2 X<regular expression> X<regex> X<regexp>
4 perlre - Perl regular expressions
8 This page describes the syntax of regular expressions in Perl.
10 If you haven't used regular expressions before, a quick-start
11 introduction is available in L<perlrequick>, and a longer tutorial
12 introduction is available in L<perlretut>.
14 For reference on how regular expressions are used in matching
15 operations, plus various examples of the same, see discussions of
16 C<m//>, C<s///>, C<qr//> and C<??> in L<perlop/"Regexp Quote-Like
22 Matching operations can have various modifiers. Modifiers
23 that relate to the interpretation of the regular expression inside
24 are listed below. Modifiers that alter the way a regular expression
25 is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and
26 L<perlop/"Gory details of parsing quoted constructs">.
31 X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline>
33 Treat string as multiple lines. That is, change "^" and "$" from matching
34 the start or end of the string to matching the start or end of any
35 line anywhere within the string.
38 X</s> X<regex, single-line> X<regexp, single-line>
39 X<regular expression, single-line>
41 Treat string as single line. That is, change "." to match any character
42 whatsoever, even a newline, which normally it would not match.
44 Used together, as C</ms>, they let the "." match any character whatsoever,
45 while still allowing "^" and "$" to match, respectively, just after
46 and just before newlines within the string.
49 X</i> X<regex, case-insensitive> X<regexp, case-insensitive>
50 X<regular expression, case-insensitive>
52 Do case-insensitive pattern matching.
54 If locale matching rules are in effect, the case map is taken from the
56 locale for code points less than 255, and from Unicode rules for larger
57 code points. However, matches that would cross the Unicode
58 rules/non-Unicode rules boundary (ords 255/256) will not succeed. See
61 There are a number of Unicode characters that match multiple characters
62 under C</i>. For example, C<LATIN SMALL LIGATURE FI>
63 should match the sequence C<fi>. Perl is not
64 currently able to do this when the multiple characters are in the pattern and
65 are split between groupings, or when one or more are quantified. Thus
67 "\N{LATIN SMALL LIGATURE FI}" =~ /fi/i; # Matches
68 "\N{LATIN SMALL LIGATURE FI}" =~ /[fi][fi]/i; # Doesn't match!
69 "\N{LATIN SMALL LIGATURE FI}" =~ /fi*/i; # Doesn't match!
71 # The below doesn't match, and it isn't clear what $1 and $2 would
73 "\N{LATIN SMALL LIGATURE FI}" =~ /(f)(i)/i; # Doesn't match!
75 Perl doesn't match multiple characters in an inverted bracketed
76 character class, which otherwise could be highly confusing. See
77 L<perlrecharclass/Negation>.
79 Also, Perl matching doesn't fully conform to the current Unicode C</i>
80 recommendations, which ask that the matching be made upon the NFD
81 (Normalization Form Decomposed) of the text. However, Unicode is
82 in the process of reconsidering and revising their recommendations.
87 Extend your pattern's legibility by permitting whitespace and comments.
91 X</p> X<regex, preserve> X<regexp, preserve>
93 Preserve the string matched such that ${^PREMATCH}, ${^MATCH}, and
94 ${^POSTMATCH} are available for use after matching.
99 Global matching, and keep the Current position after failed matching.
100 Unlike i, m, s and x, these two flags affect the way the regex is used
101 rather than the regex itself. See
102 L<perlretut/"Using regular expressions in Perl"> for further explanation
103 of the g and c modifiers.
106 X</a> X</d> X</l> X</u>
108 These modifiers, new in 5.14, affect which character-set semantics
109 (Unicode, ASCII, etc.) are used, as described below in
110 L</Character set modifiers>.
114 These are usually written as "the C</x> modifier", even though the delimiter
115 in question might not really be a slash. The modifiers C</imsxadlup>
116 may also be embedded within the regular expression itself using
117 the C<(?...)> construct, see L</Extended Patterns> below.
