2 X<regular expression> X<regex> X<regexp>
4 perlre - Perl regular expressions
8 This page describes the syntax of regular expressions in Perl.
10 If you haven't used regular expressions before, a quick-start
11 introduction is available in L<perlrequick>, and a longer tutorial
12 introduction is available in L<perlretut>.
14 For reference on how regular expressions are used in matching
15 operations, plus various examples of the same, see discussions of
16 C<m//>, C<s///>, C<qr//> and C<"??"> in L<perlop/"Regexp Quote-Like
19 New in v5.22, L<C<use re 'strict'>|re/'strict' mode> applies stricter
20 rules than otherwise when compiling regular expression patterns. It can
21 find things that, while legal, may not be what you intended.
27 Matching operations can have various modifiers. Modifiers
28 that relate to the interpretation of the regular expression inside
29 are listed below. Modifiers that alter the way a regular expression
30 is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and
31 L<perlop/"Gory details of parsing quoted constructs">.
36 X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline>
38 Treat the string as multiple lines. That is, change C<"^"> and C<"$"> from matching
39 the start of the string's first line and the end of its last line to
40 matching the start and end of each line within the string.
43 X</s> X<regex, single-line> X<regexp, single-line>
44 X<regular expression, single-line>
46 Treat the string as single line. That is, change C<"."> to match any character
47 whatsoever, even a newline, which normally it would not match.
49 Used together, as C</ms>, they let the C<"."> match any character whatsoever,
50 while still allowing C<"^"> and C<"$"> to match, respectively, just after
51 and just before newlines within the string.
54 X</i> X<regex, case-insensitive> X<regexp, case-insensitive>
55 X<regular expression, case-insensitive>
57 Do case-insensitive pattern matching. For example, "A" will match "a"
60 If locale matching rules are in effect, the case map is taken from the
62 locale for code points less than 255, and from Unicode rules for larger
63 code points. However, matches that would cross the Unicode
64 rules/non-Unicode rules boundary (ords 255/256) will not succeed, unless
65 the locale is a UTF-8 one. See L<perllocale>.
67 There are a number of Unicode characters that match a sequence of
68 multiple characters under C</i>. For example,
69 C<LATIN SMALL LIGATURE FI> should match the sequence C<fi>. Perl is not
70 currently able to do this when the multiple characters are in the pattern and
71 are split between groupings, or when one or more are quantified. Thus
73 "\N{LATIN SMALL LIGATURE FI}" =~ /fi/i; # Matches
74 "\N{LATIN SMALL LIGATURE FI}" =~ /[fi][fi]/i; # Doesn't match!
75 "\N{LATIN SMALL LIGATURE FI}" =~ /fi*/i; # Doesn't match!
77 # The below doesn't match, and it isn't clear what $1 and $2 would
79 "\N{LATIN SMALL LIGATURE FI}" =~ /(f)(i)/i; # Doesn't match!
81 Perl doesn't match multiple characters in a bracketed
82 character class unless the character that maps to them is explicitly
83 mentioned, and it doesn't match them at all if the character class is
84 inverted, which otherwise could be highly confusing. See
85 L<perlrecharclass/Bracketed Character Classes>, and
86 L<perlrecharclass/Negation>.
91 Extend your pattern's legibility by permitting whitespace and comments.
95 X</p> X<regex, preserve> X<regexp, preserve>
97 Preserve the string matched such that C<${^PREMATCH}>, C<${^MATCH}>, and
98 C<${^POSTMATCH}> are available for use after matching.
100 In Perl 5.20 and higher this is ignored. Due to a new copy-on-write
101 mechanism, C<${^PREMATCH}>, C<${^MATCH}>, and C<${^POSTMATCH}> will be available
102 after the match regardless of the modifier.
104 =item B<C<a>>, B<C<d>>, B<C<l>>, and B<C<u>>
105 X</a> X</d> X</l> X</u>
107 These modifiers, all new in 5.14, affect which character-set rules
108 (Unicode, etc.) are used, as described below in
109 L</Character set modifiers>.
112 X</n> X<regex, non-capture> X<regexp, non-capture>
113 X<regular expression, non-capture>
115 Prevent the grouping metacharacters C<()> from capturing. This modifier,
116 new in 5.22, will stop C<$1>, C<$2>, etc... from being filled in.
118 "hello" =~ /(hi|hello)/; # $1 is "hello"
119 "hello" =~ /(hi|hello)/n; # $1 is undef
121 This is equivalent to putting C<?:> at the beginning of every capturing group:
123 "hello" =~ /(?:hi|hello)/; # $1 is undef
125 C</n> can be negated on a per-group basis. Alternatively, named captures
128 "hello" =~ /(?-n:(hi|hello))/n; # $1 is "hello"
129 "hello" =~ /(?<greet>hi|hello)/n; # $1 is "hello", $+{greet} is
132 =item Other Modifiers
134 There are a number of flags that can be found at the end of regular
135 expression constructs that are I<not> generic regular expression flags, but
136 apply to the operation being performed, like matching or substitution (C<m//>
137 or C<s///> respectively).
139 Flags described further in
140 L<perlretut/"Using regular expressions in Perl"> are:
142 c - keep the current position during repeated matching
143 g - globally match the pattern repeatedly in the string
145 Substitution-specific modifiers described in
147 L<perlop/"s/PATTERN/REPLACEMENT/msixpodualngcer"> are:
149 e - evaluate the right-hand side as an expression
150 ee - evaluate the right side as a string then eval the result
151 o - pretend to optimize your code, but actually introduce bugs
152 r - perform non-destructive substitution and return the new value
156 Regular expression modifiers are usually written in documentation
157 as e.g., "the C</x> modifier", even though the delimiter
158 in question might not really be a slash. The modifiers C</imnsxadlup>
159 may also be embedded within the regular expression itself using
160 the C<(?...)> construct, see L</Extended Patterns> below.
162 =head3 Details on some modifiers
164 Some of the modifiers require more explanation than given in the
170 the regular expression parser to ignore most whitespace that is neither
171 backslashed nor within a bracketed character class. You can use this to
172 break up your regular expression into (slightly) more readable parts.
173 Also, the C<"#"> character is treated as a metacharacter introducing a
174 comment that runs up to the pattern's closing delimiter, or to the end
175 of the current line if the pattern extends onto the next line. Hence,
176 this is very much like an ordinary Perl code comment. (You can include
177 the closing delimiter within the comment only if you precede it with a
178 backslash, so be careful!)
180 Use of C</x> means that if you want real
181 whitespace or C<"#"> characters in the pattern (outside a bracketed character
182 class, which is unaffected by C</x>), then you'll either have to
183 escape them (using backslashes or C<\Q...\E>) or encode them using octal,
184 hex, or C<\N{}> escapes.
185 It is ineffective to try to continue a comment onto the next line by
186 escaping the C<\n> with a backslash or C<\Q>.
188 You can use L</(?#text)> to create a comment that ends earlier than the
189 end of the current line, but C<text> also can't contain the closing
190 delimiter unless escaped with a backslash.
192 Taken together, these features go a long way towards
193 making Perl's regular expressions more readable. Here's an example:
195 # Delete (most) C comments.
197 /\* # Match the opening delimiter.
198 .*? # Match a minimal number of characters.
199 \*/ # Match the closing delimiter.
202 Note that anything inside
203 a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect
204 space interpretation within a single multi-character construct. For
205 example in C<\x{...}>, regardless of the C</x> modifier, there can be no
206 spaces. Same for a L<quantifier|/Quantifiers> such as C<{3}> or
207 C<{5,}>. Similarly, C<(?:...)> can't have a space between the C<"{">,
208 C<"?">, and C<":">. Within any delimiters for such a
209 construct, allowed spaces are not affected by C</x>, and depend on the
210 construct. For example, C<\x{...}> can't have spaces because hexadecimal
211 numbers don't have spaces in them. But, Unicode properties can have spaces, so
212 in C<\p{...}> there can be spaces that follow the Unicode rules, for which see
213 L<perluniprops/Properties accessible through \p{} and \P{}>.
216 The set of characters that are deemed whitespace are those that Unicode
217 calls "Pattern White Space", namely:
219 U+0009 CHARACTER TABULATION
221 U+000B LINE TABULATION
223 U+000D CARRIAGE RETURN
226 U+200E LEFT-TO-RIGHT MARK
227 U+200F RIGHT-TO-LEFT MARK
228 U+2028 LINE SEPARATOR
229 U+2029 PARAGRAPH SEPARATOR
231 =head4 Character set modifiers
233 C</d>, C</u>, C</a>, and C</l>, available starting in 5.14, are called
234 the character set modifiers; they affect the character set rules
235 used for the regular expression.
237 The C</d>, C</u>, and C</l> modifiers are not likely to be of much use
238 to you, and so you need not worry about them very much. They exist for
239 Perl's internal use, so that complex regular expression data structures
240 can be automatically serialized and later exactly reconstituted,
241 including all their nuances. But, since Perl can't keep a secret, and
242 there may be rare instances where they are useful, they are documented
245 The C</a> modifier, on the other hand, may be useful. Its purpose is to
246 allow code that is to work mostly on ASCII data to not have to concern
249 Briefly, C</l> sets the character set to that of whatever B<L>ocale is in
250 effect at the time of the execution of the pattern match.
252 C</u> sets the character set to B<U>nicode.
254 C</a> also sets the character set to Unicode, BUT adds several
255 restrictions for B<A>SCII-safe matching.
257 C</d> is the old, problematic, pre-5.14 B<D>efault character set
258 behavior. Its only use is to force that old behavior.
260 At any given time, exactly one of these modifiers is in effect. Their
261 existence allows Perl to keep the originally compiled behavior of a
262 regular expression, regardless of what rules are in effect when it is
263 actually executed. And if it is interpolated into a larger regex, the
264 original's rules continue to apply to it, and only it.
266 The C</l> and C</u> modifiers are automatically selected for
267 regular expressions compiled within the scope of various pragmas,
268 and we recommend that in general, you use those pragmas instead of
269 specifying these modifiers explicitly. For one thing, the modifiers
270 affect only pattern matching, and do not extend to even any replacement
271 done, whereas using the pragmas gives consistent results for all
272 appropriate operations within their scopes. For example,
276 will match "foo" using the locale's rules for case-insensitive matching,
277 but the C</l> does not affect how the C<\U> operates. Most likely you
278 want both of them to use locale rules. To do this, instead compile the
279 regular expression within the scope of C<use locale>. This both
280 implicitly adds the C</l>, and applies locale rules to the C<\U>. The
281 lesson is to C<use locale>, and not C</l> explicitly.
283 Similarly, it would be better to use C<use feature 'unicode_strings'>
288 to get Unicode rules, as the C<\L> in the former (but not necessarily
289 the latter) would also use Unicode rules.
291 More detail on each of the modifiers follows. Most likely you don't
292 need to know this detail for C</l>, C</u>, and C</d>, and can skip ahead
293 to L<E<sol>a|/E<sol>a (and E<sol>aa)>.
297 means to use the current locale's rules (see L<perllocale>) when pattern
298 matching. For example, C<\w> will match the "word" characters of that
299 locale, and C<"/i"> case-insensitive matching will match according to
300 the locale's case folding rules. The locale used will be the one in
301 effect at the time of execution of the pattern match. This may not be
302 the same as the compilation-time locale, and can differ from one match
303 to another if there is an intervening call of the
304 L<setlocale() function|perllocale/The setlocale function>.
306 Prior to v5.20, Perl did not support multi-byte locales. Starting then,
307 UTF-8 locales are supported. No other multi byte locales are ever
308 likely to be supported. However, in all locales, one can have code
309 points above 255 and these will always be treated as Unicode no matter
310 what locale is in effect.
