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 "$" from matching
34 the start or end of the string to matching the start or end of any
35 line anywhere within the string.
38 X</s> X<regex, single-line> X<regexp, single-line>
39 X<regular expression, single-line>
41 Treat string as single line. That is, change "." to match any character
42 whatsoever, even a newline, which normally it would not match.
44 Used together, as C</ms>, they let the "." match any character whatsoever,
45 while still allowing "^" and "$" 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>. And note that C</x> doesn't affect
99 whether space interpretation within a single multi-character construct. For
100 example in C<\x{...}>, regardless of the C</x> modifier, there can be no
101 spaces. Same for a L<quantifier|/Quantifiers> such as C<{3}> or
102 C<{5,}>. Similarly, C<(?:...)> can't have a space between the C<?> and C<:>,
103 but can between the C<(> and C<?>. Within any delimiters for such a
104 construct, allowed spaces are not affected by C</x>, and depend on the
105 construct. For example, C<\x{...}> can't have spaces because hexadecimal
106 numbers don't have spaces in them. But, Unicode properties can have spaces, so
107 in C<\p{...}> there can be spaces that follow the Unicode rules, for which see
108 L<perluniprops/Properties accessible through \p{} and \P{}>.
111 =head2 Regular Expressions
113 =head3 Metacharacters
115 The patterns used in Perl pattern matching evolved from those supplied in
116 the Version 8 regex routines. (The routines are derived
117 (distantly) from Henry Spencer's freely redistributable reimplementation
118 of the V8 routines.) See L<Version 8 Regular Expressions> for
121 In particular the following metacharacters have their standard I<egrep>-ish
124 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
127 \ Quote the next metacharacter
128 ^ Match the beginning of the line
129 . Match any character (except newline)
130 $ Match the end of the line (or before newline at the end)
133 [] Bracketed Character class
135 By default, the "^" character is guaranteed to match only the
136 beginning of the string, the "$" character only the end (or before the
137 newline at the end), and Perl does certain optimizations with the
138 assumption that the string contains only one line. Embedded newlines
139 will not be matched by "^" or "$". You may, however, wish to treat a
140 string as a multi-line buffer, such that the "^" will match after any
141 newline within the string (except if the newline is the last character in
142 the string), and "$" will match before any newline. At the
143 cost of a little more overhead, you can do this by using the /m modifier
144 on the pattern match operator. (Older programs did this by setting C<$*>,
145 but this practice has been removed in perl 5.9.)
148 To simplify multi-line substitutions, the "." character never matches a
149 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
150 the string is a single line--even if it isn't.
155 The following standard quantifiers are recognized:
156 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
158 * Match 0 or more times
159 + Match 1 or more times
161 {n} Match exactly n times
162 {n,} Match at least n times
163 {n,m} Match at least n but not more than m times
165 (If a curly bracket occurs in any other context, it is treated
166 as a regular character. In particular, the lower bound
167 is not optional.) The "*" quantifier is equivalent to C<{0,}>, the "+"
168 quantifier to C<{1,}>, and the "?" quantifier to C<{0,1}>. n and m are limited
169 to non-negative integral values less than a preset limit defined when perl is built.
170 This is usually 32766 on the most common platforms. The actual limit can
171 be seen in the error message generated by code such as this:
173 $_ **= $_ , / {$_} / for 2 .. 42;
175 By default, a quantified subpattern is "greedy", that is, it will match as
176 many times as possible (given a particular starting location) while still
177 allowing the rest of the pattern to match. If you want it to match the
178 minimum number of times possible, follow the quantifier with a "?". Note
179 that the meanings don't change, just the "greediness":
180 X<metacharacter> X<greedy> X<greediness>
181 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
183 *? Match 0 or more times, not greedily
184 +? Match 1 or more times, not greedily
185 ?? Match 0 or 1 time, not greedily
186 {n}? Match exactly n times, not greedily
187 {n,}? Match at least n times, not greedily
188 {n,m}? Match at least n but not more than m times, not greedily
190 By default, when a quantified subpattern does not allow the rest of the
191 overall pattern to match, Perl will backtrack. However, this behaviour is
192 sometimes undesirable. Thus Perl provides the "possessive" quantifier form
195 *+ Match 0 or more times and give nothing back
196 ++ Match 1 or more times and give nothing back
197 ?+ Match 0 or 1 time and give nothing back
198 {n}+ Match exactly n times and give nothing back (redundant)
199 {n,}+ Match at least n times and give nothing back
200 {n,m}+ Match at least n but not more than m times and give nothing back
206 will never match, as the C<a++> will gobble up all the C<a>'s in the
207 string and won't leave any for the remaining part of the pattern. This
208 feature can be extremely useful to give perl hints about where it
209 shouldn't backtrack. For instance, the typical "match a double-quoted
210 string" problem can be most efficiently performed when written as:
212 /"(?:[^"\\]++|\\.)*+"/
214 as we know that if the final quote does not match, backtracking will not
215 help. See the independent subexpression C<< (?>...) >> for more details;
216 possessive quantifiers are just syntactic sugar for that construct. For
217 instance the above example could also be written as follows:
219 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
221 =head3 Escape sequences
223 Because patterns are processed as double quoted strings, the following
230 \a alarm (bell) (BEL)
231 \e escape (think troff) (ESC)
232 \cK control char (example: VT)
233 \x{}, \x00 character whose ordinal is the given hexadecimal number
234 \N{name} named Unicode character or character sequence
235 \N{U+263D} Unicode character (example: FIRST QUARTER MOON)
236 \o{}, \000 character whose ordinal is the given octal number
237 \l lowercase next char (think vi)
238 \u uppercase next char (think vi)
239 \L lowercase till \E (think vi)
240 \U uppercase till \E (think vi)
241 \Q quote (disable) pattern metacharacters till \E
242 \E end either case modification or quoted section, think vi
244 Details are in L<perlop/Quote and Quote-like Operators>.
246 =head3 Character Classes and other Special Escapes
248 In addition, Perl defines the following:
249 X<\g> X<\k> X<\K> X<backreference>
251 Sequence Note Description
252 [...] [1] Match a character according to the rules of the
253 bracketed character class defined by the "...".
254 Example: [a-z] matches "a" or "b" or "c" ... or "z"
255 [[:...:]] [2] Match a character according to the rules of the POSIX
256 character class "..." within the outer bracketed
257 character class. Example: [[:upper:]] matches any
259 \w [3] Match a "word" character (alphanumeric plus "_", plus
260 other connector punctuation chars plus Unicode
262 \W [3] Match a non-"word" character
263 \s [3] Match a whitespace character
264 \S [3] Match a non-whitespace character
265 \d [3] Match a decimal digit character
266 \D [3] Match a non-digit character
267 \pP [3] Match P, named property. Use \p{Prop} for longer names
269 \X [4] Match Unicode "eXtended grapheme cluster"
270 \C Match a single C-language char (octet) even if that is
271 part of a larger UTF-8 character. Thus it breaks up
272 characters into their UTF-8 bytes, so you may end up
273 with malformed pieces of UTF-8. Unsupported in
275 \1 [5] Backreference to a specific capture group or buffer.
276 '1' may actually be any positive integer.
277 \g1 [5] Backreference to a specific or previous group,
278 \g{-1} [5] The number may be negative indicating a relative
279 previous group and may optionally be wrapped in
280 curly brackets for safer parsing.
281 \g{name} [5] Named backreference
282 \k<name> [5] Named backreference
283 \K [6] Keep the stuff left of the \K, don't include it in $&
284 \N [7] Any character but \n (experimental). Not affected by
286 \v [3] Vertical whitespace
287 \V [3] Not vertical whitespace
288 \h [3] Horizontal whitespace
289 \H [3] Not horizontal whitespace
296 See L<perlrecharclass/Bracketed Character Classes> for details.
300 See L<perlrecharclass/POSIX Character Classes> for details.
304 See L<perlrecharclass/Backslash sequences> for details.
308 See L<perlrebackslash/Misc> for details.
312 See L</Capture groups> below for details.
316 See L</Extended Patterns> below for details.
320 Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the
321 character or character sequence whose name is C<NAME>; and similarly
322 when of the form C<\N{U+I<hex>}>, it matches the character whose Unicode
323 code point is I<hex>. Otherwise it matches any character but C<\n>.