119 The C</x>, C</l>, C</u>, C</a> and C</d> modifiers need a little more
125 the regular expression parser to ignore most whitespace that is neither
126 backslashed nor within a character class. You can use this to break up
127 your regular expression into (slightly) more readable parts. The C<#>
128 character is also treated as a metacharacter introducing a comment,
129 just as in ordinary Perl code. This also means that if you want real
130 whitespace or C<#> characters in the pattern (outside a character
131 class, where they are unaffected by C</x>), then you'll either have to
132 escape them (using backslashes or C<\Q...\E>) or encode them using octal,
133 hex, or C<\N{}> escapes. Taken together, these features go a long way towards
134 making Perl's regular expressions more readable. Note that you have to
135 be careful not to include the pattern delimiter in the comment--perl has
136 no way of knowing you did not intend to close the pattern early. See
137 the C-comment deletion code in L<perlop>. Also note that anything inside
138 a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect
139 space interpretation within a single multi-character construct. For
140 example in C<\x{...}>, regardless of the C</x> modifier, there can be no
141 spaces. Same for a L<quantifier|/Quantifiers> such as C<{3}> or
142 C<{5,}>. Similarly, C<(?:...)> can't have a space between the C<?> and C<:>,
143 but can between the C<(> and C<?>. Within any delimiters for such a
144 construct, allowed spaces are not affected by C</x>, and depend on the
145 construct. For example, C<\x{...}> can't have spaces because hexadecimal
146 numbers don't have spaces in them. But, Unicode properties can have spaces, so
147 in C<\p{...}> there can be spaces that follow the Unicode rules, for which see
148 L<perluniprops/Properties accessible through \p{} and \P{}>.
151 =head3 Character set modifiers
153 C</d>, C</u>, C</a>, and C</l>, available starting in 5.14, are called
154 the character set modifiers; they affect the character set semantics
155 used for the regular expression.
157 At any given time, exactly one of these modifiers is in effect. Once
158 compiled, the behavior doesn't change regardless of what rules are in
159 effect when the regular expression is executed. And if a regular
160 expression is interpolated into a larger one, the original's rules
161 continue to apply to it, and only it.
163 Note that the modifiers affect only pattern matching, and do not extend
164 to any replacement done. For example,
168 will uppercase "bar", but the C</l> does not affect how the C<\U>
169 operates. If C<use locale> is in effect, the C<\U> will use locale
170 rules; if C<use feature 'unicode_strings'> is in effect, it will
171 use Unicode rules, etc.
175 means to use the current locale's rules (see L<perllocale>) when pattern
176 matching. For example, C<\w> will match the "word" characters of that
177 locale, and C<"/i"> case-insensitive matching will match according to
178 the locale's case folding rules. The locale used will be the one in
179 effect at the time of execution of the pattern match. This may not be
180 the same as the compilation-time locale, and can differ from one match
181 to another if there is an intervening call of the
182 L<setlocale() function|perllocale/The setlocale function>.
184 Perl only supports single-byte locales. This means that code points
185 above 255 are treated as Unicode no matter what locale is in effect.
186 Under Unicode rules, there are a few case-insensitive matches that cross
187 the 255/256 boundary. These are disallowed under C</l>. For example,
188 0xFF does not caselessly match the character at 0x178, C<LATIN CAPITAL
189 LETTER Y WITH DIAERESIS>, because 0xFF may not be C<LATIN SMALL LETTER Y
190 WITH DIAERESIS> in the current locale, and Perl has no way of knowing if
191 that character even exists in the locale, much less what code point it
194 This modifier may be specified to be the default by C<use locale>, but
195 see L</Which character set modifier is in effect?>.
200 means to use Unicode rules when pattern matching. On ASCII platforms,
201 this means that the code points between 128 and 255 take on their
202 Latin-1 (ISO-8859-1) meanings (which are the same as Unicode's), whereas
203 in strict ASCII their meanings are undefined. Thus the platform
204 effectively becomes a Unicode platform, hence, for example, C<\w> will
205 match any of the more than 100_000 word characters in Unicode.
207 Unlike most locales, which are specific to a language and country pair,
208 Unicode classifies all the characters that are letters I<somewhere> as
209 C<\w>. For example, your locale might not think that C<LATIN SMALL
210 LETTER ETH> is a letter (unless you happen to speak Icelandic), but
211 Unicode does. Similarly, all the characters that are decimal digits
212 somewhere in the world will match C<\d>; this is hundreds, not 10,
213 possible matches. And some of those digits look like some of the 10
214 ASCII digits, but mean a different number, so a human could easily think
215 a number is a different quantity than it really is. For example,
216 C<BENGALI DIGIT FOUR> (U+09EA) looks very much like an
217 C<ASCII DIGIT EIGHT> (U+0038). And, C<\d+>, may match strings of digits
218 that are a mixture from different writing systems, creating a security
219 issue. L<Unicode::UCD/num()> can be used to sort this out.
