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. The modifiers C</imsx>
81 may also be embedded within the regular expression itself using
82 the C<(?...)> construct. Also are new (in 5.14) character set semantics
83 modifiers B<C<<"a">>, B<C<"d">>, B<C<"l">> and B<C<"u">>, which, in 5.14
84 only, must be used embedded in the regular expression, and not after the
85 trailing delimiter. All this is discussed below in
86 L</Extended Patterns>.
87 X</a> X</d> X</l> X</u>
89 The C</x> modifier itself needs a little more explanation. It tells
90 the regular expression parser to ignore most whitespace that is neither
91 backslashed nor within a character class. You can use this to break up
92 your regular expression into (slightly) more readable parts. The C<#>
93 character is also treated as a metacharacter introducing a comment,
94 just as in ordinary Perl code. This also means that if you want real
95 whitespace or C<#> characters in the pattern (outside a character
96 class, where they are unaffected by C</x>), then you'll either have to
97 escape them (using backslashes or C<\Q...\E>) or encode them using octal,
98 hex, or C<\N{}> escapes. Taken together, these features go a long way towards
99 making Perl's regular expressions more readable. Note that you have to
100 be careful not to include the pattern delimiter in the comment--perl has
101 no way of knowing you did not intend to close the pattern early. See
102 the C-comment deletion code in L<perlop>. Also note that anything inside
103 a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect
104 whether space interpretation within a single multi-character construct. For
105 example in C<\x{...}>, regardless of the C</x> modifier, there can be no
106 spaces. Same for a L<quantifier|/Quantifiers> such as C<{3}> or
107 C<{5,}>. Similarly, C<(?:...)> can't have a space between the C<?> and C<:>,
108 but can between the C<(> and C<?>. Within any delimiters for such a
109 construct, allowed spaces are not affected by C</x>, and depend on the
110 construct. For example, C<\x{...}> can't have spaces because hexadecimal
111 numbers don't have spaces in them. But, Unicode properties can have spaces, so
112 in C<\p{...}> there can be spaces that follow the Unicode rules, for which see
113 L<perluniprops/Properties accessible through \p{} and \P{}>.
116 =head2 Regular Expressions
118 =head3 Metacharacters
120 The patterns used in Perl pattern matching evolved from those supplied in
121 the Version 8 regex routines. (The routines are derived
122 (distantly) from Henry Spencer's freely redistributable reimplementation
123 of the V8 routines.) See L<Version 8 Regular Expressions> for
126 In particular the following metacharacters have their standard I<egrep>-ish
129 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
132 \ Quote the next metacharacter
133 ^ Match the beginning of the line
134 . Match any character (except newline)
135 $ Match the end of the line (or before newline at the end)
138 [] Bracketed Character class
140 By default, the "^" character is guaranteed to match only the
141 beginning of the string, the "$" character only the end (or before the
142 newline at the end), and Perl does certain optimizations with the
143 assumption that the string contains only one line. Embedded newlines
144 will not be matched by "^" or "$". You may, however, wish to treat a
145 string as a multi-line buffer, such that the "^" will match after any
146 newline within the string (except if the newline is the last character in
147 the string), and "$" will match before any newline. At the
148 cost of a little more overhead, you can do this by using the /m modifier
149 on the pattern match operator. (Older programs did this by setting C<$*>,
150 but this practice has been removed in perl 5.9.)
153 To simplify multi-line substitutions, the "." character never matches a
154 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
155 the string is a single line--even if it isn't.
160 The following standard quantifiers are recognized:
161 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
163 * Match 0 or more times
164 + Match 1 or more times
166 {n} Match exactly n times
167 {n,} Match at least n times
168 {n,m} Match at least n but not more than m times
170 (If a curly bracket occurs in any other context, it is treated
171 as a regular character. In particular, the lower bound
172 is not optional.) The "*" quantifier is equivalent to C<{0,}>, the "+"
173 quantifier to C<{1,}>, and the "?" quantifier to C<{0,1}>. n and m are limited
174 to non-negative integral values less than a preset limit defined when perl is built.
175 This is usually 32766 on the most common platforms. The actual limit can
176 be seen in the error message generated by code such as this:
178 $_ **= $_ , / {$_} / for 2 .. 42;
180 By default, a quantified subpattern is "greedy", that is, it will match as
181 many times as possible (given a particular starting location) while still
182 allowing the rest of the pattern to match. If you want it to match the
183 minimum number of times possible, follow the quantifier with a "?". Note
184 that the meanings don't change, just the "greediness":
185 X<metacharacter> X<greedy> X<greediness>
186 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
188 *? Match 0 or more times, not greedily
189 +? Match 1 or more times, not greedily
190 ?? Match 0 or 1 time, not greedily
191 {n}? Match exactly n times, not greedily
192 {n,}? Match at least n times, not greedily
193 {n,m}? Match at least n but not more than m times, not greedily
195 By default, when a quantified subpattern does not allow the rest of the
196 overall pattern to match, Perl will backtrack. However, this behaviour is
197 sometimes undesirable. Thus Perl provides the "possessive" quantifier form
200 *+ Match 0 or more times and give nothing back
201 ++ Match 1 or more times and give nothing back
202 ?+ Match 0 or 1 time and give nothing back
203 {n}+ Match exactly n times and give nothing back (redundant)
204 {n,}+ Match at least n times and give nothing back
205 {n,m}+ Match at least n but not more than m times and give nothing back
211 will never match, as the C<a++> will gobble up all the C<a>'s in the
212 string and won't leave any for the remaining part of the pattern. This
213 feature can be extremely useful to give perl hints about where it
214 shouldn't backtrack. For instance, the typical "match a double-quoted
215 string" problem can be most efficiently performed when written as:
217 /"(?:[^"\\]++|\\.)*+"/
219 as we know that if the final quote does not match, backtracking will not
220 help. See the independent subexpression C<< (?>...) >> for more details;
221 possessive quantifiers are just syntactic sugar for that construct. For
222 instance the above example could also be written as follows:
224 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
226 =head3 Escape sequences
228 Because patterns are processed as double quoted strings, the following
235 \a alarm (bell) (BEL)
236 \e escape (think troff) (ESC)
237 \cK control char (example: VT)
238 \x{}, \x00 character whose ordinal is the given hexadecimal number
239 \N{name} named Unicode character or character sequence
240 \N{U+263D} Unicode character (example: FIRST QUARTER MOON)
241 \o{}, \000 character whose ordinal is the given octal number
242 \l lowercase next char (think vi)
243 \u uppercase next char (think vi)
244 \L lowercase till \E (think vi)
245 \U uppercase till \E (think vi)
246 \Q quote (disable) pattern metacharacters till \E
247 \E end either case modification or quoted section, think vi
249 Details are in L<perlop/Quote and Quote-like Operators>.
251 =head3 Character Classes and other Special Escapes
253 In addition, Perl defines the following:
254 X<\g> X<\k> X<\K> X<backreference>
256 Sequence Note Description
257 [...] [1] Match a character according to the rules of the
258 bracketed character class defined by the "...".
259 Example: [a-z] matches "a" or "b" or "c" ... or "z"
260 [[:...:]] [2] Match a character according to the rules of the POSIX
261 character class "..." within the outer bracketed
262 character class. Example: [[:upper:]] matches any
264 \w [3] Match a "word" character (alphanumeric plus "_", plus
265 other connector punctuation chars plus Unicode
267 \W [3] Match a non-"word" character
268 \s [3] Match a whitespace character
269 \S [3] Match a non-whitespace character
270 \d [3] Match a decimal digit character
271 \D [3] Match a non-digit character
272 \pP [3] Match P, named property. Use \p{Prop} for longer names
274 \X [4] Match Unicode "eXtended grapheme cluster"
275 \C Match a single C-language char (octet) even if that is
276 part of a larger UTF-8 character. Thus it breaks up
277 characters into their UTF-8 bytes, so you may end up
278 with malformed pieces of UTF-8. Unsupported in
280 \1 [5] Backreference to a specific capture group or buffer.
281 '1' may actually be any positive integer.
282 \g1 [5] Backreference to a specific or previous group,
283 \g{-1} [5] The number may be negative indicating a relative
284 previous group and may optionally be wrapped in
285 curly brackets for safer parsing.