329 Perl defines the following zero-width assertions:
330 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
331 X<regexp, zero-width assertion>
332 X<regular expression, zero-width assertion>
333 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
335 \b Match a word boundary
336 \B Match except at a word boundary
337 \A Match only at beginning of string
338 \Z Match only at end of string, or before newline at the end
339 \z Match only at end of string
340 \G Match only at pos() (e.g. at the end-of-match position
343 A word boundary (C<\b>) is a spot between two characters
344 that has a C<\w> on one side of it and a C<\W> on the other side
345 of it (in either order), counting the imaginary characters off the
346 beginning and end of the string as matching a C<\W>. (Within
347 character classes C<\b> represents backspace rather than a word
348 boundary, just as it normally does in any double-quoted string.)
349 The C<\A> and C<\Z> are just like "^" and "$", except that they
350 won't match multiple times when the C</m> modifier is used, while
351 "^" and "$" will match at every internal line boundary. To match
352 the actual end of the string and not ignore an optional trailing
354 X<\b> X<\A> X<\Z> X<\z> X</m>
356 The C<\G> assertion can be used to chain global matches (using
357 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
358 It is also useful when writing C<lex>-like scanners, when you have
359 several patterns that you want to match against consequent substrings
360 of your string, see the previous reference. The actual location
361 where C<\G> will match can also be influenced by using C<pos()> as
362 an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
363 matches is modified somewhat, in that contents to the left of C<\G> is
364 not counted when determining the length of the match. Thus the following
365 will not match forever:
370 while ($string =~ /(.\G)/g) {
374 It will print 'A' and then terminate, as it considers the match to
375 be zero-width, and thus will not match at the same position twice in a
378 It is worth noting that C<\G> improperly used can result in an infinite
379 loop. Take care when using patterns that include C<\G> in an alternation.
381 =head3 Capture groups
383 The bracketing construct C<( ... )> creates capture groups (also referred to as
384 capture buffers). To refer to the current contents of a group later on, within
385 the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>)
386 for the second, and so on.
387 This is called a I<backreference>.
388 X<regex, capture buffer> X<regexp, capture buffer>
389 X<regex, capture group> X<regexp, capture group>
390 X<regular expression, capture buffer> X<backreference>
391 X<regular expression, capture group> X<backreference>
392 X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
393 X<named capture buffer> X<regular expression, named capture buffer>
394 X<named capture group> X<regular expression, named capture group>
395 X<%+> X<$+{name}> X<< \k<name> >>
396 There is no limit to the number of captured substrings that you may use.
397 Groups are numbered with the leftmost open parenthesis being number 1, etc. If
398 a group did not match, the associated backreference won't match either. (This
399 can happen if the group is optional, or in a different branch of an
401 You can omit the C<"g">, and write C<"\1">, etc, but there are some issues with
402 this form, described below.
404 You can also refer to capture groups relatively, by using a negative number, so
405 that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture
406 group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For
413 \g{-1} # backref to group 3
414 \g{-3} # backref to group 1
418 would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to
419 interpolate regexes into larger regexes and not have to worry about the
420 capture groups being renumbered.
422 You can dispense with numbers altogether and create named capture groups.
423 The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to
424 reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may
425 also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.)
426 I<name> must not begin with a number, nor contain hyphens.
427 When different groups within the same pattern have the same name, any reference
428 to that name assumes the leftmost defined group. Named groups count in
429 absolute and relative numbering, and so can also be referred to by those
431 (It's possible to do things with named capture groups that would otherwise
434 Capture group contents are dynamically scoped and available to you outside the
435 pattern until the end of the enclosing block or until the next successful
436 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
437 You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">,
438 etc); or by name via the C<%+> hash, using C<"$+{I<name>}">.
440 Braces are required in referring to named capture groups, but are optional for
441 absolute or relative numbered ones. Braces are safer when creating a regex by
442 concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a>
443 contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which
444 is probably not what you intended.
446 The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that
447 there were no named nor relative numbered capture groups. Absolute numbered
448 groups were referred to using C<\1>, C<\2>, etc, and this notation is still
449 accepted (and likely always will be). But it leads to some ambiguities if
450 there are more than 9 capture groups, as C<\10> could mean either the tenth
451 capture group, or the character whose ordinal in octal is 010 (a backspace in
452 ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference
453 only if at least 10 left parentheses have opened before it. Likewise C<\11> is
454 a backreference only if at least 11 left parentheses have opened before it.
455 And so on. C<\1> through C<\9> are always interpreted as backreferences.
456 There are several examples below that illustrate these perils. You can avoid
457 the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups;
458 and for octal constants always using C<\o{}>, or for C<\077> and below, using 3
459 digits padded with leading zeros, since a leading zero implies an octal
462 The C<\I<digit>> notation also works in certain circumstances outside
463 the pattern. See L</Warning on \1 Instead of $1> below for details.)
467 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
469 /(.)\g1/ # find first doubled char
470 and print "'$1' is the first doubled character\n";
472 /(?<char>.)\k<char>/ # ... a different way
473 and print "'$+{char}' is the first doubled character\n";
475 /(?'char'.)\g1/ # ... mix and match
476 and print "'$1' is the first doubled character\n";
478 if (/Time: (..):(..):(..)/) { # parse out values
484 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference
485 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal
486 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference
487 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal
489 $a = '(.)\1'; # Creates problems when concatenated.
490 $b = '(.)\g{1}'; # Avoids the problems.
491 "aa" =~ /${a}/; # True
492 "aa" =~ /${b}/; # True
493 "aa0" =~ /${a}0/; # False!
494 "aa0" =~ /${b}0/; # True
495 "aa\x08" =~ /${a}0/; # True!
496 "aa\x08" =~ /${b}0/; # False
498 Several special variables also refer back to portions of the previous
499 match. C<$+> returns whatever the last bracket match matched.
500 C<$&> returns the entire matched string. (At one point C<$0> did
501 also, but now it returns the name of the program.) C<$`> returns
502 everything before the matched string. C<$'> returns everything
503 after the matched string. And C<$^N> contains whatever was matched by
504 the most-recently closed group (submatch). C<$^N> can be used in
505 extended patterns (see below), for example to assign a submatch to a
507 X<$+> X<$^N> X<$&> X<$`> X<$'>
509 These special variables, like the C<%+> hash and the numbered match variables
510 (C<$1>, C<$2>, C<$3>, etc.) are dynamically scoped
511 until the end of the enclosing block or until the next successful
512 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
513 X<$+> X<$^N> X<$&> X<$`> X<$'>
514 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
516 B<NOTE>: Failed matches in Perl do not reset the match variables,
517 which makes it easier to write code that tests for a series of more
518 specific cases and remembers the best match.
520 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
521 C<$'> anywhere in the program, it has to provide them for every
522 pattern match. This may substantially slow your program. Perl
523 uses the same mechanism to produce C<$1>, C<$2>, etc, so you also pay a
524 price for each pattern that contains capturing parentheses. (To
525 avoid this cost while retaining the grouping behaviour, use the
526 extended regular expression C<(?: ... )> instead.) But if you never
527 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
528 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
529 if you can, but if you can't (and some algorithms really appreciate
530 them), once you've used them once, use them at will, because you've
531 already paid the price. As of 5.005, C<$&> is not so costly as the
535 As a workaround for this problem, Perl 5.10.0 introduces C<${^PREMATCH}>,
536 C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&>
537 and C<$'>, B<except> that they are only guaranteed to be defined after a
538 successful match that was executed with the C</p> (preserve) modifier.
539 The use of these variables incurs no global performance penalty, unlike
540 their punctuation char equivalents, however at the trade-off that you
541 have to tell perl when you want to use them.
544 =head2 Quoting metacharacters
546 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
547 C<\w>, C<\n>. Unlike some other regular expression languages, there
548 are no backslashed symbols that aren't alphanumeric. So anything
549 that looks like \\, \(, \), \<, \>, \{, or \} is always
550 interpreted as a literal character, not a metacharacter. This was
551 once used in a common idiom to disable or quote the special meanings
552 of regular expression metacharacters in a string that you want to
553 use for a pattern. Simply quote all non-"word" characters:
555 $pattern =~ s/(\W)/\\$1/g;
557 (If C<use locale> is set, then this depends on the current locale.)
