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.
25 Matching operations can have various modifiers. Modifiers
26 that relate to the interpretation of the regular expression inside
27 are listed below. Modifiers that alter the way a regular expression
28 is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and
29 L<perlop/"Gory details of parsing quoted constructs">.
34 X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline>
36 Treat string as multiple lines. That is, change "^" and "$" from matching
37 the start of the string's first line and the end of its last line to
38 matching the start and end of each line within the string.
41 X</s> X<regex, single-line> X<regexp, single-line>
42 X<regular expression, single-line>
44 Treat string as single line. That is, change "." to match any character
45 whatsoever, even a newline, which normally it would not match.
47 Used together, as C</ms>, they let the "." match any character whatsoever,
48 while still allowing "^" and "$" to match, respectively, just after
49 and just before newlines within the string.
52 X</i> X<regex, case-insensitive> X<regexp, case-insensitive>
53 X<regular expression, case-insensitive>
55 Do case-insensitive pattern matching.
57 If locale matching rules are in effect, the case map is taken from the
59 locale for code points less than 255, and from Unicode rules for larger
60 code points. However, matches that would cross the Unicode
61 rules/non-Unicode rules boundary (ords 255/256) will not succeed. See
64 There are a number of Unicode characters that match multiple characters
65 under C</i>. For example, C<LATIN SMALL LIGATURE FI>
66 should match the sequence C<fi>. Perl is not
67 currently able to do this when the multiple characters are in the pattern and
68 are split between groupings, or when one or more are quantified. Thus
70 "\N{LATIN SMALL LIGATURE FI}" =~ /fi/i; # Matches
71 "\N{LATIN SMALL LIGATURE FI}" =~ /[fi][fi]/i; # Doesn't match!
72 "\N{LATIN SMALL LIGATURE FI}" =~ /fi*/i; # Doesn't match!
74 # The below doesn't match, and it isn't clear what $1 and $2 would
76 "\N{LATIN SMALL LIGATURE FI}" =~ /(f)(i)/i; # Doesn't match!
78 Perl doesn't match multiple characters in a bracketed
79 character class unless the character that maps to them is explicitly
80 mentioned, and it doesn't match them at all if the character class is
81 inverted, which otherwise could be highly confusing. See
82 L<perlrecharclass/Bracketed Character Classes>, and
83 L<perlrecharclass/Negation>.
88 Extend your pattern's legibility by permitting whitespace and comments.
92 X</p> X<regex, preserve> X<regexp, preserve>
94 Preserve the string matched such that ${^PREMATCH}, ${^MATCH}, and
95 ${^POSTMATCH} are available for use after matching.
97 In Perl 5.20 and higher this is ignored. Due to a new copy-on-write
98 mechanism, ${^PREMATCH}, ${^MATCH}, and ${^POSTMATCH} will be available
99 after the match regardless of the modifier.
102 X</a> X</d> X</l> X</u>
104 These modifiers, all new in 5.14, affect which character-set rules
105 (Unicode, etc.) are used, as described below in
106 L</Character set modifiers>.
109 X</n> X<regex, non-capture> X<regexp, non-capture>
110 X<regular expression, non-capture>
112 Prevent the grouping metacharacters C<()> from capturing. This modifier,
113 new in 5.22, will stop C<$1>, C<$2>, etc... from being filled in.
115 "hello" =~ /(hi|hello)/; # $1 is "hello"
116 "hello" =~ /(hi|hello)/n; # $1 is undef
118 This is equivalent to putting C<?:> at the beginning of every capturing group:
120 "hello" =~ /(?:hi|hello)/; # $1 is undef
122 C</n> can be negated on a per-group basis. Alternatively, named captures
125 "hello" =~ /(?-n:(hi|hello))/n; # $1 is "hello"
126 "hello" =~ /(?<greet>hi|hello)/n; # $1 is "hello", $+{greet} is
129 =item Other Modifiers
131 There are a number of flags that can be found at the end of regular
132 expression constructs that are I<not> generic regular expression flags, but
133 apply to the operation being performed, like matching or substitution (C<m//>
134 or C<s///> respectively).
136 Flags described further in
137 L<perlretut/"Using regular expressions in Perl"> are:
139 c - keep the current position during repeated matching
140 g - globally match the pattern repeatedly in the string
142 Substitution-specific modifiers described in
144 L<perlop/"s/PATTERN/REPLACEMENT/msixpodualngcer"> are:
146 e - evaluate the right-hand side as an expression
147 ee - evaluate the right side as a string then eval the result
148 o - pretend to optimize your code, but actually introduce bugs
149 r - perform non-destructive substitution and return the new value
153 Regular expression modifiers are usually written in documentation
154 as e.g., "the C</x> modifier", even though the delimiter
155 in question might not really be a slash. The modifiers C</imsxadlup>
156 may also be embedded within the regular expression itself using
157 the C<(?...)> construct, see L</Extended Patterns> below.
162 the regular expression parser to ignore most whitespace that is neither
163 backslashed nor within a bracketed character class. You can use this to
164 break up your regular expression into (slightly) more readable parts.
165 Also, the C<#> character is treated as a metacharacter introducing a
166 comment that runs up to the pattern's closing delimiter, or to the end
167 of the current line if the pattern extends onto the next line. Hence,
168 this is very much like an ordinary Perl code comment. (You can include
169 the closing delimiter within the comment only if you precede it with a
170 backslash, so be careful!)
172 Use of C</x> means that if you want real
173 whitespace or C<#> characters in the pattern (outside a bracketed character
174 class, which is unaffected by C</x>), then you'll either have to
175 escape them (using backslashes or C<\Q...\E>) or encode them using octal,
176 hex, or C<\N{}> escapes.
177 It is ineffective to try to continue a comment onto the next line by
178 escaping the C<\n> with a backslash or C<\Q>.
180 You can use L</(?#text)> to create a comment that ends earlier than the
181 end of the current line, but C<text> also can't contain the closing
182 delimiter unless escaped with a backslash.
184 Taken together, these features go a long way towards
185 making Perl's regular expressions more readable. Here's an example:
187 # Delete (most) C comments.
189 /\* # Match the opening delimiter.
190 .*? # Match a minimal number of characters.
191 \*/ # Match the closing delimiter.
194 Note that anything inside
195 a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect
196 space interpretation within a single multi-character construct. For
197 example in C<\x{...}>, regardless of the C</x> modifier, there can be no
198 spaces. Same for a L<quantifier|/Quantifiers> such as C<{3}> or
199 C<{5,}>. Similarly, C<(?:...)> can't have a space between the C<(>,
200 C<?>, and C<:>. Within any delimiters for such a
201 construct, allowed spaces are not affected by C</x>, and depend on the
202 construct. For example, C<\x{...}> can't have spaces because hexadecimal
203 numbers don't have spaces in them. But, Unicode properties can have spaces, so
204 in C<\p{...}> there can be spaces that follow the Unicode rules, for which see
205 L<perluniprops/Properties accessible through \p{} and \P{}>.
208 The set of characters that are deemed whitespace are those that Unicode
209 calls "Pattern White Space", namely:
211 U+0009 CHARACTER TABULATION
213 U+000B LINE TABULATION
215 U+000D CARRIAGE RETURN
218 U+200E LEFT-TO-RIGHT MARK
219 U+200F RIGHT-TO-LEFT MARK
220 U+2028 LINE SEPARATOR
221 U+2029 PARAGRAPH SEPARATOR
223 =head3 Character set modifiers
225 C</d>, C</u>, C</a>, and C</l>, available starting in 5.14, are called
226 the character set modifiers; they affect the character set rules
227 used for the regular expression.
229 The C</d>, C</u>, and C</l> modifiers are not likely to be of much use
230 to you, and so you need not worry about them very much. They exist for
231 Perl's internal use, so that complex regular expression data structures
232 can be automatically serialized and later exactly reconstituted,
233 including all their nuances. But, since Perl can't keep a secret, and
234 there may be rare instances where they are useful, they are documented
237 The C</a> modifier, on the other hand, may be useful. Its purpose is to
238 allow code that is to work mostly on ASCII data to not have to concern
241 Briefly, C</l> sets the character set to that of whatever B<L>ocale is in
242 effect at the time of the execution of the pattern match.
244 C</u> sets the character set to B<U>nicode.
246 C</a> also sets the character set to Unicode, BUT adds several
247 restrictions for B<A>SCII-safe matching.
249 C</d> is the old, problematic, pre-5.14 B<D>efault character set
250 behavior. Its only use is to force that old behavior.
252 At any given time, exactly one of these modifiers is in effect. Their
253 existence allows Perl to keep the originally compiled behavior of a
254 regular expression, regardless of what rules are in effect when it is
255 actually executed. And if it is interpolated into a larger regex, the
256 original's rules continue to apply to it, and only it.
258 The C</l> and C</u> modifiers are automatically selected for
259 regular expressions compiled within the scope of various pragmas,
260 and we recommend that in general, you use those pragmas instead of
261 specifying these modifiers explicitly. For one thing, the modifiers
262 affect only pattern matching, and do not extend to even any replacement
263 done, whereas using the pragmas give consistent results for all
264 appropriate operations within their scopes. For example,
268 will match "foo" using the locale's rules for case-insensitive matching,
269 but the C</l> does not affect how the C<\U> operates. Most likely you
270 want both of them to use locale rules. To do this, instead compile the
271 regular expression within the scope of C<use locale>. This both
272 implicitly adds the C</l> and applies locale rules to the C<\U>. The
273 lesson is to C<use locale> and not C</l> explicitly.
275 Similarly, it would be better to use C<use feature 'unicode_strings'>
280 to get Unicode rules, as the C<\L> in the former (but not necessarily
281 the latter) would also use Unicode rules.
283 More detail on each of the modifiers follows. Most likely you don't
284 need to know this detail for C</l>, C</u>, and C</d>, and can skip ahead
285 to L<E<sol>a|/E<sol>a (and E<sol>aa)>.
289 means to use the current locale's rules (see L<perllocale>) when pattern
290 matching. For example, C<\w> will match the "word" characters of that
291 locale, and C<"/i"> case-insensitive matching will match according to
292 the locale's case folding rules. The locale used will be the one in
293 effect at the time of execution of the pattern match. This may not be
294 the same as the compilation-time locale, and can differ from one match
295 to another if there is an intervening call of the
296 L<setlocale() function|perllocale/The setlocale function>.
298 The only non-single-byte locale Perl supports is (starting in v5.20)
299 UTF-8. This means that code points above 255 are treated as Unicode no
300 matter what locale is in effect (since UTF-8 implies Unicode).
302 Under Unicode rules, there are a few case-insensitive matches that cross
303 the 255/256 boundary. Except for UTF-8 locales in Perls v5.20 and
304 later, these are disallowed under C</l>. For example, 0xFF (on ASCII
305 platforms) does not caselessly match the character at 0x178, C<LATIN
306 CAPITAL LETTER Y WITH DIAERESIS>, because 0xFF may not be C<LATIN SMALL
307 LETTER Y WITH DIAERESIS> in the current locale, and Perl has no way of
308 knowing if that character even exists in the locale, much less what code
311 In a UTF-8 locale in v5.20 and later, the only visible difference
312 between locale and non-locale in regular expressions should be tainting
315 This modifier may be specified to be the default by C<use locale>, but
316 see L</Which character set modifier is in effect?>.