286 \g{name} [5] Named backreference
287 \k<name> [5] Named backreference
288 \K [6] Keep the stuff left of the \K, don't include it in $&
289 \N [7] Any character but \n (experimental). Not affected by
291 \v [3] Vertical whitespace
292 \V [3] Not vertical whitespace
293 \h [3] Horizontal whitespace
294 \H [3] Not horizontal whitespace
301 See L<perlrecharclass/Bracketed Character Classes> for details.
305 See L<perlrecharclass/POSIX Character Classes> for details.
309 See L<perlrecharclass/Backslash sequences> for details.
313 See L<perlrebackslash/Misc> for details.
317 See L</Capture groups> below for details.
321 See L</Extended Patterns> below for details.
325 Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the
326 character or character sequence whose name is C<NAME>; and similarly
327 when of the form C<\N{U+I<hex>}>, it matches the character whose Unicode
328 code point is I<hex>. Otherwise it matches any character but C<\n>.
334 Perl defines the following zero-width assertions:
335 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
336 X<regexp, zero-width assertion>
337 X<regular expression, zero-width assertion>
338 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
340 \b Match a word boundary
341 \B Match except at a word boundary
342 \A Match only at beginning of string
343 \Z Match only at end of string, or before newline at the end
344 \z Match only at end of string
345 \G Match only at pos() (e.g. at the end-of-match position
348 A word boundary (C<\b>) is a spot between two characters
349 that has a C<\w> on one side of it and a C<\W> on the other side
350 of it (in either order), counting the imaginary characters off the
351 beginning and end of the string as matching a C<\W>. (Within
352 character classes C<\b> represents backspace rather than a word
353 boundary, just as it normally does in any double-quoted string.)
354 The C<\A> and C<\Z> are just like "^" and "$", except that they
355 won't match multiple times when the C</m> modifier is used, while
356 "^" and "$" will match at every internal line boundary. To match
357 the actual end of the string and not ignore an optional trailing
359 X<\b> X<\A> X<\Z> X<\z> X</m>
361 The C<\G> assertion can be used to chain global matches (using
362 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
363 It is also useful when writing C<lex>-like scanners, when you have
364 several patterns that you want to match against consequent substrings
365 of your string, see the previous reference. The actual location
366 where C<\G> will match can also be influenced by using C<pos()> as
367 an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
368 matches is modified somewhat, in that contents to the left of C<\G> is
369 not counted when determining the length of the match. Thus the following
370 will not match forever:
375 while ($string =~ /(.\G)/g) {
379 It will print 'A' and then terminate, as it considers the match to
380 be zero-width, and thus will not match at the same position twice in a
383 It is worth noting that C<\G> improperly used can result in an infinite
384 loop. Take care when using patterns that include C<\G> in an alternation.
386 =head3 Capture groups
388 The bracketing construct C<( ... )> creates capture groups (also referred to as
389 capture buffers). To refer to the current contents of a group later on, within
390 the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>)
391 for the second, and so on.
392 This is called a I<backreference>.
393 X<regex, capture buffer> X<regexp, capture buffer>
394 X<regex, capture group> X<regexp, capture group>
395 X<regular expression, capture buffer> X<backreference>
396 X<regular expression, capture group> X<backreference>
397 X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
398 X<named capture buffer> X<regular expression, named capture buffer>
399 X<named capture group> X<regular expression, named capture group>
400 X<%+> X<$+{name}> X<< \k<name> >>
401 There is no limit to the number of captured substrings that you may use.
402 Groups are numbered with the leftmost open parenthesis being number 1, etc. If
403 a group did not match, the associated backreference won't match either. (This
404 can happen if the group is optional, or in a different branch of an
406 You can omit the C<"g">, and write C<"\1">, etc, but there are some issues with
407 this form, described below.
409 You can also refer to capture groups relatively, by using a negative number, so
410 that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture
411 group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For
418 \g{-1} # backref to group 3
419 \g{-3} # backref to group 1
423 would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to
424 interpolate regexes into larger regexes and not have to worry about the
425 capture groups being renumbered.
427 You can dispense with numbers altogether and create named capture groups.
428 The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to
429 reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may
430 also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.)
431 I<name> must not begin with a number, nor contain hyphens.
432 When different groups within the same pattern have the same name, any reference
433 to that name assumes the leftmost defined group. Named groups count in
434 absolute and relative numbering, and so can also be referred to by those
436 (It's possible to do things with named capture groups that would otherwise
439 Capture group contents are dynamically scoped and available to you outside the
440 pattern until the end of the enclosing block or until the next successful
441 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
442 You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">,
443 etc); or by name via the C<%+> hash, using C<"$+{I<name>}">.
445 Braces are required in referring to named capture groups, but are optional for
446 absolute or relative numbered ones. Braces are safer when creating a regex by
447 concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a>
448 contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which
449 is probably not what you intended.
451 The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that
452 there were no named nor relative numbered capture groups. Absolute numbered
453 groups were referred to using C<\1>, C<\2>, etc, and this notation is still
454 accepted (and likely always will be). But it leads to some ambiguities if
455 there are more than 9 capture groups, as C<\10> could mean either the tenth
456 capture group, or the character whose ordinal in octal is 010 (a backspace in
457 ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference
458 only if at least 10 left parentheses have opened before it. Likewise C<\11> is
459 a backreference only if at least 11 left parentheses have opened before it.
460 And so on. C<\1> through C<\9> are always interpreted as backreferences.
461 There are several examples below that illustrate these perils. You can avoid
462 the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups;
463 and for octal constants always using C<\o{}>, or for C<\077> and below, using 3
464 digits padded with leading zeros, since a leading zero implies an octal
467 The C<\I<digit>> notation also works in certain circumstances outside
468 the pattern. See L</Warning on \1 Instead of $1> below for details.)
472 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
474 /(.)\g1/ # find first doubled char
475 and print "'$1' is the first doubled character\n";
477 /(?<char>.)\k<char>/ # ... a different way
478 and print "'$+{char}' is the first doubled character\n";
480 /(?'char'.)\g1/ # ... mix and match
481 and print "'$1' is the first doubled character\n";
483 if (/Time: (..):(..):(..)/) { # parse out values
489 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference
490 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal
491 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference
492 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal
494 $a = '(.)\1'; # Creates problems when concatenated.
495 $b = '(.)\g{1}'; # Avoids the problems.
496 "aa" =~ /${a}/; # True
497 "aa" =~ /${b}/; # True
498 "aa0" =~ /${a}0/; # False!
499 "aa0" =~ /${b}0/; # True
500 "aa\x08" =~ /${a}0/; # True!
501 "aa\x08" =~ /${b}0/; # False
503 Several special variables also refer back to portions of the previous
504 match. C<$+> returns whatever the last bracket match matched.
505 C<$&> returns the entire matched string. (At one point C<$0> did
506 also, but now it returns the name of the program.) C<$`> returns
507 everything before the matched string. C<$'> returns everything
508 after the matched string. And C<$^N> contains whatever was matched by
509 the most-recently closed group (submatch). C<$^N> can be used in
510 extended patterns (see below), for example to assign a submatch to a
512 X<$+> X<$^N> X<$&> X<$`> X<$'>
514 These special variables, like the C<%+> hash and the numbered match variables
515 (C<$1>, C<$2>, C<$3>, etc.) are dynamically scoped
516 until the end of the enclosing block or until the next successful
517 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
518 X<$+> X<$^N> X<$&> X<$`> X<$'>
519 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
521 B<NOTE>: Failed matches in Perl do not reset the match variables,
522 which makes it easier to write code that tests for a series of more
523 specific cases and remembers the best match.
525 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
526 C<$'> anywhere in the program, it has to provide them for every
527 pattern match. This may substantially slow your program. Perl
528 uses the same mechanism to produce C<$1>, C<$2>, etc, so you also pay a
529 price for each pattern that contains capturing parentheses. (To
530 avoid this cost while retaining the grouping behaviour, use the
531 extended regular expression C<(?: ... )> instead.) But if you never
532 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
533 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
534 if you can, but if you can't (and some algorithms really appreciate
535 them), once you've used them once, use them at will, because you've
536 already paid the price. As of 5.005, C<$&> is not so costly as the
540 As a workaround for this problem, Perl 5.10.0 introduces C<${^PREMATCH}>,
541 C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&>
542 and C<$'>, B<except> that they are only guaranteed to be defined after a
543 successful match that was executed with the C</p> (preserve) modifier.
544 The use of these variables incurs no global performance penalty, unlike
545 their punctuation char equivalents, however at the trade-off that you
546 have to tell perl when you want to use them.
549 =head2 Quoting metacharacters
551 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
552 C<\w>, C<\n>. Unlike some other regular expression languages, there
553 are no backslashed symbols that aren't alphanumeric. So anything
554 that looks like \\, \(, \), \<, \>, \{, or \} is always
555 interpreted as a literal character, not a metacharacter. This was
556 once used in a common idiom to disable or quote the special meanings
557 of regular expression metacharacters in a string that you want to
558 use for a pattern. Simply quote all non-"word" characters:
560 $pattern =~ s/(\W)/\\$1/g;
562 (If C<use locale> is set, then this depends on the current locale.)