312 Under Unicode rules, there are a few case-insensitive matches that cross
313 the 255/256 boundary. Except for UTF-8 locales in Perls v5.20 and
314 later, these are disallowed under C</l>. For example, 0xFF (on ASCII
315 platforms) does not caselessly match the character at 0x178, C<LATIN
316 CAPITAL LETTER Y WITH DIAERESIS>, because 0xFF may not be C<LATIN SMALL
317 LETTER Y WITH DIAERESIS> in the current locale, and Perl has no way of
318 knowing if that character even exists in the locale, much less what code
321 In a UTF-8 locale in v5.20 and later, the only visible difference
322 between locale and non-locale in regular expressions should be tainting
325 This modifier may be specified to be the default by C<use locale>, but
326 see L</Which character set modifier is in effect?>.
331 means to use Unicode rules when pattern matching. On ASCII platforms,
332 this means that the code points between 128 and 255 take on their
333 Latin-1 (ISO-8859-1) meanings (which are the same as Unicode's).
334 (Otherwise Perl considers their meanings to be undefined.) Thus,
335 under this modifier, the ASCII platform effectively becomes a Unicode
336 platform; and hence, for example, C<\w> will match any of the more than
337 100_000 word characters in Unicode.
339 Unlike most locales, which are specific to a language and country pair,
340 Unicode classifies all the characters that are letters I<somewhere> in
342 C<\w>. For example, your locale might not think that C<LATIN SMALL
343 LETTER ETH> is a letter (unless you happen to speak Icelandic), but
344 Unicode does. Similarly, all the characters that are decimal digits
345 somewhere in the world will match C<\d>; this is hundreds, not 10,
346 possible matches. And some of those digits look like some of the 10
347 ASCII digits, but mean a different number, so a human could easily think
348 a number is a different quantity than it really is. For example,
349 C<BENGALI DIGIT FOUR> (U+09EA) looks very much like an
350 C<ASCII DIGIT EIGHT> (U+0038). And, C<\d+>, may match strings of digits
351 that are a mixture from different writing systems, creating a security
352 issue. L<Unicode::UCD/num()> can be used to sort
353 this out. Or the C</a> modifier can be used to force C<\d> to match
354 just the ASCII 0 through 9.
356 Also, under this modifier, case-insensitive matching works on the full
358 characters. The C<KELVIN SIGN>, for example matches the letters "k" and
359 "K"; and C<LATIN SMALL LIGATURE FF> matches the sequence "ff", which,
360 if you're not prepared, might make it look like a hexadecimal constant,
361 presenting another potential security issue. See
362 L<http://unicode.org/reports/tr36> for a detailed discussion of Unicode
365 This modifier may be specified to be the default by C<use feature
366 'unicode_strings>, C<use locale ':not_characters'>, or
367 C<L<use 5.012|perlfunc/use VERSION>> (or higher),
368 but see L</Which character set modifier is in effect?>.
373 This modifier means to use the "Default" native rules of the platform
374 except when there is cause to use Unicode rules instead, as follows:
380 the target string is encoded in UTF-8; or
384 the pattern is encoded in UTF-8; or
388 the pattern explicitly mentions a code point that is above 255 (say by
393 the pattern uses a Unicode name (C<\N{...}>); or
397 the pattern uses a Unicode property (C<\p{...}> or C<\P{...}>); or
401 the pattern uses a Unicode break (C<\b{...}> or C<\B{...}>); or
405 the pattern uses L</C<(?[ ])>>
409 Another mnemonic for this modifier is "Depends", as the rules actually
410 used depend on various things, and as a result you can get unexpected
411 results. See L<perlunicode/The "Unicode Bug">. The Unicode Bug has
412 become rather infamous, leading to yet another (printable) name for this
415 Unless the pattern or string are encoded in UTF-8, only ASCII characters
416 can match positively.
418 Here are some examples of how that works on an ASCII platform:
420 $str = "\xDF"; # $str is not in UTF-8 format.
421 $str =~ /^\w/; # No match, as $str isn't in UTF-8 format.
422 $str .= "\x{0e0b}"; # Now $str is in UTF-8 format.
423 $str =~ /^\w/; # Match! $str is now in UTF-8 format.
425 $str =~ /^\w/; # Still a match! $str remains in UTF-8 format.
427 This modifier is automatically selected by default when none of the
428 others are, so yet another name for it is "Default".
430 Because of the unexpected behaviors associated with this modifier, you
431 probably should only explicitly use it to maintain weird backward
436 This modifier stands for ASCII-restrict (or ASCII-safe). This modifier,
437 unlike the others, may be doubled-up to increase its effect.
439 When it appears singly, it causes the sequences C<\d>, C<\s>, C<\w>, and
440 the Posix character classes to match only in the ASCII range. They thus
441 revert to their pre-5.6, pre-Unicode meanings. Under C</a>, C<\d>
442 always means precisely the digits C<"0"> to C<"9">; C<\s> means the five
443 characters C<[ \f\n\r\t]>, and starting in Perl v5.18, the vertical tab;
444 C<\w> means the 63 characters
445 C<[A-Za-z0-9_]>; and likewise, all the Posix classes such as
446 C<[[:print:]]> match only the appropriate ASCII-range characters.
448 This modifier is useful for people who only incidentally use Unicode,
449 and who do not wish to be burdened with its complexities and security
452 With C</a>, one can write C<\d> with confidence that it will only match
453 ASCII characters, and should the need arise to match beyond ASCII, you
454 can instead use C<\p{Digit}> (or C<\p{Word}> for C<\w>). There are
455 similar C<\p{...}> constructs that can match beyond ASCII both white
456 space (see L<perlrecharclass/Whitespace>), and Posix classes (see
457 L<perlrecharclass/POSIX Character Classes>). Thus, this modifier
458 doesn't mean you can't use Unicode, it means that to get Unicode
459 matching you must explicitly use a construct (C<\p{}>, C<\P{}>) that
462 As you would expect, this modifier causes, for example, C<\D> to mean
463 the same thing as C<[^0-9]>; in fact, all non-ASCII characters match
464 C<\D>, C<\S>, and C<\W>. C<\b> still means to match at the boundary
465 between C<\w> and C<\W>, using the C</a> definitions of them (similarly
468 Otherwise, C</a> behaves like the C</u> modifier, in that
469 case-insensitive matching uses Unicode rules; for example, "k" will
470 match the Unicode C<\N{KELVIN SIGN}> under C</i> matching, and code
471 points in the Latin1 range, above ASCII will have Unicode rules when it
472 comes to case-insensitive matching.
474 To forbid ASCII/non-ASCII matches (like "k" with C<\N{KELVIN SIGN}>),
475 specify the C<"a"> twice, for example C</aai> or C</aia>. (The first
476 occurrence of C<"a"> restricts the C<\d>, etc., and the second occurrence
477 adds the C</i> restrictions.) But, note that code points outside the
478 ASCII range will use Unicode rules for C</i> matching, so the modifier
479 doesn't really restrict things to just ASCII; it just forbids the
480 intermixing of ASCII and non-ASCII.
482 To summarize, this modifier provides protection for applications that
483 don't wish to be exposed to all of Unicode. Specifying it twice
484 gives added protection.
486 This modifier may be specified to be the default by C<use re '/a'>
487 or C<use re '/aa'>. If you do so, you may actually have occasion to use
488 the C</u> modifier explicitly if there are a few regular expressions
489 where you do want full Unicode rules (but even here, it's best if
490 everything were under feature C<"unicode_strings">, along with the
491 C<use re '/aa'>). Also see L</Which character set modifier is in
496 =head4 Which character set modifier is in effect?
498 Which of these modifiers is in effect at any given point in a regular
499 expression depends on a fairly complex set of interactions. These have
500 been designed so that in general you don't have to worry about it, but
501 this section gives the gory details. As
502 explained below in L</Extended Patterns> it is possible to explicitly
503 specify modifiers that apply only to portions of a regular expression.
504 The innermost always has priority over any outer ones, and one applying
505 to the whole expression has priority over any of the default settings that are
506 described in the remainder of this section.
508 The C<L<use re 'E<sol>foo'|re/"'/flags' mode">> pragma can be used to set
509 default modifiers (including these) for regular expressions compiled
510 within its scope. This pragma has precedence over the other pragmas
511 listed below that also change the defaults.
513 Otherwise, C<L<use locale|perllocale>> sets the default modifier to C</l>;
514 and C<L<use feature 'unicode_strings|feature>>, or
515 C<L<use 5.012|perlfunc/use VERSION>> (or higher) set the default to
516 C</u> when not in the same scope as either C<L<use locale|perllocale>>
517 or C<L<use bytes|bytes>>.
518 (C<L<use locale ':not_characters'|perllocale/Unicode and UTF-8>> also
519 sets the default to C</u>, overriding any plain C<use locale>.)
520 Unlike the mechanisms mentioned above, these
521 affect operations besides regular expressions pattern matching, and so
522 give more consistent results with other operators, including using
523 C<\U>, C<\l>, etc. in substitution replacements.
525 If none of the above apply, for backwards compatibility reasons, the
526 C</d> modifier is the one in effect by default. As this can lead to
527 unexpected results, it is best to specify which other rule set should be
530 =head4 Character set modifier behavior prior to Perl 5.14
532 Prior to 5.14, there were no explicit modifiers, but C</l> was implied
533 for regexes compiled within the scope of C<use locale>, and C</d> was
534 implied otherwise. However, interpolating a regex into a larger regex
535 would ignore the original compilation in favor of whatever was in effect
536 at the time of the second compilation. There were a number of
537 inconsistencies (bugs) with the C</d> modifier, where Unicode rules
538 would be used when inappropriate, and vice versa. C<\p{}> did not imply
539 Unicode rules, and neither did all occurrences of C<\N{}>, until 5.12.
541 =head2 Regular Expressions
543 =head3 Metacharacters
545 The patterns used in Perl pattern matching evolved from those supplied in
546 the Version 8 regex routines. (The routines are derived
547 (distantly) from Henry Spencer's freely redistributable reimplementation
548 of the V8 routines.) See L<Version 8 Regular Expressions> for
551 In particular the following metacharacters have their standard I<egrep>-ish
554 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
557 \ Quote the next metacharacter
558 ^ Match the beginning of the line
559 . Match any character (except newline)
560 $ Match the end of the string (or before newline at the end
564 [] Bracketed Character class
566 By default, the C<"^"> character is guaranteed to match only the
567 beginning of the string, the C<"$"> character only the end (or before the
568 newline at the end), and Perl does certain optimizations with the
569 assumption that the string contains only one line. Embedded newlines
570 will not be matched by C<"^"> or C<"$">. You may, however, wish to treat a
571 string as a multi-line buffer, such that the C<"^"> will match after any
572 newline within the string (except if the newline is the last character in
573 the string), and C<"$"> will match before any newline. At the
574 cost of a little more overhead, you can do this by using the /m modifier
575 on the pattern match operator. (Older programs did this by setting C<$*>,
576 but this option was removed in perl 5.10.)
579 To simplify multi-line substitutions, the C<"."> character never matches a
580 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
581 the string is a single line--even if it isn't.
586 The following standard quantifiers are recognized:
587 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
589 * Match 0 or more times
590 + Match 1 or more times
592 {n} Match exactly n times
593 {n,} Match at least n times
594 {n,m} Match at least n but not more than m times
596 (If a curly bracket occurs in a context other than one of the
597 quantifiers listed above, where it does not form part of a backslashed
598 sequence like C<\x{...}>, it is treated as a regular character.
599 However, a deprecation warning is raised for these
600 occurrences, and in Perl v5.26, literal uses of a curly bracket will be
601 required to be escaped, say by preceding them with a backslash (C<"\{">)
602 or enclosing them within square brackets (C<"[{]">). This change will
603 allow for future syntax extensions (like making the lower bound of a
604 quantifier optional), and better error checking of quantifiers.)
606 The C<"*"> quantifier is equivalent to C<{0,}>, the C<"+">
607 quantifier to C<{1,}>, and the C<"?"> quantifier to C<{0,1}>. I<n> and I<m> are limited
608 to non-negative integral values less than a preset limit defined when perl is built.