321 means to use Unicode rules when pattern matching. On ASCII platforms,
322 this means that the code points between 128 and 255 take on their
323 Latin-1 (ISO-8859-1) meanings (which are the same as Unicode's).
324 (Otherwise Perl considers their meanings to be undefined.) Thus,
325 under this modifier, the ASCII platform effectively becomes a Unicode
326 platform; and hence, for example, C<\w> will match any of the more than
327 100_000 word characters in Unicode.
329 Unlike most locales, which are specific to a language and country pair,
330 Unicode classifies all the characters that are letters I<somewhere> in
332 C<\w>. For example, your locale might not think that C<LATIN SMALL
333 LETTER ETH> is a letter (unless you happen to speak Icelandic), but
334 Unicode does. Similarly, all the characters that are decimal digits
335 somewhere in the world will match C<\d>; this is hundreds, not 10,
336 possible matches. And some of those digits look like some of the 10
337 ASCII digits, but mean a different number, so a human could easily think
338 a number is a different quantity than it really is. For example,
339 C<BENGALI DIGIT FOUR> (U+09EA) looks very much like an
340 C<ASCII DIGIT EIGHT> (U+0038). And, C<\d+>, may match strings of digits
341 that are a mixture from different writing systems, creating a security
342 issue. L<Unicode::UCD/num()> can be used to sort
343 this out. Or the C</a> modifier can be used to force C<\d> to match
344 just the ASCII 0 through 9.
346 Also, under this modifier, case-insensitive matching works on the full
348 characters. The C<KELVIN SIGN>, for example matches the letters "k" and
349 "K"; and C<LATIN SMALL LIGATURE FF> matches the sequence "ff", which,
350 if you're not prepared, might make it look like a hexadecimal constant,
351 presenting another potential security issue. See
352 L<http://unicode.org/reports/tr36> for a detailed discussion of Unicode
355 This modifier may be specified to be the default by C<use feature
356 'unicode_strings>, C<use locale ':not_characters'>, or
357 C<L<use 5.012|perlfunc/use VERSION>> (or higher),
358 but see L</Which character set modifier is in effect?>.
363 This modifier means to use the "Default" native rules of the platform
364 except when there is cause to use Unicode rules instead, as follows:
370 the target string is encoded in UTF-8; or
374 the pattern is encoded in UTF-8; or
378 the pattern explicitly mentions a code point that is above 255 (say by
383 the pattern uses a Unicode name (C<\N{...}>); or
387 the pattern uses a Unicode property (C<\p{...}>); or
391 the pattern uses L</C<(?[ ])>>
395 Another mnemonic for this modifier is "Depends", as the rules actually
396 used depend on various things, and as a result you can get unexpected
397 results. See L<perlunicode/The "Unicode Bug">. The Unicode Bug has
398 become rather infamous, leading to yet another (printable) name for this
401 Unless the pattern or string are encoded in UTF-8, only ASCII characters
402 can match positively.
404 Here are some examples of how that works on an ASCII platform:
406 $str = "\xDF"; # $str is not in UTF-8 format.
407 $str =~ /^\w/; # No match, as $str isn't in UTF-8 format.
408 $str .= "\x{0e0b}"; # Now $str is in UTF-8 format.
409 $str =~ /^\w/; # Match! $str is now in UTF-8 format.
411 $str =~ /^\w/; # Still a match! $str remains in UTF-8 format.
413 This modifier is automatically selected by default when none of the
414 others are, so yet another name for it is "Default".
416 Because of the unexpected behaviors associated with this modifier, you
417 probably should only use it to maintain weird backward compatibilities.
421 This modifier stands for ASCII-restrict (or ASCII-safe). This modifier,
422 unlike the others, may be doubled-up to increase its effect.
424 When it appears singly, it causes the sequences C<\d>, C<\s>, C<\w>, and
425 the Posix character classes to match only in the ASCII range. They thus
426 revert to their pre-5.6, pre-Unicode meanings. Under C</a>, C<\d>
427 always means precisely the digits C<"0"> to C<"9">; C<\s> means the five
428 characters C<[ \f\n\r\t]>, and starting in Perl v5.18, experimentally,
429 the vertical tab; C<\w> means the 63 characters
430 C<[A-Za-z0-9_]>; and likewise, all the Posix classes such as
431 C<[[:print:]]> match only the appropriate ASCII-range characters.
433 This modifier is useful for people who only incidentally use Unicode,
434 and who do not wish to be burdened with its complexities and security
437 With C</a>, one can write C<\d> with confidence that it will only match
438 ASCII characters, and should the need arise to match beyond ASCII, you
439 can instead use C<\p{Digit}> (or C<\p{Word}> for C<\w>). There are
440 similar C<\p{...}> constructs that can match beyond ASCII both white
441 space (see L<perlrecharclass/Whitespace>), and Posix classes (see
442 L<perlrecharclass/POSIX Character Classes>). Thus, this modifier
443 doesn't mean you can't use Unicode, it means that to get Unicode
444 matching you must explicitly use a construct (C<\p{}>, C<\P{}>) that
447 As you would expect, this modifier causes, for example, C<\D> to mean
448 the same thing as C<[^0-9]>; in fact, all non-ASCII characters match
449 C<\D>, C<\S>, and C<\W>. C<\b> still means to match at the boundary
450 between C<\w> and C<\W>, using the C</a> definitions of them (similarly
453 Otherwise, C</a> behaves like the C</u> modifier, in that
454 case-insensitive matching uses Unicode rules; for example, "k" will
455 match the Unicode C<\N{KELVIN SIGN}> under C</i> matching, and code
456 points in the Latin1 range, above ASCII will have Unicode rules when it
457 comes to case-insensitive matching.
459 To forbid ASCII/non-ASCII matches (like "k" with C<\N{KELVIN SIGN}>),
460 specify the "a" twice, for example C</aai> or C</aia>. (The first
461 occurrence of "a" restricts the C<\d>, etc., and the second occurrence
462 adds the C</i> restrictions.) But, note that code points outside the
463 ASCII range will use Unicode rules for C</i> matching, so the modifier
464 doesn't really restrict things to just ASCII; it just forbids the
465 intermixing of ASCII and non-ASCII.
467 To summarize, this modifier provides protection for applications that
468 don't wish to be exposed to all of Unicode. Specifying it twice
469 gives added protection.
471 This modifier may be specified to be the default by C<use re '/a'>
472 or C<use re '/aa'>. If you do so, you may actually have occasion to use
473 the C</u> modifier explicitly if there are a few regular expressions
474 where you do want full Unicode rules (but even here, it's best if
475 everything were under feature C<"unicode_strings">, along with the
476 C<use re '/aa'>). Also see L</Which character set modifier is in
481 =head4 Which character set modifier is in effect?
483 Which of these modifiers is in effect at any given point in a regular
484 expression depends on a fairly complex set of interactions. These have
485 been designed so that in general you don't have to worry about it, but
486 this section gives the gory details. As
487 explained below in L</Extended Patterns> it is possible to explicitly
488 specify modifiers that apply only to portions of a regular expression.
489 The innermost always has priority over any outer ones, and one applying
490 to the whole expression has priority over any of the default settings that are
491 described in the remainder of this section.
493 The C<L<use re 'E<sol>foo'|re/"'/flags' mode">> pragma can be used to set
494 default modifiers (including these) for regular expressions compiled
495 within its scope. This pragma has precedence over the other pragmas
496 listed below that also change the defaults.
498 Otherwise, C<L<use locale|perllocale>> sets the default modifier to C</l>;
499 and C<L<use feature 'unicode_strings|feature>>, or
500 C<L<use 5.012|perlfunc/use VERSION>> (or higher) set the default to
501 C</u> when not in the same scope as either C<L<use locale|perllocale>>
502 or C<L<use bytes|bytes>>.
503 (C<L<use locale ':not_characters'|perllocale/Unicode and UTF-8>> also
504 sets the default to C</u>, overriding any plain C<use locale>.)
505 Unlike the mechanisms mentioned above, these
506 affect operations besides regular expressions pattern matching, and so
507 give more consistent results with other operators, including using
508 C<\U>, C<\l>, etc. in substitution replacements.
510 If none of the above apply, for backwards compatibility reasons, the
511 C</d> modifier is the one in effect by default. As this can lead to
512 unexpected results, it is best to specify which other rule set should be
515 =head4 Character set modifier behavior prior to Perl 5.14
517 Prior to 5.14, there were no explicit modifiers, but C</l> was implied
518 for regexes compiled within the scope of C<use locale>, and C</d> was
519 implied otherwise. However, interpolating a regex into a larger regex
520 would ignore the original compilation in favor of whatever was in effect
521 at the time of the second compilation. There were a number of
522 inconsistencies (bugs) with the C</d> modifier, where Unicode rules
523 would be used when inappropriate, and vice versa. C<\p{}> did not imply
524 Unicode rules, and neither did all occurrences of C<\N{}>, until 5.12.
526 =head2 Regular Expressions
528 =head3 Metacharacters
530 The patterns used in Perl pattern matching evolved from those supplied in
531 the Version 8 regex routines. (The routines are derived
532 (distantly) from Henry Spencer's freely redistributable reimplementation
533 of the V8 routines.) See L<Version 8 Regular Expressions> for
536 In particular the following metacharacters have their standard I<egrep>-ish
539 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
542 \ Quote the next metacharacter
543 ^ Match the beginning of the line
544 . Match any character (except newline)
545 $ Match the end of the string (or before newline at the end
549 [] Bracketed Character class
551 By default, the "^" character is guaranteed to match only the
552 beginning of the string, the "$" character only the end (or before the
553 newline at the end), and Perl does certain optimizations with the
554 assumption that the string contains only one line. Embedded newlines
555 will not be matched by "^" or "$". You may, however, wish to treat a
556 string as a multi-line buffer, such that the "^" will match after any
557 newline within the string (except if the newline is the last character in
558 the string), and "$" will match before any newline. At the
559 cost of a little more overhead, you can do this by using the /m modifier
560 on the pattern match operator. (Older programs did this by setting C<$*>,
561 but this option was removed in perl 5.10.)
564 To simplify multi-line substitutions, the "." character never matches a
565 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
566 the string is a single line--even if it isn't.
571 The following standard quantifiers are recognized:
572 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
574 * Match 0 or more times
575 + Match 1 or more times
577 {n} Match exactly n times
578 {n,} Match at least n times
579 {n,m} Match at least n but not more than m times
581 (If a curly bracket occurs in any other context and does not form part of
582 a backslashed sequence like C<\x{...}>, it is treated as a regular
583 character. However, a deprecation warning is raised for all such
584 occurrences, and in Perl v5.26, literal uses of a curly bracket will be
585 required to be escaped, say by preceding them with a backslash (C<"\{">)
586 or enclosing them within square brackets (C<"[{]">). This change will
587 allow for future syntax extensions (like making the lower bound of a
588 quantifier optional), and better error checking of quantifiers.)
590 The "*" quantifier is equivalent to C<{0,}>, the "+"
591 quantifier to C<{1,}>, and the "?" quantifier to C<{0,1}>. n and m are limited
592 to non-negative integral values less than a preset limit defined when perl is built.