563 Today it is more common to use the quotemeta() function or the C<\Q>
564 metaquoting escape sequence to disable all metacharacters' special
567 /$unquoted\Q$quoted\E$unquoted/
569 Beware that if you put literal backslashes (those not inside
570 interpolated variables) between C<\Q> and C<\E>, double-quotish
571 backslash interpolation may lead to confusing results. If you
572 I<need> to use literal backslashes within C<\Q...\E>,
573 consult L<perlop/"Gory details of parsing quoted constructs">.
575 =head2 Extended Patterns
577 Perl also defines a consistent extension syntax for features not
578 found in standard tools like B<awk> and B<lex>. The syntax is a
579 pair of parentheses with a question mark as the first thing within
580 the parentheses. The character after the question mark indicates
583 The stability of these extensions varies widely. Some have been
584 part of the core language for many years. Others are experimental
585 and may change without warning or be completely removed. Check
586 the documentation on an individual feature to verify its current
589 A question mark was chosen for this and for the minimal-matching
590 construct because 1) question marks are rare in older regular
591 expressions, and 2) whenever you see one, you should stop and
592 "question" exactly what is going on. That's psychology...
599 A comment. The text is ignored. If the C</x> modifier enables
600 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
601 the comment as soon as it sees a C<)>, so there is no way to put a literal
604 =item C<(?adlupimsx-imsx)>
606 =item C<(?^alupimsx)>
609 One or more embedded pattern-match modifiers, to be turned on (or
610 turned off, if preceded by C<->) for the remainder of the pattern or
611 the remainder of the enclosing pattern group (if any).
613 This is particularly useful for dynamic patterns, such as those read in from a
614 configuration file, taken from an argument, or specified in a table
615 somewhere. Consider the case where some patterns want to be case
616 sensitive and some do not: The case insensitive ones merely need to
617 include C<(?i)> at the front of the pattern. For example:
620 if ( /$pattern/i ) { }
624 $pattern = "(?i)foobar";
625 if ( /$pattern/ ) { }
627 These modifiers are restored at the end of the enclosing group. For example,
629 ( (?i) blah ) \s+ \g1
631 will match C<blah> in any case, some spaces, and an exact (I<including the case>!)
632 repetition of the previous word, assuming the C</x> modifier, and no C</i>
633 modifier outside this group.
635 These modifiers do not carry over into named subpatterns called in the
636 enclosing group. In other words, a pattern such as C<((?i)(&NAME))> does not
637 change the case-sensitivity of the "NAME" pattern.
639 Any of these modifiers can be set to apply globally to all regular
640 expressions compiled within the scope of a C<use re>. See
641 L<re/"'/flags' mode">.
643 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
644 after the C<"?"> is a shorthand equivalent to C<d-imsx>. Flags (except
645 C<"d">) may follow the caret to override it.
646 But a minus sign is not legal with it.
648 Also, starting in Perl 5.14, are modifiers C<"a">, C<"d">, C<"l">, and
649 C<"u">, which for 5.14 may not be used as suffix modifiers.
651 C<"l"> means to use a locale (see L<perllocale>) when pattern matching.
652 The locale used will be the one in effect at the time of execution of
653 the pattern match. This may not be the same as the compilation-time
654 locale, and can differ from one match to another if there is an
655 intervening call of the
656 L<setlocale() function|perllocale/The setlocale function>.
657 This modifier is automatically set if the regular expression is compiled
658 within the scope of a C<"use locale"> pragma. Results are not
659 well-defined when using this and matching against a utf8-encoded string.
661 C<"u"> means to use Unicode semantics when pattern matching. It is
662 automatically set if the regular expression is encoded in utf8, or is
663 compiled within the scope of a
664 L<C<"use feature 'unicode_strings">|feature> pragma (and isn't also in
665 the scope of L<C<"use locale">|locale> nor L<C<"use bytes">|bytes>
666 pragmas. On ASCII platforms, the code points between 128 and 255 take on their
667 Latin-1 (ISO-8859-1) meanings (which are the same as Unicode's), whereas
668 in strict ASCII their meanings are undefined. Thus the platform
669 effectively becomes a Unicode platform. The ASCII characters remain as
670 ASCII characters (since ASCII is a subset of Latin-1 and Unicode). For
671 example, when this option is not on, on a non-utf8 string, C<"\w">
672 matches precisely C<[A-Za-z0-9_]>. When the option is on, it matches
673 not just those, but all the Latin-1 word characters (such as an "n" with
674 a tilde). On EBCDIC platforms, which already are equivalent to Latin-1,
675 this modifier changes behavior only when the C<"/i"> modifier is also
676 specified, and affects only two characters, giving them full Unicode
677 semantics: the C<MICRO SIGN> will match the Greek capital and
678 small letters C<MU>; otherwise not; and the C<LATIN CAPITAL LETTER SHARP
679 S> will match any of C<SS>, C<Ss>, C<sS>, and C<ss>, otherwise not.
680 (This last case is buggy, however.)
682 C<"a"> is the same as C<"u">, except that C<\d>, C<\s>, C<\w>, and the
683 Posix character classes are restricted to matching in the ASCII range
684 only. That is, with this modifier, C<\d> always means precisely the
685 digits C<"0"> to C<"9">; C<\s> means the five characters C<[ \f\n\r\t]>;
686 C<\w> means the 63 characters C<[A-Za-z0-9_]>; and likewise, all the
687 Posix classes such as C<[[:print:]]> match only the appropriate
688 ASCII-range characters. As you would expect, this modifier causes, for
689 example, C<\D> to mean the same thing as C<[^0-9]>; in fact, all
690 non-ASCII characters match C<\D>, C<\S>, and C<\W>. C<\b> still means
691 to match at the boundary between C<\w> and C<\W>, using the C<"a">
692 definitions of them (similarly for C<\B>). Otherwise, C<"a"> behaves
693 like the C<"u"> modifier, in that case-insensitive matching uses Unicode
694 semantics; for example, "k" will match the Unicode C<\N{KELVIN SIGN}>
695 under C</i> matching, and code points in the Latin1 range, above ASCII
696 will have Unicode semantics when it comes to case-insensitive matching.
697 But writing two in "a"'s in a row will increase its effect, causing the
698 Kelvin sign and all other non-ASCII characters to not match any ASCII
699 character under C</i> matching.
701 C<"d"> means to use the traditional Perl pattern matching behavior.
702 This is dualistic (hence the name C<"d">, which also could stand for
703 "depends"). When this is in effect, Perl matches according to the
704 platform's native character set rules unless there is something that
705 indicates to use Unicode rules. If either the target string or the
706 pattern itself is encoded in UTF-8, Unicode rules are used. Also, if
707 the pattern contains Unicode-only features, such as code points above
708 255, C<\p()> Unicode properties or C<\N{}> Unicode names, Unicode rules
709 will be used. It is automatically selected by default if the regular
710 expression is compiled neither within the scope of a C<"use locale">
711 pragma nor a <C<"use feature 'unicode_strings"> pragma.
712 This behavior causes a number of glitches, see
713 L<perlunicode/The "Unicode Bug">.
715 Note that the C<a>, C<d>, C<l>, C<p>, and C<u> modifiers are special in
716 that they can only be enabled, not disabled, and the C<a>, C<d>, C<l>, and
717 C<u> modifiers are mutually exclusive: specifying one de-specifies the
718 others, and a maximum of one may appear in the construct. Thus, for
719 example, C<(?-p)>, C<(?-d:...)>, and C<(?dl:...)> will warn when
720 compiled under C<use warnings>.
722 Note also that the C<p> modifier is special in that its presence
723 anywhere in a pattern has a global effect.
728 =item C<(?adluimsx-imsx:pattern)>
730 =item C<(?^aluimsx:pattern)>
733 This is for clustering, not capturing; it groups subexpressions like
734 "()", but doesn't make backreferences as "()" does. So
736 @fields = split(/\b(?:a|b|c)\b/)
740 @fields = split(/\b(a|b|c)\b/)
742 but doesn't spit out extra fields. It's also cheaper not to capture
743 characters if you don't need to.
745 Any letters between C<?> and C<:> act as flags modifiers as with
746 C<(?adluimsx-imsx)>. For example,
748 /(?s-i:more.*than).*million/i
750 is equivalent to the more verbose
752 /(?:(?s-i)more.*than).*million/i
754 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
755 after the C<"?"> is a shorthand equivalent to C<d-imsx>. Any positive
756 flags (except C<"d">) may follow the caret, so
764 The caret tells Perl that this cluster doesn't inherit the flags of any
765 surrounding pattern, but to go back to the system defaults (C<d-imsx>),
766 modified by any flags specified.