558 Today it is more common to use the quotemeta() function or the C<\Q>
559 metaquoting escape sequence to disable all metacharacters' special
562 /$unquoted\Q$quoted\E$unquoted/
564 Beware that if you put literal backslashes (those not inside
565 interpolated variables) between C<\Q> and C<\E>, double-quotish
566 backslash interpolation may lead to confusing results. If you
567 I<need> to use literal backslashes within C<\Q...\E>,
568 consult L<perlop/"Gory details of parsing quoted constructs">.
570 =head2 Extended Patterns
572 Perl also defines a consistent extension syntax for features not
573 found in standard tools like B<awk> and B<lex>. The syntax is a
574 pair of parentheses with a question mark as the first thing within
575 the parentheses. The character after the question mark indicates
578 The stability of these extensions varies widely. Some have been
579 part of the core language for many years. Others are experimental
580 and may change without warning or be completely removed. Check
581 the documentation on an individual feature to verify its current
584 A question mark was chosen for this and for the minimal-matching
585 construct because 1) question marks are rare in older regular
586 expressions, and 2) whenever you see one, you should stop and
587 "question" exactly what is going on. That's psychology...
594 A comment. The text is ignored. If the C</x> modifier enables
595 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
596 the comment as soon as it sees a C<)>, so there is no way to put a literal
599 =item C<(?dlupimsx-imsx)>
604 One or more embedded pattern-match modifiers, to be turned on (or
605 turned off, if preceded by C<->) for the remainder of the pattern or
606 the remainder of the enclosing pattern group (if any).
608 This is particularly useful for dynamic patterns, such as those read in from a
609 configuration file, taken from an argument, or specified in a table
610 somewhere. Consider the case where some patterns want to be case
611 sensitive and some do not: The case insensitive ones merely need to
612 include C<(?i)> at the front of the pattern. For example:
615 if ( /$pattern/i ) { }
619 $pattern = "(?i)foobar";
620 if ( /$pattern/ ) { }
622 These modifiers are restored at the end of the enclosing group. For example,
624 ( (?i) blah ) \s+ \g1
626 will match C<blah> in any case, some spaces, and an exact (I<including the case>!)
627 repetition of the previous word, assuming the C</x> modifier, and no C</i>
628 modifier outside this group.
630 These modifiers do not carry over into named subpatterns called in the
631 enclosing group. In other words, a pattern such as C<((?i)(&NAME))> does not
632 change the case-sensitivity of the "NAME" pattern.
634 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
635 after the C<"?"> is a shorthand equivalent to C<d-imsx>. Flags (except
636 C<"d">) may follow the caret to override it.
637 But a minus sign is not legal with it.
639 Also, starting in Perl 5.14, are modifiers C<"d">, C<"l">, and C<"u">,
640 which for 5.14 may not be used as suffix modifiers.
642 C<"l"> means to use a locale (see L<perllocale>) when pattern matching.
643 The locale used will be the one in effect at the time of execution of
644 the pattern match. This may not be the same as the compilation-time
645 locale, and can differ from one match to another if there is an
646 intervening call of the
647 L<setlocale() function|perllocale/The setlocale function>.
648 This modifier is automatically set if the regular expression is compiled
649 within the scope of a C<"use locale"> pragma.
651 C<"u"> means to use Unicode semantics when pattern matching. It is
652 automatically set if the regular expression is compiled within the scope
653 of a L<C<"use feature 'unicode_strings">|feature> pragma (and isn't
654 also in the scope of L<C<"use locale">|locale> nor
655 L<C<"use bytes">|bytes> pragmas. It is not fully implemented at the
656 time of this writing, but work is being done to complete the job. On
657 EBCDIC platforms this currently has no effect, but on ASCII platforms,
658 it effectively turns them into Latin-1 platforms. That is, the ASCII
659 characters remain as ASCII characters (since ASCII is a subset of
660 Latin-1), but the non-ASCII code points are treated as Latin-1
661 characters. Right now, this only applies to the C<"\b">, C<"\s">, and
662 C<"\w"> pattern matching operators, plus their complements. For
663 example, when this option is not on, C<"\w"> matches precisely
664 C<[A-Za-z0-9_]> (on a non-utf8 string). When the option is on, it
665 matches not just those, but all the Latin-1 word characters (such as an
666 "n" with a tilde). It thus matches exactly the same set of code points
667 from 0 to 255 as it would if the string were encoded in utf8.
669 C<"d"> means to use the traditional Perl pattern matching behavior.
670 This is dualistic (hence the name C<"d">, which also could stand for
671 "default"). When this is in effect, Perl matches utf8-encoded strings
672 using Unicode rules, and matches non-utf8-encoded strings using the
673 platform's native character set rules.
674 See L<perlunicode/The "Unicode Bug">. It is automatically selected by
675 default if the regular expression is compiled neither within the scope
676 of a C<"use locale"> pragma nor a <C<"use feature 'unicode_strings">
679 Note that the C<d>, C<l>, C<p>, and C<u> modifiers are special in that
680 they can only be enabled, not disabled, and the C<d>, C<l>, and C<u>
681 modifiers are mutually exclusive: specifying one de-specifies the
682 others, and a maximum of one may appear in the construct. Thus, for
683 example, C<(?-p)>, C<(?-d:...)>, and C<(?-dl:...)> will warn when
684 compiled under C<use warnings>.
686 Note also that the C<p> modifier is special in that its presence
687 anywhere in a pattern has a global effect.
692 =item C<(?dluimsx-imsx:pattern)>
694 =item C<(?^luimsx:pattern)>
697 This is for clustering, not capturing; it groups subexpressions like
698 "()", but doesn't make backreferences as "()" does. So
700 @fields = split(/\b(?:a|b|c)\b/)
704 @fields = split(/\b(a|b|c)\b/)
706 but doesn't spit out extra fields. It's also cheaper not to capture
707 characters if you don't need to.
709 Any letters between C<?> and C<:> act as flags modifiers as with
710 C<(?dluimsx-imsx)>. For example,
712 /(?s-i:more.*than).*million/i
714 is equivalent to the more verbose
716 /(?:(?s-i)more.*than).*million/i
718 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
719 after the C<"?"> is a shorthand equivalent to C<d-imsx>. Any positive
720 flags (except C<"d">) may follow the caret, so
728 The caret tells Perl that this cluster doesn't inherit the flags of any
729 surrounding pattern, but to go back to the system defaults (C<d-imsx>),
730 modified by any flags specified.
732 The caret allows for simpler stringification of compiled regular
733 expressions. These look like
737 with any non-default flags appearing between the caret and the colon.
738 A test that looks at such stringification thus doesn't need to have the
739 system default flags hard-coded in it, just the caret. If new flags are
740 added to Perl, the meaning of the caret's expansion will change to include
741 the default for those flags, so the test will still work, unchanged.
743 Specifying a negative flag after the caret is an error, as the flag is
746 Mnemonic for C<(?^...)>: A fresh beginning since the usual use of a caret is
747 to match at the beginning.
750 X<(?|)> X<Branch reset>
752 This is the "branch reset" pattern, which has the special property
753 that the capture groups are numbered from the same starting point
754 in each alternation branch. It is available starting from perl 5.10.0.
756 Capture groups are numbered from left to right, but inside this
757 construct the numbering is restarted for each branch.
759 The numbering within each branch will be as normal, and any groups
760 following this construct will be numbered as though the construct
761 contained only one branch, that being the one with the most capture
764 This construct will be useful when you want to capture one of a
765 number of alternative matches.
767 Consider the following pattern. The numbers underneath show in
768 which group the captured content will be stored.
771 # before ---------------branch-reset----------- after
772 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
775 Be careful when using the branch reset pattern in combination with
776 named captures. Named captures are implemented as being aliases to
777 numbered groups holding the captures, and that interferes with the
778 implementation of the branch reset pattern. If you are using named
779 captures in a branch reset pattern, it's best to use the same names,
780 in the same order, in each of the alternations:
782 /(?| (?<a> x ) (?<b> y )
783 | (?<a> z ) (?<b> w )) /x
785 Not doing so may lead to surprises:
787 "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x;
788 say $+ {a}; # Prints '12'
789 say $+ {b}; # *Also* prints '12'.