609 This is usually 32766 on the most common platforms. The actual limit can
610 be seen in the error message generated by code such as this:
612 $_ **= $_ , / {$_} / for 2 .. 42;
614 By default, a quantified subpattern is "greedy", that is, it will match as
615 many times as possible (given a particular starting location) while still
616 allowing the rest of the pattern to match. If you want it to match the
617 minimum number of times possible, follow the quantifier with a C<"?">. Note
618 that the meanings don't change, just the "greediness":
619 X<metacharacter> X<greedy> X<greediness>
620 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
622 *? Match 0 or more times, not greedily
623 +? Match 1 or more times, not greedily
624 ?? Match 0 or 1 time, not greedily
625 {n}? Match exactly n times, not greedily (redundant)
626 {n,}? Match at least n times, not greedily
627 {n,m}? Match at least n but not more than m times, not greedily
629 Normally when a quantified subpattern does not allow the rest of the
630 overall pattern to match, Perl will backtrack. However, this behaviour is
631 sometimes undesirable. Thus Perl provides the "possessive" quantifier form
634 *+ Match 0 or more times and give nothing back
635 ++ Match 1 or more times and give nothing back
636 ?+ Match 0 or 1 time and give nothing back
637 {n}+ Match exactly n times and give nothing back (redundant)
638 {n,}+ Match at least n times and give nothing back
639 {n,m}+ Match at least n but not more than m times and give nothing back
645 will never match, as the C<a++> will gobble up all the C<a>'s in the
646 string and won't leave any for the remaining part of the pattern. This
647 feature can be extremely useful to give perl hints about where it
648 shouldn't backtrack. For instance, the typical "match a double-quoted
649 string" problem can be most efficiently performed when written as:
651 /"(?:[^"\\]++|\\.)*+"/
653 as we know that if the final quote does not match, backtracking will not
654 help. See the independent subexpression
655 L</C<< (?>pattern) >>> for more details;
656 possessive quantifiers are just syntactic sugar for that construct. For
657 instance the above example could also be written as follows:
659 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
661 Note that the possessive quantifier modifier can not be be combined
662 with the non-greedy modifier. This is because it would make no sense.
663 Consider the follow equivalency table:
671 =head3 Escape sequences
673 Because patterns are processed as double-quoted strings, the following
680 \a alarm (bell) (BEL)
681 \e escape (think troff) (ESC)
682 \cK control char (example: VT)
683 \x{}, \x00 character whose ordinal is the given hexadecimal number
684 \N{name} named Unicode character or character sequence
685 \N{U+263D} Unicode character (example: FIRST QUARTER MOON)
686 \o{}, \000 character whose ordinal is the given octal number
687 \l lowercase next char (think vi)
688 \u uppercase next char (think vi)
689 \L lowercase until \E (think vi)
690 \U uppercase until \E (think vi)
691 \Q quote (disable) pattern metacharacters until \E
692 \E end either case modification or quoted section, think vi
694 Details are in L<perlop/Quote and Quote-like Operators>.
696 =head3 Character Classes and other Special Escapes
698 In addition, Perl defines the following:
699 X<\g> X<\k> X<\K> X<backreference>
701 Sequence Note Description
702 [...] [1] Match a character according to the rules of the
703 bracketed character class defined by the "...".
704 Example: [a-z] matches "a" or "b" or "c" ... or "z"
705 [[:...:]] [2] Match a character according to the rules of the POSIX
706 character class "..." within the outer bracketed
707 character class. Example: [[:upper:]] matches any
709 (?[...]) [8] Extended bracketed character class
710 \w [3] Match a "word" character (alphanumeric plus "_", plus
711 other connector punctuation chars plus Unicode
713 \W [3] Match a non-"word" character
714 \s [3] Match a whitespace character
715 \S [3] Match a non-whitespace character
716 \d [3] Match a decimal digit character
717 \D [3] Match a non-digit character
718 \pP [3] Match P, named property. Use \p{Prop} for longer names
720 \X [4] Match Unicode "eXtended grapheme cluster"
721 \1 [5] Backreference to a specific capture group or buffer.
722 '1' may actually be any positive integer.
723 \g1 [5] Backreference to a specific or previous group,
724 \g{-1} [5] The number may be negative indicating a relative
725 previous group and may optionally be wrapped in
726 curly brackets for safer parsing.
727 \g{name} [5] Named backreference
728 \k<name> [5] Named backreference
729 \K [6] Keep the stuff left of the \K, don't include it in $&
730 \N [7] Any character but \n. Not affected by /s modifier
731 \v [3] Vertical whitespace
732 \V [3] Not vertical whitespace
733 \h [3] Horizontal whitespace
734 \H [3] Not horizontal whitespace
741 See L<perlrecharclass/Bracketed Character Classes> for details.
745 See L<perlrecharclass/POSIX Character Classes> for details.
749 See L<perlrecharclass/Backslash sequences> for details.
753 See L<perlrebackslash/Misc> for details.
757 See L</Capture groups> below for details.
761 See L</Extended Patterns> below for details.
765 Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the
766 character or character sequence whose name is C<NAME>; and similarly
767 when of the form C<\N{U+I<hex>}>, it matches the character whose Unicode
768 code point is I<hex>. Otherwise it matches any character but C<\n>.
772 See L<perlrecharclass/Extended Bracketed Character Classes> for details.
778 Perl defines the following zero-width assertions:
779 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
780 X<regexp, zero-width assertion>
781 X<regular expression, zero-width assertion>
782 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
784 \b{} Match at Unicode boundary of specified type
785 \B{} Match where corresponding \b{} doesn't match
786 \b Match a word boundary
787 \B Match except at a word boundary
788 \A Match only at beginning of string
789 \Z Match only at end of string, or before newline at the end
790 \z Match only at end of string
791 \G Match only at pos() (e.g. at the end-of-match position
794 A Unicode boundary (C<\b{}>), available starting in v5.22, is a spot
795 between two characters, or before the first character in the string, or
796 after the final character in the string where certain criteria defined
797 by Unicode are met. See L<perlrebackslash/\b{}, \b, \B{}, \B> for
800 A word boundary (C<\b>) is a spot between two characters
801 that has a C<\w> on one side of it and a C<\W> on the other side
802 of it (in either order), counting the imaginary characters off the
803 beginning and end of the string as matching a C<\W>. (Within
804 character classes C<\b> represents backspace rather than a word
805 boundary, just as it normally does in any double-quoted string.)
806 The C<\A> and C<\Z> are just like C<"^"> and C<"$">, except that they
807 won't match multiple times when the C</m> modifier is used, while
808 C<"^"> and C<"$"> will match at every internal line boundary. To match
809 the actual end of the string and not ignore an optional trailing
811 X<\b> X<\A> X<\Z> X<\z> X</m>
813 The C<\G> assertion can be used to chain global matches (using
814 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
815 It is also useful when writing C<lex>-like scanners, when you have
816 several patterns that you want to match against consequent substrings
817 of your string; see the previous reference. The actual location
818 where C<\G> will match can also be influenced by using C<pos()> as
819 an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
820 matches (see L</"Repeated Patterns Matching a Zero-length Substring">)
821 is modified somewhat, in that contents to the left of C<\G> are
822 not counted when determining the length of the match. Thus the following
823 will not match forever:
828 while ($string =~ /(.\G)/g) {
832 It will print 'A' and then terminate, as it considers the match to
833 be zero-width, and thus will not match at the same position twice in a
836 It is worth noting that C<\G> improperly used can result in an infinite
837 loop. Take care when using patterns that include C<\G> in an alternation.
839 Note also that C<s///> will refuse to overwrite part of a substitution
840 that has already been replaced; so for example this will stop after the
841 first iteration, rather than iterating its way backwards through the
847 print; # prints 1234X6789, not XXXXX6789
850 =head3 Capture groups
852 The bracketing construct C<( ... )> creates capture groups (also referred to as
853 capture buffers). To refer to the current contents of a group later on, within
854 the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>)
855 for the second, and so on.
856 This is called a I<backreference>.
857 X<regex, capture buffer> X<regexp, capture buffer>
858 X<regex, capture group> X<regexp, capture group>
859 X<regular expression, capture buffer> X<backreference>
860 X<regular expression, capture group> X<backreference>
861 X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
862 X<named capture buffer> X<regular expression, named capture buffer>
863 X<named capture group> X<regular expression, named capture group>
864 X<%+> X<$+{name}> X<< \k<name> >>
865 There is no limit to the number of captured substrings that you may use.
866 Groups are numbered with the leftmost open parenthesis being number 1, etc. If
867 a group did not match, the associated backreference won't match either. (This
868 can happen if the group is optional, or in a different branch of an
870 You can omit the C<"g">, and write C<"\1">, etc, but there are some issues with
871 this form, described below.
873 You can also refer to capture groups relatively, by using a negative number, so
874 that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture
875 group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For
882 \g{-1} # backref to group 3
883 \g{-3} # backref to group 1
887 would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to
888 interpolate regexes into larger regexes and not have to worry about the
889 capture groups being renumbered.
891 You can dispense with numbers altogether and create named capture groups.
892 The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to
893 reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may
894 also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.)
895 I<name> must not begin with a number, nor contain hyphens.
896 When different groups within the same pattern have the same name, any reference
897 to that name assumes the leftmost defined group. Named groups count in
898 absolute and relative numbering, and so can also be referred to by those
900 (It's possible to do things with named capture groups that would otherwise
903 Capture group contents are dynamically scoped and available to you outside the
904 pattern until the end of the enclosing block or until the next successful
905 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
906 You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">,
907 etc); or by name via the C<%+> hash, using C<"$+{I<name>}">.
909 Braces are required in referring to named capture groups, but are optional for
910 absolute or relative numbered ones. Braces are safer when creating a regex by
911 concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a>
912 contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which
913 is probably not what you intended.
915 The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that
916 there were no named nor relative numbered capture groups. Absolute numbered
917 groups were referred to using C<\1>,
918 C<\2>, etc., and this notation is still
919 accepted (and likely always will be). But it leads to some ambiguities if
920 there are more than 9 capture groups, as C<\10> could mean either the tenth
921 capture group, or the character whose ordinal in octal is 010 (a backspace in
922 ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference
923 only if at least 10 left parentheses have opened before it. Likewise C<\11> is
924 a backreference only if at least 11 left parentheses have opened before it.
925 And so on. C<\1> through C<\9> are always interpreted as backreferences.
926 There are several examples below that illustrate these perils. You can avoid
927 the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups;
928 and for octal constants always using C<\o{}>, or for C<\077> and below, using 3
929 digits padded with leading zeros, since a leading zero implies an octal
932 The C<\I<digit>> notation also works in certain circumstances outside
933 the pattern. See L</Warning on \1 Instead of $1> below for details.
937 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
939 /(.)\g1/ # find first doubled char
940 and print "'$1' is the first doubled character\n";
942 /(?<char>.)\k<char>/ # ... a different way
943 and print "'$+{char}' is the first doubled character\n";
945 /(?'char'.)\g1/ # ... mix and match
946 and print "'$1' is the first doubled character\n";
948 if (/Time: (..):(..):(..)/) { # parse out values
954 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference
955 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal
956 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference
957 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal
959 $a = '(.)\1'; # Creates problems when concatenated.
960 $b = '(.)\g{1}'; # Avoids the problems.
961 "aa" =~ /${a}/; # True
962 "aa" =~ /${b}/; # True
963 "aa0" =~ /${a}0/; # False!
964 "aa0" =~ /${b}0/; # True
965 "aa\x08" =~ /${a}0/; # True!
966 "aa\x08" =~ /${b}0/; # False
968 Several special variables also refer back to portions of the previous
969 match. C<$+> returns whatever the last bracket match matched.