768 The caret allows for simpler stringification of compiled regular
769 expressions. These look like
773 with any non-default flags appearing between the caret and the colon.
774 A test that looks at such stringification thus doesn't need to have the
775 system default flags hard-coded in it, just the caret. If new flags are
776 added to Perl, the meaning of the caret's expansion will change to include
777 the default for those flags, so the test will still work, unchanged.
779 Specifying a negative flag after the caret is an error, as the flag is
782 Mnemonic for C<(?^...)>: A fresh beginning since the usual use of a caret is
783 to match at the beginning.
786 X<(?|)> X<Branch reset>
788 This is the "branch reset" pattern, which has the special property
789 that the capture groups are numbered from the same starting point
790 in each alternation branch. It is available starting from perl 5.10.0.
792 Capture groups are numbered from left to right, but inside this
793 construct the numbering is restarted for each branch.
795 The numbering within each branch will be as normal, and any groups
796 following this construct will be numbered as though the construct
797 contained only one branch, that being the one with the most capture
800 This construct will be useful when you want to capture one of a
801 number of alternative matches.
803 Consider the following pattern. The numbers underneath show in
804 which group the captured content will be stored.
807 # before ---------------branch-reset----------- after
808 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
811 Be careful when using the branch reset pattern in combination with
812 named captures. Named captures are implemented as being aliases to
813 numbered groups holding the captures, and that interferes with the
814 implementation of the branch reset pattern. If you are using named
815 captures in a branch reset pattern, it's best to use the same names,
816 in the same order, in each of the alternations:
818 /(?| (?<a> x ) (?<b> y )
819 | (?<a> z ) (?<b> w )) /x
821 Not doing so may lead to surprises:
823 "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x;
824 say $+ {a}; # Prints '12'
825 say $+ {b}; # *Also* prints '12'.
827 The problem here is that both the group named C<< a >> and the group
828 named C<< b >> are aliases for the group belonging to C<< $1 >>.
830 =item Look-Around Assertions
831 X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
833 Look-around assertions are zero width patterns which match a specific
834 pattern without including it in C<$&>. Positive assertions match when
835 their subpattern matches, negative assertions match when their subpattern
836 fails. Look-behind matches text up to the current match position,
837 look-ahead matches text following the current match position.
842 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
844 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
845 matches a word followed by a tab, without including the tab in C<$&>.
848 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
850 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
851 matches any occurrence of "foo" that isn't followed by "bar". Note
852 however that look-ahead and look-behind are NOT the same thing. You cannot
853 use this for look-behind.
855 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
856 will not do what you want. That's because the C<(?!foo)> is just saying that
857 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
858 match. You would have to do something like C</(?!foo)...bar/> for that. We
859 say "like" because there's the case of your "bar" not having three characters
860 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
861 Sometimes it's still easier just to say:
863 if (/bar/ && $` !~ /foo$/)
865 For look-behind see below.
867 =item C<(?<=pattern)> C<\K>
868 X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K>
870 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
871 matches a word that follows a tab, without including the tab in C<$&>.
872 Works only for fixed-width look-behind.
874 There is a special form of this construct, called C<\K>, which causes the
875 regex engine to "keep" everything it had matched prior to the C<\K> and
876 not include it in C<$&>. This effectively provides variable length
877 look-behind. The use of C<\K> inside of another look-around assertion
878 is allowed, but the behaviour is currently not well defined.
880 For various reasons C<\K> may be significantly more efficient than the
881 equivalent C<< (?<=...) >> construct, and it is especially useful in
882 situations where you want to efficiently remove something following
883 something else in a string. For instance
887 can be rewritten as the much more efficient
891 =item C<(?<!pattern)>
892 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
894 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
895 matches any occurrence of "foo" that does not follow "bar". Works
896 only for fixed-width look-behind.
900 =item C<(?'NAME'pattern)>
902 =item C<< (?<NAME>pattern) >>
903 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
905 A named capture group. Identical in every respect to normal capturing
906 parentheses C<()> but for the additional fact that C<%+> or C<%-> may be
907 used after a successful match to refer to a named group. See L<perlvar>
908 for more details on the C<%+> and C<%-> hashes.
910 If multiple distinct capture groups have the same name then the
911 $+{NAME} will refer to the leftmost defined group in the match.
913 The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent.
915 B<NOTE:> While the notation of this construct is the same as the similar
916 function in .NET regexes, the behavior is not. In Perl the groups are
917 numbered sequentially regardless of being named or not. Thus in the
922 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
923 the opposite which is what a .NET regex hacker might expect.
925 Currently NAME is restricted to simple identifiers only.
926 In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
927 its Unicode extension (see L<utf8>),
928 though it isn't extended by the locale (see L<perllocale>).
930 B<NOTE:> In order to make things easier for programmers with experience
931 with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >>
932 may be used instead of C<< (?<NAME>pattern) >>; however this form does not
933 support the use of single quotes as a delimiter for the name.
935 =item C<< \k<NAME> >>
937 =item C<< \k'NAME' >>
939 Named backreference. Similar to numeric backreferences, except that
940 the group is designated by name and not number. If multiple groups
941 have the same name then it refers to the leftmost defined group in
944 It is an error to refer to a name not defined by a C<< (?<NAME>) >>
945 earlier in the pattern.
947 Both forms are equivalent.
949 B<NOTE:> In order to make things easier for programmers with experience
950 with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >>
951 may be used instead of C<< \k<NAME> >>.
954 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
956 B<WARNING>: This extended regular expression feature is considered
957 experimental, and may be changed without notice. Code executed that
958 has side effects may not perform identically from version to version
959 due to the effect of future optimisations in the regex engine.
961 This zero-width assertion evaluates any embedded Perl code. It
962 always succeeds, and its C<code> is not interpolated. Currently,
963 the rules to determine where the C<code> ends are somewhat convoluted.
965 This feature can be used together with the special variable C<$^N> to
966 capture the results of submatches in variables without having to keep
967 track of the number of nested parentheses. For example:
969 $_ = "The brown fox jumps over the lazy dog";
970 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
971 print "color = $color, animal = $animal\n";
973 Inside the C<(?{...})> block, C<$_> refers to the string the regular
974 expression is matching against. You can also use C<pos()> to know what is
975 the current position of matching within this string.
977 The C<code> is properly scoped in the following sense: If the assertion
978 is backtracked (compare L<"Backtracking">), all changes introduced after
979 C<local>ization are undone, so that
983 (?{ $cnt = 0 }) # Initialize $cnt.
987 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
991 (?{ $res = $cnt }) # On success copy to
992 # non-localized location.
995 will set C<$res = 4>. Note that after the match, C<$cnt> returns to the globally
996 introduced value, because the scopes that restrict C<local> operators
999 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
1000 switch. If I<not> used in this way, the result of evaluation of
1001 C<code> is put into the special variable C<$^R>. This happens
1002 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
1003 inside the same regular expression.
1005 The assignment to C<$^R> above is properly localized, so the old
1006 value of C<$^R> is restored if the assertion is backtracked; compare
1009 For reasons of security, this construct is forbidden if the regular
1010 expression involves run-time interpolation of variables, unless the
1011 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
1012 variables contain results of C<qr//> operator (see
1013 L<perlop/"qr/STRINGE<sol>msixpo">).
1015 This restriction is due to the wide-spread and remarkably convenient
1016 custom of using run-time determined strings as patterns. For example:
1022 Before Perl knew how to execute interpolated code within a pattern,
1023 this operation was completely safe from a security point of view,
1024 although it could raise an exception from an illegal pattern. If
1025 you turn on the C<use re 'eval'>, though, it is no longer secure,
1026 so you should only do so if you are also using taint checking.
1027 Better yet, use the carefully constrained evaluation within a Safe
1028 compartment. See L<perlsec> for details about both these mechanisms.
1030 B<WARNING>: Use of lexical (C<my>) variables in these blocks is
1031 broken. The result is unpredictable and will make perl unstable. The
1032 workaround is to use global (C<our>) variables.
1034 B<WARNING>: In perl 5.12.x and earlier, the regex engine
1035 was not re-entrant, so interpolated code could not
1036 safely invoke the regex engine either directly with
1037 C<m//> or C<s///>), or indirectly with functions such as
1038 C<split>. Invoking the regex engine in these blocks would make perl
1041 =item C<(??{ code })>
1043 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
1045 B<WARNING>: This extended regular expression feature is considered
1046 experimental, and may be changed without notice. Code executed that
1047 has side effects may not perform identically from version to version
1048 due to the effect of future optimisations in the regex engine.