221 Also, case-insensitive matching works on the full set of Unicode
222 characters. The C<KELVIN SIGN>, for example matches the letters "k" and
223 "K"; and C<LATIN SMALL LIGATURE FF> matches the sequence "ff", which,
224 if you're not prepared, might make it look like a hexadecimal constant,
225 presenting another potential security issue. See
226 L<http://unicode.org/reports/tr36> for a detailed discussion of Unicode
229 On the EBCDIC platforms that Perl handles, the native character set is
230 equivalent to Latin-1. Thus this modifier changes behavior only when
231 the C<"/i"> modifier is also specified, and it turns out it affects only
232 two characters, giving them full Unicode semantics: the C<MICRO SIGN>
233 will match the Greek capital and small letters C<MU>, otherwise not; and
234 the C<LATIN CAPITAL LETTER SHARP S> will match any of C<SS>, C<Ss>,
235 C<sS>, and C<ss>, otherwise not.
237 This modifier may be specified to be the default by C<use feature
238 'unicode_strings>, but see
239 L</Which character set modifier is in effect?>.
244 is the same as C</u>, except that C<\d>, C<\s>, C<\w>, and the
245 Posix character classes are restricted to matching in the ASCII range
246 only. That is, with this modifier, C<\d> always means precisely the
247 digits C<"0"> to C<"9">; C<\s> means the five characters C<[ \f\n\r\t]>;
248 C<\w> means the 63 characters C<[A-Za-z0-9_]>; and likewise, all the
249 Posix classes such as C<[[:print:]]> match only the appropriate
250 ASCII-range characters.
252 This modifier is useful for people who only incidentally use Unicode.
253 With it, one can write C<\d> with confidence that it will only match
254 ASCII characters, and should the need arise to match beyond ASCII, you
255 can use C<\p{Digit}>, or C<\p{Word}> for C<\w>. There are similar
256 C<\p{...}> constructs that can match white space and Posix classes
257 beyond ASCII. See L<perlrecharclass/POSIX Character Classes>.
259 As you would expect, this modifier causes, for example, C<\D> to mean
260 the same thing as C<[^0-9]>; in fact, all non-ASCII characters match
261 C<\D>, C<\S>, and C<\W>. C<\b> still means to match at the boundary
262 between C<\w> and C<\W>, using the C</a> definitions of them (similarly
265 Otherwise, C</a> behaves like the C</u> modifier, in that
266 case-insensitive matching uses Unicode semantics; for example, "k" will
267 match the Unicode C<\N{KELVIN SIGN}> under C</i> matching, and code
268 points in the Latin1 range, above ASCII will have Unicode rules when it
269 comes to case-insensitive matching.
271 To forbid ASCII/non-ASCII matches (like "k" with C<\N{KELVIN SIGN}>),
272 specify the "a" twice, for example C</aai> or C</aia>
274 To reiterate, this modifier provides protection for applications that
275 don't wish to be exposed to all of Unicode. Specifying it twice
276 gives added protection.
278 This modifier may be specified to be the default by C<use re '/a'>
279 or C<use re '/aa'>, but see
280 L</Which character set modifier is in effect?>.
286 This modifier means to use the "Default" native rules of the platform
287 except when there is cause to use Unicode rules instead, as follows:
293 the target string is encoded in UTF-8; or
297 the pattern is encoded in UTF-8; or
301 the pattern explicitly mentions a code point that is above 255 (say by
306 the pattern uses a Unicode name (C<\N{...}>); or
310 the pattern uses a Unicode property (C<\p{...}>)
314 Another mnemonic for this modifier is "Depends", as the rules actually
315 used depend on various things, and as a result you can get unexpected
316 results. See L<perlunicode/The "Unicode Bug">.
318 On ASCII platforms, the native rules are ASCII, and on EBCDIC platforms
319 (at least the ones that Perl handles), they are Latin-1.
321 Here are some examples of how that works on an ASCII platform:
323 $str = "\xDF"; # $str is not in UTF-8 format.
324 $str =~ /^\w/; # No match, as $str isn't in UTF-8 format.
325 $str .= "\x{0e0b}"; # Now $str is in UTF-8 format.
326 $str =~ /^\w/; # Match! $str is now in UTF-8 format.
328 $str =~ /^\w/; # Still a match! $str remains in UTF-8 format.
330 =head4 Which character set modifier is in effect?
332 Which of these modifiers is in effect at any given point in a regular
333 expression depends on a fairly complex set of interactions. As
334 explained below in L</Extended Patterns> it is possible to explicitly
335 specify modifiers that apply only to portions of a regular expression.
336 The innermost always has priority over any outer ones, and one applying
337 to the whole expression has priority over any of the default settings that are
338 described in the remainder of this section.
340 The C<L<use re 'E<sol>foo'|re/'E<sol>flags' mode">> pragma can be used to set
341 default modifiers (including these) for regular expressions compiled
342 within its scope. This pragma has precedence over the other pragmas
343 listed below that change the defaults.