970 C<$&> returns the entire matched string. (At one point C<$0> did
971 also, but now it returns the name of the program.) C<$`> returns
972 everything before the matched string. C<$'> returns everything
973 after the matched string. And C<$^N> contains whatever was matched by
974 the most-recently closed group (submatch). C<$^N> can be used in
975 extended patterns (see below), for example to assign a submatch to a
977 X<$+> X<$^N> X<$&> X<$`> X<$'>
979 These special variables, like the C<%+> hash and the numbered match variables
980 (C<$1>, C<$2>, C<$3>, etc.) are dynamically scoped
981 until the end of the enclosing block or until the next successful
982 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
983 X<$+> X<$^N> X<$&> X<$`> X<$'>
984 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
986 B<NOTE>: Failed matches in Perl do not reset the match variables,
987 which makes it easier to write code that tests for a series of more
988 specific cases and remembers the best match.
990 B<WARNING>: If your code is to run on Perl 5.16 or earlier,
991 beware that once Perl sees that you need one of C<$&>, C<$`>, or
992 C<$'> anywhere in the program, it has to provide them for every
993 pattern match. This may substantially slow your program.
995 Perl uses the same mechanism to produce C<$1>, C<$2>, etc, so you also
996 pay a price for each pattern that contains capturing parentheses.
997 (To avoid this cost while retaining the grouping behaviour, use the
998 extended regular expression C<(?: ... )> instead.) But if you never
999 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
1000 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
1001 if you can, but if you can't (and some algorithms really appreciate
1002 them), once you've used them once, use them at will, because you've
1003 already paid the price.
1006 Perl 5.16 introduced a slightly more efficient mechanism that notes
1007 separately whether each of C<$`>, C<$&>, and C<$'> have been seen, and
1008 thus may only need to copy part of the string. Perl 5.20 introduced a
1009 much more efficient copy-on-write mechanism which eliminates any slowdown.
1011 As another workaround for this problem, Perl 5.10.0 introduced C<${^PREMATCH}>,
1012 C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&>
1013 and C<$'>, B<except> that they are only guaranteed to be defined after a
1014 successful match that was executed with the C</p> (preserve) modifier.
1015 The use of these variables incurs no global performance penalty, unlike
1016 their punctuation character equivalents, however at the trade-off that you
1017 have to tell perl when you want to use them. As of Perl 5.20, these three
1018 variables are equivalent to C<$`>, C<$&> and C<$'>, and C</p> is ignored.
1021 =head2 Quoting metacharacters
1023 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
1024 C<\w>, C<\n>. Unlike some other regular expression languages, there
1025 are no backslashed symbols that aren't alphanumeric. So anything
1026 that looks like C<\\>, C<\(>, C<\)>, C<\[>, C<\]>, C<\{>, or C<\}> is
1028 interpreted as a literal character, not a metacharacter. This was
1029 once used in a common idiom to disable or quote the special meanings
1030 of regular expression metacharacters in a string that you want to
1031 use for a pattern. Simply quote all non-"word" characters:
1033 $pattern =~ s/(\W)/\\$1/g;
1035 (If C<use locale> is set, then this depends on the current locale.)
1036 Today it is more common to use the C<L<quotemeta()|perlfunc/quotemeta>>
1037 function or the C<\Q> metaquoting escape sequence to disable all
1038 metacharacters' special meanings like this:
1040 /$unquoted\Q$quoted\E$unquoted/
1042 Beware that if you put literal backslashes (those not inside
1043 interpolated variables) between C<\Q> and C<\E>, double-quotish
1044 backslash interpolation may lead to confusing results. If you
1045 I<need> to use literal backslashes within C<\Q...\E>,
1046 consult L<perlop/"Gory details of parsing quoted constructs">.
1048 C<quotemeta()> and C<\Q> are fully described in L<perlfunc/quotemeta>.
1050 =head2 Extended Patterns
1052 Perl also defines a consistent extension syntax for features not
1053 found in standard tools like B<awk> and
1054 B<lex>. The syntax for most of these is a
1055 pair of parentheses with a question mark as the first thing within
1056 the parentheses. The character after the question mark indicates
1059 The stability of these extensions varies widely. Some have been
1060 part of the core language for many years. Others are experimental
1061 and may change without warning or be completely removed. Check
1062 the documentation on an individual feature to verify its current
1065 A question mark was chosen for this and for the minimal-matching
1066 construct because 1) question marks are rare in older regular
1067 expressions, and 2) whenever you see one, you should stop and
1068 "question" exactly what is going on. That's psychology....
1075 A comment. The text is ignored.
1076 Note that Perl closes
1077 the comment as soon as it sees a C<")">, so there is no way to put a literal
1078 C<")"> in the comment. The pattern's closing delimiter must be escaped by
1079 a backslash if it appears in the comment.
1081 See L</E<sol>x> for another way to have comments in patterns.
1083 =item C<(?adlupimnsx-imnsx)>
1085 =item C<(?^alupimnsx)>
1088 One or more embedded pattern-match modifiers, to be turned on (or
1089 turned off, if preceded by C<"-">) for the remainder of the pattern or
1090 the remainder of the enclosing pattern group (if any).
1092 This is particularly useful for dynamic patterns, such as those read in from a
1093 configuration file, taken from an argument, or specified in a table
1094 somewhere. Consider the case where some patterns want to be
1095 case-sensitive and some do not: The case-insensitive ones merely need to
1096 include C<(?i)> at the front of the pattern. For example:
1098 $pattern = "foobar";
1099 if ( /$pattern/i ) { }
1103 $pattern = "(?i)foobar";
1104 if ( /$pattern/ ) { }
1106 These modifiers are restored at the end of the enclosing group. For example,
1108 ( (?i) blah ) \s+ \g1
1110 will match C<blah> in any case, some spaces, and an exact (I<including the case>!)
1111 repetition of the previous word, assuming the C</x> modifier, and no C</i>
1112 modifier outside this group.
1114 These modifiers do not carry over into named subpatterns called in the
1115 enclosing group. In other words, a pattern such as C<((?i)(?&NAME))> does not
1116 change the case-sensitivity of the C<"NAME"> pattern.
1118 Any of these modifiers can be set to apply globally to all regular
1119 expressions compiled within the scope of a C<use re>. See
1120 L<re/"'/flags' mode">.
1122 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
1123 after the C<"?"> is a shorthand equivalent to C<d-imnsx>. Flags (except
1124 C<"d">) may follow the caret to override it.
1125 But a minus sign is not legal with it.
1127 Note that the C<a>, C<d>, C<l>, C<p>, and C<u> modifiers are special in
1128 that they can only be enabled, not disabled, and the C<a>, C<d>, C<l>, and
1129 C<u> modifiers are mutually exclusive: specifying one de-specifies the
1130 others, and a maximum of one (or two C<a>'s) may appear in the
1131 construct. Thus, for
1132 example, C<(?-p)> will warn when compiled under C<use warnings>;
1133 C<(?-d:...)> and C<(?dl:...)> are fatal errors.
1135 Note also that the C<p> modifier is special in that its presence
1136 anywhere in a pattern has a global effect.
1138 =item C<(?:pattern)>
1141 =item C<(?adluimnsx-imnsx:pattern)>
1143 =item C<(?^aluimnsx:pattern)>
1146 This is for clustering, not capturing; it groups subexpressions like
1147 C<"()">, but doesn't make backreferences as C<"()"> does. So
1149 @fields = split(/\b(?:a|b|c)\b/)
1153 @fields = split(/\b(a|b|c)\b/)
1155 but doesn't spit out extra fields. It's also cheaper not to capture
1156 characters if you don't need to.
1158 Any letters between C<"?"> and C<":"> act as flags modifiers as with
1159 C<(?adluimnsx-imnsx)>. For example,
1161 /(?s-i:more.*than).*million/i
1163 is equivalent to the more verbose
1165 /(?:(?s-i)more.*than).*million/i
1167 Note that any C<()> constructs enclosed within this one will still
1168 capture unless the C</n> modifier is in effect.
1170 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
1171 after the C<"?"> is a shorthand equivalent to C<d-imnsx>. Any positive
1172 flags (except C<"d">) may follow the caret, so
1180 The caret tells Perl that this cluster doesn't inherit the flags of any
1181 surrounding pattern, but uses the system defaults (C<d-imnsx>),
1182 modified by any flags specified.
1184 The caret allows for simpler stringification of compiled regular
1185 expressions. These look like
1189 with any non-default flags appearing between the caret and the colon.
1190 A test that looks at such stringification thus doesn't need to have the
1191 system default flags hard-coded in it, just the caret. If new flags are
1192 added to Perl, the meaning of the caret's expansion will change to include
1193 the default for those flags, so the test will still work, unchanged.
1195 Specifying a negative flag after the caret is an error, as the flag is
1198 Mnemonic for C<(?^...)>: A fresh beginning since the usual use of a caret is
1199 to match at the beginning.
1201 =item C<(?|pattern)>
1202 X<(?|)> X<Branch reset>
1204 This is the "branch reset" pattern, which has the special property
1205 that the capture groups are numbered from the same starting point
1206 in each alternation branch. It is available starting from perl 5.10.0.
1208 Capture groups are numbered from left to right, but inside this
1209 construct the numbering is restarted for each branch.
1211 The numbering within each branch will be as normal, and any groups
1212 following this construct will be numbered as though the construct
1213 contained only one branch, that being the one with the most capture
1216 This construct is useful when you want to capture one of a
1217 number of alternative matches.
1219 Consider the following pattern. The numbers underneath show in
1220 which group the captured content will be stored.
1223 # before ---------------branch-reset----------- after
1224 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1227 Be careful when using the branch reset pattern in combination with
1228 named captures. Named captures are implemented as being aliases to
1229 numbered groups holding the captures, and that interferes with the
1230 implementation of the branch reset pattern. If you are using named
1231 captures in a branch reset pattern, it's best to use the same names,
1232 in the same order, in each of the alternations:
1234 /(?| (?<a> x ) (?<b> y )
1235 | (?<a> z ) (?<b> w )) /x
1237 Not doing so may lead to surprises:
1239 "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x;
1240 say $+ {a}; # Prints '12'
1241 say $+ {b}; # *Also* prints '12'.
1243 The problem here is that both the group named C<< a >> and the group
1244 named C<< b >> are aliases for the group belonging to C<< $1 >>.
1246 =item Lookaround Assertions
1247 X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
1249 Lookaround assertions are zero-width patterns which match a specific
1250 pattern without including it in C<$&>. Positive assertions match when
1251 their subpattern matches, negative assertions match when their subpattern
1252 fails. Lookbehind matches text up to the current match position,
1253 lookahead matches text following the current match position.
1257 =item C<(?=pattern)>
1258 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
1260 A zero-width positive lookahead assertion. For example, C</\w+(?=\t)/>
1261 matches a word followed by a tab, without including the tab in C<$&>.
1263 =item C<(?!pattern)>
1264 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
1266 A zero-width negative lookahead assertion. For example C</foo(?!bar)/>
1267 matches any occurrence of "foo" that isn't followed by "bar". Note
1268 however that lookahead and lookbehind are NOT the same thing. You cannot
1269 use this for lookbehind.
1271 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
1272 will not do what you want. That's because the C<(?!foo)> is just saying that
1273 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
1274 match. Use lookbehind instead (see below).
1276 =item C<(?<=pattern)> C<\K>
1277 X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K>
1279 A zero-width positive lookbehind assertion. For example, C</(?<=\t)\w+/>
1280 matches a word that follows a tab, without including the tab in C<$&>.
1281 Works only for fixed-width lookbehind.
1283 There is a special form of this construct, called C<\K> (available since
1284 Perl 5.10.0), which causes the
1285 regex engine to "keep" everything it had matched prior to the C<\K> and
1286 not include it in C<$&>. This effectively provides variable-length
1287 lookbehind. The use of C<\K> inside of another lookaround assertion
1288 is allowed, but the behaviour is currently not well defined.