593 This is usually 32766 on the most common platforms. The actual limit can
594 be seen in the error message generated by code such as this:
596 $_ **= $_ , / {$_} / for 2 .. 42;
598 By default, a quantified subpattern is "greedy", that is, it will match as
599 many times as possible (given a particular starting location) while still
600 allowing the rest of the pattern to match. If you want it to match the
601 minimum number of times possible, follow the quantifier with a "?". Note
602 that the meanings don't change, just the "greediness":
603 X<metacharacter> X<greedy> X<greediness>
604 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
606 *? Match 0 or more times, not greedily
607 +? Match 1 or more times, not greedily
608 ?? Match 0 or 1 time, not greedily
609 {n}? Match exactly n times, not greedily (redundant)
610 {n,}? Match at least n times, not greedily
611 {n,m}? Match at least n but not more than m times, not greedily
613 Normally when a quantified subpattern does not allow the rest of the
614 overall pattern to match, Perl will backtrack. However, this behaviour is
615 sometimes undesirable. Thus Perl provides the "possessive" quantifier form
618 *+ Match 0 or more times and give nothing back
619 ++ Match 1 or more times and give nothing back
620 ?+ Match 0 or 1 time and give nothing back
621 {n}+ Match exactly n times and give nothing back (redundant)
622 {n,}+ Match at least n times and give nothing back
623 {n,m}+ Match at least n but not more than m times and give nothing back
629 will never match, as the C<a++> will gobble up all the C<a>'s in the
630 string and won't leave any for the remaining part of the pattern. This
631 feature can be extremely useful to give perl hints about where it
632 shouldn't backtrack. For instance, the typical "match a double-quoted
633 string" problem can be most efficiently performed when written as:
635 /"(?:[^"\\]++|\\.)*+"/
637 as we know that if the final quote does not match, backtracking will not
638 help. See the independent subexpression
639 L</C<< (?>pattern) >>> for more details;
640 possessive quantifiers are just syntactic sugar for that construct. For
641 instance the above example could also be written as follows:
643 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
645 Note that the possessive quantifier modifier can not be be combined
646 with the non-greedy modifier. This is because it would make no sense.
647 Consider the follow equivalency table:
655 =head3 Escape sequences
657 Because patterns are processed as double-quoted strings, the following
664 \a alarm (bell) (BEL)
665 \e escape (think troff) (ESC)
666 \cK control char (example: VT)
667 \x{}, \x00 character whose ordinal is the given hexadecimal number
668 \N{name} named Unicode character or character sequence
669 \N{U+263D} Unicode character (example: FIRST QUARTER MOON)
670 \o{}, \000 character whose ordinal is the given octal number
671 \l lowercase next char (think vi)
672 \u uppercase next char (think vi)
673 \L lowercase until \E (think vi)
674 \U uppercase until \E (think vi)
675 \Q quote (disable) pattern metacharacters until \E
676 \E end either case modification or quoted section, think vi
678 Details are in L<perlop/Quote and Quote-like Operators>.
680 =head3 Character Classes and other Special Escapes
682 In addition, Perl defines the following:
683 X<\g> X<\k> X<\K> X<backreference>
685 Sequence Note Description
686 [...] [1] Match a character according to the rules of the
687 bracketed character class defined by the "...".
688 Example: [a-z] matches "a" or "b" or "c" ... or "z"
689 [[:...:]] [2] Match a character according to the rules of the POSIX
690 character class "..." within the outer bracketed
691 character class. Example: [[:upper:]] matches any
693 (?[...]) [8] Extended bracketed character class
694 \w [3] Match a "word" character (alphanumeric plus "_", plus
695 other connector punctuation chars plus Unicode
697 \W [3] Match a non-"word" character
698 \s [3] Match a whitespace character
699 \S [3] Match a non-whitespace character
700 \d [3] Match a decimal digit character
701 \D [3] Match a non-digit character
702 \pP [3] Match P, named property. Use \p{Prop} for longer names
704 \X [4] Match Unicode "eXtended grapheme cluster"
705 \C Match a single C-language char (octet) even if that is
706 part of a larger UTF-8 character. Thus it breaks up
707 characters into their UTF-8 bytes, so you may end up
708 with malformed pieces of UTF-8. Unsupported in
709 lookbehind. (Deprecated.)
710 \1 [5] Backreference to a specific capture group or buffer.
711 '1' may actually be any positive integer.
712 \g1 [5] Backreference to a specific or previous group,
713 \g{-1} [5] The number may be negative indicating a relative
714 previous group and may optionally be wrapped in
715 curly brackets for safer parsing.
716 \g{name} [5] Named backreference
717 \k<name> [5] Named backreference
718 \K [6] Keep the stuff left of the \K, don't include it in $&
719 \N [7] Any character but \n. Not affected by /s modifier
720 \v [3] Vertical whitespace
721 \V [3] Not vertical whitespace
722 \h [3] Horizontal whitespace
723 \H [3] Not horizontal whitespace
730 See L<perlrecharclass/Bracketed Character Classes> for details.
734 See L<perlrecharclass/POSIX Character Classes> for details.
738 See L<perlrecharclass/Backslash sequences> for details.
742 See L<perlrebackslash/Misc> for details.
746 See L</Capture groups> below for details.
750 See L</Extended Patterns> below for details.
754 Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the
755 character or character sequence whose name is C<NAME>; and similarly
756 when of the form C<\N{U+I<hex>}>, it matches the character whose Unicode
757 code point is I<hex>. Otherwise it matches any character but C<\n>.
761 See L<perlrecharclass/Extended Bracketed Character Classes> for details.
767 Perl defines the following zero-width assertions:
768 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
769 X<regexp, zero-width assertion>
770 X<regular expression, zero-width assertion>
771 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
773 \b Match a word boundary
774 \B Match except at a word boundary
775 \A Match only at beginning of string
776 \Z Match only at end of string, or before newline at the end
777 \z Match only at end of string
778 \G Match only at pos() (e.g. at the end-of-match position
781 A word boundary (C<\b>) is a spot between two characters
782 that has a C<\w> on one side of it and a C<\W> on the other side
783 of it (in either order), counting the imaginary characters off the
784 beginning and end of the string as matching a C<\W>. (Within
785 character classes C<\b> represents backspace rather than a word
786 boundary, just as it normally does in any double-quoted string.)
787 The C<\A> and C<\Z> are just like "^" and "$", except that they
788 won't match multiple times when the C</m> modifier is used, while
789 "^" and "$" will match at every internal line boundary. To match
790 the actual end of the string and not ignore an optional trailing
792 X<\b> X<\A> X<\Z> X<\z> X</m>
794 The C<\G> assertion can be used to chain global matches (using
795 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
796 It is also useful when writing C<lex>-like scanners, when you have
797 several patterns that you want to match against consequent substrings
798 of your string; see the previous reference. The actual location
799 where C<\G> will match can also be influenced by using C<pos()> as
800 an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
801 matches (see L</"Repeated Patterns Matching a Zero-length Substring">)
802 is modified somewhat, in that contents to the left of C<\G> are
803 not counted when determining the length of the match. Thus the following
804 will not match forever:
809 while ($string =~ /(.\G)/g) {
813 It will print 'A' and then terminate, as it considers the match to
814 be zero-width, and thus will not match at the same position twice in a
817 It is worth noting that C<\G> improperly used can result in an infinite
818 loop. Take care when using patterns that include C<\G> in an alternation.
820 Note also that C<s///> will refuse to overwrite part of a substitution
821 that has already been replaced; so for example this will stop after the
822 first iteration, rather than iterating its way backwards through the
828 print; # prints 1234X6789, not XXXXX6789
831 =head3 Capture groups
833 The bracketing construct C<( ... )> creates capture groups (also referred to as
834 capture buffers). To refer to the current contents of a group later on, within
835 the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>)
836 for the second, and so on.
837 This is called a I<backreference>.
838 X<regex, capture buffer> X<regexp, capture buffer>
839 X<regex, capture group> X<regexp, capture group>
840 X<regular expression, capture buffer> X<backreference>
841 X<regular expression, capture group> X<backreference>
842 X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
843 X<named capture buffer> X<regular expression, named capture buffer>
844 X<named capture group> X<regular expression, named capture group>
845 X<%+> X<$+{name}> X<< \k<name> >>
846 There is no limit to the number of captured substrings that you may use.
847 Groups are numbered with the leftmost open parenthesis being number 1, etc. If
848 a group did not match, the associated backreference won't match either. (This
849 can happen if the group is optional, or in a different branch of an
851 You can omit the C<"g">, and write C<"\1">, etc, but there are some issues with
852 this form, described below.
854 You can also refer to capture groups relatively, by using a negative number, so
855 that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture
856 group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For
863 \g{-1} # backref to group 3
864 \g{-3} # backref to group 1
868 would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to
869 interpolate regexes into larger regexes and not have to worry about the
870 capture groups being renumbered.
872 You can dispense with numbers altogether and create named capture groups.
873 The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to
874 reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may
875 also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.)
876 I<name> must not begin with a number, nor contain hyphens.
877 When different groups within the same pattern have the same name, any reference
878 to that name assumes the leftmost defined group. Named groups count in
879 absolute and relative numbering, and so can also be referred to by those
881 (It's possible to do things with named capture groups that would otherwise
884 Capture group contents are dynamically scoped and available to you outside the
885 pattern until the end of the enclosing block or until the next successful
886 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
887 You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">,
888 etc); or by name via the C<%+> hash, using C<"$+{I<name>}">.
890 Braces are required in referring to named capture groups, but are optional for
891 absolute or relative numbered ones. Braces are safer when creating a regex by
892 concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a>
893 contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which
894 is probably not what you intended.
896 The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that
897 there were no named nor relative numbered capture groups. Absolute numbered
898 groups were referred to using C<\1>,
899 C<\2>, etc., and this notation is still
900 accepted (and likely always will be). But it leads to some ambiguities if
901 there are more than 9 capture groups, as C<\10> could mean either the tenth
902 capture group, or the character whose ordinal in octal is 010 (a backspace in
903 ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference
904 only if at least 10 left parentheses have opened before it. Likewise C<\11> is
905 a backreference only if at least 11 left parentheses have opened before it.
906 And so on. C<\1> through C<\9> are always interpreted as backreferences.
907 There are several examples below that illustrate these perils. You can avoid
908 the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups;
909 and for octal constants always using C<\o{}>, or for C<\077> and below, using 3
910 digits padded with leading zeros, since a leading zero implies an octal
913 The C<\I<digit>> notation also works in certain circumstances outside
914 the pattern. See L</Warning on \1 Instead of $1> below for details.
918 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
920 /(.)\g1/ # find first doubled char
921 and print "'$1' is the first doubled character\n";
923 /(?<char>.)\k<char>/ # ... a different way
924 and print "'$+{char}' is the first doubled character\n";
926 /(?'char'.)\g1/ # ... mix and match
927 and print "'$1' is the first doubled character\n";
929 if (/Time: (..):(..):(..)/) { # parse out values
935 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference
936 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal
937 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference
938 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal
940 $a = '(.)\1'; # Creates problems when concatenated.
941 $b = '(.)\g{1}'; # Avoids the problems.
942 "aa" =~ /${a}/; # True
943 "aa" =~ /${b}/; # True
944 "aa0" =~ /${a}0/; # False!
945 "aa0" =~ /${b}0/; # True
946 "aa\x08" =~ /${a}0/; # True!
947 "aa\x08" =~ /${b}0/; # False
949 Several special variables also refer back to portions of the previous
950 match. C<$+> returns whatever the last bracket match matched.