1050 This is a "postponed" regular subexpression. The C<code> is evaluated
1051 at run time, at the moment this subexpression may match. The result
1052 of evaluation is considered as a regular expression and matched as
1053 if it were inserted instead of this construct. Note that this means
1054 that the contents of capture groups defined inside an eval'ed pattern
1055 are not available outside of the pattern, and vice versa, there is no
1056 way for the inner pattern to refer to a capture group defined outside.
1059 ('a' x 100)=~/(??{'(.)' x 100})/
1061 B<will> match, it will B<not> set $1.
1063 The C<code> is not interpolated. As before, the rules to determine
1064 where the C<code> ends are currently somewhat convoluted.
1066 The following pattern matches a parenthesized group:
1071 (?> [^()]+ ) # Non-parens without backtracking
1073 (??{ $re }) # Group with matching parens
1078 See also C<(?PARNO)> for a different, more efficient way to accomplish
1081 For reasons of security, this construct is forbidden if the regular
1082 expression involves run-time interpolation of variables, unless the
1083 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
1084 variables contain results of C<qr//> operator (see
1085 L<perlop/"qrE<sol>STRINGE<sol>msixpo">).
1087 In perl 5.12.x and earlier, because the regex engine was not re-entrant,
1088 delayed code could not safely invoke the regex engine either directly with
1089 C<m//> or C<s///>), or indirectly with functions such as C<split>.
1091 Recursing deeper than 50 times without consuming any input string will
1092 result in a fatal error. The maximum depth is compiled into perl, so
1093 changing it requires a custom build.
1095 =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)>
1096 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
1097 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
1098 X<regex, relative recursion>
1100 Similar to C<(??{ code })> except it does not involve compiling any code,
1101 instead it treats the contents of a capture group as an independent
1102 pattern that must match at the current position. Capture groups
1103 contained by the pattern will have the value as determined by the
1104 outermost recursion.
1106 PARNO is a sequence of digits (not starting with 0) whose value reflects
1107 the paren-number of the capture group to recurse to. C<(?R)> recurses to
1108 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
1109 C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed
1110 to be relative, with negative numbers indicating preceding capture groups
1111 and positive ones following. Thus C<(?-1)> refers to the most recently
1112 declared group, and C<(?+1)> indicates the next group to be declared.
1113 Note that the counting for relative recursion differs from that of
1114 relative backreferences, in that with recursion unclosed groups B<are>
1117 The following pattern matches a function foo() which may contain
1118 balanced parentheses as the argument.
1120 $re = qr{ ( # paren group 1 (full function)
1122 ( # paren group 2 (parens)
1124 ( # paren group 3 (contents of parens)
1126 (?> [^()]+ ) # Non-parens without backtracking
1128 (?2) # Recurse to start of paren group 2
1136 If the pattern was used as follows
1138 'foo(bar(baz)+baz(bop))'=~/$re/
1139 and print "\$1 = $1\n",
1143 the output produced should be the following:
1145 $1 = foo(bar(baz)+baz(bop))
1146 $2 = (bar(baz)+baz(bop))
1147 $3 = bar(baz)+baz(bop)
1149 If there is no corresponding capture group defined, then it is a
1150 fatal error. Recursing deeper than 50 times without consuming any input
1151 string will also result in a fatal error. The maximum depth is compiled
1152 into perl, so changing it requires a custom build.
1154 The following shows how using negative indexing can make it
1155 easier to embed recursive patterns inside of a C<qr//> construct
1158 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
1159 if (/foo $parens \s+ + \s+ bar $parens/x) {
1160 # do something here...
1163 B<Note> that this pattern does not behave the same way as the equivalent
1164 PCRE or Python construct of the same form. In Perl you can backtrack into
1165 a recursed group, in PCRE and Python the recursed into group is treated
1166 as atomic. Also, modifiers are resolved at compile time, so constructs
1167 like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
1173 Recurse to a named subpattern. Identical to C<(?PARNO)> except that the
1174 parenthesis to recurse to is determined by name. If multiple parentheses have
1175 the same name, then it recurses to the leftmost.
1177 It is an error to refer to a name that is not declared somewhere in the
1180 B<NOTE:> In order to make things easier for programmers with experience
1181 with the Python or PCRE regex engines the pattern C<< (?P>NAME) >>
1182 may be used instead of C<< (?&NAME) >>.
1184 =item C<(?(condition)yes-pattern|no-pattern)>
1187 =item C<(?(condition)yes-pattern)>
1189 Conditional expression. C<(condition)> should be either an integer in
1190 parentheses (which is valid if the corresponding pair of parentheses
1191 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
1192 name in angle brackets or single quotes (which is valid if a group
1193 with the given name matched), or the special symbol (R) (true when
1194 evaluated inside of recursion or eval). Additionally the R may be
1195 followed by a number, (which will be true when evaluated when recursing
1196 inside of the appropriate group), or by C<&NAME>, in which case it will
1197 be true only when evaluated during recursion in the named group.
1199 Here's a summary of the possible predicates:
1205 Checks if the numbered capturing group has matched something.
1207 =item (<NAME>) ('NAME')
1209 Checks if a group with the given name has matched something.
1211 =item (?=...) (?!...) (?<=...) (?<!...)
1213 Checks whether the pattern matches (or does not match, for the '!'
1218 Treats the return value of the code block as the condition.
1222 Checks if the expression has been evaluated inside of recursion.
1226 Checks if the expression has been evaluated while executing directly
1227 inside of the n-th capture group. This check is the regex equivalent of
1229 if ((caller(0))[3] eq 'subname') { ... }
1231 In other words, it does not check the full recursion stack.
1235 Similar to C<(R1)>, this predicate checks to see if we're executing
1236 directly inside of the leftmost group with a given name (this is the same
1237 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1238 stack, but only the name of the innermost active recursion.
1242 In this case, the yes-pattern is never directly executed, and no
1243 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1244 See below for details.
1255 matches a chunk of non-parentheses, possibly included in parentheses
1258 A special form is the C<(DEFINE)> predicate, which never executes directly
1259 its yes-pattern, and does not allow a no-pattern. This allows to define
1260 subpatterns which will be executed only by using the recursion mechanism.
1261 This way, you can define a set of regular expression rules that can be
1262 bundled into any pattern you choose.
1264 It is recommended that for this usage you put the DEFINE block at the
1265 end of the pattern, and that you name any subpatterns defined within it.
1267 Also, it's worth noting that patterns defined this way probably will
1268 not be as efficient, as the optimiser is not very clever about
1271 An example of how this might be used is as follows:
1273 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
1279 Note that capture groups matched inside of recursion are not accessible
1280 after the recursion returns, so the extra layer of capturing groups is
1281 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1282 C<$+{NAME}> would be.
1284 =item C<< (?>pattern) >>
1285 X<backtrack> X<backtracking> X<atomic> X<possessive>
1287 An "independent" subexpression, one which matches the substring
1288 that a I<standalone> C<pattern> would match if anchored at the given
1289 position, and it matches I<nothing other than this substring>. This
1290 construct is useful for optimizations of what would otherwise be
1291 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1292 It may also be useful in places where the "grab all you can, and do not
1293 give anything back" semantic is desirable.
1295 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1296 (anchored at the beginning of string, as above) will match I<all>
1297 characters C<a> at the beginning of string, leaving no C<a> for
1298 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1299 since the match of the subgroup C<a*> is influenced by the following
1300 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1301 C<a*ab> will match fewer characters than a standalone C<a*>, since
1302 this makes the tail match.
1304 An effect similar to C<< (?>pattern) >> may be achieved by writing
1305 C<(?=(pattern))\g1>. This matches the same substring as a standalone
1306 C<a+>, and the following C<\g1> eats the matched string; it therefore
1307 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1308 (The difference between these two constructs is that the second one
1309 uses a capturing group, thus shifting ordinals of backreferences
1310 in the rest of a regular expression.)
1312 Consider this pattern:
1323 That will efficiently match a nonempty group with matching parentheses
1324 two levels deep or less. However, if there is no such group, it
1325 will take virtually forever on a long string. That's because there
1326 are so many different ways to split a long string into several
1327 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1328 to a subpattern of the above pattern. Consider how the pattern
1329 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1330 seconds, but that each extra letter doubles this time. This
1331 exponential performance will make it appear that your program has
1332 hung. However, a tiny change to this pattern
1336 (?> [^()]+ ) # change x+ above to (?> x+ )
1343 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1344 this yourself would be a productive exercise), but finishes in a fourth
1345 the time when used on a similar string with 1000000 C<a>s. Be aware,
1346 however, that this pattern currently triggers a warning message under
1347 the C<use warnings> pragma or B<-w> switch saying it
1348 C<"matches null string many times in regex">.
1350 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1351 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1352 This was only 4 times slower on a string with 1000000 C<a>s.