1290 For various reasons C<\K> may be significantly more efficient than the
1291 equivalent C<< (?<=...) >> construct, and it is especially useful in
1292 situations where you want to efficiently remove something following
1293 something else in a string. For instance
1297 can be rewritten as the much more efficient
1301 =item C<(?<!pattern)>
1302 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
1304 A zero-width negative lookbehind assertion. For example C</(?<!bar)foo/>
1305 matches any occurrence of "foo" that does not follow "bar". Works
1306 only for fixed-width lookbehind.
1310 =item C<(?'NAME'pattern)>
1312 =item C<< (?<NAME>pattern) >>
1313 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
1315 A named capture group. Identical in every respect to normal capturing
1316 parentheses C<()> but for the additional fact that the group
1317 can be referred to by name in various regular expression
1318 constructs (like C<\g{NAME}>) and can be accessed by name
1319 after a successful match via C<%+> or C<%->. See L<perlvar>
1320 for more details on the C<%+> and C<%-> hashes.
1322 If multiple distinct capture groups have the same name then the
1323 C<$+{NAME}> will refer to the leftmost defined group in the match.
1325 The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent.
1327 B<NOTE:> While the notation of this construct is the same as the similar
1328 function in .NET regexes, the behavior is not. In Perl the groups are
1329 numbered sequentially regardless of being named or not. Thus in the
1334 C<$+{I<foo>}> will be the same as C<$2>, and C<$3> will contain 'z' instead of
1335 the opposite which is what a .NET regex hacker might expect.
1337 Currently I<NAME> is restricted to simple identifiers only.
1338 In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
1339 its Unicode extension (see L<utf8>),
1340 though it isn't extended by the locale (see L<perllocale>).
1342 B<NOTE:> In order to make things easier for programmers with experience
1343 with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >>
1344 may be used instead of C<< (?<NAME>pattern) >>; however this form does not
1345 support the use of single quotes as a delimiter for the name.
1347 =item C<< \k<NAME> >>
1349 =item C<< \k'NAME' >>
1351 Named backreference. Similar to numeric backreferences, except that
1352 the group is designated by name and not number. If multiple groups
1353 have the same name then it refers to the leftmost defined group in
1356 It is an error to refer to a name not defined by a C<< (?<NAME>) >>
1357 earlier in the pattern.
1359 Both forms are equivalent.
1361 B<NOTE:> In order to make things easier for programmers with experience
1362 with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >>
1363 may be used instead of C<< \k<NAME> >>.
1365 =item C<(?{ code })>
1366 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
1368 B<WARNING>: Using this feature safely requires that you understand its
1369 limitations. Code executed that has side effects may not perform identically
1370 from version to version due to the effect of future optimisations in the regex
1371 engine. For more information on this, see L</Embedded Code Execution
1374 This zero-width assertion executes any embedded Perl code. It always
1375 succeeds, and its return value is set as C<$^R>.
1377 In literal patterns, the code is parsed at the same time as the
1378 surrounding code. While within the pattern, control is passed temporarily
1379 back to the perl parser, until the logically-balancing closing brace is
1380 encountered. This is similar to the way that an array index expression in
1381 a literal string is handled, for example
1383 "abc$array[ 1 + f('[') + g()]def"
1385 In particular, braces do not need to be balanced:
1387 s/abc(?{ f('{'); })/def/
1389 Even in a pattern that is interpolated and compiled at run-time, literal
1390 code blocks will be compiled once, at perl compile time; the following
1394 my $qr = qr/(?{ BEGIN { print "A" } })/;
1396 /$foo$qr(?{ BEGIN { print "B" } })/;
1399 In patterns where the text of the code is derived from run-time
1400 information rather than appearing literally in a source code /pattern/,
1401 the code is compiled at the same time that the pattern is compiled, and
1402 for reasons of security, C<use re 'eval'> must be in scope. This is to
1403 stop user-supplied patterns containing code snippets from being
1406 In situations where you need to enable this with C<use re 'eval'>, you should
1407 also have taint checking enabled. Better yet, use the carefully
1408 constrained evaluation within a Safe compartment. See L<perlsec> for
1409 details about both these mechanisms.
1411 From the viewpoint of parsing, lexical variable scope and closures,
1415 behaves approximately like
1417 /AAA/ && do { BBB } && /CCC/
1421 qr/AAA(?{ BBB })CCC/
1423 behaves approximately like
1425 sub { /AAA/ && do { BBB } && /CCC/ }
1429 { my $i = 1; $r = qr/(?{ print $i })/ }
1433 Inside a C<(?{...})> block, C<$_> refers to the string the regular
1434 expression is matching against. You can also use C<pos()> to know what is
1435 the current position of matching within this string.
1437 The code block introduces a new scope from the perspective of lexical
1438 variable declarations, but B<not> from the perspective of C<local> and
1439 similar localizing behaviours. So later code blocks within the same
1440 pattern will still see the values which were localized in earlier blocks.
1441 These accumulated localizations are undone either at the end of a
1442 successful match, or if the assertion is backtracked (compare
1443 L<"Backtracking">). For example,
1447 (?{ $cnt = 0 }) # Initialize $cnt.
1451 local $cnt = $cnt + 1; # Update $cnt,
1452 # backtracking-safe.
1456 (?{ $res = $cnt }) # On success copy to
1457 # non-localized location.
1460 will initially increment C<$cnt> up to 8; then during backtracking, its
1461 value will be unwound back to 4, which is the value assigned to C<$res>.
1462 At the end of the regex execution, C<$cnt> will be wound back to its initial
1465 This assertion may be used as the condition in a
1467 (?(condition)yes-pattern|no-pattern)
1469 switch. If I<not> used in this way, the result of evaluation of C<code>
1470 is put into the special variable C<$^R>. This happens immediately, so
1471 C<$^R> can be used from other C<(?{ code })> assertions inside the same
1474 The assignment to C<$^R> above is properly localized, so the old
1475 value of C<$^R> is restored if the assertion is backtracked; compare
1478 Note that the special variable C<$^N> is particularly useful with code
1479 blocks to capture the results of submatches in variables without having to
1480 keep track of the number of nested parentheses. For example:
1482 $_ = "The brown fox jumps over the lazy dog";
1483 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
1484 print "color = $color, animal = $animal\n";
1487 =item C<(??{ code })>
1489 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
1491 B<WARNING>: Using this feature safely requires that you understand its
1492 limitations. Code executed that has side effects may not perform
1493 identically from version to version due to the effect of future
1494 optimisations in the regex engine. For more information on this, see
1495 L</Embedded Code Execution Frequency>.
1497 This is a "postponed" regular subexpression. It behaves in I<exactly> the
1498 same way as a C<(?{ code })> code block as described above, except that
1499 its return value, rather than being assigned to C<$^R>, is treated as a
1500 pattern, compiled if it's a string (or used as-is if its a qr// object),
1501 then matched as if it were inserted instead of this construct.
1503 During the matching of this sub-pattern, it has its own set of
1504 captures which are valid during the sub-match, but are discarded once
1505 control returns to the main pattern. For example, the following matches,
1506 with the inner pattern capturing "B" and matching "BB", while the outer
1507 pattern captures "A";
1509 my $inner = '(.)\1';
1510 "ABBA" =~ /^(.)(??{ $inner })\1/;
1511 print $1; # prints "A";
1513 Note that this means that there is no way for the inner pattern to refer
1514 to a capture group defined outside. (The code block itself can use C<$1>,
1515 etc., to refer to the enclosing pattern's capture groups.) Thus, although
1517 ('a' x 100)=~/(??{'(.)' x 100})/
1519 I<will> match, it will I<not> set C<$1> on exit.
1521 The following pattern matches a parenthesized group:
1526 (?> [^()]+ ) # Non-parens without backtracking
1528 (??{ $re }) # Group with matching parens
1534 L<C<(?I<PARNO>)>|/(?PARNO) (?-PARNO) (?+PARNO) (?R) (?0)>
1535 for a different, more efficient way to accomplish
1538 Executing a postponed regular expression 50 times without consuming any
1539 input string will result in a fatal error. The maximum depth is compiled
1540 into perl, so changing it requires a custom build.
1542 =item C<(?I<PARNO>)> C<(?-I<PARNO>)> C<(?+I<PARNO>)> C<(?R)> C<(?0)>
1543 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
1544 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
1545 X<regex, relative recursion> X<GOSUB> X<GOSTART>
1547 Recursive subpattern. Treat the contents of a given capture buffer in the
1548 current pattern as an independent subpattern and attempt to match it at
1549 the current position in the string. Information about capture state from
1550 the caller for things like backreferences is available to the subpattern,
1551 but capture buffers set by the subpattern are not visible to the caller.
1553 Similar to C<(??{ code })> except that it does not involve executing any
1554 code or potentially compiling a returned pattern string; instead it treats
1555 the part of the current pattern contained within a specified capture group
1556 as an independent pattern that must match at the current position. Also
1557 different is the treatment of capture buffers, unlike C<(??{ code })>
1558 recursive patterns have access to their caller's match state, so one can
1559 use backreferences safely.
1561 I<PARNO> is a sequence of digits (not starting with 0) whose value reflects
1562 the paren-number of the capture group to recurse to. C<(?R)> recurses to
1563 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
1564 C<(?R)>. If I<PARNO> is preceded by a plus or minus sign then it is assumed
1565 to be relative, with negative numbers indicating preceding capture groups
1566 and positive ones following. Thus C<(?-1)> refers to the most recently
1567 declared group, and C<(?+1)> indicates the next group to be declared.
1568 Note that the counting for relative recursion differs from that of
1569 relative backreferences, in that with recursion unclosed groups B<are>
1572 The following pattern matches a function C<foo()> which may contain
1573 balanced parentheses as the argument.
1575 $re = qr{ ( # paren group 1 (full function)
1577 ( # paren group 2 (parens)
1579 ( # paren group 3 (contents of parens)
1581 (?> [^()]+ ) # Non-parens without backtracking
1583 (?2) # Recurse to start of paren group 2
1591 If the pattern was used as follows
1593 'foo(bar(baz)+baz(bop))'=~/$re/
1594 and print "\$1 = $1\n",
1598 the output produced should be the following:
1600 $1 = foo(bar(baz)+baz(bop))
1601 $2 = (bar(baz)+baz(bop))
1602 $3 = bar(baz)+baz(bop)
1604 If there is no corresponding capture group defined, then it is a
1605 fatal error. Recursing deeper than 50 times without consuming any input
1606 string will also result in a fatal error. The maximum depth is compiled
1607 into perl, so changing it requires a custom build.
1609 The following shows how using negative indexing can make it
1610 easier to embed recursive patterns inside of a C<qr//> construct
1613 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
1614 if (/foo $parens \s+ \+ \s+ bar $parens/x) {
1615 # do something here...
1618 B<Note> that this pattern does not behave the same way as the equivalent
1619 PCRE or Python construct of the same form. In Perl you can backtrack into
1620 a recursed group, in PCRE and Python the recursed into group is treated
1621 as atomic. Also, modifiers are resolved at compile time, so constructs
1622 like C<(?i:(?1))> or C<(?:(?i)(?1))> do not affect how the sub-pattern will
1628 Recurse to a named subpattern. Identical to C<(?I<PARNO>)> except that the
1629 parenthesis to recurse to is determined by name. If multiple parentheses have
1630 the same name, then it recurses to the leftmost.
1632 It is an error to refer to a name that is not declared somewhere in the
1635 B<NOTE:> In order to make things easier for programmers with experience
1636 with the Python or PCRE regex engines the pattern C<< (?P>NAME) >>
1637 may be used instead of C<< (?&NAME) >>.