951 C<$&> returns the entire matched string. (At one point C<$0> did
952 also, but now it returns the name of the program.) C<$`> returns
953 everything before the matched string. C<$'> returns everything
954 after the matched string. And C<$^N> contains whatever was matched by
955 the most-recently closed group (submatch). C<$^N> can be used in
956 extended patterns (see below), for example to assign a submatch to a
958 X<$+> X<$^N> X<$&> X<$`> X<$'>
960 These special variables, like the C<%+> hash and the numbered match variables
961 (C<$1>, C<$2>, C<$3>, etc.) are dynamically scoped
962 until the end of the enclosing block or until the next successful
963 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
964 X<$+> X<$^N> X<$&> X<$`> X<$'>
965 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
967 B<NOTE>: Failed matches in Perl do not reset the match variables,
968 which makes it easier to write code that tests for a series of more
969 specific cases and remembers the best match.
971 B<WARNING>: If your code is to run on Perl 5.16 or earlier,
972 beware that once Perl sees that you need one of C<$&>, C<$`>, or
973 C<$'> anywhere in the program, it has to provide them for every
974 pattern match. This may substantially slow your program.
976 Perl uses the same mechanism to produce C<$1>, C<$2>, etc, so you also
977 pay a price for each pattern that contains capturing parentheses.
978 (To avoid this cost while retaining the grouping behaviour, use the
979 extended regular expression C<(?: ... )> instead.) But if you never
980 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
981 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
982 if you can, but if you can't (and some algorithms really appreciate
983 them), once you've used them once, use them at will, because you've
984 already paid the price.
987 Perl 5.16 introduced a slightly more efficient mechanism that notes
988 separately whether each of C<$`>, C<$&>, and C<$'> have been seen, and
989 thus may only need to copy part of the string. Perl 5.20 introduced a
990 much more efficient copy-on-write mechanism which eliminates any slowdown.
992 As another workaround for this problem, Perl 5.10.0 introduced C<${^PREMATCH}>,
993 C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&>
994 and C<$'>, B<except> that they are only guaranteed to be defined after a
995 successful match that was executed with the C</p> (preserve) modifier.
996 The use of these variables incurs no global performance penalty, unlike
997 their punctuation char equivalents, however at the trade-off that you
998 have to tell perl when you want to use them. As of Perl 5.20, these three
999 variables are equivalent to C<$`>, C<$&> and C<$'>, and C</p> is ignored.
1002 =head2 Quoting metacharacters
1004 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
1005 C<\w>, C<\n>. Unlike some other regular expression languages, there
1006 are no backslashed symbols that aren't alphanumeric. So anything
1007 that looks like \\, \(, \), \[, \], \{, or \} is always
1008 interpreted as a literal character, not a metacharacter. This was
1009 once used in a common idiom to disable or quote the special meanings
1010 of regular expression metacharacters in a string that you want to
1011 use for a pattern. Simply quote all non-"word" characters:
1013 $pattern =~ s/(\W)/\\$1/g;
1015 (If C<use locale> is set, then this depends on the current locale.)
1016 Today it is more common to use the quotemeta() function or the C<\Q>
1017 metaquoting escape sequence to disable all metacharacters' special
1020 /$unquoted\Q$quoted\E$unquoted/
1022 Beware that if you put literal backslashes (those not inside
1023 interpolated variables) between C<\Q> and C<\E>, double-quotish
1024 backslash interpolation may lead to confusing results. If you
1025 I<need> to use literal backslashes within C<\Q...\E>,
1026 consult L<perlop/"Gory details of parsing quoted constructs">.
1028 C<quotemeta()> and C<\Q> are fully described in L<perlfunc/quotemeta>.
1030 =head2 Extended Patterns
1032 Perl also defines a consistent extension syntax for features not
1033 found in standard tools like B<awk> and
1034 B<lex>. The syntax for most of these is a
1035 pair of parentheses with a question mark as the first thing within
1036 the parentheses. The character after the question mark indicates
1039 The stability of these extensions varies widely. Some have been
1040 part of the core language for many years. Others are experimental
1041 and may change without warning or be completely removed. Check
1042 the documentation on an individual feature to verify its current
1045 A question mark was chosen for this and for the minimal-matching
1046 construct because 1) question marks are rare in older regular
1047 expressions, and 2) whenever you see one, you should stop and
1048 "question" exactly what is going on. That's psychology....
1055 A comment. The text is ignored.
1056 Note that Perl closes
1057 the comment as soon as it sees a C<)>, so there is no way to put a literal
1058 C<)> in the comment. The pattern's closing delimiter must be escaped by
1059 a backslash if it appears in the comment.
1061 See L</E<sol>x> for another way to have comments in patterns.
1063 =item C<(?adlupimsx-imsx)>
1065 =item C<(?^alupimsx)>
1068 One or more embedded pattern-match modifiers, to be turned on (or
1069 turned off, if preceded by C<->) for the remainder of the pattern or
1070 the remainder of the enclosing pattern group (if any).
1072 This is particularly useful for dynamic patterns, such as those read in from a
1073 configuration file, taken from an argument, or specified in a table
1074 somewhere. Consider the case where some patterns want to be
1075 case-sensitive and some do not: The case-insensitive ones merely need to
1076 include C<(?i)> at the front of the pattern. For example:
1078 $pattern = "foobar";
1079 if ( /$pattern/i ) { }
1083 $pattern = "(?i)foobar";
1084 if ( /$pattern/ ) { }
1086 These modifiers are restored at the end of the enclosing group. For example,
1088 ( (?i) blah ) \s+ \g1
1090 will match C<blah> in any case, some spaces, and an exact (I<including the case>!)
1091 repetition of the previous word, assuming the C</x> modifier, and no C</i>
1092 modifier outside this group.
1094 These modifiers do not carry over into named subpatterns called in the
1095 enclosing group. In other words, a pattern such as C<((?i)(?&NAME))> does not
1096 change the case-sensitivity of the "NAME" pattern.
1098 Any of these modifiers can be set to apply globally to all regular
1099 expressions compiled within the scope of a C<use re>. See
1100 L<re/"'/flags' mode">.
1102 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
1103 after the C<"?"> is a shorthand equivalent to C<d-imsx>. Flags (except
1104 C<"d">) may follow the caret to override it.
1105 But a minus sign is not legal with it.
1107 Note that the C<a>, C<d>, C<l>, C<p>, and C<u> modifiers are special in
1108 that they can only be enabled, not disabled, and the C<a>, C<d>, C<l>, and
1109 C<u> modifiers are mutually exclusive: specifying one de-specifies the
1110 others, and a maximum of one (or two C<a>'s) may appear in the
1111 construct. Thus, for
1112 example, C<(?-p)> will warn when compiled under C<use warnings>;
1113 C<(?-d:...)> and C<(?dl:...)> are fatal errors.
1115 Note also that the C<p> modifier is special in that its presence
1116 anywhere in a pattern has a global effect.
1118 =item C<(?:pattern)>
1121 =item C<(?adluimsx-imsx:pattern)>
1123 =item C<(?^aluimsx:pattern)>
1126 This is for clustering, not capturing; it groups subexpressions like
1127 "()", but doesn't make backreferences as "()" does. So
1129 @fields = split(/\b(?:a|b|c)\b/)
1133 @fields = split(/\b(a|b|c)\b/)
1135 but doesn't spit out extra fields. It's also cheaper not to capture
1136 characters if you don't need to.
1138 Any letters between C<?> and C<:> act as flags modifiers as with
1139 C<(?adluimsx-imsx)>. For example,
1141 /(?s-i:more.*than).*million/i
1143 is equivalent to the more verbose
1145 /(?:(?s-i)more.*than).*million/i
1147 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
1148 after the C<"?"> is a shorthand equivalent to C<d-imsx>. Any positive
1149 flags (except C<"d">) may follow the caret, so
1157 The caret tells Perl that this cluster doesn't inherit the flags of any
1158 surrounding pattern, but uses the system defaults (C<d-imsx>),
1159 modified by any flags specified.
1161 The caret allows for simpler stringification of compiled regular
1162 expressions. These look like
1166 with any non-default flags appearing between the caret and the colon.
1167 A test that looks at such stringification thus doesn't need to have the
1168 system default flags hard-coded in it, just the caret. If new flags are
1169 added to Perl, the meaning of the caret's expansion will change to include
1170 the default for those flags, so the test will still work, unchanged.
1172 Specifying a negative flag after the caret is an error, as the flag is
1175 Mnemonic for C<(?^...)>: A fresh beginning since the usual use of a caret is
1176 to match at the beginning.
1178 =item C<(?|pattern)>
1179 X<(?|)> X<Branch reset>
1181 This is the "branch reset" pattern, which has the special property
1182 that the capture groups are numbered from the same starting point
1183 in each alternation branch. It is available starting from perl 5.10.0.
1185 Capture groups are numbered from left to right, but inside this
1186 construct the numbering is restarted for each branch.
1188 The numbering within each branch will be as normal, and any groups
1189 following this construct will be numbered as though the construct
1190 contained only one branch, that being the one with the most capture
1193 This construct is useful when you want to capture one of a
1194 number of alternative matches.
1196 Consider the following pattern. The numbers underneath show in
1197 which group the captured content will be stored.
1200 # before ---------------branch-reset----------- after
1201 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1204 Be careful when using the branch reset pattern in combination with
1205 named captures. Named captures are implemented as being aliases to
1206 numbered groups holding the captures, and that interferes with the
1207 implementation of the branch reset pattern. If you are using named
1208 captures in a branch reset pattern, it's best to use the same names,
1209 in the same order, in each of the alternations:
1211 /(?| (?<a> x ) (?<b> y )
1212 | (?<a> z ) (?<b> w )) /x
1214 Not doing so may lead to surprises:
1216 "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x;
1217 say $+ {a}; # Prints '12'
1218 say $+ {b}; # *Also* prints '12'.
1220 The problem here is that both the group named C<< a >> and the group
1221 named C<< b >> are aliases for the group belonging to C<< $1 >>.
1223 =item Look-Around Assertions
1224 X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
1226 Look-around assertions are zero-width patterns which match a specific
1227 pattern without including it in C<$&>. Positive assertions match when
1228 their subpattern matches, negative assertions match when their subpattern
1229 fails. Look-behind matches text up to the current match position,
1230 look-ahead matches text following the current match position.
1234 =item C<(?=pattern)>
1235 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
1237 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
1238 matches a word followed by a tab, without including the tab in C<$&>.
1240 =item C<(?!pattern)>
1241 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
1243 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
1244 matches any occurrence of "foo" that isn't followed by "bar". Note
1245 however that look-ahead and look-behind are NOT the same thing. You cannot
1246 use this for look-behind.
1248 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
1249 will not do what you want. That's because the C<(?!foo)> is just saying that
1250 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
1251 match. Use look-behind instead (see below).
1253 =item C<(?<=pattern)> C<\K>
1254 X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K>
1256 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
1257 matches a word that follows a tab, without including the tab in C<$&>.
1258 Works only for fixed-width look-behind.
1260 There is a special form of this construct, called C<\K> (available since
1261 Perl 5.10.0), which causes the
1262 regex engine to "keep" everything it had matched prior to the C<\K> and
1263 not include it in C<$&>. This effectively provides variable-length
1264 look-behind. The use of C<\K> inside of another look-around assertion
1265 is allowed, but the behaviour is currently not well defined.