1354 The "grab all you can, and do not give anything back" semantic is desirable
1355 in many situations where on the first sight a simple C<()*> looks like
1356 the correct solution. Suppose we parse text with comments being delimited
1357 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1358 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1359 the comment delimiter, because it may "give up" some whitespace if
1360 the remainder of the pattern can be made to match that way. The correct
1361 answer is either one of these:
1366 For example, to grab non-empty comments into $1, one should use either
1369 / (?> \# [ \t]* ) ( .+ ) /x;
1370 / \# [ \t]* ( [^ \t] .* ) /x;
1372 Which one you pick depends on which of these expressions better reflects
1373 the above specification of comments.
1375 In some literature this construct is called "atomic matching" or
1376 "possessive matching".
1378 Possessive quantifiers are equivalent to putting the item they are applied
1379 to inside of one of these constructs. The following equivalences apply:
1381 Quantifier Form Bracketing Form
1382 --------------- ---------------
1386 PAT{min,max}+ (?>PAT{min,max})
1390 =head2 Special Backtracking Control Verbs
1392 B<WARNING:> These patterns are experimental and subject to change or
1393 removal in a future version of Perl. Their usage in production code should
1394 be noted to avoid problems during upgrades.
1396 These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1397 otherwise stated the ARG argument is optional; in some cases, it is
1400 Any pattern containing a special backtracking verb that allows an argument
1401 has the special behaviour that when executed it sets the current package's
1402 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1405 On failure, the C<$REGERROR> variable will be set to the ARG value of the
1406 verb pattern, if the verb was involved in the failure of the match. If the
1407 ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1408 name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1409 none. Also, the C<$REGMARK> variable will be set to FALSE.
1411 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1412 the C<$REGMARK> variable will be set to the name of the last
1413 C<(*MARK:NAME)> pattern executed. See the explanation for the
1414 C<(*MARK:NAME)> verb below for more details.
1416 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1417 and most other regex related variables. They are not local to a scope, nor
1418 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1419 Use C<local> to localize changes to them to a specific scope if necessary.
1421 If a pattern does not contain a special backtracking verb that allows an
1422 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
1426 =item Verbs that take an argument
1430 =item C<(*PRUNE)> C<(*PRUNE:NAME)>
1431 X<(*PRUNE)> X<(*PRUNE:NAME)>
1433 This zero-width pattern prunes the backtracking tree at the current point
1434 when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1435 where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1436 A may backtrack as necessary to match. Once it is reached, matching
1437 continues in B, which may also backtrack as necessary; however, should B
1438 not match, then no further backtracking will take place, and the pattern
1439 will fail outright at the current starting position.
1441 The following example counts all the possible matching strings in a
1442 pattern (without actually matching any of them).
1444 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1445 print "Count=$count\n";
1460 If we add a C<(*PRUNE)> before the count like the following
1462 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
1463 print "Count=$count\n";
1465 we prevent backtracking and find the count of the longest matching
1466 at each matching starting point like so:
1473 Any number of C<(*PRUNE)> assertions may be used in a pattern.
1475 See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1476 control backtracking. In some cases, the use of C<(*PRUNE)> can be
1477 replaced with a C<< (?>pattern) >> with no functional difference; however,
1478 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1479 C<< (?>pattern) >> alone.
1482 =item C<(*SKIP)> C<(*SKIP:NAME)>
1485 This zero-width pattern is similar to C<(*PRUNE)>, except that on
1486 failure it also signifies that whatever text that was matched leading up
1487 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1488 of this pattern. This effectively means that the regex engine "skips" forward
1489 to this position on failure and tries to match again, (assuming that
1490 there is sufficient room to match).
1492 The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1493 C<(*MARK:NAME)> was encountered while matching, then it is that position
1494 which is used as the "skip point". If no C<(*MARK)> of that name was
1495 encountered, then the C<(*SKIP)> operator has no effect. When used
1496 without a name the "skip point" is where the match point was when
1497 executing the (*SKIP) pattern.
1499 Compare the following to the examples in C<(*PRUNE)>, note the string
1502 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
1503 print "Count=$count\n";
1511 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
1512 executed, the next starting point will be where the cursor was when the
1513 C<(*SKIP)> was executed.
1515 =item C<(*MARK:NAME)> C<(*:NAME)>
1516 X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)>
1518 This zero-width pattern can be used to mark the point reached in a string
1519 when a certain part of the pattern has been successfully matched. This
1520 mark may be given a name. A later C<(*SKIP)> pattern will then skip
1521 forward to that point if backtracked into on failure. Any number of
1522 C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated.
1524 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1525 can be used to "label" a pattern branch, so that after matching, the
1526 program can determine which branches of the pattern were involved in the
1529 When a match is successful, the C<$REGMARK> variable will be set to the
1530 name of the most recently executed C<(*MARK:NAME)> that was involved
1533 This can be used to determine which branch of a pattern was matched
1534 without using a separate capture group for each branch, which in turn
1535 can result in a performance improvement, as perl cannot optimize
1536 C</(?:(x)|(y)|(z))/> as efficiently as something like
1537 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1539 When a match has failed, and unless another verb has been involved in
1540 failing the match and has provided its own name to use, the C<$REGERROR>
1541 variable will be set to the name of the most recently executed
1544 See C<(*SKIP)> for more details.
1546 As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
1548 =item C<(*THEN)> C<(*THEN:NAME)>
1550 This is similar to the "cut group" operator C<::> from Perl 6. Like
1551 C<(*PRUNE)>, this verb always matches, and when backtracked into on
1552 failure, it causes the regex engine to try the next alternation in the
1553 innermost enclosing group (capturing or otherwise).
1555 Its name comes from the observation that this operation combined with the
1556 alternation operator (C<|>) can be used to create what is essentially a
1557 pattern-based if/then/else block:
1559 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
1561 Note that if this operator is used and NOT inside of an alternation then
1562 it acts exactly like the C<(*PRUNE)> operator.
1572 / ( A (*THEN) B | C (*THEN) D ) /
1576 / ( A (*PRUNE) B | C (*PRUNE) D ) /
1578 as after matching the A but failing on the B the C<(*THEN)> verb will
1579 backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
1584 This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a
1585 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
1586 into on failure it causes the match to fail outright. No further attempts
1587 to find a valid match by advancing the start pointer will occur again.
1590 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
1591 print "Count=$count\n";
1598 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1599 does not match, the regex engine will not try any further matching on the
1604 =item Verbs without an argument
1608 =item C<(*FAIL)> C<(*F)>
1611 This pattern matches nothing and always fails. It can be used to force the
1612 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1613 fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
1615 It is probably useful only when combined with C<(?{})> or C<(??{})>.
1620 B<WARNING:> This feature is highly experimental. It is not recommended
1621 for production code.
1623 This pattern matches nothing and causes the end of successful matching at
1624 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1625 whether there is actually more to match in the string. When inside of a
1626 nested pattern, such as recursion, or in a subpattern dynamically generated
1627 via C<(??{})>, only the innermost pattern is ended immediately.
1629 If the C<(*ACCEPT)> is inside of capturing groups then the groups are
1630 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1633 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
1635 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
1636 be set. If another branch in the inner parentheses were matched, such as in the
1637 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
1644 X<backtrack> X<backtracking>
1646 NOTE: This section presents an abstract approximation of regular
1647 expression behavior. For a more rigorous (and complicated) view of
1648 the rules involved in selecting a match among possible alternatives,
1649 see L<Combining RE Pieces>.
1651 A fundamental feature of regular expression matching involves the
1652 notion called I<backtracking>, which is currently used (when needed)
1653 by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>,
1654 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1655 internally, but the general principle outlined here is valid.
1657 For a regular expression to match, the I<entire> regular expression must
1658 match, not just part of it. So if the beginning of a pattern containing a
1659 quantifier succeeds in a way that causes later parts in the pattern to
1660 fail, the matching engine backs up and recalculates the beginning
1661 part--that's why it's called backtracking.
1663 Here is an example of backtracking: Let's say you want to find the
1664 word following "foo" in the string "Food is on the foo table.":
1666 $_ = "Food is on the foo table.";
1667 if ( /\b(foo)\s+(\w+)/i ) {
1668 print "$2 follows $1.\n";
1671 When the match runs, the first part of the regular expression (C<\b(foo)>)
1672 finds a possible match right at the beginning of the string, and loads up
1673 $1 with "Foo". However, as soon as the matching engine sees that there's
1674 no whitespace following the "Foo" that it had saved in $1, it realizes its
1675 mistake and starts over again one character after where it had the
1676 tentative match. This time it goes all the way until the next occurrence
1677 of "foo". The complete regular expression matches this time, and you get
1678 the expected output of "table follows foo."