1639 =item C<(?(condition)yes-pattern|no-pattern)>
1642 =item C<(?(condition)yes-pattern)>
1644 Conditional expression. Matches C<yes-pattern> if C<condition> yields
1645 a true value, matches C<no-pattern> otherwise. A missing pattern always
1648 C<(condition)> should be one of:
1652 =item an integer in parentheses
1654 (which is valid if the corresponding pair of parentheses
1657 =item a lookahead/lookbehind/evaluate zero-width assertion;
1659 =item a name in angle brackets or single quotes
1661 (which is valid if a group with the given name matched);
1663 =item the special symbol C<(R)>
1665 (true when evaluated inside of recursion or eval). Additionally the
1667 followed by a number, (which will be true when evaluated when recursing
1668 inside of the appropriate group), or by C<&NAME>, in which case it will
1669 be true only when evaluated during recursion in the named group.
1673 Here's a summary of the possible predicates:
1677 =item C<(1)> C<(2)> ...
1679 Checks if the numbered capturing group has matched something.
1681 =item C<(E<lt>I<NAME>E<gt>)> C<('I<NAME>')>
1683 Checks if a group with the given name has matched something.
1685 =item C<(?=...)> C<(?!...)> C<(?<=...)> C<(?<!...)>
1687 Checks whether the pattern matches (or does not match, for the C<"!">
1690 =item C<(?{ I<CODE> })>
1692 Treats the return value of the code block as the condition.
1696 Checks if the expression has been evaluated inside of recursion.
1698 =item C<(R1)> C<(R2)> ...
1700 Checks if the expression has been evaluated while executing directly
1701 inside of the n-th capture group. This check is the regex equivalent of
1703 if ((caller(0))[3] eq 'subname') { ... }
1705 In other words, it does not check the full recursion stack.
1707 =item C<(R&I<NAME>)>
1709 Similar to C<(R1)>, this predicate checks to see if we're executing
1710 directly inside of the leftmost group with a given name (this is the same
1711 logic used by C<(?&I<NAME>)> to disambiguate). It does not check the full
1712 stack, but only the name of the innermost active recursion.
1716 In this case, the yes-pattern is never directly executed, and no
1717 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1718 See below for details.
1729 matches a chunk of non-parentheses, possibly included in parentheses
1732 A special form is the C<(DEFINE)> predicate, which never executes its
1733 yes-pattern directly, and does not allow a no-pattern. This allows one to
1734 define subpatterns which will be executed only by the recursion mechanism.
1735 This way, you can define a set of regular expression rules that can be
1736 bundled into any pattern you choose.
1738 It is recommended that for this usage you put the DEFINE block at the
1739 end of the pattern, and that you name any subpatterns defined within it.
1741 Also, it's worth noting that patterns defined this way probably will
1742 not be as efficient, as the optimizer is not very clever about
1745 An example of how this might be used is as follows:
1747 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
1750 (?<ADDRESS_PAT>....)
1753 Note that capture groups matched inside of recursion are not accessible
1754 after the recursion returns, so the extra layer of capturing groups is
1755 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1756 C<$+{NAME}> would be.
1758 Finally, keep in mind that subpatterns created inside a DEFINE block
1759 count towards the absolute and relative number of captures, so this:
1761 my @captures = "a" =~ /(.) # First capture
1763 (?<EXAMPLE> 1 ) # Second capture
1765 say scalar @captures;
1767 Will output 2, not 1. This is particularly important if you intend to
1768 compile the definitions with the C<qr//> operator, and later
1769 interpolate them in another pattern.
1771 =item C<< (?>pattern) >>
1772 X<backtrack> X<backtracking> X<atomic> X<possessive>
1774 An "independent" subexpression, one which matches the substring
1775 that a I<standalone> C<pattern> would match if anchored at the given
1776 position, and it matches I<nothing other than this substring>. This
1777 construct is useful for optimizations of what would otherwise be
1778 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1779 It may also be useful in places where the "grab all you can, and do not
1780 give anything back" semantic is desirable.
1782 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1783 (anchored at the beginning of string, as above) will match I<all>
1784 characters C<a> at the beginning of string, leaving no C<a> for
1785 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1786 since the match of the subgroup C<a*> is influenced by the following
1787 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1788 C<a*ab> will match fewer characters than a standalone C<a*>, since
1789 this makes the tail match.
1791 C<< (?>pattern) >> does not disable backtracking altogether once it has
1792 matched. It is still possible to backtrack past the construct, but not
1793 into it. So C<< ((?>a*)|(?>b*))ar >> will still match "bar".
1795 An effect similar to C<< (?>pattern) >> may be achieved by writing
1796 C<(?=(pattern))\g{-1}>. This matches the same substring as a standalone
1797 C<a+>, and the following C<\g{-1}> eats the matched string; it therefore
1798 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1799 (The difference between these two constructs is that the second one
1800 uses a capturing group, thus shifting ordinals of backreferences
1801 in the rest of a regular expression.)
1803 Consider this pattern:
1814 That will efficiently match a nonempty group with matching parentheses
1815 two levels deep or less. However, if there is no such group, it
1816 will take virtually forever on a long string. That's because there
1817 are so many different ways to split a long string into several
1818 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1819 to a subpattern of the above pattern. Consider how the pattern
1820 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1821 seconds, but that each extra letter doubles this time. This
1822 exponential performance will make it appear that your program has
1823 hung. However, a tiny change to this pattern
1827 (?> [^()]+ ) # change x+ above to (?> x+ )
1834 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1835 this yourself would be a productive exercise), but finishes in a fourth
1836 the time when used on a similar string with 1000000 C<a>s. Be aware,
1837 however, that, when this construct is followed by a
1838 quantifier, it currently triggers a warning message under
1839 the C<use warnings> pragma or B<-w> switch saying it
1840 C<"matches null string many times in regex">.
1842 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1843 effect may be achieved by negative lookahead, as in C<[^()]+ (?! [^()] )>.
1844 This was only 4 times slower on a string with 1000000 C<a>s.
1846 The "grab all you can, and do not give anything back" semantic is desirable
1847 in many situations where on the first sight a simple C<()*> looks like
1848 the correct solution. Suppose we parse text with comments being delimited
1849 by C<"#"> followed by some optional (horizontal) whitespace. Contrary to
1850 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1851 the comment delimiter, because it may "give up" some whitespace if
1852 the remainder of the pattern can be made to match that way. The correct
1853 answer is either one of these:
1858 For example, to grab non-empty comments into C<$1>, one should use either
1861 / (?> \# [ \t]* ) ( .+ ) /x;
1862 / \# [ \t]* ( [^ \t] .* ) /x;
1864 Which one you pick depends on which of these expressions better reflects
1865 the above specification of comments.
1867 In some literature this construct is called "atomic matching" or
1868 "possessive matching".
1870 Possessive quantifiers are equivalent to putting the item they are applied
1871 to inside of one of these constructs. The following equivalences apply:
1873 Quantifier Form Bracketing Form
1874 --------------- ---------------
1878 PAT{min,max}+ (?>PAT{min,max})
1882 See L<perlrecharclass/Extended Bracketed Character Classes>.
1886 =head2 Special Backtracking Control Verbs
1888 These special patterns are generally of the form C<(*I<VERB>:I<ARG>)>. Unless
1889 otherwise stated the I<ARG> argument is optional; in some cases, it is
1892 Any pattern containing a special backtracking verb that allows an argument
1893 has the special behaviour that when executed it sets the current package's
1894 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1897 On failure, the C<$REGERROR> variable will be set to the I<ARG> value of the
1898 verb pattern, if the verb was involved in the failure of the match. If the
1899 I<ARG> part of the pattern was omitted, then C<$REGERROR> will be set to the
1900 name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1901 none. Also, the C<$REGMARK> variable will be set to FALSE.
1903 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1904 the C<$REGMARK> variable will be set to the name of the last
1905 C<(*MARK:NAME)> pattern executed. See the explanation for the
1906 C<(*MARK:NAME)> verb below for more details.
1908 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1909 and most other regex-related variables. They are not local to a scope, nor
1910 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1911 Use C<local> to localize changes to them to a specific scope if necessary.
1913 If a pattern does not contain a special backtracking verb that allows an
1914 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
1922 =item C<(*PRUNE)> C<(*PRUNE:NAME)>
1923 X<(*PRUNE)> X<(*PRUNE:NAME)>
1925 This zero-width pattern prunes the backtracking tree at the current point
1926 when backtracked into on failure. Consider the pattern C<I<A> (*PRUNE) I<B>>,
1927 where I<A> and I<B> are complex patterns. Until the C<(*PRUNE)> verb is reached,
1928 I<A> may backtrack as necessary to match. Once it is reached, matching
1929 continues in I<B>, which may also backtrack as necessary; however, should B
1930 not match, then no further backtracking will take place, and the pattern
1931 will fail outright at the current starting position.
1933 The following example counts all the possible matching strings in a
1934 pattern (without actually matching any of them).
1936 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1937 print "Count=$count\n";
1952 If we add a C<(*PRUNE)> before the count like the following
1954 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
1955 print "Count=$count\n";
1957 we prevent backtracking and find the count of the longest matching string
1958 at each matching starting point like so:
1965 Any number of C<(*PRUNE)> assertions may be used in a pattern.
1967 See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1968 control backtracking. In some cases, the use of C<(*PRUNE)> can be
1969 replaced with a C<< (?>pattern) >> with no functional difference; however,
1970 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1971 C<< (?>pattern) >> alone.
1973 =item C<(*SKIP)> C<(*SKIP:NAME)>
1976 This zero-width pattern is similar to C<(*PRUNE)>, except that on
1977 failure it also signifies that whatever text that was matched leading up
1978 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1979 of this pattern. This effectively means that the regex engine "skips" forward
1980 to this position on failure and tries to match again, (assuming that
1981 there is sufficient room to match).
1983 The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1984 C<(*MARK:NAME)> was encountered while matching, then it is that position
1985 which is used as the "skip point". If no C<(*MARK)> of that name was
1986 encountered, then the C<(*SKIP)> operator has no effect. When used
1987 without a name the "skip point" is where the match point was when
1988 executing the C<(*SKIP)> pattern.
1990 Compare the following to the examples in C<(*PRUNE)>; note the string
1993 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
1994 print "Count=$count\n";
2002 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
2003 executed, the next starting point will be where the cursor was when the
2004 C<(*SKIP)> was executed.
2006 =item C<(*MARK:NAME)> C<(*:NAME)>
2007 X<(*MARK)> X<(*MARK:NAME)> X<(*:NAME)>
2009 This zero-width pattern can be used to mark the point reached in a string
2010 when a certain part of the pattern has been successfully matched. This
2011 mark may be given a name. A later C<(*SKIP)> pattern will then skip
2012 forward to that point if backtracked into on failure. Any number of
2013 C<(*MARK)> patterns are allowed, and the I<NAME> portion may be duplicated.
2015 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
2016 can be used to "label" a pattern branch, so that after matching, the
2017 program can determine which branches of the pattern were involved in the
2020 When a match is successful, the C<$REGMARK> variable will be set to the
2021 name of the most recently executed C<(*MARK:NAME)> that was involved
2024 This can be used to determine which branch of a pattern was matched
2025 without using a separate capture group for each branch, which in turn
2026 can result in a performance improvement, as perl cannot optimize
2027 C</(?:(x)|(y)|(z))/> as efficiently as something like
2028 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
2030 When a match has failed, and unless another verb has been involved in
2031 failing the match and has provided its own name to use, the C<$REGERROR>
2032 variable will be set to the name of the most recently executed
2035 See L</(*SKIP)> for more details.
2037 As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
2039 =item C<(*THEN)> C<(*THEN:NAME)>
2041 This is similar to the "cut group" operator C<::> from Perl 6. Like
2042 C<(*PRUNE)>, this verb always matches, and when backtracked into on
2043 failure, it causes the regex engine to try the next alternation in the
2044 innermost enclosing group (capturing or otherwise) that has alternations.
2045 The two branches of a C<(?(condition)yes-pattern|no-pattern)> do not
2046 count as an alternation, as far as C<(*THEN)> is concerned.