791 The problem here is that both the group named C<< a >> and the group
792 named C<< b >> are aliases for the group belonging to C<< $1 >>.
794 =item Look-Around Assertions
795 X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
797 Look-around assertions are zero width patterns which match a specific
798 pattern without including it in C<$&>. Positive assertions match when
799 their subpattern matches, negative assertions match when their subpattern
800 fails. Look-behind matches text up to the current match position,
801 look-ahead matches text following the current match position.
806 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
808 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
809 matches a word followed by a tab, without including the tab in C<$&>.
812 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
814 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
815 matches any occurrence of "foo" that isn't followed by "bar". Note
816 however that look-ahead and look-behind are NOT the same thing. You cannot
817 use this for look-behind.
819 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
820 will not do what you want. That's because the C<(?!foo)> is just saying that
821 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
822 match. You would have to do something like C</(?!foo)...bar/> for that. We
823 say "like" because there's the case of your "bar" not having three characters
824 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
825 Sometimes it's still easier just to say:
827 if (/bar/ && $` !~ /foo$/)
829 For look-behind see below.
831 =item C<(?<=pattern)> C<\K>
832 X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K>
834 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
835 matches a word that follows a tab, without including the tab in C<$&>.
836 Works only for fixed-width look-behind.
838 There is a special form of this construct, called C<\K>, which causes the
839 regex engine to "keep" everything it had matched prior to the C<\K> and
840 not include it in C<$&>. This effectively provides variable length
841 look-behind. The use of C<\K> inside of another look-around assertion
842 is allowed, but the behaviour is currently not well defined.
844 For various reasons C<\K> may be significantly more efficient than the
845 equivalent C<< (?<=...) >> construct, and it is especially useful in
846 situations where you want to efficiently remove something following
847 something else in a string. For instance
851 can be rewritten as the much more efficient
855 =item C<(?<!pattern)>
856 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
858 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
859 matches any occurrence of "foo" that does not follow "bar". Works
860 only for fixed-width look-behind.
864 =item C<(?'NAME'pattern)>
866 =item C<< (?<NAME>pattern) >>
867 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
869 A named capture group. Identical in every respect to normal capturing
870 parentheses C<()> but for the additional fact that C<%+> or C<%-> may be
871 used after a successful match to refer to a named group. See C<perlvar>
872 for more details on the C<%+> and C<%-> hashes.
874 If multiple distinct capture groups have the same name then the
875 $+{NAME} will refer to the leftmost defined group in the match.
877 The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent.
879 B<NOTE:> While the notation of this construct is the same as the similar
880 function in .NET regexes, the behavior is not. In Perl the groups are
881 numbered sequentially regardless of being named or not. Thus in the
886 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
887 the opposite which is what a .NET regex hacker might expect.
889 Currently NAME is restricted to simple identifiers only.
890 In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
891 its Unicode extension (see L<utf8>),
892 though it isn't extended by the locale (see L<perllocale>).
894 B<NOTE:> In order to make things easier for programmers with experience
895 with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >>
896 may be used instead of C<< (?<NAME>pattern) >>; however this form does not
897 support the use of single quotes as a delimiter for the name.
899 =item C<< \k<NAME> >>
901 =item C<< \k'NAME' >>
903 Named backreference. Similar to numeric backreferences, except that
904 the group is designated by name and not number. If multiple groups
905 have the same name then it refers to the leftmost defined group in
908 It is an error to refer to a name not defined by a C<< (?<NAME>) >>
909 earlier in the pattern.
911 Both forms are equivalent.
913 B<NOTE:> In order to make things easier for programmers with experience
914 with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >>
915 may be used instead of C<< \k<NAME> >>.
918 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
920 B<WARNING>: This extended regular expression feature is considered
921 experimental, and may be changed without notice. Code executed that
922 has side effects may not perform identically from version to version
923 due to the effect of future optimisations in the regex engine.
925 This zero-width assertion evaluates any embedded Perl code. It
926 always succeeds, and its C<code> is not interpolated. Currently,
927 the rules to determine where the C<code> ends are somewhat convoluted.
929 This feature can be used together with the special variable C<$^N> to
930 capture the results of submatches in variables without having to keep
931 track of the number of nested parentheses. For example:
933 $_ = "The brown fox jumps over the lazy dog";
934 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
935 print "color = $color, animal = $animal\n";
937 Inside the C<(?{...})> block, C<$_> refers to the string the regular
938 expression is matching against. You can also use C<pos()> to know what is
939 the current position of matching within this string.
941 The C<code> is properly scoped in the following sense: If the assertion
942 is backtracked (compare L<"Backtracking">), all changes introduced after
943 C<local>ization are undone, so that
947 (?{ $cnt = 0 }) # Initialize $cnt.
951 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
955 (?{ $res = $cnt }) # On success copy to
956 # non-localized location.
959 will set C<$res = 4>. Note that after the match, C<$cnt> returns to the globally
960 introduced value, because the scopes that restrict C<local> operators
963 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
964 switch. If I<not> used in this way, the result of evaluation of
965 C<code> is put into the special variable C<$^R>. This happens
966 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
967 inside the same regular expression.
969 The assignment to C<$^R> above is properly localized, so the old
970 value of C<$^R> is restored if the assertion is backtracked; compare
973 For reasons of security, this construct is forbidden if the regular
974 expression involves run-time interpolation of variables, unless the
975 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
976 variables contain results of C<qr//> operator (see
977 L<perlop/"qr/STRINGE<sol>msixpo">).
979 This restriction is due to the wide-spread and remarkably convenient
980 custom of using run-time determined strings as patterns. For example:
986 Before Perl knew how to execute interpolated code within a pattern,
987 this operation was completely safe from a security point of view,
988 although it could raise an exception from an illegal pattern. If
989 you turn on the C<use re 'eval'>, though, it is no longer secure,
990 so you should only do so if you are also using taint checking.
991 Better yet, use the carefully constrained evaluation within a Safe
992 compartment. See L<perlsec> for details about both these mechanisms.
994 B<WARNING>: Use of lexical (C<my>) variables in these blocks is
995 broken. The result is unpredictable and will make perl unstable. The
996 workaround is to use global (C<our>) variables.
998 B<WARNING>: In perl 5.12.x and earlier, the regex engine
999 was not re-entrant, so interpolated code could not
1000 safely invoke the regex engine either directly with
1001 C<m//> or C<s///>), or indirectly with functions such as
1002 C<split>. Invoking the regex engine in these blocks would make perl
1005 =item C<(??{ code })>
1007 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
1009 B<WARNING>: This extended regular expression feature is considered
1010 experimental, and may be changed without notice. Code executed that
1011 has side effects may not perform identically from version to version
1012 due to the effect of future optimisations in the regex engine.
1014 This is a "postponed" regular subexpression. The C<code> is evaluated
1015 at run time, at the moment this subexpression may match. The result
1016 of evaluation is considered as a regular expression and matched as
1017 if it were inserted instead of this construct. Note that this means
1018 that the contents of capture groups defined inside an eval'ed pattern
1019 are not available outside of the pattern, and vice versa, there is no
1020 way for the inner pattern to refer to a capture group defined outside.
1023 ('a' x 100)=~/(??{'(.)' x 100})/
1025 B<will> match, it will B<not> set $1.
1027 The C<code> is not interpolated. As before, the rules to determine
1028 where the C<code> ends are currently somewhat convoluted.
1030 The following pattern matches a parenthesized group:
1035 (?> [^()]+ ) # Non-parens without backtracking
1037 (??{ $re }) # Group with matching parens
1042 See also C<(?PARNO)> for a different, more efficient way to accomplish
1045 For reasons of security, this construct is forbidden if the regular
1046 expression involves run-time interpolation of variables, unless the
1047 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
1048 variables contain results of C<qr//> operator (see
1049 L<perlop/"qrE<sol>STRINGE<sol>msixpo">).
1051 In perl 5.12.x and earlier, because the regex engine was not re-entrant,
1052 delayed code could not safely invoke the regex engine either directly with
1053 C<m//> or C<s///>), or indirectly with functions such as C<split>.
1055 Recursing deeper than 50 times without consuming any input string will
1056 result in a fatal error. The maximum depth is compiled into perl, so
1057 changing it requires a custom build.