1680 Sometimes minimal matching can help a lot. Imagine you'd like to match
1681 everything between "foo" and "bar". Initially, you write something
1684 $_ = "The food is under the bar in the barn.";
1685 if ( /foo(.*)bar/ ) {
1689 Which perhaps unexpectedly yields:
1691 got <d is under the bar in the >
1693 That's because C<.*> was greedy, so you get everything between the
1694 I<first> "foo" and the I<last> "bar". Here it's more effective
1695 to use minimal matching to make sure you get the text between a "foo"
1696 and the first "bar" thereafter.
1698 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1699 got <d is under the >
1701 Here's another example. Let's say you'd like to match a number at the end
1702 of a string, and you also want to keep the preceding part of the match.
1705 $_ = "I have 2 numbers: 53147";
1706 if ( /(.*)(\d*)/ ) { # Wrong!
1707 print "Beginning is <$1>, number is <$2>.\n";
1710 That won't work at all, because C<.*> was greedy and gobbled up the
1711 whole string. As C<\d*> can match on an empty string the complete
1712 regular expression matched successfully.
1714 Beginning is <I have 2 numbers: 53147>, number is <>.
1716 Here are some variants, most of which don't work:
1718 $_ = "I have 2 numbers: 53147";
1731 printf "%-12s ", $pat;
1733 print "<$1> <$2>\n";
1739 That will print out:
1741 (.*)(\d*) <I have 2 numbers: 53147> <>
1742 (.*)(\d+) <I have 2 numbers: 5314> <7>
1744 (.*?)(\d+) <I have > <2>
1745 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1746 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1747 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1748 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1750 As you see, this can be a bit tricky. It's important to realize that a
1751 regular expression is merely a set of assertions that gives a definition
1752 of success. There may be 0, 1, or several different ways that the
1753 definition might succeed against a particular string. And if there are
1754 multiple ways it might succeed, you need to understand backtracking to
1755 know which variety of success you will achieve.
1757 When using look-ahead assertions and negations, this can all get even
1758 trickier. Imagine you'd like to find a sequence of non-digits not
1759 followed by "123". You might try to write that as
1762 if ( /^\D*(?!123)/ ) { # Wrong!
1763 print "Yup, no 123 in $_\n";
1766 But that isn't going to match; at least, not the way you're hoping. It
1767 claims that there is no 123 in the string. Here's a clearer picture of
1768 why that pattern matches, contrary to popular expectations:
1773 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1774 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1776 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1777 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1785 You might have expected test 3 to fail because it seems to a more
1786 general purpose version of test 1. The important difference between
1787 them is that test 3 contains a quantifier (C<\D*>) and so can use
1788 backtracking, whereas test 1 will not. What's happening is
1789 that you've asked "Is it true that at the start of $x, following 0 or more
1790 non-digits, you have something that's not 123?" If the pattern matcher had
1791 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1794 The search engine will initially match C<\D*> with "ABC". Then it will
1795 try to match C<(?!123> with "123", which fails. But because
1796 a quantifier (C<\D*>) has been used in the regular expression, the
1797 search engine can backtrack and retry the match differently
1798 in the hope of matching the complete regular expression.
1800 The pattern really, I<really> wants to succeed, so it uses the
1801 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1802 time. Now there's indeed something following "AB" that is not
1803 "123". It's "C123", which suffices.
1805 We can deal with this by using both an assertion and a negation.
1806 We'll say that the first part in $1 must be followed both by a digit
1807 and by something that's not "123". Remember that the look-aheads
1808 are zero-width expressions--they only look, but don't consume any
1809 of the string in their match. So rewriting this way produces what
1810 you'd expect; that is, case 5 will fail, but case 6 succeeds:
1812 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1813 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
1817 In other words, the two zero-width assertions next to each other work as though
1818 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
1819 matches only if you're at the beginning of the line AND the end of the
1820 line simultaneously. The deeper underlying truth is that juxtaposition in
1821 regular expressions always means AND, except when you write an explicit OR
1822 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1823 although the attempted matches are made at different positions because "a"
1824 is not a zero-width assertion, but a one-width assertion.
1826 B<WARNING>: Particularly complicated regular expressions can take
1827 exponential time to solve because of the immense number of possible
1828 ways they can use backtracking to try for a match. For example, without
1829 internal optimizations done by the regular expression engine, this will
1830 take a painfully long time to run:
1832 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1834 And if you used C<*>'s in the internal groups instead of limiting them
1835 to 0 through 5 matches, then it would take forever--or until you ran
1836 out of stack space. Moreover, these internal optimizations are not
1837 always applicable. For example, if you put C<{0,5}> instead of C<*>
1838 on the external group, no current optimization is applicable, and the
1839 match takes a long time to finish.
1841 A powerful tool for optimizing such beasts is what is known as an
1842 "independent group",
1843 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
1844 zero-length look-ahead/look-behind assertions will not backtrack to make
1845 the tail match, since they are in "logical" context: only
1846 whether they match is considered relevant. For an example
1847 where side-effects of look-ahead I<might> have influenced the
1848 following match, see L<C<< (?>pattern) >>>.
1850 =head2 Version 8 Regular Expressions
1851 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
1853 In case you're not familiar with the "regular" Version 8 regex
1854 routines, here are the pattern-matching rules not described above.
1856 Any single character matches itself, unless it is a I<metacharacter>
1857 with a special meaning described here or above. You can cause
1858 characters that normally function as metacharacters to be interpreted
1859 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
1860 character; "\\" matches a "\"). This escape mechanism is also required
1861 for the character used as the pattern delimiter.
1863 A series of characters matches that series of characters in the target
1864 string, so the pattern C<blurfl> would match "blurfl" in the target
1867 You can specify a character class, by enclosing a list of characters
1868 in C<[]>, which will match any character from the list. If the
1869 first character after the "[" is "^", the class matches any character not
1870 in the list. Within a list, the "-" character specifies a
1871 range, so that C<a-z> represents all characters between "a" and "z",
1872 inclusive. If you want either "-" or "]" itself to be a member of a
1873 class, put it at the start of the list (possibly after a "^"), or
1874 escape it with a backslash. "-" is also taken literally when it is
1875 at the end of the list, just before the closing "]". (The
1876 following all specify the same class of three characters: C<[-az]>,
1877 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
1878 specifies a class containing twenty-six characters, even on EBCDIC-based
1879 character sets.) Also, if you try to use the character
1880 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1881 a range, the "-" is understood literally.
1883 Note also that the whole range idea is rather unportable between
1884 character sets--and even within character sets they may cause results
1885 you probably didn't expect. A sound principle is to use only ranges
1886 that begin from and end at either alphabetics of equal case ([a-e],
1887 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1888 spell out the character sets in full.
1890 Characters may be specified using a metacharacter syntax much like that
1891 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1892 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
1893 of three octal digits, matches the character whose coded character set value
1894 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1895 matches the character whose ordinal is I<nn>. The expression \cI<x>
1896 matches the character control-I<x>. Finally, the "." metacharacter
1897 matches any character except "\n" (unless you use C</s>).
1899 You can specify a series of alternatives for a pattern using "|" to
1900 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
1901 or "foe" in the target string (as would C<f(e|i|o)e>). The
1902 first alternative includes everything from the last pattern delimiter
1903 ("(", "[", or the beginning of the pattern) up to the first "|", and
1904 the last alternative contains everything from the last "|" to the next
1905 pattern delimiter. That's why it's common practice to include
1906 alternatives in parentheses: to minimize confusion about where they
1909 Alternatives are tried from left to right, so the first
1910 alternative found for which the entire expression matches, is the one that
1911 is chosen. This means that alternatives are not necessarily greedy. For
1912 example: when matching C<foo|foot> against "barefoot", only the "foo"
1913 part will match, as that is the first alternative tried, and it successfully
1914 matches the target string. (This might not seem important, but it is
1915 important when you are capturing matched text using parentheses.)
1917 Also remember that "|" is interpreted as a literal within square brackets,
1918 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
1920 Within a pattern, you may designate subpatterns for later reference
1921 by enclosing them in parentheses, and you may refer back to the
1922 I<n>th subpattern later in the pattern using the metacharacter
1923 \I<n>. Subpatterns are numbered based on the left to right order
1924 of their opening parenthesis. A backreference matches whatever
1925 actually matched the subpattern in the string being examined, not
1926 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\g1\d*> will
1927 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
1928 1 matched "0x", even though the rule C<0|0x> could potentially match
1929 the leading 0 in the second number.