1267 For various reasons C<\K> may be significantly more efficient than the
1268 equivalent C<< (?<=...) >> construct, and it is especially useful in
1269 situations where you want to efficiently remove something following
1270 something else in a string. For instance
1274 can be rewritten as the much more efficient
1278 =item C<(?<!pattern)>
1279 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
1281 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
1282 matches any occurrence of "foo" that does not follow "bar". Works
1283 only for fixed-width look-behind.
1287 =item C<(?'NAME'pattern)>
1289 =item C<< (?<NAME>pattern) >>
1290 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
1292 A named capture group. Identical in every respect to normal capturing
1293 parentheses C<()> but for the additional fact that the group
1294 can be referred to by name in various regular expression
1295 constructs (like C<\g{NAME}>) and can be accessed by name
1296 after a successful match via C<%+> or C<%->. See L<perlvar>
1297 for more details on the C<%+> and C<%-> hashes.
1299 If multiple distinct capture groups have the same name then the
1300 $+{NAME} will refer to the leftmost defined group in the match.
1302 The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent.
1304 B<NOTE:> While the notation of this construct is the same as the similar
1305 function in .NET regexes, the behavior is not. In Perl the groups are
1306 numbered sequentially regardless of being named or not. Thus in the
1311 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
1312 the opposite which is what a .NET regex hacker might expect.
1314 Currently NAME is restricted to simple identifiers only.
1315 In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
1316 its Unicode extension (see L<utf8>),
1317 though it isn't extended by the locale (see L<perllocale>).
1319 B<NOTE:> In order to make things easier for programmers with experience
1320 with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >>
1321 may be used instead of C<< (?<NAME>pattern) >>; however this form does not
1322 support the use of single quotes as a delimiter for the name.
1324 =item C<< \k<NAME> >>
1326 =item C<< \k'NAME' >>
1328 Named backreference. Similar to numeric backreferences, except that
1329 the group is designated by name and not number. If multiple groups
1330 have the same name then it refers to the leftmost defined group in
1333 It is an error to refer to a name not defined by a C<< (?<NAME>) >>
1334 earlier in the pattern.
1336 Both forms are equivalent.
1338 B<NOTE:> In order to make things easier for programmers with experience
1339 with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >>
1340 may be used instead of C<< \k<NAME> >>.
1342 =item C<(?{ code })>
1343 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
1345 B<WARNING>: Using this feature safely requires that you understand its
1346 limitations. Code executed that has side effects may not perform identically
1347 from version to version due to the effect of future optimisations in the regex
1348 engine. For more information on this, see L</Embedded Code Execution
1351 This zero-width assertion executes any embedded Perl code. It always
1352 succeeds, and its return value is set as C<$^R>.
1354 In literal patterns, the code is parsed at the same time as the
1355 surrounding code. While within the pattern, control is passed temporarily
1356 back to the perl parser, until the logically-balancing closing brace is
1357 encountered. This is similar to the way that an array index expression in
1358 a literal string is handled, for example
1360 "abc$array[ 1 + f('[') + g()]def"
1362 In particular, braces do not need to be balanced:
1364 s/abc(?{ f('{'); })/def/
1366 Even in a pattern that is interpolated and compiled at run-time, literal
1367 code blocks will be compiled once, at perl compile time; the following
1371 my $qr = qr/(?{ BEGIN { print "A" } })/;
1373 /$foo$qr(?{ BEGIN { print "B" } })/;
1376 In patterns where the text of the code is derived from run-time
1377 information rather than appearing literally in a source code /pattern/,
1378 the code is compiled at the same time that the pattern is compiled, and
1379 for reasons of security, C<use re 'eval'> must be in scope. This is to
1380 stop user-supplied patterns containing code snippets from being
1383 In situations where you need to enable this with C<use re 'eval'>, you should
1384 also have taint checking enabled. Better yet, use the carefully
1385 constrained evaluation within a Safe compartment. See L<perlsec> for
1386 details about both these mechanisms.
1388 From the viewpoint of parsing, lexical variable scope and closures,
1392 behaves approximately like
1394 /AAA/ && do { BBB } && /CCC/
1398 qr/AAA(?{ BBB })CCC/
1400 behaves approximately like
1402 sub { /AAA/ && do { BBB } && /CCC/ }
1406 { my $i = 1; $r = qr/(?{ print $i })/ }
1410 Inside a C<(?{...})> block, C<$_> refers to the string the regular
1411 expression is matching against. You can also use C<pos()> to know what is
1412 the current position of matching within this string.
1414 The code block introduces a new scope from the perspective of lexical
1415 variable declarations, but B<not> from the perspective of C<local> and
1416 similar localizing behaviours. So later code blocks within the same
1417 pattern will still see the values which were localized in earlier blocks.
1418 These accumulated localizations are undone either at the end of a
1419 successful match, or if the assertion is backtracked (compare
1420 L<"Backtracking">). For example,
1424 (?{ $cnt = 0 }) # Initialize $cnt.
1428 local $cnt = $cnt + 1; # Update $cnt,
1429 # backtracking-safe.
1433 (?{ $res = $cnt }) # On success copy to
1434 # non-localized location.
1437 will initially increment C<$cnt> up to 8; then during backtracking, its
1438 value will be unwound back to 4, which is the value assigned to C<$res>.
1439 At the end of the regex execution, $cnt will be wound back to its initial
1442 This assertion may be used as the condition in a
1444 (?(condition)yes-pattern|no-pattern)
1446 switch. If I<not> used in this way, the result of evaluation of C<code>
1447 is put into the special variable C<$^R>. This happens immediately, so
1448 C<$^R> can be used from other C<(?{ code })> assertions inside the same
1451 The assignment to C<$^R> above is properly localized, so the old
1452 value of C<$^R> is restored if the assertion is backtracked; compare
1455 Note that the special variable C<$^N> is particularly useful with code
1456 blocks to capture the results of submatches in variables without having to
1457 keep track of the number of nested parentheses. For example:
1459 $_ = "The brown fox jumps over the lazy dog";
1460 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
1461 print "color = $color, animal = $animal\n";
1464 =item C<(??{ code })>
1466 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
1468 B<WARNING>: Using this feature safely requires that you understand its
1469 limitations. Code executed that has side effects may not perform
1470 identically from version to version due to the effect of future
1471 optimisations in the regex engine. For more information on this, see
1472 L</Embedded Code Execution Frequency>.
1474 This is a "postponed" regular subexpression. It behaves in I<exactly> the
1475 same way as a C<(?{ code })> code block as described above, except that
1476 its return value, rather than being assigned to C<$^R>, is treated as a
1477 pattern, compiled if it's a string (or used as-is if its a qr// object),
1478 then matched as if it were inserted instead of this construct.
1480 During the matching of this sub-pattern, it has its own set of
1481 captures which are valid during the sub-match, but are discarded once
1482 control returns to the main pattern. For example, the following matches,
1483 with the inner pattern capturing "B" and matching "BB", while the outer
1484 pattern captures "A";
1486 my $inner = '(.)\1';
1487 "ABBA" =~ /^(.)(??{ $inner })\1/;
1488 print $1; # prints "A";
1490 Note that this means that there is no way for the inner pattern to refer
1491 to a capture group defined outside. (The code block itself can use C<$1>,
1492 etc., to refer to the enclosing pattern's capture groups.) Thus, although
1494 ('a' x 100)=~/(??{'(.)' x 100})/
1496 I<will> match, it will I<not> set $1 on exit.
1498 The following pattern matches a parenthesized group:
1503 (?> [^()]+ ) # Non-parens without backtracking
1505 (??{ $re }) # Group with matching parens
1511 L<C<(?I<PARNO>)>|/(?PARNO) (?-PARNO) (?+PARNO) (?R) (?0)>
1512 for a different, more efficient way to accomplish
1515 Executing a postponed regular expression 50 times without consuming any
1516 input string will result in a fatal error. The maximum depth is compiled
1517 into perl, so changing it requires a custom build.
1519 =item C<(?I<PARNO>)> C<(?-I<PARNO>)> C<(?+I<PARNO>)> C<(?R)> C<(?0)>
1520 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
1521 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
1522 X<regex, relative recursion> X<GOSUB> X<GOSTART>
1524 Recursive subpattern. Treat the contents of a given capture buffer in the
1525 current pattern as an independent subpattern and attempt to match it at
1526 the current position in the string. Information about capture state from
1527 the caller for things like backreferences is available to the subpattern,
1528 but capture buffers set by the subpattern are not visible to the caller.
1530 Similar to C<(??{ code })> except that it does not involve executing any
1531 code or potentially compiling a returned pattern string; instead it treats
1532 the part of the current pattern contained within a specified capture group
1533 as an independent pattern that must match at the current position. Also
1534 different is the treatment of capture buffers, unlike C<(??{ code })>
1535 recursive patterns have access to their callers match state, so one can
1536 use backreferences safely.
1538 I<PARNO> is a sequence of digits (not starting with 0) whose value reflects
1539 the paren-number of the capture group to recurse to. C<(?R)> recurses to
1540 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
1541 C<(?R)>. If I<PARNO> is preceded by a plus or minus sign then it is assumed
1542 to be relative, with negative numbers indicating preceding capture groups
1543 and positive ones following. Thus C<(?-1)> refers to the most recently
1544 declared group, and C<(?+1)> indicates the next group to be declared.
1545 Note that the counting for relative recursion differs from that of
1546 relative backreferences, in that with recursion unclosed groups B<are>
1549 The following pattern matches a function foo() which may contain
1550 balanced parentheses as the argument.
1552 $re = qr{ ( # paren group 1 (full function)
1554 ( # paren group 2 (parens)
1556 ( # paren group 3 (contents of parens)
1558 (?> [^()]+ ) # Non-parens without backtracking
1560 (?2) # Recurse to start of paren group 2
1568 If the pattern was used as follows
1570 'foo(bar(baz)+baz(bop))'=~/$re/
1571 and print "\$1 = $1\n",
1575 the output produced should be the following:
1577 $1 = foo(bar(baz)+baz(bop))
1578 $2 = (bar(baz)+baz(bop))
1579 $3 = bar(baz)+baz(bop)
1581 If there is no corresponding capture group defined, then it is a
1582 fatal error. Recursing deeper than 50 times without consuming any input
1583 string will also result in a fatal error. The maximum depth is compiled
1584 into perl, so changing it requires a custom build.
1586 The following shows how using negative indexing can make it
1587 easier to embed recursive patterns inside of a C<qr//> construct
1590 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
1591 if (/foo $parens \s+ \+ \s+ bar $parens/x) {
1592 # do something here...
1595 B<Note> that this pattern does not behave the same way as the equivalent
1596 PCRE or Python construct of the same form. In Perl you can backtrack into
1597 a recursed group, in PCRE and Python the recursed into group is treated
1598 as atomic. Also, modifiers are resolved at compile time, so constructs
1599 like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
1605 Recurse to a named subpattern. Identical to C<(?I<PARNO>)> except that the
1606 parenthesis to recurse to is determined by name. If multiple parentheses have
1607 the same name, then it recurses to the leftmost.
1609 It is an error to refer to a name that is not declared somewhere in the
1612 B<NOTE:> In order to make things easier for programmers with experience
1613 with the Python or PCRE regex engines the pattern C<< (?P>NAME) >>
1614 may be used instead of C<< (?&NAME) >>.