1059 =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)>
1060 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
1061 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
1062 X<regex, relative recursion>
1064 Similar to C<(??{ code })> except it does not involve compiling any code,
1065 instead it treats the contents of a capture group as an independent
1066 pattern that must match at the current position. Capture groups
1067 contained by the pattern will have the value as determined by the
1068 outermost recursion.
1070 PARNO is a sequence of digits (not starting with 0) whose value reflects
1071 the paren-number of the capture group to recurse to. C<(?R)> recurses to
1072 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
1073 C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed
1074 to be relative, with negative numbers indicating preceding capture groups
1075 and positive ones following. Thus C<(?-1)> refers to the most recently
1076 declared group, and C<(?+1)> indicates the next group to be declared.
1077 Note that the counting for relative recursion differs from that of
1078 relative backreferences, in that with recursion unclosed groups B<are>
1081 The following pattern matches a function foo() which may contain
1082 balanced parentheses as the argument.
1084 $re = qr{ ( # paren group 1 (full function)
1086 ( # paren group 2 (parens)
1088 ( # paren group 3 (contents of parens)
1090 (?> [^()]+ ) # Non-parens without backtracking
1092 (?2) # Recurse to start of paren group 2
1100 If the pattern was used as follows
1102 'foo(bar(baz)+baz(bop))'=~/$re/
1103 and print "\$1 = $1\n",
1107 the output produced should be the following:
1109 $1 = foo(bar(baz)+baz(bop))
1110 $2 = (bar(baz)+baz(bop))
1111 $3 = bar(baz)+baz(bop)
1113 If there is no corresponding capture group defined, then it is a
1114 fatal error. Recursing deeper than 50 times without consuming any input
1115 string will also result in a fatal error. The maximum depth is compiled
1116 into perl, so changing it requires a custom build.
1118 The following shows how using negative indexing can make it
1119 easier to embed recursive patterns inside of a C<qr//> construct
1122 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
1123 if (/foo $parens \s+ + \s+ bar $parens/x) {
1124 # do something here...
1127 B<Note> that this pattern does not behave the same way as the equivalent
1128 PCRE or Python construct of the same form. In Perl you can backtrack into
1129 a recursed group, in PCRE and Python the recursed into group is treated
1130 as atomic. Also, modifiers are resolved at compile time, so constructs
1131 like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
1137 Recurse to a named subpattern. Identical to C<(?PARNO)> except that the
1138 parenthesis to recurse to is determined by name. If multiple parentheses have
1139 the same name, then it recurses to the leftmost.
1141 It is an error to refer to a name that is not declared somewhere in the
1144 B<NOTE:> In order to make things easier for programmers with experience
1145 with the Python or PCRE regex engines the pattern C<< (?P>NAME) >>
1146 may be used instead of C<< (?&NAME) >>.
1148 =item C<(?(condition)yes-pattern|no-pattern)>
1151 =item C<(?(condition)yes-pattern)>
1153 Conditional expression. C<(condition)> should be either an integer in
1154 parentheses (which is valid if the corresponding pair of parentheses
1155 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
1156 name in angle brackets or single quotes (which is valid if a group
1157 with the given name matched), or the special symbol (R) (true when
1158 evaluated inside of recursion or eval). Additionally the R may be
1159 followed by a number, (which will be true when evaluated when recursing
1160 inside of the appropriate group), or by C<&NAME>, in which case it will
1161 be true only when evaluated during recursion in the named group.
1163 Here's a summary of the possible predicates:
1169 Checks if the numbered capturing group has matched something.
1171 =item (<NAME>) ('NAME')
1173 Checks if a group with the given name has matched something.
1177 Treats the code block as the condition.
1181 Checks if the expression has been evaluated inside of recursion.
1185 Checks if the expression has been evaluated while executing directly
1186 inside of the n-th capture group. This check is the regex equivalent of
1188 if ((caller(0))[3] eq 'subname') { ... }
1190 In other words, it does not check the full recursion stack.
1194 Similar to C<(R1)>, this predicate checks to see if we're executing
1195 directly inside of the leftmost group with a given name (this is the same
1196 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1197 stack, but only the name of the innermost active recursion.
1201 In this case, the yes-pattern is never directly executed, and no
1202 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1203 See below for details.
1214 matches a chunk of non-parentheses, possibly included in parentheses
1217 A special form is the C<(DEFINE)> predicate, which never executes directly
1218 its yes-pattern, and does not allow a no-pattern. This allows to define
1219 subpatterns which will be executed only by using the recursion mechanism.
1220 This way, you can define a set of regular expression rules that can be
1221 bundled into any pattern you choose.
1223 It is recommended that for this usage you put the DEFINE block at the
1224 end of the pattern, and that you name any subpatterns defined within it.
1226 Also, it's worth noting that patterns defined this way probably will
1227 not be as efficient, as the optimiser is not very clever about
1230 An example of how this might be used is as follows:
1232 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
1238 Note that capture groups matched inside of recursion are not accessible
1239 after the recursion returns, so the extra layer of capturing groups is
1240 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1241 C<$+{NAME}> would be.
1243 =item C<< (?>pattern) >>
1244 X<backtrack> X<backtracking> X<atomic> X<possessive>
1246 An "independent" subexpression, one which matches the substring
1247 that a I<standalone> C<pattern> would match if anchored at the given
1248 position, and it matches I<nothing other than this substring>. This
1249 construct is useful for optimizations of what would otherwise be
1250 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1251 It may also be useful in places where the "grab all you can, and do not
1252 give anything back" semantic is desirable.
1254 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1255 (anchored at the beginning of string, as above) will match I<all>
1256 characters C<a> at the beginning of string, leaving no C<a> for
1257 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1258 since the match of the subgroup C<a*> is influenced by the following
1259 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1260 C<a*ab> will match fewer characters than a standalone C<a*>, since
1261 this makes the tail match.
1263 An effect similar to C<< (?>pattern) >> may be achieved by writing
1264 C<(?=(pattern))\g1>. This matches the same substring as a standalone
1265 C<a+>, and the following C<\g1> eats the matched string; it therefore
1266 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1267 (The difference between these two constructs is that the second one
1268 uses a capturing group, thus shifting ordinals of backreferences
1269 in the rest of a regular expression.)
1271 Consider this pattern:
1282 That will efficiently match a nonempty group with matching parentheses
1283 two levels deep or less. However, if there is no such group, it
1284 will take virtually forever on a long string. That's because there
1285 are so many different ways to split a long string into several
1286 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1287 to a subpattern of the above pattern. Consider how the pattern
1288 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1289 seconds, but that each extra letter doubles this time. This
1290 exponential performance will make it appear that your program has
1291 hung. However, a tiny change to this pattern
1295 (?> [^()]+ ) # change x+ above to (?> x+ )
1302 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1303 this yourself would be a productive exercise), but finishes in a fourth
1304 the time when used on a similar string with 1000000 C<a>s. Be aware,
1305 however, that this pattern currently triggers a warning message under
1306 the C<use warnings> pragma or B<-w> switch saying it
1307 C<"matches null string many times in regex">.
1309 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1310 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1311 This was only 4 times slower on a string with 1000000 C<a>s.
1313 The "grab all you can, and do not give anything back" semantic is desirable
1314 in many situations where on the first sight a simple C<()*> looks like
1315 the correct solution. Suppose we parse text with comments being delimited
1316 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1317 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1318 the comment delimiter, because it may "give up" some whitespace if
1319 the remainder of the pattern can be made to match that way. The correct
1320 answer is either one of these:
1325 For example, to grab non-empty comments into $1, one should use either
1328 / (?> \# [ \t]* ) ( .+ ) /x;
1329 / \# [ \t]* ( [^ \t] .* ) /x;
1331 Which one you pick depends on which of these expressions better reflects
1332 the above specification of comments.
1334 In some literature this construct is called "atomic matching" or
1335 "possessive matching".
