab">? One way to describe which substring is
1989 actually matched is the concept of backtracking (see L<"Backtracking">).
1990 However, this description is too low-level and makes you think
1991 in terms of a particular implementation.
1993 Another description starts with notions of "better"/"worse". All the
1994 substrings which may be matched by the given regular expression can be
1995 sorted from the "best" match to the "worst" match, and it is the "best"
1996 match which is chosen. This substitutes the question of "what is chosen?"
1997 by the question of "which matches are better, and which are worse?".
1999 Again, for elementary pieces there is no such question, since at most
2000 one match at a given position is possible. This section describes the
2001 notion of better/worse for combining operators. In the description
2002 below C<S> and C<T> are regular subexpressions.
2008 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
2009 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
2010 which can be matched by C<T>.
2012 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
2015 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
2016 C<B> is better match for C<T> than C<B'>.
2020 When C<S> can match, it is a better match than when only C<T> can match.
2022 Ordering of two matches for C<S> is the same as for C<S>. Similar for
2023 two matches for C<T>.
2025 =item C<S{REPEAT_COUNT}>
2027 Matches as C<SSS...S> (repeated as many times as necessary).
2031 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
2033 =item C<S{min,max}?>
2035 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
2037 =item C<S?>, C<S*>, C<S+>
2039 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
2041 =item C<S??>, C<S*?>, C<S+?>
2043 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
2047 Matches the best match for C<S> and only that.
2049 =item C<(?=S)>, C<(?<=S)>
2051 Only the best match for C<S> is considered. (This is important only if
2052 C<S> has capturing parentheses, and backreferences are used somewhere
2053 else in the whole regular expression.)
2055 =item C<(?!S)>, C<(?<!S)>
2057 For this grouping operator there is no need to describe the ordering, since
2058 only whether or not C<S> can match is important.
2060 =item C<(??{ EXPR })>, C<(?PARNO)>
2062 The ordering is the same as for the regular expression which is
2063 the result of EXPR, or the pattern contained by capture buffer PARNO.
2065 =item C<(?(condition)yes-pattern|no-pattern)>
2067 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
2068 already determined. The ordering of the matches is the same as for the
2069 chosen subexpression.
2073 The above recipes describe the ordering of matches I<at a given position>.
2074 One more rule is needed to understand how a match is determined for the
2075 whole regular expression: a match at an earlier position is always better
2076 than a match at a later position.
2078 =head2 Creating Custom RE Engines
2080 Overloaded constants (see L<overload>) provide a simple way to extend
2081 the functionality of the RE engine.
2083 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
2084 matches at a boundary between whitespace characters and non-whitespace
2085 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
2086 at these positions, so we want to have each C<\Y|> in the place of the
2087 more complicated version. We can create a module C<customre> to do
2095 die "No argument to customre::import allowed" if @_;
2096 overload::constant 'qr' => \&convert;
2099 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
2101 # We must also take care of not escaping the legitimate \\Y|
2102 # sequence, hence the presence of '\\' in the conversion rules.
2103 my %rules = ( '\\' => '\\\\',
2104 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
2110 { $rules{$1} or invalid($re,$1) }sgex;
2114 Now C<use customre> enables the new escape in constant regular
2115 expressions, i.e., those without any runtime variable interpolations.
2116 As documented in L<overload>, this conversion will work only over
2117 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
2118 part of this regular expression needs to be converted explicitly
2119 (but only if the special meaning of C<\Y|> should be enabled inside $re):
2124 $re = customre::convert $re;
2127 =head1 PCRE/Python Support
2129 As of Perl 5.10.0, Perl supports several Python/PCRE specific extensions
2130 to the regex syntax. While Perl programmers are encouraged to use the
2131 Perl specific syntax, the following are also accepted:
2135 =item C<< (?PE<lt>NAMEE<gt>pattern) >>
2137 Define a named capture buffer. Equivalent to C<< (?<NAME>pattern) >>.
2139 =item C<< (?P=NAME) >>
2141 Backreference to a named capture buffer. Equivalent to C<< \g{NAME} >>.
2143 =item C<< (?P>NAME) >>
2145 Subroutine call to a named capture buffer. Equivalent to C<< (?&NAME) >>.
2151 This document varies from difficult to understand to completely
2152 and utterly opaque. The wandering prose riddled with jargon is
2153 hard to fathom in several places.
2155 This document needs a rewrite that separates the tutorial content
2156 from the reference content.
2164 L<perlop/"Regexp Quote-Like Operators">.
2166 L<perlop/"Gory details of parsing quoted constructs">.
2176 I<Mastering Regular Expressions> by Jeffrey Friedl, published
2177 by O'Reilly and Associates.