1616 =item C<(?(condition)yes-pattern|no-pattern)>
1619 =item C<(?(condition)yes-pattern)>
1621 Conditional expression. Matches C<yes-pattern> if C<condition> yields
1622 a true value, matches C<no-pattern> otherwise. A missing pattern always
1625 C<(condition)> should be one of: 1) an integer in
1626 parentheses (which is valid if the corresponding pair of parentheses
1627 matched); 2) a look-ahead/look-behind/evaluate zero-width assertion; 3) a
1628 name in angle brackets or single quotes (which is valid if a group
1629 with the given name matched); or 4) the special symbol (R) (true when
1630 evaluated inside of recursion or eval). Additionally the R may be
1631 followed by a number, (which will be true when evaluated when recursing
1632 inside of the appropriate group), or by C<&NAME>, in which case it will
1633 be true only when evaluated during recursion in the named group.
1635 Here's a summary of the possible predicates:
1641 Checks if the numbered capturing group has matched something.
1643 =item (<NAME>) ('NAME')
1645 Checks if a group with the given name has matched something.
1647 =item (?=...) (?!...) (?<=...) (?<!...)
1649 Checks whether the pattern matches (or does not match, for the '!'
1654 Treats the return value of the code block as the condition.
1658 Checks if the expression has been evaluated inside of recursion.
1662 Checks if the expression has been evaluated while executing directly
1663 inside of the n-th capture group. This check is the regex equivalent of
1665 if ((caller(0))[3] eq 'subname') { ... }
1667 In other words, it does not check the full recursion stack.
1671 Similar to C<(R1)>, this predicate checks to see if we're executing
1672 directly inside of the leftmost group with a given name (this is the same
1673 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1674 stack, but only the name of the innermost active recursion.
1678 In this case, the yes-pattern is never directly executed, and no
1679 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1680 See below for details.
1691 matches a chunk of non-parentheses, possibly included in parentheses
1694 A special form is the C<(DEFINE)> predicate, which never executes its
1695 yes-pattern directly, and does not allow a no-pattern. This allows one to
1696 define subpatterns which will be executed only by the recursion mechanism.
1697 This way, you can define a set of regular expression rules that can be
1698 bundled into any pattern you choose.
1700 It is recommended that for this usage you put the DEFINE block at the
1701 end of the pattern, and that you name any subpatterns defined within it.
1703 Also, it's worth noting that patterns defined this way probably will
1704 not be as efficient, as the optimizer is not very clever about
1707 An example of how this might be used is as follows:
1709 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
1712 (?<ADDRESS_PAT>....)
1715 Note that capture groups matched inside of recursion are not accessible
1716 after the recursion returns, so the extra layer of capturing groups is
1717 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1718 C<$+{NAME}> would be.
1720 Finally, keep in mind that subpatterns created inside a DEFINE block
1721 count towards the absolute and relative number of captures, so this:
1723 my @captures = "a" =~ /(.) # First capture
1725 (?<EXAMPLE> 1 ) # Second capture
1727 say scalar @captures;
1729 Will output 2, not 1. This is particularly important if you intend to
1730 compile the definitions with the C<qr//> operator, and later
1731 interpolate them in another pattern.
1733 =item C<< (?>pattern) >>
1734 X<backtrack> X<backtracking> X<atomic> X<possessive>
1736 An "independent" subexpression, one which matches the substring
1737 that a I<standalone> C<pattern> would match if anchored at the given
1738 position, and it matches I<nothing other than this substring>. This
1739 construct is useful for optimizations of what would otherwise be
1740 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1741 It may also be useful in places where the "grab all you can, and do not
1742 give anything back" semantic is desirable.
1744 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1745 (anchored at the beginning of string, as above) will match I<all>
1746 characters C<a> at the beginning of string, leaving no C<a> for
1747 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1748 since the match of the subgroup C<a*> is influenced by the following
1749 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1750 C<a*ab> will match fewer characters than a standalone C<a*>, since
1751 this makes the tail match.
1753 C<< (?>pattern) >> does not disable backtracking altogether once it has
1754 matched. It is still possible to backtrack past the construct, but not
1755 into it. So C<< ((?>a*)|(?>b*))ar >> will still match "bar".
1757 An effect similar to C<< (?>pattern) >> may be achieved by writing
1758 C<(?=(pattern))\g{-1}>. This matches the same substring as a standalone
1759 C<a+>, and the following C<\g{-1}> eats the matched string; it therefore
1760 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1761 (The difference between these two constructs is that the second one
1762 uses a capturing group, thus shifting ordinals of backreferences
1763 in the rest of a regular expression.)
1765 Consider this pattern:
1776 That will efficiently match a nonempty group with matching parentheses
1777 two levels deep or less. However, if there is no such group, it
1778 will take virtually forever on a long string. That's because there
1779 are so many different ways to split a long string into several
1780 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1781 to a subpattern of the above pattern. Consider how the pattern
1782 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1783 seconds, but that each extra letter doubles this time. This
1784 exponential performance will make it appear that your program has
1785 hung. However, a tiny change to this pattern
1789 (?> [^()]+ ) # change x+ above to (?> x+ )
1796 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1797 this yourself would be a productive exercise), but finishes in a fourth
1798 the time when used on a similar string with 1000000 C<a>s. Be aware,
1799 however, that, when this construct is followed by a
1800 quantifier, it currently triggers a warning message under
1801 the C<use warnings> pragma or B<-w> switch saying it
1802 C<"matches null string many times in regex">.
1804 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1805 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1806 This was only 4 times slower on a string with 1000000 C<a>s.
1808 The "grab all you can, and do not give anything back" semantic is desirable
1809 in many situations where on the first sight a simple C<()*> looks like
1810 the correct solution. Suppose we parse text with comments being delimited
1811 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1812 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1813 the comment delimiter, because it may "give up" some whitespace if
1814 the remainder of the pattern can be made to match that way. The correct
1815 answer is either one of these:
1820 For example, to grab non-empty comments into $1, one should use either
1823 / (?> \# [ \t]* ) ( .+ ) /x;
1824 / \# [ \t]* ( [^ \t] .* ) /x;
1826 Which one you pick depends on which of these expressions better reflects
1827 the above specification of comments.
1829 In some literature this construct is called "atomic matching" or
1830 "possessive matching".
1832 Possessive quantifiers are equivalent to putting the item they are applied
1833 to inside of one of these constructs. The following equivalences apply:
1835 Quantifier Form Bracketing Form
1836 --------------- ---------------
1840 PAT{min,max}+ (?>PAT{min,max})
1844 See L<perlrecharclass/Extended Bracketed Character Classes>.
1848 =head2 Special Backtracking Control Verbs
1850 These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1851 otherwise stated the ARG argument is optional; in some cases, it is
1854 Any pattern containing a special backtracking verb that allows an argument
1855 has the special behaviour that when executed it sets the current package's
1856 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1859 On failure, the C<$REGERROR> variable will be set to the ARG value of the
1860 verb pattern, if the verb was involved in the failure of the match. If the
1861 ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1862 name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1863 none. Also, the C<$REGMARK> variable will be set to FALSE.
1865 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1866 the C<$REGMARK> variable will be set to the name of the last
1867 C<(*MARK:NAME)> pattern executed. See the explanation for the
1868 C<(*MARK:NAME)> verb below for more details.
1870 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1871 and most other regex-related variables. They are not local to a scope, nor
1872 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1873 Use C<local> to localize changes to them to a specific scope if necessary.
1875 If a pattern does not contain a special backtracking verb that allows an
1876 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
1880 =item Verbs that take an argument
1884 =item C<(*PRUNE)> C<(*PRUNE:NAME)>
1885 X<(*PRUNE)> X<(*PRUNE:NAME)>
1887 This zero-width pattern prunes the backtracking tree at the current point
1888 when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1889 where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1890 A may backtrack as necessary to match. Once it is reached, matching
1891 continues in B, which may also backtrack as necessary; however, should B
1892 not match, then no further backtracking will take place, and the pattern
1893 will fail outright at the current starting position.
1895 The following example counts all the possible matching strings in a
1896 pattern (without actually matching any of them).
1898 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1899 print "Count=$count\n";
1914 If we add a C<(*PRUNE)> before the count like the following
1916 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
1917 print "Count=$count\n";
1919 we prevent backtracking and find the count of the longest matching string
1920 at each matching starting point like so:
1927 Any number of C<(*PRUNE)> assertions may be used in a pattern.
1929 See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1930 control backtracking. In some cases, the use of C<(*PRUNE)> can be
1931 replaced with a C<< (?>pattern) >> with no functional difference; however,
1932 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1933 C<< (?>pattern) >> alone.
1935 =item C<(*SKIP)> C<(*SKIP:NAME)>
1938 This zero-width pattern is similar to C<(*PRUNE)>, except that on
1939 failure it also signifies that whatever text that was matched leading up
1940 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1941 of this pattern. This effectively means that the regex engine "skips" forward
1942 to this position on failure and tries to match again, (assuming that
1943 there is sufficient room to match).
1945 The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1946 C<(*MARK:NAME)> was encountered while matching, then it is that position
1947 which is used as the "skip point". If no C<(*MARK)> of that name was
1948 encountered, then the C<(*SKIP)> operator has no effect. When used
1949 without a name the "skip point" is where the match point was when
1950 executing the (*SKIP) pattern.
1952 Compare the following to the examples in C<(*PRUNE)>; note the string
1955 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
1956 print "Count=$count\n";
1964 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
1965 executed, the next starting point will be where the cursor was when the
1966 C<(*SKIP)> was executed.
1968 =item C<(*MARK:NAME)> C<(*:NAME)>
1969 X<(*MARK)> X<(*MARK:NAME)> X<(*:NAME)>
1971 This zero-width pattern can be used to mark the point reached in a string
1972 when a certain part of the pattern has been successfully matched. This
1973 mark may be given a name. A later C<(*SKIP)> pattern will then skip
1974 forward to that point if backtracked into on failure. Any number of
1975 C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated.
1977 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1978 can be used to "label" a pattern branch, so that after matching, the
1979 program can determine which branches of the pattern were involved in the
1982 When a match is successful, the C<$REGMARK> variable will be set to the
1983 name of the most recently executed C<(*MARK:NAME)> that was involved
1986 This can be used to determine which branch of a pattern was matched
1987 without using a separate capture group for each branch, which in turn
1988 can result in a performance improvement, as perl cannot optimize
1989 C</(?:(x)|(y)|(z))/> as efficiently as something like
1990 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1992 When a match has failed, and unless another verb has been involved in
1993 failing the match and has provided its own name to use, the C<$REGERROR>
1994 variable will be set to the name of the most recently executed
1997 See L</(*SKIP)> for more details.
1999 As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
2001 =item C<(*THEN)> C<(*THEN:NAME)>
2003 This is similar to the "cut group" operator C<::> from Perl 6. Like
2004 C<(*PRUNE)>, this verb always matches, and when backtracked into on
2005 failure, it causes the regex engine to try the next alternation in the
2006 innermost enclosing group (capturing or otherwise) that has alternations.
2007 The two branches of a C<(?(condition)yes-pattern|no-pattern)> do not
2008 count as an alternation, as far as C<(*THEN)> is concerned.
2010 Its name comes from the observation that this operation combined with the
2011 alternation operator (C<|>) can be used to create what is essentially a
2012 pattern-based if/then/else block:
2014 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
2016 Note that if this operator is used and NOT inside of an alternation then
2017 it acts exactly like the C<(*PRUNE)> operator.