1337 Possessive quantifiers are equivalent to putting the item they are applied
1338 to inside of one of these constructs. The following equivalences apply:
1340 Quantifier Form Bracketing Form
1341 --------------- ---------------
1345 PAT{min,max}+ (?>PAT{min,max})
1349 =head2 Special Backtracking Control Verbs
1351 B<WARNING:> These patterns are experimental and subject to change or
1352 removal in a future version of Perl. Their usage in production code should
1353 be noted to avoid problems during upgrades.
1355 These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1356 otherwise stated the ARG argument is optional; in some cases, it is
1359 Any pattern containing a special backtracking verb that allows an argument
1360 has the special behaviour that when executed it sets the current package's
1361 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1364 On failure, the C<$REGERROR> variable will be set to the ARG value of the
1365 verb pattern, if the verb was involved in the failure of the match. If the
1366 ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1367 name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1368 none. Also, the C<$REGMARK> variable will be set to FALSE.
1370 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1371 the C<$REGMARK> variable will be set to the name of the last
1372 C<(*MARK:NAME)> pattern executed. See the explanation for the
1373 C<(*MARK:NAME)> verb below for more details.
1375 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1376 and most other regex related variables. They are not local to a scope, nor
1377 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1378 Use C<local> to localize changes to them to a specific scope if necessary.
1380 If a pattern does not contain a special backtracking verb that allows an
1381 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
1385 =item Verbs that take an argument
1389 =item C<(*PRUNE)> C<(*PRUNE:NAME)>
1390 X<(*PRUNE)> X<(*PRUNE:NAME)>
1392 This zero-width pattern prunes the backtracking tree at the current point
1393 when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1394 where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1395 A may backtrack as necessary to match. Once it is reached, matching
1396 continues in B, which may also backtrack as necessary; however, should B
1397 not match, then no further backtracking will take place, and the pattern
1398 will fail outright at the current starting position.
1400 The following example counts all the possible matching strings in a
1401 pattern (without actually matching any of them).
1403 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1404 print "Count=$count\n";
1419 If we add a C<(*PRUNE)> before the count like the following
1421 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
1422 print "Count=$count\n";
1424 we prevent backtracking and find the count of the longest matching
1425 at each matching starting point like so:
1432 Any number of C<(*PRUNE)> assertions may be used in a pattern.
1434 See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1435 control backtracking. In some cases, the use of C<(*PRUNE)> can be
1436 replaced with a C<< (?>pattern) >> with no functional difference; however,
1437 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1438 C<< (?>pattern) >> alone.
1441 =item C<(*SKIP)> C<(*SKIP:NAME)>
1444 This zero-width pattern is similar to C<(*PRUNE)>, except that on
1445 failure it also signifies that whatever text that was matched leading up
1446 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1447 of this pattern. This effectively means that the regex engine "skips" forward
1448 to this position on failure and tries to match again, (assuming that
1449 there is sufficient room to match).
1451 The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1452 C<(*MARK:NAME)> was encountered while matching, then it is that position
1453 which is used as the "skip point". If no C<(*MARK)> of that name was
1454 encountered, then the C<(*SKIP)> operator has no effect. When used
1455 without a name the "skip point" is where the match point was when
1456 executing the (*SKIP) pattern.
1458 Compare the following to the examples in C<(*PRUNE)>, note the string
1461 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
1462 print "Count=$count\n";
1470 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
1471 executed, the next starting point will be where the cursor was when the
1472 C<(*SKIP)> was executed.
1474 =item C<(*MARK:NAME)> C<(*:NAME)>
1475 X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)>
1477 This zero-width pattern can be used to mark the point reached in a string
1478 when a certain part of the pattern has been successfully matched. This
1479 mark may be given a name. A later C<(*SKIP)> pattern will then skip
1480 forward to that point if backtracked into on failure. Any number of
1481 C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated.
1483 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1484 can be used to "label" a pattern branch, so that after matching, the
1485 program can determine which branches of the pattern were involved in the
1488 When a match is successful, the C<$REGMARK> variable will be set to the
1489 name of the most recently executed C<(*MARK:NAME)> that was involved
1492 This can be used to determine which branch of a pattern was matched
1493 without using a separate capture group for each branch, which in turn
1494 can result in a performance improvement, as perl cannot optimize
1495 C</(?:(x)|(y)|(z))/> as efficiently as something like
1496 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1498 When a match has failed, and unless another verb has been involved in
1499 failing the match and has provided its own name to use, the C<$REGERROR>
1500 variable will be set to the name of the most recently executed
1503 See C<(*SKIP)> for more details.
1505 As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
1507 =item C<(*THEN)> C<(*THEN:NAME)>
1509 This is similar to the "cut group" operator C<::> from Perl 6. Like
1510 C<(*PRUNE)>, this verb always matches, and when backtracked into on
1511 failure, it causes the regex engine to try the next alternation in the
1512 innermost enclosing group (capturing or otherwise).
1514 Its name comes from the observation that this operation combined with the
1515 alternation operator (C<|>) can be used to create what is essentially a
1516 pattern-based if/then/else block:
1518 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
1520 Note that if this operator is used and NOT inside of an alternation then
1521 it acts exactly like the C<(*PRUNE)> operator.
1531 / ( A (*THEN) B | C (*THEN) D ) /
1535 / ( A (*PRUNE) B | C (*PRUNE) D ) /
1537 as after matching the A but failing on the B the C<(*THEN)> verb will
1538 backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
1543 This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a
1544 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
1545 into on failure it causes the match to fail outright. No further attempts
1546 to find a valid match by advancing the start pointer will occur again.
1549 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
1550 print "Count=$count\n";
1557 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1558 does not match, the regex engine will not try any further matching on the
1563 =item Verbs without an argument
1567 =item C<(*FAIL)> C<(*F)>
1570 This pattern matches nothing and always fails. It can be used to force the
1571 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1572 fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
1574 It is probably useful only when combined with C<(?{})> or C<(??{})>.
1579 B<WARNING:> This feature is highly experimental. It is not recommended
1580 for production code.
1582 This pattern matches nothing and causes the end of successful matching at
1583 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1584 whether there is actually more to match in the string. When inside of a
1585 nested pattern, such as recursion, or in a subpattern dynamically generated
1586 via C<(??{})>, only the innermost pattern is ended immediately.
1588 If the C<(*ACCEPT)> is inside of capturing groups then the groups are
1589 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1592 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
1594 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
1595 be set. If another branch in the inner parentheses were matched, such as in the
1596 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
1603 X<backtrack> X<backtracking>
1605 NOTE: This section presents an abstract approximation of regular
1606 expression behavior. For a more rigorous (and complicated) view of
1607 the rules involved in selecting a match among possible alternatives,
1608 see L<Combining RE Pieces>.
1610 A fundamental feature of regular expression matching involves the
1611 notion called I<backtracking>, which is currently used (when needed)
1612 by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>,
1613 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1614 internally, but the general principle outlined here is valid.
1616 For a regular expression to match, the I<entire> regular expression must
1617 match, not just part of it. So if the beginning of a pattern containing a
1618 quantifier succeeds in a way that causes later parts in the pattern to
1619 fail, the matching engine backs up and recalculates the beginning
1620 part--that's why it's called backtracking.
1622 Here is an example of backtracking: Let's say you want to find the
1623 word following "foo" in the string "Food is on the foo table.":
1625 $_ = "Food is on the foo table.";
1626 if ( /\b(foo)\s+(\w+)/i ) {
1627 print "$2 follows $1.\n";
1630 When the match runs, the first part of the regular expression (C<\b(foo)>)
1631 finds a possible match right at the beginning of the string, and loads up
1632 $1 with "Foo". However, as soon as the matching engine sees that there's
1633 no whitespace following the "Foo" that it had saved in $1, it realizes its
1634 mistake and starts over again one character after where it had the
1635 tentative match. This time it goes all the way until the next occurrence
1636 of "foo". The complete regular expression matches this time, and you get
1637 the expected output of "table follows foo."
1639 Sometimes minimal matching can help a lot. Imagine you'd like to match
1640 everything between "foo" and "bar". Initially, you write something
1643 $_ = "The food is under the bar in the barn.";
1644 if ( /foo(.*)bar/ ) {
1648 Which perhaps unexpectedly yields:
1650 got <d is under the bar in the >
1652 That's because C<.*> was greedy, so you get everything between the
1653 I<first> "foo" and the I<last> "bar". Here it's more effective
1654 to use minimal matching to make sure you get the text between a "foo"
1655 and the first "bar" thereafter.