1931 =head2 Warning on \1 Instead of $1
1933 Some people get too used to writing things like:
1935 $pattern =~ s/(\W)/\\\1/g;
1937 This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid
1939 B<sed> addicts, but it's a dirty habit to get into. That's because in
1940 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
1941 the usual double-quoted string means a control-A. The customary Unix
1942 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1943 of doing that, you get yourself into trouble if you then add an C</e>
1946 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1952 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1953 C<${1}000>. The operation of interpolation should not be confused
1954 with the operation of matching a backreference. Certainly they mean two
1955 different things on the I<left> side of the C<s///>.
1957 =head2 Repeated Patterns Matching a Zero-length Substring
1959 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1961 Regular expressions provide a terse and powerful programming language. As
1962 with most other power tools, power comes together with the ability
1965 A common abuse of this power stems from the ability to make infinite
1966 loops using regular expressions, with something as innocuous as:
1968 'foo' =~ m{ ( o? )* }x;
1970 The C<o?> matches at the beginning of C<'foo'>, and since the position
1971 in the string is not moved by the match, C<o?> would match again and again
1972 because of the C<*> quantifier. Another common way to create a similar cycle
1973 is with the looping modifier C<//g>:
1975 @matches = ( 'foo' =~ m{ o? }xg );
1979 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1981 or the loop implied by split().
1983 However, long experience has shown that many programming tasks may
1984 be significantly simplified by using repeated subexpressions that
1985 may match zero-length substrings. Here's a simple example being:
1987 @chars = split //, $string; # // is not magic in split
1988 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1990 Thus Perl allows such constructs, by I<forcefully breaking
1991 the infinite loop>. The rules for this are different for lower-level
1992 loops given by the greedy quantifiers C<*+{}>, and for higher-level
1993 ones like the C</g> modifier or split() operator.
1995 The lower-level loops are I<interrupted> (that is, the loop is
1996 broken) when Perl detects that a repeated expression matched a
1997 zero-length substring. Thus
1999 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
2001 is made equivalent to
2003 m{ (?: NON_ZERO_LENGTH )*
2008 The higher level-loops preserve an additional state between iterations:
2009 whether the last match was zero-length. To break the loop, the following
2010 match after a zero-length match is prohibited to have a length of zero.
2011 This prohibition interacts with backtracking (see L<"Backtracking">),
2012 and so the I<second best> match is chosen if the I<best> match is of
2020 results in C<< <><b><><a><><r><> >>. At each position of the string the best
2021 match given by non-greedy C<??> is the zero-length match, and the I<second
2022 best> match is what is matched by C<\w>. Thus zero-length matches
2023 alternate with one-character-long matches.
2025 Similarly, for repeated C<m/()/g> the second-best match is the match at the
2026 position one notch further in the string.
2028 The additional state of being I<matched with zero-length> is associated with
2029 the matched string, and is reset by each assignment to pos().
2030 Zero-length matches at the end of the previous match are ignored
2033 =head2 Combining RE Pieces
2035 Each of the elementary pieces of regular expressions which were described
2036 before (such as C<ab> or C<\Z>) could match at most one substring
2037 at the given position of the input string. However, in a typical regular
2038 expression these elementary pieces are combined into more complicated
2039 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
2040 (in these examples C<S> and C<T> are regular subexpressions).
2042 Such combinations can include alternatives, leading to a problem of choice:
2043 if we match a regular expression C<a|ab> against C<"abc">, will it match
2044 substring C<"a"> or C<"ab">? One way to describe which substring is
2045 actually matched is the concept of backtracking (see L<"Backtracking">).
2046 However, this description is too low-level and makes you think
2047 in terms of a particular implementation.
2049 Another description starts with notions of "better"/"worse". All the
2050 substrings which may be matched by the given regular expression can be
2051 sorted from the "best" match to the "worst" match, and it is the "best"
2052 match which is chosen. This substitutes the question of "what is chosen?"
2053 by the question of "which matches are better, and which are worse?".
2055 Again, for elementary pieces there is no such question, since at most
2056 one match at a given position is possible. This section describes the
2057 notion of better/worse for combining operators. In the description
2058 below C<S> and C<T> are regular subexpressions.
2064 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
2065 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
2066 which can be matched by C<T>.
2068 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
2071 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
2072 C<B> is better match for C<T> than C<B'>.
2076 When C<S> can match, it is a better match than when only C<T> can match.
2078 Ordering of two matches for C<S> is the same as for C<S>. Similar for
2079 two matches for C<T>.
2081 =item C<S{REPEAT_COUNT}>
2083 Matches as C<SSS...S> (repeated as many times as necessary).
2087 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
2089 =item C<S{min,max}?>
2091 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
2093 =item C<S?>, C<S*>, C<S+>
2095 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
2097 =item C<S??>, C<S*?>, C<S+?>
2099 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
2103 Matches the best match for C<S> and only that.
2105 =item C<(?=S)>, C<(?<=S)>
2107 Only the best match for C<S> is considered. (This is important only if
2108 C<S> has capturing parentheses, and backreferences are used somewhere
2109 else in the whole regular expression.)
2111 =item C<(?!S)>, C<(?<!S)>
2113 For this grouping operator there is no need to describe the ordering, since
2114 only whether or not C<S> can match is important.
2116 =item C<(??{ EXPR })>, C<(?PARNO)>
2118 The ordering is the same as for the regular expression which is
2119 the result of EXPR, or the pattern contained by capture group PARNO.
2121 =item C<(?(condition)yes-pattern|no-pattern)>
2123 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
2124 already determined. The ordering of the matches is the same as for the
2125 chosen subexpression.
2129 The above recipes describe the ordering of matches I<at a given position>.
2130 One more rule is needed to understand how a match is determined for the
2131 whole regular expression: a match at an earlier position is always better
2132 than a match at a later position.
2134 =head2 Creating Custom RE Engines
2136 Overloaded constants (see L<overload>) provide a simple way to extend
2137 the functionality of the RE engine.
2139 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
2140 matches at a boundary between whitespace characters and non-whitespace
2141 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
2142 at these positions, so we want to have each C<\Y|> in the place of the
2143 more complicated version. We can create a module C<customre> to do
2151 die "No argument to customre::import allowed" if @_;
2152 overload::constant 'qr' => \&convert;
2155 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
2157 # We must also take care of not escaping the legitimate \\Y|
2158 # sequence, hence the presence of '\\' in the conversion rules.
2159 my %rules = ( '\\' => '\\\\',
2160 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
2166 { $rules{$1} or invalid($re,$1) }sgex;
2170 Now C<use customre> enables the new escape in constant regular
2171 expressions, i.e., those without any runtime variable interpolations.
2172 As documented in L<overload>, this conversion will work only over
2173 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
2174 part of this regular expression needs to be converted explicitly
2175 (but only if the special meaning of C<\Y|> should be enabled inside $re):
2180 $re = customre::convert $re;
2183 =head1 PCRE/Python Support
2185 As of Perl 5.10.0, Perl supports several Python/PCRE specific extensions
2186 to the regex syntax. While Perl programmers are encouraged to use the
2187 Perl specific syntax, the following are also accepted:
2191 =item C<< (?PE<lt>NAMEE<gt>pattern) >>
2193 Define a named capture group. Equivalent to C<< (?<NAME>pattern) >>.
2195 =item C<< (?P=NAME) >>
2197 Backreference to a named capture group. Equivalent to C<< \g{NAME} >>.
2199 =item C<< (?P>NAME) >>
2201 Subroutine call to a named capture group. Equivalent to C<< (?&NAME) >>.
2207 There are numerous problems with case insensitive matching of characters
2208 outside the ASCII range, especially with those whose folds are multiple
2209 characters, such as ligatures like C<LATIN SMALL LIGATURE FF>.
2211 In a bracketed character class with case insensitive matching, ranges only work
2212 for ASCII characters. For example,
2213 C<m/[\N{CYRILLIC CAPITAL LETTER A}-\N{CYRILLIC CAPITAL LETTER YA}]/i>
2214 doesn't match all the Russian upper and lower case letters.
2216 Many regular expression constructs don't work on EBCDIC platforms.
2218 This document varies from difficult to understand to completely
2219 and utterly opaque. The wandering prose riddled with jargon is
2220 hard to fathom in several places.
2222 This document needs a rewrite that separates the tutorial content
2223 from the reference content.
2231 L<perlop/"Regexp Quote-Like Operators">.
2233 L<perlop/"Gory details of parsing quoted constructs">.
2243 I<Mastering Regular Expressions> by Jeffrey Friedl, published
2244 by O'Reilly and Associates.