2048 Its name comes from the observation that this operation combined with the
2049 alternation operator (C<"|">) can be used to create what is essentially a
2050 pattern-based if/then/else block:
2052 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
2054 Note that if this operator is used and NOT inside of an alternation then
2055 it acts exactly like the C<(*PRUNE)> operator.
2065 / ( A (*THEN) B | C ) /
2069 / ( A (*PRUNE) B | C ) /
2071 as after matching the I<A> but failing on the I<B> the C<(*THEN)> verb will
2072 backtrack and try I<C>; but the C<(*PRUNE)> verb will simply fail.
2074 =item C<(*COMMIT)> C<(*COMMIT:args)>
2077 This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a
2078 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
2079 into on failure it causes the match to fail outright. No further attempts
2080 to find a valid match by advancing the start pointer will occur again.
2083 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
2084 print "Count=$count\n";
2091 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
2092 does not match, the regex engine will not try any further matching on the
2095 =item C<(*FAIL)> C<(*F)> C<(*FAIL:arg)>
2098 This pattern matches nothing and always fails. It can be used to force the
2099 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
2100 fact, C<(?!)> gets optimised into C<(*FAIL)> internally. You can provide
2101 an argument so that if the match fails because of this C<FAIL> directive
2102 the argument can be obtained from C<$REGERROR>.
2104 It is probably useful only when combined with C<(?{})> or C<(??{})>.
2106 =item C<(*ACCEPT)> C<(*ACCEPT:arg)>
2109 This pattern matches nothing and causes the end of successful matching at
2110 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
2111 whether there is actually more to match in the string. When inside of a
2112 nested pattern, such as recursion, or in a subpattern dynamically generated
2113 via C<(??{})>, only the innermost pattern is ended immediately.
2115 If the C<(*ACCEPT)> is inside of capturing groups then the groups are
2116 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
2119 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
2121 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
2122 be set. If another branch in the inner parentheses was matched, such as in the
2123 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
2125 You can provide an argument, which will be available in the var
2126 C<$REGMARK> after the match completes.
2133 X<backtrack> X<backtracking>
2135 NOTE: This section presents an abstract approximation of regular
2136 expression behavior. For a more rigorous (and complicated) view of
2137 the rules involved in selecting a match among possible alternatives,
2138 see L<Combining RE Pieces>.
2140 A fundamental feature of regular expression matching involves the
2141 notion called I<backtracking>, which is currently used (when needed)
2142 by all regular non-possessive expression quantifiers, namely C<"*">, C<"*?">, C<"+">,
2143 C<"+?">, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
2144 internally, but the general principle outlined here is valid.
2146 For a regular expression to match, the I<entire> regular expression must
2147 match, not just part of it. So if the beginning of a pattern containing a
2148 quantifier succeeds in a way that causes later parts in the pattern to
2149 fail, the matching engine backs up and recalculates the beginning
2150 part--that's why it's called backtracking.
2152 Here is an example of backtracking: Let's say you want to find the
2153 word following "foo" in the string "Food is on the foo table.":
2155 $_ = "Food is on the foo table.";
2156 if ( /\b(foo)\s+(\w+)/i ) {
2157 print "$2 follows $1.\n";
2160 When the match runs, the first part of the regular expression (C<\b(foo)>)
2161 finds a possible match right at the beginning of the string, and loads up
2162 C<$1> with "Foo". However, as soon as the matching engine sees that there's
2163 no whitespace following the "Foo" that it had saved in C<$1>, it realizes its
2164 mistake and starts over again one character after where it had the
2165 tentative match. This time it goes all the way until the next occurrence
2166 of "foo". The complete regular expression matches this time, and you get
2167 the expected output of "table follows foo."
2169 Sometimes minimal matching can help a lot. Imagine you'd like to match
2170 everything between "foo" and "bar". Initially, you write something
2173 $_ = "The food is under the bar in the barn.";
2174 if ( /foo(.*)bar/ ) {
2178 Which perhaps unexpectedly yields:
2180 got <d is under the bar in the >
2182 That's because C<.*> was greedy, so you get everything between the
2183 I<first> "foo" and the I<last> "bar". Here it's more effective
2184 to use minimal matching to make sure you get the text between a "foo"
2185 and the first "bar" thereafter.
2187 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
2188 got <d is under the >
2190 Here's another example. Let's say you'd like to match a number at the end
2191 of a string, and you also want to keep the preceding part of the match.
2194 $_ = "I have 2 numbers: 53147";
2195 if ( /(.*)(\d*)/ ) { # Wrong!
2196 print "Beginning is <$1>, number is <$2>.\n";
2199 That won't work at all, because C<.*> was greedy and gobbled up the
2200 whole string. As C<\d*> can match on an empty string the complete
2201 regular expression matched successfully.
2203 Beginning is <I have 2 numbers: 53147>, number is <>.
2205 Here are some variants, most of which don't work:
2207 $_ = "I have 2 numbers: 53147";
2220 printf "%-12s ", $pat;
2222 print "<$1> <$2>\n";
2228 That will print out:
2230 (.*)(\d*) <I have 2 numbers: 53147> <>
2231 (.*)(\d+) <I have 2 numbers: 5314> <7>
2233 (.*?)(\d+) <I have > <2>
2234 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
2235 (.*?)(\d+)$ <I have 2 numbers: > <53147>
2236 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
2237 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
2239 As you see, this can be a bit tricky. It's important to realize that a
2240 regular expression is merely a set of assertions that gives a definition
2241 of success. There may be 0, 1, or several different ways that the
2242 definition might succeed against a particular string. And if there are
2243 multiple ways it might succeed, you need to understand backtracking to
2244 know which variety of success you will achieve.
2246 When using lookahead assertions and negations, this can all get even
2247 trickier. Imagine you'd like to find a sequence of non-digits not
2248 followed by "123". You might try to write that as
2251 if ( /^\D*(?!123)/ ) { # Wrong!
2252 print "Yup, no 123 in $_\n";
2255 But that isn't going to match; at least, not the way you're hoping. It
2256 claims that there is no 123 in the string. Here's a clearer picture of
2257 why that pattern matches, contrary to popular expectations:
2262 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
2263 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
2265 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
2266 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
2274 You might have expected test 3 to fail because it seems to a more
2275 general purpose version of test 1. The important difference between
2276 them is that test 3 contains a quantifier (C<\D*>) and so can use
2277 backtracking, whereas test 1 will not. What's happening is
2278 that you've asked "Is it true that at the start of C<$x>, following 0 or more
2279 non-digits, you have something that's not 123?" If the pattern matcher had
2280 let C<\D*> expand to "ABC", this would have caused the whole pattern to
2283 The search engine will initially match C<\D*> with "ABC". Then it will
2284 try to match C<(?!123)> with "123", which fails. But because
2285 a quantifier (C<\D*>) has been used in the regular expression, the
2286 search engine can backtrack and retry the match differently
2287 in the hope of matching the complete regular expression.
2289 The pattern really, I<really> wants to succeed, so it uses the
2290 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
2291 time. Now there's indeed something following "AB" that is not
2292 "123". It's "C123", which suffices.
2294 We can deal with this by using both an assertion and a negation.
2295 We'll say that the first part in C<$1> must be followed both by a digit
2296 and by something that's not "123". Remember that the lookaheads
2297 are zero-width expressions--they only look, but don't consume any
2298 of the string in their match. So rewriting this way produces what
2299 you'd expect; that is, case 5 will fail, but case 6 succeeds:
2301 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
2302 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
2306 In other words, the two zero-width assertions next to each other work as though
2307 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
2308 matches only if you're at the beginning of the line AND the end of the
2309 line simultaneously. The deeper underlying truth is that juxtaposition in
2310 regular expressions always means AND, except when you write an explicit OR
2311 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
2312 although the attempted matches are made at different positions because "a"
2313 is not a zero-width assertion, but a one-width assertion.
2315 B<WARNING>: Particularly complicated regular expressions can take
2316 exponential time to solve because of the immense number of possible
2317 ways they can use backtracking to try for a match. For example, without
2318 internal optimizations done by the regular expression engine, this will
2319 take a painfully long time to run:
2321 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
2323 And if you used C<"*">'s in the internal groups instead of limiting them
2324 to 0 through 5 matches, then it would take forever--or until you ran
2325 out of stack space. Moreover, these internal optimizations are not
2326 always applicable. For example, if you put C<{0,5}> instead of C<"*">
2327 on the external group, no current optimization is applicable, and the
2328 match takes a long time to finish.
2330 A powerful tool for optimizing such beasts is what is known as an
2331 "independent group",
2332 which does not backtrack (see L</C<< (?>pattern) >>>). Note also that
2333 zero-length lookahead/lookbehind assertions will not backtrack to make
2334 the tail match, since they are in "logical" context: only
2335 whether they match is considered relevant. For an example
2336 where side-effects of lookahead I<might> have influenced the
2337 following match, see L</C<< (?>pattern) >>>.
2339 =head2 Version 8 Regular Expressions
2340 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
2342 In case you're not familiar with the "regular" Version 8 regex
2343 routines, here are the pattern-matching rules not described above.
2345 Any single character matches itself, unless it is a I<metacharacter>
2346 with a special meaning described here or above. You can cause
2347 characters that normally function as metacharacters to be interpreted
2348 literally by prefixing them with a C<"\"> (e.g., C<"\."> matches a C<".">, not any
2349 character; "\\" matches a C<"\">). This escape mechanism is also required
2350 for the character used as the pattern delimiter.
2352 A series of characters matches that series of characters in the target
2353 string, so the pattern C<blurfl> would match "blurfl" in the target
2356 You can specify a character class, by enclosing a list of characters
2357 in C<[]>, which will match any character from the list. If the
2358 first character after the C<"["> is C<"^">, the class matches any character not
2359 in the list. Within a list, the C<"-"> character specifies a
2360 range, so that C<a-z> represents all characters between "a" and "z",
2361 inclusive. If you want either C<"-"> or C<"]"> itself to be a member of a
2362 class, put it at the start of the list (possibly after a C<"^">), or
2363 escape it with a backslash. C<"-"> is also taken literally when it is
2364 at the end of the list, just before the closing C<"]">. (The
2365 following all specify the same class of three characters: C<[-az]>,
2366 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
2367 specifies a class containing twenty-six characters, even on EBCDIC-based
2368 character sets.) Also, if you try to use the character
2369 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
2370 a range, the C<"-"> is understood literally.
2372 Note also that the whole range idea is rather unportable between
2373 character sets, except for four situations that Perl handles specially.
2374 Any subset of the ranges C<[A-Z]>, C<[a-z]>, and C<[0-9]> are guaranteed
2375 to match the expected subset of ASCII characters, no matter what
2376 character set the platform is running. The fourth portable way to
2377 specify ranges is to use the C<\N{...}> syntax to specify either end
2378 point of the range. For example, C<[\N{U+04}-\N{U+07}]> means to match
2379 the Unicode code points C<\N{U+04}>, C<\N{U+05}>, C<\N{U+06}>, and
2380 C<\N{U+07}>, whatever their native values may be on the platform. Under
2381 L<use re 'strict'|re/'strict' mode> or within a L</C<(?[ ])>>, a warning
2382 is raised, if enabled, and the other end point of a range which has a
2383 C<\N{...}> endpoint is not portably specified. For example,
2385 [\N{U+00}-\x06] # Warning under "use re 'strict'".
2387 It is hard to understand without digging what exactly matches ranges
2388 other than subsets of C<[A-Z]>, C<[a-z]>, and C<[0-9]>. A sound
2389 principle is to use only ranges that begin from and end at either
2390 alphabetics of equal case ([a-e], [A-E]), or digits ([0-9]). Anything
2391 else is unsafe or unclear. If in doubt, spell out the range in full.