1657 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1658 got <d is under the >
1660 Here's another example. Let's say you'd like to match a number at the end
1661 of a string, and you also want to keep the preceding part of the match.
1664 $_ = "I have 2 numbers: 53147";
1665 if ( /(.*)(\d*)/ ) { # Wrong!
1666 print "Beginning is <$1>, number is <$2>.\n";
1669 That won't work at all, because C<.*> was greedy and gobbled up the
1670 whole string. As C<\d*> can match on an empty string the complete
1671 regular expression matched successfully.
1673 Beginning is <I have 2 numbers: 53147>, number is <>.
1675 Here are some variants, most of which don't work:
1677 $_ = "I have 2 numbers: 53147";
1690 printf "%-12s ", $pat;
1692 print "<$1> <$2>\n";
1698 That will print out:
1700 (.*)(\d*) <I have 2 numbers: 53147> <>
1701 (.*)(\d+) <I have 2 numbers: 5314> <7>
1703 (.*?)(\d+) <I have > <2>
1704 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1705 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1706 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1707 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1709 As you see, this can be a bit tricky. It's important to realize that a
1710 regular expression is merely a set of assertions that gives a definition
1711 of success. There may be 0, 1, or several different ways that the
1712 definition might succeed against a particular string. And if there are
1713 multiple ways it might succeed, you need to understand backtracking to
1714 know which variety of success you will achieve.
1716 When using look-ahead assertions and negations, this can all get even
1717 trickier. Imagine you'd like to find a sequence of non-digits not
1718 followed by "123". You might try to write that as
1721 if ( /^\D*(?!123)/ ) { # Wrong!
1722 print "Yup, no 123 in $_\n";
1725 But that isn't going to match; at least, not the way you're hoping. It
1726 claims that there is no 123 in the string. Here's a clearer picture of
1727 why that pattern matches, contrary to popular expectations:
1732 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1733 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1735 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1736 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1744 You might have expected test 3 to fail because it seems to a more
1745 general purpose version of test 1. The important difference between
1746 them is that test 3 contains a quantifier (C<\D*>) and so can use
1747 backtracking, whereas test 1 will not. What's happening is
1748 that you've asked "Is it true that at the start of $x, following 0 or more
1749 non-digits, you have something that's not 123?" If the pattern matcher had
1750 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1753 The search engine will initially match C<\D*> with "ABC". Then it will
1754 try to match C<(?!123> with "123", which fails. But because
1755 a quantifier (C<\D*>) has been used in the regular expression, the
1756 search engine can backtrack and retry the match differently
1757 in the hope of matching the complete regular expression.
1759 The pattern really, I<really> wants to succeed, so it uses the
1760 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1761 time. Now there's indeed something following "AB" that is not
1762 "123". It's "C123", which suffices.
1764 We can deal with this by using both an assertion and a negation.
1765 We'll say that the first part in $1 must be followed both by a digit
1766 and by something that's not "123". Remember that the look-aheads
1767 are zero-width expressions--they only look, but don't consume any
1768 of the string in their match. So rewriting this way produces what
1769 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1771 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1772 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1776 In other words, the two zero-width assertions next to each other work as though
1777 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1778 matches only if you're at the beginning of the line AND the end of the
1779 line simultaneously. The deeper underlying truth is that juxtaposition in
1780 regular expressions always means AND, except when you write an explicit OR
1781 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1782 although the attempted matches are made at different positions because "a"
1783 is not a zero-width assertion, but a one-width assertion.
1785 B<WARNING>: Particularly complicated regular expressions can take
1786 exponential time to solve because of the immense number of possible
1787 ways they can use backtracking to try for a match. For example, without
1788 internal optimizations done by the regular expression engine, this will
1789 take a painfully long time to run:
1791 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1793 And if you used C<*>'s in the internal groups instead of limiting them
1794 to 0 through 5 matches, then it would take forever--or until you ran
1795 out of stack space. Moreover, these internal optimizations are not
1796 always applicable. For example, if you put C<{0,5}> instead of C<*>
1797 on the external group, no current optimization is applicable, and the
1798 match takes a long time to finish.
1800 A powerful tool for optimizing such beasts is what is known as an
1801 "independent group",
1802 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1803 zero-length look-ahead/look-behind assertions will not backtrack to make
1804 the tail match, since they are in "logical" context: only
1805 whether they match is considered relevant. For an example
1806 where side-effects of look-ahead I<might> have influenced the
1807 following match, see L<C<< (?>pattern) >>>.
1809 =head2 Version 8 Regular Expressions
1810 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1812 In case you're not familiar with the "regular" Version 8 regex
1813 routines, here are the pattern-matching rules not described above.
1815 Any single character matches itself, unless it is a I<metacharacter>
1816 with a special meaning described here or above. You can cause
1817 characters that normally function as metacharacters to be interpreted
1818 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1819 character; "\\" matches a "\"). This escape mechanism is also required
1820 for the character used as the pattern delimiter.
1822 A series of characters matches that series of characters in the target
1823 string, so the pattern C<blurfl> would match "blurfl" in the target
1826 You can specify a character class, by enclosing a list of characters
1827 in C<[]>, which will match any character from the list. If the
1828 first character after the "[" is "^", the class matches any character not
1829 in the list. Within a list, the "-" character specifies a
1830 range, so that C<a-z> represents all characters between "a" and "z",
1831 inclusive. If you want either "-" or "]" itself to be a member of a
1832 class, put it at the start of the list (possibly after a "^"), or
1833 escape it with a backslash. "-" is also taken literally when it is
1834 at the end of the list, just before the closing "]". (The
1835 following all specify the same class of three characters: C<[-az]>,
1836 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1837 specifies a class containing twenty-six characters, even on EBCDIC-based
1838 character sets.) Also, if you try to use the character
1839 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1840 a range, the "-" is understood literally.
1842 Note also that the whole range idea is rather unportable between
1843 character sets--and even within character sets they may cause results
1844 you probably didn't expect. A sound principle is to use only ranges
1845 that begin from and end at either alphabetics of equal case ([a-e],
1846 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1847 spell out the character sets in full.
1849 Characters may be specified using a metacharacter syntax much like that
1850 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1851 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1852 of three octal digits, matches the character whose coded character set value
1853 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1854 matches the character whose ordinal is I<nn>. The expression \cI<x>
1855 matches the character control-I<x>. Finally, the "." metacharacter
1856 matches any character except "\n" (unless you use C</s>).
1858 You can specify a series of alternatives for a pattern using "|" to
1859 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1860 or "foe" in the target string (as would C<f(e|i|o)e>). The
1861 first alternative includes everything from the last pattern delimiter
1862 ("(", "[", or the beginning of the pattern) up to the first "|", and
1863 the last alternative contains everything from the last "|" to the next
1864 pattern delimiter. That's why it's common practice to include
1865 alternatives in parentheses: to minimize confusion about where they
1868 Alternatives are tried from left to right, so the first
1869 alternative found for which the entire expression matches, is the one that
1870 is chosen. This means that alternatives are not necessarily greedy. For
1871 example: when matching C<foo|foot> against "barefoot", only the "foo"
1872 part will match, as that is the first alternative tried, and it successfully
1873 matches the target string. (This might not seem important, but it is
1874 important when you are capturing matched text using parentheses.)
1876 Also remember that "|" is interpreted as a literal within square brackets,
1877 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1879 Within a pattern, you may designate subpatterns for later reference
1880 by enclosing them in parentheses, and you may refer back to the
1881 I<n>th subpattern later in the pattern using the metacharacter
1882 \I<n>. Subpatterns are numbered based on the left to right order
1883 of their opening parenthesis. A backreference matches whatever
1884 actually matched the subpattern in the string being examined, not
1885 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\g1\d*> will
1886 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1887 1 matched "0x", even though the rule C<0|0x> could potentially match
1888 the leading 0 in the second number.
1890 =head2 Warning on \1 Instead of $1
1892 Some people get too used to writing things like:
1894 $pattern =~ s/(\W)/\\\1/g;
1896 This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid
1898 B<sed> addicts, but it's a dirty habit to get into. That's because in
1899 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1900 the usual double-quoted string means a control-A. The customary Unix
1901 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1902 of doing that, you get yourself into trouble if you then add an C</e>
1905 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1911 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1912 C<${1}000>. The operation of interpolation should not be confused
1913 with the operation of matching a backreference. Certainly they mean two
1914 different things on the I<left> side of the C<s///>.