2027 / ( A (*THEN) B | C ) /
2031 / ( A (*PRUNE) B | C ) /
2033 as after matching the A but failing on the B the C<(*THEN)> verb will
2034 backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
2038 =item Verbs without an argument
2045 This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a
2046 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
2047 into on failure it causes the match to fail outright. No further attempts
2048 to find a valid match by advancing the start pointer will occur again.
2051 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
2052 print "Count=$count\n";
2059 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
2060 does not match, the regex engine will not try any further matching on the
2063 =item C<(*FAIL)> C<(*F)>
2066 This pattern matches nothing and always fails. It can be used to force the
2067 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
2068 fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
2070 It is probably useful only when combined with C<(?{})> or C<(??{})>.
2075 This pattern matches nothing and causes the end of successful matching at
2076 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
2077 whether there is actually more to match in the string. When inside of a
2078 nested pattern, such as recursion, or in a subpattern dynamically generated
2079 via C<(??{})>, only the innermost pattern is ended immediately.
2081 If the C<(*ACCEPT)> is inside of capturing groups then the groups are
2082 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
2085 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
2087 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
2088 be set. If another branch in the inner parentheses was matched, such as in the
2089 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
2096 X<backtrack> X<backtracking>
2098 NOTE: This section presents an abstract approximation of regular
2099 expression behavior. For a more rigorous (and complicated) view of
2100 the rules involved in selecting a match among possible alternatives,
2101 see L<Combining RE Pieces>.
2103 A fundamental feature of regular expression matching involves the
2104 notion called I<backtracking>, which is currently used (when needed)
2105 by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>,
2106 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
2107 internally, but the general principle outlined here is valid.
2109 For a regular expression to match, the I<entire> regular expression must
2110 match, not just part of it. So if the beginning of a pattern containing a
2111 quantifier succeeds in a way that causes later parts in the pattern to
2112 fail, the matching engine backs up and recalculates the beginning
2113 part--that's why it's called backtracking.
2115 Here is an example of backtracking: Let's say you want to find the
2116 word following "foo" in the string "Food is on the foo table.":
2118 $_ = "Food is on the foo table.";
2119 if ( /\b(foo)\s+(\w+)/i ) {
2120 print "$2 follows $1.\n";
2123 When the match runs, the first part of the regular expression (C<\b(foo)>)
2124 finds a possible match right at the beginning of the string, and loads up
2125 $1 with "Foo". However, as soon as the matching engine sees that there's
2126 no whitespace following the "Foo" that it had saved in $1, it realizes its
2127 mistake and starts over again one character after where it had the
2128 tentative match. This time it goes all the way until the next occurrence
2129 of "foo". The complete regular expression matches this time, and you get
2130 the expected output of "table follows foo."
2132 Sometimes minimal matching can help a lot. Imagine you'd like to match
2133 everything between "foo" and "bar". Initially, you write something
2136 $_ = "The food is under the bar in the barn.";
2137 if ( /foo(.*)bar/ ) {
2141 Which perhaps unexpectedly yields:
2143 got <d is under the bar in the >
2145 That's because C<.*> was greedy, so you get everything between the
2146 I<first> "foo" and the I<last> "bar". Here it's more effective
2147 to use minimal matching to make sure you get the text between a "foo"
2148 and the first "bar" thereafter.
2150 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
2151 got <d is under the >
2153 Here's another example. Let's say you'd like to match a number at the end
2154 of a string, and you also want to keep the preceding part of the match.
2157 $_ = "I have 2 numbers: 53147";
2158 if ( /(.*)(\d*)/ ) { # Wrong!
2159 print "Beginning is <$1>, number is <$2>.\n";
2162 That won't work at all, because C<.*> was greedy and gobbled up the
2163 whole string. As C<\d*> can match on an empty string the complete
2164 regular expression matched successfully.
2166 Beginning is <I have 2 numbers: 53147>, number is <>.
2168 Here are some variants, most of which don't work:
2170 $_ = "I have 2 numbers: 53147";
2183 printf "%-12s ", $pat;
2185 print "<$1> <$2>\n";
2191 That will print out:
2193 (.*)(\d*) <I have 2 numbers: 53147> <>
2194 (.*)(\d+) <I have 2 numbers: 5314> <7>
2196 (.*?)(\d+) <I have > <2>
2197 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
2198 (.*?)(\d+)$ <I have 2 numbers: > <53147>
2199 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
2200 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
2202 As you see, this can be a bit tricky. It's important to realize that a
2203 regular expression is merely a set of assertions that gives a definition
2204 of success. There may be 0, 1, or several different ways that the
2205 definition might succeed against a particular string. And if there are
2206 multiple ways it might succeed, you need to understand backtracking to
2207 know which variety of success you will achieve.
2209 When using look-ahead assertions and negations, this can all get even
2210 trickier. Imagine you'd like to find a sequence of non-digits not
2211 followed by "123". You might try to write that as
2214 if ( /^\D*(?!123)/ ) { # Wrong!
2215 print "Yup, no 123 in $_\n";
2218 But that isn't going to match; at least, not the way you're hoping. It
2219 claims that there is no 123 in the string. Here's a clearer picture of
2220 why that pattern matches, contrary to popular expectations:
2225 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
2226 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
2228 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
2229 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
2237 You might have expected test 3 to fail because it seems to a more
2238 general purpose version of test 1. The important difference between
2239 them is that test 3 contains a quantifier (C<\D*>) and so can use
2240 backtracking, whereas test 1 will not. What's happening is
2241 that you've asked "Is it true that at the start of $x, following 0 or more
2242 non-digits, you have something that's not 123?" If the pattern matcher had
2243 let C<\D*> expand to "ABC", this would have caused the whole pattern to
2246 The search engine will initially match C<\D*> with "ABC". Then it will
2247 try to match C<(?!123)> with "123", which fails. But because
2248 a quantifier (C<\D*>) has been used in the regular expression, the
2249 search engine can backtrack and retry the match differently
2250 in the hope of matching the complete regular expression.
2252 The pattern really, I<really> wants to succeed, so it uses the
2253 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
2254 time. Now there's indeed something following "AB" that is not
2255 "123". It's "C123", which suffices.
2257 We can deal with this by using both an assertion and a negation.
2258 We'll say that the first part in $1 must be followed both by a digit
2259 and by something that's not "123". Remember that the look-aheads
2260 are zero-width expressions--they only look, but don't consume any
2261 of the string in their match. So rewriting this way produces what
2262 you'd expect; that is, case 5 will fail, but case 6 succeeds:
2264 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
2265 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
2269 In other words, the two zero-width assertions next to each other work as though
2270 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
2271 matches only if you're at the beginning of the line AND the end of the
2272 line simultaneously. The deeper underlying truth is that juxtaposition in
2273 regular expressions always means AND, except when you write an explicit OR
2274 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
2275 although the attempted matches are made at different positions because "a"
2276 is not a zero-width assertion, but a one-width assertion.
2278 B<WARNING>: Particularly complicated regular expressions can take
2279 exponential time to solve because of the immense number of possible
2280 ways they can use backtracking to try for a match. For example, without
2281 internal optimizations done by the regular expression engine, this will
2282 take a painfully long time to run:
2284 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
2286 And if you used C<*>'s in the internal groups instead of limiting them
2287 to 0 through 5 matches, then it would take forever--or until you ran
2288 out of stack space. Moreover, these internal optimizations are not
2289 always applicable. For example, if you put C<{0,5}> instead of C<*>
2290 on the external group, no current optimization is applicable, and the
2291 match takes a long time to finish.
2293 A powerful tool for optimizing such beasts is what is known as an
2294 "independent group",
2295 which does not backtrack (see L</C<< (?>pattern) >>>). Note also that
2296 zero-length look-ahead/look-behind assertions will not backtrack to make
2297 the tail match, since they are in "logical" context: only
2298 whether they match is considered relevant. For an example
2299 where side-effects of look-ahead I<might> have influenced the
2300 following match, see L</C<< (?>pattern) >>>.
2302 =head2 Version 8 Regular Expressions
2303 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
2305 In case you're not familiar with the "regular" Version 8 regex
2306 routines, here are the pattern-matching rules not described above.
2308 Any single character matches itself, unless it is a I<metacharacter>
2309 with a special meaning described here or above. You can cause
2310 characters that normally function as metacharacters to be interpreted
2311 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
2312 character; "\\" matches a "\"). This escape mechanism is also required
2313 for the character used as the pattern delimiter.
2315 A series of characters matches that series of characters in the target
2316 string, so the pattern C<blurfl> would match "blurfl" in the target
2319 You can specify a character class, by enclosing a list of characters
2320 in C<[]>, which will match any character from the list. If the
2321 first character after the "[" is "^", the class matches any character not
2322 in the list. Within a list, the "-" character specifies a
2323 range, so that C<a-z> represents all characters between "a" and "z",
2324 inclusive. If you want either "-" or "]" itself to be a member of a
2325 class, put it at the start of the list (possibly after a "^"), or
2326 escape it with a backslash. "-" is also taken literally when it is
2327 at the end of the list, just before the closing "]". (The
2328 following all specify the same class of three characters: C<[-az]>,
2329 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
2330 specifies a class containing twenty-six characters, even on EBCDIC-based
2331 character sets.) Also, if you try to use the character
2332 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
2333 a range, the "-" is understood literally.
2335 Note also that the whole range idea is rather unportable between
2336 character sets, except for four situations that Perl handles specially.
2337 Any subset of the ranges C<[A-Z]>, C<[a-z]>, and C<[0-9]> are guaranteed
2338 to match the expected subset of ASCII characters, no matter what
2339 character set the platform is running. The fourth portable way to
2340 specify ranges is to use the C<\N{...}> syntax to specify either end
2341 point of the range. For example, C<[\N{U+04}-\N{U+07}]> means to match
2342 the Unicode code points C<\N{U+04}>, C<\N{U+05}>, C<\N{U+06}>, and
2343 C<\N{U+07}>, whatever their native values may be on the platform. Under
2344 L<use re 'strict'|re/'strict' mode> or within a L</C<(?[ ])>>, a warning
2345 is raised, if enabled, and the other end point of a range which has a
2346 C<\N{...}> endpoint is not portably specified. For example,
2348 [\N{U+00}-\x06] # Warning under "use re 'strict'".
2350 It is hard to understand without digging what exactly matches ranges
2351 other than subsets of C<[A-Z]>, C<[a-z]>, and C<[0-9]>. A sound
2352 principle is to use only ranges that begin from and end at either
2353 alphabetics of equal case ([a-e], [A-E]), or digits ([0-9]). Anything
2354 else is unsafe or unclear. If in doubt, spell out the range in full.
2356 Characters may be specified using a metacharacter syntax much like that
2357 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
2358 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
2359 of three octal digits, matches the character whose coded character set value
2360 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
2361 matches the character whose ordinal is I<nn>. The expression \cI<x>
2362 matches the character control-I<x>. Finally, the "." metacharacter
2363 matches any character except "\n" (unless you use C</s>).