2393 Characters may be specified using a metacharacter syntax much like that
2394 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
2395 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
2396 of three octal digits, matches the character whose coded character set value
2397 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
2398 matches the character whose ordinal is I<nn>. The expression \cI<x>
2399 matches the character control-I<x>. Finally, the C<"."> metacharacter
2400 matches any character except "\n" (unless you use C</s>).
2402 You can specify a series of alternatives for a pattern using C<"|"> to
2403 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
2404 or "foe" in the target string (as would C<f(e|i|o)e>). The
2405 first alternative includes everything from the last pattern delimiter
2406 (C<"(">, "(?:", etc. or the beginning of the pattern) up to the first C<"|">, and
2407 the last alternative contains everything from the last C<"|"> to the next
2408 closing pattern delimiter. That's why it's common practice to include
2409 alternatives in parentheses: to minimize confusion about where they
2412 Alternatives are tried from left to right, so the first
2413 alternative found for which the entire expression matches, is the one that
2414 is chosen. This means that alternatives are not necessarily greedy. For
2415 example: when matching C<foo|foot> against "barefoot", only the "foo"
2416 part will match, as that is the first alternative tried, and it successfully
2417 matches the target string. (This might not seem important, but it is
2418 important when you are capturing matched text using parentheses.)
2420 Also remember that C<"|"> is interpreted as a literal within square brackets,
2421 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
2423 Within a pattern, you may designate subpatterns for later reference
2424 by enclosing them in parentheses, and you may refer back to the
2425 I<n>th subpattern later in the pattern using the metacharacter
2426 \I<n> or \gI<n>. Subpatterns are numbered based on the left to right order
2427 of their opening parenthesis. A backreference matches whatever
2428 actually matched the subpattern in the string being examined, not
2429 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\g1\d*> will
2430 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
2431 1 matched "0x", even though the rule C<0|0x> could potentially match
2432 the leading 0 in the second number.
2434 =head2 Warning on C<\1> Instead of C<$1>
2436 Some people get too used to writing things like:
2438 $pattern =~ s/(\W)/\\\1/g;
2440 This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid
2442 B<sed> addicts, but it's a dirty habit to get into. That's because in
2443 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
2444 the usual double-quoted string means a control-A. The customary Unix
2445 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
2446 of doing that, you get yourself into trouble if you then add an C</e>
2449 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
2455 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
2456 C<${1}000>. The operation of interpolation should not be confused
2457 with the operation of matching a backreference. Certainly they mean two
2458 different things on the I<left> side of the C<s///>.
2460 =head2 Repeated Patterns Matching a Zero-length Substring
2462 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
2464 Regular expressions provide a terse and powerful programming language. As
2465 with most other power tools, power comes together with the ability
2468 A common abuse of this power stems from the ability to make infinite
2469 loops using regular expressions, with something as innocuous as:
2471 'foo' =~ m{ ( o? )* }x;
2473 The C<o?> matches at the beginning of C<'foo'>, and since the position
2474 in the string is not moved by the match, C<o?> would match again and again
2475 because of the C<"*"> quantifier. Another common way to create a similar cycle
2476 is with the looping modifier C<//g>:
2478 @matches = ( 'foo' =~ m{ o? }xg );
2482 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
2484 or the loop implied by C<split()>.
2486 However, long experience has shown that many programming tasks may
2487 be significantly simplified by using repeated subexpressions that
2488 may match zero-length substrings. Here's a simple example being:
2490 @chars = split //, $string; # // is not magic in split
2491 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
2493 Thus Perl allows such constructs, by I<forcefully breaking
2494 the infinite loop>. The rules for this are different for lower-level
2495 loops given by the greedy quantifiers C<*+{}>, and for higher-level
2496 ones like the C</g> modifier or C<split()> operator.
2498 The lower-level loops are I<interrupted> (that is, the loop is
2499 broken) when Perl detects that a repeated expression matched a
2500 zero-length substring. Thus
2502 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
2504 is made equivalent to
2506 m{ (?: NON_ZERO_LENGTH )* (?: ZERO_LENGTH )? }x;
2508 For example, this program
2515 (?{print "hello"}) # print hello whenever this
2517 (?=(b)) # zero-width assertion
2518 )* # any number of times
2529 Notice that "hello" is only printed once, as when Perl sees that the sixth
2530 iteration of the outermost C<(?:)*> matches a zero-length string, it stops
2533 The higher-level loops preserve an additional state between iterations:
2534 whether the last match was zero-length. To break the loop, the following
2535 match after a zero-length match is prohibited to have a length of zero.
2536 This prohibition interacts with backtracking (see L<"Backtracking">),
2537 and so the I<second best> match is chosen if the I<best> match is of
2545 results in C<< <><b><><a><><r><> >>. At each position of the string the best
2546 match given by non-greedy C<??> is the zero-length match, and the I<second
2547 best> match is what is matched by C<\w>. Thus zero-length matches
2548 alternate with one-character-long matches.
2550 Similarly, for repeated C<m/()/g> the second-best match is the match at the
2551 position one notch further in the string.
2553 The additional state of being I<matched with zero-length> is associated with
2554 the matched string, and is reset by each assignment to C<pos()>.
2555 Zero-length matches at the end of the previous match are ignored
2558 =head2 Combining RE Pieces
2560 Each of the elementary pieces of regular expressions which were described
2561 before (such as C<ab> or C<\Z>) could match at most one substring
2562 at the given position of the input string. However, in a typical regular
2563 expression these elementary pieces are combined into more complicated
2564 patterns using combining operators C<ST>, C<S|T>, C<S*> etc.
2565 (in these examples C<S> and C<T> are regular subexpressions).
2567 Such combinations can include alternatives, leading to a problem of choice:
2568 if we match a regular expression C<a|ab> against C<"abc">, will it match
2569 substring C<"a"> or C<"ab">? One way to describe which substring is
2570 actually matched is the concept of backtracking (see L<"Backtracking">).
2571 However, this description is too low-level and makes you think
2572 in terms of a particular implementation.
2574 Another description starts with notions of "better"/"worse". All the
2575 substrings which may be matched by the given regular expression can be
2576 sorted from the "best" match to the "worst" match, and it is the "best"
2577 match which is chosen. This substitutes the question of "what is chosen?"
2578 by the question of "which matches are better, and which are worse?".
2580 Again, for elementary pieces there is no such question, since at most
2581 one match at a given position is possible. This section describes the
2582 notion of better/worse for combining operators. In the description
2583 below C<S> and C<T> are regular subexpressions.
2589 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
2590 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
2591 which can be matched by C<T>.
2593 If C<A> is a better match for C<S> than C<A'>, C<AB> is a better
2596 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
2597 C<B> is a better match for C<T> than C<B'>.
2601 When C<S> can match, it is a better match than when only C<T> can match.
2603 Ordering of two matches for C<S> is the same as for C<S>. Similar for
2604 two matches for C<T>.
2606 =item C<S{REPEAT_COUNT}>
2608 Matches as C<SSS...S> (repeated as many times as necessary).
2612 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
2614 =item C<S{min,max}?>
2616 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
2618 =item C<S?>, C<S*>, C<S+>
2620 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
2622 =item C<S??>, C<S*?>, C<S+?>
2624 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
2628 Matches the best match for C<S> and only that.
2630 =item C<(?=S)>, C<(?<=S)>
2632 Only the best match for C<S> is considered. (This is important only if
2633 C<S> has capturing parentheses, and backreferences are used somewhere
2634 else in the whole regular expression.)
2636 =item C<(?!S)>, C<(?<!S)>
2638 For this grouping operator there is no need to describe the ordering, since
2639 only whether or not C<S> can match is important.
2641 =item C<(??{ EXPR })>, C<(?I<PARNO>)>
2643 The ordering is the same as for the regular expression which is
2644 the result of EXPR, or the pattern contained by capture group I<PARNO>.
2646 =item C<(?(condition)yes-pattern|no-pattern)>
2648 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
2649 already determined. The ordering of the matches is the same as for the
2650 chosen subexpression.
2654 The above recipes describe the ordering of matches I<at a given position>.
2655 One more rule is needed to understand how a match is determined for the
2656 whole regular expression: a match at an earlier position is always better
2657 than a match at a later position.
2659 =head2 Creating Custom RE Engines
2661 As of Perl 5.10.0, one can create custom regular expression engines. This
2662 is not for the faint of heart, as they have to plug in at the C level. See
2663 L<perlreapi> for more details.
2665 As an alternative, overloaded constants (see L<overload>) provide a simple
2666 way to extend the functionality of the RE engine, by substituting one
2667 pattern for another.
2669 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
2670 matches at a boundary between whitespace characters and non-whitespace
2671 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
2672 at these positions, so we want to have each C<\Y|> in the place of the
2673 more complicated version. We can create a module C<customre> to do
2681 die "No argument to customre::import allowed" if @_;
2682 overload::constant 'qr' => \&convert;
2685 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
2687 # We must also take care of not escaping the legitimate \\Y|
2688 # sequence, hence the presence of '\\' in the conversion rules.
2689 my %rules = ( '\\' => '\\\\',
2690 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
2696 { $rules{$1} or invalid($re,$1) }sgex;
2700 Now C<use customre> enables the new escape in constant regular
2701 expressions, i.e., those without any runtime variable interpolations.
2702 As documented in L<overload>, this conversion will work only over
2703 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
2704 part of this regular expression needs to be converted explicitly
2705 (but only if the special meaning of C<\Y|> should be enabled inside C<$re>):
2710 $re = customre::convert $re;
2713 =head2 Embedded Code Execution Frequency
2715 The exact rules for how often (??{}) and (?{}) are executed in a pattern
2716 are unspecified. In the case of a successful match you can assume that
2717 they DWIM and will be executed in left to right order the appropriate
2718 number of times in the accepting path of the pattern as would any other
2719 meta-pattern. How non-accepting pathways and match failures affect the
2720 number of times a pattern is executed is specifically unspecified and
2721 may vary depending on what optimizations can be applied to the pattern
2722 and is likely to change from version to version.
2726 "aaabcdeeeee"=~/a(?{print "a"})b(?{print "b"})cde/;
2728 the exact number of times "a" or "b" are printed out is unspecified for
2729 failure, but you may assume they will be printed at least once during
2730 a successful match, additionally you may assume that if "b" is printed,
2731 it will be preceded by at least one "a".
2733 In the case of branching constructs like the following:
2735 /a(b|(?{ print "a" }))c(?{ print "c" })/;
2737 you can assume that the input "ac" will output "ac", and that "abc"
2738 will output only "c".
2740 When embedded code is quantified, successful matches will call the
2741 code once for each matched iteration of the quantifier. For
2744 "good" =~ /g(?:o(?{print "o"}))*d/;
2746 will output "o" twice.
2748 =head2 PCRE/Python Support
2750 As of Perl 5.10.0, Perl supports several Python/PCRE-specific extensions
2751 to the regex syntax. While Perl programmers are encouraged to use the
2752 Perl-specific syntax, the following are also accepted:
2756 =item C<< (?PE<lt>NAMEE<gt>pattern) >>
2758 Define a named capture group. Equivalent to C<< (?<NAME>pattern) >>.
2760 =item C<< (?P=NAME) >>
2762 Backreference to a named capture group. Equivalent to C<< \g{NAME} >>.
2764 =item C<< (?P>NAME) >>
2766 Subroutine call to a named capture group. Equivalent to C<< (?&NAME) >>.
2772 There are a number of issues with regard to case-insensitive matching
2773 in Unicode rules. See C<i> under L</Modifiers> above.
2775 This document varies from difficult to understand to completely
2776 and utterly opaque. The wandering prose riddled with jargon is
2777 hard to fathom in several places.
2779 This document needs a rewrite that separates the tutorial content
2780 from the reference content.
2788 L<perlop/"Regexp Quote-Like Operators">.
2790 L<perlop/"Gory details of parsing quoted constructs">.
2800 I<Mastering Regular Expressions> by Jeffrey Friedl, published
2801 by O'Reilly and Associates.