1916 =head2 Repeated Patterns Matching a Zero-length Substring
1918 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1920 Regular expressions provide a terse and powerful programming language. As
1921 with most other power tools, power comes together with the ability
1924 A common abuse of this power stems from the ability to make infinite
1925 loops using regular expressions, with something as innocuous as:
1927 'foo' =~ m{ ( o? )* }x;
1929 The C<o?> matches at the beginning of C<'foo'>, and since the position
1930 in the string is not moved by the match, C<o?> would match again and again
1931 because of the C<*> quantifier. Another common way to create a similar cycle
1932 is with the looping modifier C<//g>:
1934 @matches = ( 'foo' =~ m{ o? }xg );
1938 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1940 or the loop implied by split().
1942 However, long experience has shown that many programming tasks may
1943 be significantly simplified by using repeated subexpressions that
1944 may match zero-length substrings. Here's a simple example being:
1946 @chars = split //, $string; # // is not magic in split
1947 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1949 Thus Perl allows such constructs, by I<forcefully breaking
1950 the infinite loop>. The rules for this are different for lower-level
1951 loops given by the greedy quantifiers C<*+{}>, and for higher-level
1952 ones like the C</g> modifier or split() operator.
1954 The lower-level loops are I<interrupted> (that is, the loop is
1955 broken) when Perl detects that a repeated expression matched a
1956 zero-length substring. Thus
1958 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1960 is made equivalent to
1962 m{ (?: NON_ZERO_LENGTH )*
1967 The higher level-loops preserve an additional state between iterations:
1968 whether the last match was zero-length. To break the loop, the following
1969 match after a zero-length match is prohibited to have a length of zero.
1970 This prohibition interacts with backtracking (see L<"Backtracking">),
1971 and so the I<second best> match is chosen if the I<best> match is of
1979 results in C<< <><b><><a><><r><> >>. At each position of the string the best
1980 match given by non-greedy C<??> is the zero-length match, and the I<second
1981 best> match is what is matched by C<\w>. Thus zero-length matches
1982 alternate with one-character-long matches.
1984 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1985 position one notch further in the string.
1987 The additional state of being I<matched with zero-length> is associated with
1988 the matched string, and is reset by each assignment to pos().
1989 Zero-length matches at the end of the previous match are ignored
1992 =head2 Combining RE Pieces
1994 Each of the elementary pieces of regular expressions which were described
1995 before (such as C<ab> or C<\Z>) could match at most one substring
1996 at the given position of the input string. However, in a typical regular
1997 expression these elementary pieces are combined into more complicated
1998 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1999 (in these examples C<S> and C<T> are regular subexpressions).
2001 Such combinations can include alternatives, leading to a problem of choice:
2002 if we match a regular expression C<a|ab> against C<"abc">, will it match
2003 substring C<"a"> or C<"ab">? One way to describe which substring is
2004 actually matched is the concept of backtracking (see L<"Backtracking">).
2005 However, this description is too low-level and makes you think
2006 in terms of a particular implementation.
2008 Another description starts with notions of "better"/"worse". All the
2009 substrings which may be matched by the given regular expression can be
2010 sorted from the "best" match to the "worst" match, and it is the "best"
2011 match which is chosen. This substitutes the question of "what is chosen?"
2012 by the question of "which matches are better, and which are worse?".
2014 Again, for elementary pieces there is no such question, since at most
2015 one match at a given position is possible. This section describes the
2016 notion of better/worse for combining operators. In the description
2017 below C<S> and C<T> are regular subexpressions.
2023 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
2024 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
2025 which can be matched by C<T>.
2027 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
2030 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
2031 C<B> is better match for C<T> than C<B'>.
2035 When C<S> can match, it is a better match than when only C<T> can match.
2037 Ordering of two matches for C<S> is the same as for C<S>. Similar for
2038 two matches for C<T>.
2040 =item C<S{REPEAT_COUNT}>
2042 Matches as C<SSS...S> (repeated as many times as necessary).
2046 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
2048 =item C<S{min,max}?>
2050 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
2052 =item C<S?>, C<S*>, C<S+>
2054 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
2056 =item C<S??>, C<S*?>, C<S+?>
2058 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
2062 Matches the best match for C<S> and only that.
2064 =item C<(?=S)>, C<(?<=S)>
2066 Only the best match for C<S> is considered. (This is important only if
2067 C<S> has capturing parentheses, and backreferences are used somewhere
2068 else in the whole regular expression.)
2070 =item C<(?!S)>, C<(?<!S)>
2072 For this grouping operator there is no need to describe the ordering, since
2073 only whether or not C<S> can match is important.
2075 =item C<(??{ EXPR })>, C<(?PARNO)>
2077 The ordering is the same as for the regular expression which is
2078 the result of EXPR, or the pattern contained by capture group PARNO.
2080 =item C<(?(condition)yes-pattern|no-pattern)>
2082 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
2083 already determined. The ordering of the matches is the same as for the
2084 chosen subexpression.
2088 The above recipes describe the ordering of matches I<at a given position>.
2089 One more rule is needed to understand how a match is determined for the
2090 whole regular expression: a match at an earlier position is always better
2091 than a match at a later position.
2093 =head2 Creating Custom RE Engines
2095 Overloaded constants (see L<overload>) provide a simple way to extend
2096 the functionality of the RE engine.
2098 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
2099 matches at a boundary between whitespace characters and non-whitespace
2100 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
2101 at these positions, so we want to have each C<\Y|> in the place of the
2102 more complicated version. We can create a module C<customre> to do
2110 die "No argument to customre::import allowed" if @_;
2111 overload::constant 'qr' => \&convert;
2114 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
2116 # We must also take care of not escaping the legitimate \\Y|
2117 # sequence, hence the presence of '\\' in the conversion rules.
2118 my %rules = ( '\\' => '\\\\',
2119 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
2125 { $rules{$1} or invalid($re,$1) }sgex;
2129 Now C<use customre> enables the new escape in constant regular
2130 expressions, i.e., those without any runtime variable interpolations.
2131 As documented in L<overload>, this conversion will work only over
2132 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
2133 part of this regular expression needs to be converted explicitly
2134 (but only if the special meaning of C<\Y|> should be enabled inside $re):
2139 $re = customre::convert $re;
2142 =head1 PCRE/Python Support
2144 As of Perl 5.10.0, Perl supports several Python/PCRE specific extensions
2145 to the regex syntax. While Perl programmers are encouraged to use the
2146 Perl specific syntax, the following are also accepted:
2150 =item C<< (?PE<lt>NAMEE<gt>pattern) >>
2152 Define a named capture group. Equivalent to C<< (?<NAME>pattern) >>.
2154 =item C<< (?P=NAME) >>
2156 Backreference to a named capture group. Equivalent to C<< \g{NAME} >>.
2158 =item C<< (?P>NAME) >>
2160 Subroutine call to a named capture group. Equivalent to C<< (?&NAME) >>.
2166 There are numerous problems with case insensitive matching of characters
2167 outside the ASCII range, especially with those whose folds are multiple
2168 characters, such as ligatures like C<LATIN SMALL LIGATURE FF>.
2170 In a bracketed character class with case insensitive matching, ranges only work
2171 for ASCII characters. For example,
2172 C<m/[\N{CYRILLIC CAPITAL LETTER A}-\N{CYRILLIC CAPITAL LETTER YA}]/i>
2173 doesn't match all the Russian upper and lower case letters.
2175 Many regular expression constructs don't work on EBCDIC platforms.
2177 This document varies from difficult to understand to completely
2178 and utterly opaque. The wandering prose riddled with jargon is
2179 hard to fathom in several places.
2181 This document needs a rewrite that separates the tutorial content
2182 from the reference content.
2190 L<perlop/"Regexp Quote-Like Operators">.
2192 L<perlop/"Gory details of parsing quoted constructs">.
2202 I<Mastering Regular Expressions> by Jeffrey Friedl, published
2203 by O'Reilly and Associates.