2365 You can specify a series of alternatives for a pattern using "|" to
2366 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
2367 or "foe" in the target string (as would C<f(e|i|o)e>). The
2368 first alternative includes everything from the last pattern delimiter
2369 ("(", "(?:", etc. or the beginning of the pattern) up to the first "|", and
2370 the last alternative contains everything from the last "|" to the next
2371 closing pattern delimiter. That's why it's common practice to include
2372 alternatives in parentheses: to minimize confusion about where they
2375 Alternatives are tried from left to right, so the first
2376 alternative found for which the entire expression matches, is the one that
2377 is chosen. This means that alternatives are not necessarily greedy. For
2378 example: when matching C<foo|foot> against "barefoot", only the "foo"
2379 part will match, as that is the first alternative tried, and it successfully
2380 matches the target string. (This might not seem important, but it is
2381 important when you are capturing matched text using parentheses.)
2383 Also remember that "|" is interpreted as a literal within square brackets,
2384 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
2386 Within a pattern, you may designate subpatterns for later reference
2387 by enclosing them in parentheses, and you may refer back to the
2388 I<n>th subpattern later in the pattern using the metacharacter
2389 \I<n> or \gI<n>. Subpatterns are numbered based on the left to right order
2390 of their opening parenthesis. A backreference matches whatever
2391 actually matched the subpattern in the string being examined, not
2392 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\g1\d*> will
2393 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
2394 1 matched "0x", even though the rule C<0|0x> could potentially match
2395 the leading 0 in the second number.
2397 =head2 Warning on \1 Instead of $1
2399 Some people get too used to writing things like:
2401 $pattern =~ s/(\W)/\\\1/g;
2403 This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid
2405 B<sed> addicts, but it's a dirty habit to get into. That's because in
2406 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
2407 the usual double-quoted string means a control-A. The customary Unix
2408 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
2409 of doing that, you get yourself into trouble if you then add an C</e>
2412 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
2418 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
2419 C<${1}000>. The operation of interpolation should not be confused
2420 with the operation of matching a backreference. Certainly they mean two
2421 different things on the I<left> side of the C<s///>.
2423 =head2 Repeated Patterns Matching a Zero-length Substring
2425 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
2427 Regular expressions provide a terse and powerful programming language. As
2428 with most other power tools, power comes together with the ability
2431 A common abuse of this power stems from the ability to make infinite
2432 loops using regular expressions, with something as innocuous as:
2434 'foo' =~ m{ ( o? )* }x;
2436 The C<o?> matches at the beginning of C<'foo'>, and since the position
2437 in the string is not moved by the match, C<o?> would match again and again
2438 because of the C<*> quantifier. Another common way to create a similar cycle
2439 is with the looping modifier C<//g>:
2441 @matches = ( 'foo' =~ m{ o? }xg );
2445 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
2447 or the loop implied by split().
2449 However, long experience has shown that many programming tasks may
2450 be significantly simplified by using repeated subexpressions that
2451 may match zero-length substrings. Here's a simple example being:
2453 @chars = split //, $string; # // is not magic in split
2454 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
2456 Thus Perl allows such constructs, by I<forcefully breaking
2457 the infinite loop>. The rules for this are different for lower-level
2458 loops given by the greedy quantifiers C<*+{}>, and for higher-level
2459 ones like the C</g> modifier or split() operator.
2461 The lower-level loops are I<interrupted> (that is, the loop is
2462 broken) when Perl detects that a repeated expression matched a
2463 zero-length substring. Thus
2465 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
2467 is made equivalent to
2469 m{ (?: NON_ZERO_LENGTH )* (?: ZERO_LENGTH )? }x;
2471 For example, this program
2478 (?{print "hello"}) # print hello whenever this
2480 (?=(b)) # zero-width assertion
2481 )* # any number of times
2492 Notice that "hello" is only printed once, as when Perl sees that the sixth
2493 iteration of the outermost C<(?:)*> matches a zero-length string, it stops
2496 The higher-level loops preserve an additional state between iterations:
2497 whether the last match was zero-length. To break the loop, the following
2498 match after a zero-length match is prohibited to have a length of zero.
2499 This prohibition interacts with backtracking (see L<"Backtracking">),
2500 and so the I<second best> match is chosen if the I<best> match is of
2508 results in C<< <><b><><a><><r><> >>. At each position of the string the best
2509 match given by non-greedy C<??> is the zero-length match, and the I<second
2510 best> match is what is matched by C<\w>. Thus zero-length matches
2511 alternate with one-character-long matches.
2513 Similarly, for repeated C<m/()/g> the second-best match is the match at the
2514 position one notch further in the string.
2516 The additional state of being I<matched with zero-length> is associated with
2517 the matched string, and is reset by each assignment to pos().
2518 Zero-length matches at the end of the previous match are ignored
2521 =head2 Combining RE Pieces
2523 Each of the elementary pieces of regular expressions which were described
2524 before (such as C<ab> or C<\Z>) could match at most one substring
2525 at the given position of the input string. However, in a typical regular
2526 expression these elementary pieces are combined into more complicated
2527 patterns using combining operators C<ST>, C<S|T>, C<S*> etc.
2528 (in these examples C<S> and C<T> are regular subexpressions).
2530 Such combinations can include alternatives, leading to a problem of choice:
2531 if we match a regular expression C<a|ab> against C<"abc">, will it match
2532 substring C<"a"> or C<"ab">? One way to describe which substring is
2533 actually matched is the concept of backtracking (see L<"Backtracking">).
2534 However, this description is too low-level and makes you think
2535 in terms of a particular implementation.
2537 Another description starts with notions of "better"/"worse". All the
2538 substrings which may be matched by the given regular expression can be
2539 sorted from the "best" match to the "worst" match, and it is the "best"
2540 match which is chosen. This substitutes the question of "what is chosen?"
2541 by the question of "which matches are better, and which are worse?".
2543 Again, for elementary pieces there is no such question, since at most
2544 one match at a given position is possible. This section describes the
2545 notion of better/worse for combining operators. In the description
2546 below C<S> and C<T> are regular subexpressions.
2552 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
2553 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
2554 which can be matched by C<T>.
2556 If C<A> is a better match for C<S> than C<A'>, C<AB> is a better
2559 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
2560 C<B> is a better match for C<T> than C<B'>.
2564 When C<S> can match, it is a better match than when only C<T> can match.
2566 Ordering of two matches for C<S> is the same as for C<S>. Similar for
2567 two matches for C<T>.
2569 =item C<S{REPEAT_COUNT}>
2571 Matches as C<SSS...S> (repeated as many times as necessary).
2575 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
2577 =item C<S{min,max}?>
2579 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
2581 =item C<S?>, C<S*>, C<S+>
2583 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
2585 =item C<S??>, C<S*?>, C<S+?>
2587 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
2591 Matches the best match for C<S> and only that.
2593 =item C<(?=S)>, C<(?<=S)>
2595 Only the best match for C<S> is considered. (This is important only if
2596 C<S> has capturing parentheses, and backreferences are used somewhere
2597 else in the whole regular expression.)
2599 =item C<(?!S)>, C<(?<!S)>
2601 For this grouping operator there is no need to describe the ordering, since
2602 only whether or not C<S> can match is important.
2604 =item C<(??{ EXPR })>, C<(?I<PARNO>)>
2606 The ordering is the same as for the regular expression which is
2607 the result of EXPR, or the pattern contained by capture group I<PARNO>.
2609 =item C<(?(condition)yes-pattern|no-pattern)>
2611 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
2612 already determined. The ordering of the matches is the same as for the
2613 chosen subexpression.
2617 The above recipes describe the ordering of matches I<at a given position>.
2618 One more rule is needed to understand how a match is determined for the
2619 whole regular expression: a match at an earlier position is always better
2620 than a match at a later position.
2622 =head2 Creating Custom RE Engines
2624 As of Perl 5.10.0, one can create custom regular expression engines. This
2625 is not for the faint of heart, as they have to plug in at the C level. See
2626 L<perlreapi> for more details.
2628 As an alternative, overloaded constants (see L<overload>) provide a simple
2629 way to extend the functionality of the RE engine, by substituting one
2630 pattern for another.
2632 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
2633 matches at a boundary between whitespace characters and non-whitespace
2634 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
2635 at these positions, so we want to have each C<\Y|> in the place of the
2636 more complicated version. We can create a module C<customre> to do
2644 die "No argument to customre::import allowed" if @_;
2645 overload::constant 'qr' => \&convert;
2648 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
2650 # We must also take care of not escaping the legitimate \\Y|
2651 # sequence, hence the presence of '\\' in the conversion rules.
2652 my %rules = ( '\\' => '\\\\',
2653 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
2659 { $rules{$1} or invalid($re,$1) }sgex;
2663 Now C<use customre> enables the new escape in constant regular
2664 expressions, i.e., those without any runtime variable interpolations.
2665 As documented in L<overload>, this conversion will work only over
2666 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
2667 part of this regular expression needs to be converted explicitly
2668 (but only if the special meaning of C<\Y|> should be enabled inside $re):
2673 $re = customre::convert $re;
2676 =head2 Embedded Code Execution Frequency
2678 The exact rules for how often (??{}) and (?{}) are executed in a pattern
2679 are unspecified. In the case of a successful match you can assume that
2680 they DWIM and will be executed in left to right order the appropriate
2681 number of times in the accepting path of the pattern as would any other
2682 meta-pattern. How non-accepting pathways and match failures affect the
2683 number of times a pattern is executed is specifically unspecified and
2684 may vary depending on what optimizations can be applied to the pattern
2685 and is likely to change from version to version.
2689 "aaabcdeeeee"=~/a(?{print "a"})b(?{print "b"})cde/;
2691 the exact number of times "a" or "b" are printed out is unspecified for
2692 failure, but you may assume they will be printed at least once during
2693 a successful match, additionally you may assume that if "b" is printed,
2694 it will be preceded by at least one "a".
2696 In the case of branching constructs like the following:
2698 /a(b|(?{ print "a" }))c(?{ print "c" })/;
2700 you can assume that the input "ac" will output "ac", and that "abc"
2701 will output only "c".
2703 When embedded code is quantified, successful matches will call the
2704 code once for each matched iteration of the quantifier. For
2707 "good" =~ /g(?:o(?{print "o"}))*d/;
2709 will output "o" twice.
2711 =head2 PCRE/Python Support
2713 As of Perl 5.10.0, Perl supports several Python/PCRE-specific extensions
2714 to the regex syntax. While Perl programmers are encouraged to use the
2715 Perl-specific syntax, the following are also accepted:
2719 =item C<< (?PE<lt>NAMEE<gt>pattern) >>
2721 Define a named capture group. Equivalent to C<< (?<NAME>pattern) >>.
2723 =item C<< (?P=NAME) >>
2725 Backreference to a named capture group. Equivalent to C<< \g{NAME} >>.
2727 =item C<< (?P>NAME) >>
2729 Subroutine call to a named capture group. Equivalent to C<< (?&NAME) >>.
2735 Many regular expression constructs don't work on EBCDIC platforms.
2737 There are a number of issues with regard to case-insensitive matching
2738 in Unicode rules. See C<i> under L</Modifiers> above.
2740 This document varies from difficult to understand to completely
2741 and utterly opaque. The wandering prose riddled with jargon is
2742 hard to fathom in several places.
2744 This document needs a rewrite that separates the tutorial content
2745 from the reference content.
2753 L<perlop/"Regexp Quote-Like Operators">.
2755 L<perlop/"Gory details of parsing quoted constructs">.
2765 I<Mastering Regular Expressions> by Jeffrey Friedl, published
2766 by O'Reilly and Associates.