345 Otherwise, C<L<use locale|perllocale>> sets the default modifier to C</l>;
346 and C<L<use feature 'unicode_strings|feature>> or
347 C<L<use 5.012|perlfunc/use VERSION>> (or higher) set the default to
348 C</u> when not in the same scope as either C<L<use locale|perllocale>>
349 or C<L<use bytes|bytes>>. Unlike the mechanisms mentioned above, these
350 affect operations besides regular expressions pattern matching, and so
351 give more consistent results with other operators, including using
352 C<\U>, C<\l>, etc. in substitution replacements.
354 If none of the above apply, for backwards compatibility reasons, the
355 C</d> modifier is the one in effect by default. As this can lead to
356 unexpected results, it is best to specify which other rule set should be
359 =head4 Character set modifier behavior prior to Perl 5.14
361 Prior to 5.14, there were no explicit modifiers, but C</l> was implied
362 for regexes compiled within the scope of C<use locale>, and C</d> was
363 implied otherwise. However, interpolating a regex into a larger regex
364 would ignore the original compilation in favor of whatever was in effect
365 at the time of the second compilation. There were a number of
366 inconsistencies (bugs) with the C</d> modifier, where Unicode rules
367 would be used when inappropriate, and vice versa. C<\p{}> did not imply
368 Unicode rules, and neither did all occurrences of C<\N{}>, until 5.12.
370 =head2 Regular Expressions
372 =head3 Metacharacters
374 The patterns used in Perl pattern matching evolved from those supplied in
375 the Version 8 regex routines. (The routines are derived
376 (distantly) from Henry Spencer's freely redistributable reimplementation
377 of the V8 routines.) See L<Version 8 Regular Expressions> for
380 In particular the following metacharacters have their standard I<egrep>-ish
383 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
386 \ Quote the next metacharacter
387 ^ Match the beginning of the line
388 . Match any character (except newline)
389 $ Match the end of the line (or before newline at the end)
392 [] Bracketed Character class
394 By default, the "^" character is guaranteed to match only the
395 beginning of the string, the "$" character only the end (or before the
396 newline at the end), and Perl does certain optimizations with the
397 assumption that the string contains only one line. Embedded newlines
398 will not be matched by "^" or "$". You may, however, wish to treat a
399 string as a multi-line buffer, such that the "^" will match after any
400 newline within the string (except if the newline is the last character in
401 the string), and "$" will match before any newline. At the
402 cost of a little more overhead, you can do this by using the /m modifier
403 on the pattern match operator. (Older programs did this by setting C<$*>,
404 but this option was removed in perl 5.9.)
407 To simplify multi-line substitutions, the "." character never matches a
408 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
409 the string is a single line--even if it isn't.
414 The following standard quantifiers are recognized:
415 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
417 * Match 0 or more times
418 + Match 1 or more times
420 {n} Match exactly n times
421 {n,} Match at least n times
422 {n,m} Match at least n but not more than m times
424 (If a curly bracket occurs in any other context and does not form part of
425 a backslashed sequence like C<\x{...}>, it is treated
426 as a regular character. In particular, the lower bound
427 is not optional.) The "*" quantifier is equivalent to C<{0,}>, the "+"
428 quantifier to C<{1,}>, and the "?" quantifier to C<{0,1}>. n and m are limited
429 to non-negative integral values less than a preset limit defined when perl is built.
430 This is usually 32766 on the most common platforms. The actual limit can
431 be seen in the error message generated by code such as this:
433 $_ **= $_ , / {$_} / for 2 .. 42;
435 By default, a quantified subpattern is "greedy", that is, it will match as
436 many times as possible (given a particular starting location) while still
437 allowing the rest of the pattern to match. If you want it to match the
438 minimum number of times possible, follow the quantifier with a "?". Note
439 that the meanings don't change, just the "greediness":
440 X<metacharacter> X<greedy> X<greediness>
441 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
443 *? Match 0 or more times, not greedily
444 +? Match 1 or more times, not greedily
445 ?? Match 0 or 1 time, not greedily
446 {n}? Match exactly n times, not greedily (redundant)
447 {n,}? Match at least n times, not greedily
448 {n,m}? Match at least n but not more than m times, not greedily
450 By default, when a quantified subpattern does not allow the rest of the
451 overall pattern to match, Perl will backtrack. However, this behaviour is
452 sometimes undesirable. Thus Perl provides the "possessive" quantifier form
455 *+ Match 0 or more times and give nothing back
456 ++ Match 1 or more times and give nothing back
457 ?+ Match 0 or 1 time and give nothing back
458 {n}+ Match exactly n times and give nothing back (redundant)
459 {n,}+ Match at least n times and give nothing back
460 {n,m}+ Match at least n but not more than m times and give nothing back
466 will never match, as the C<a++> will gobble up all the C<a>'s in the
467 string and won't leave any for the remaining part of the pattern. This
468 feature can be extremely useful to give perl hints about where it
469 shouldn't backtrack. For instance, the typical "match a double-quoted
470 string" problem can be most efficiently performed when written as:
472 /"(?:[^"\\]++|\\.)*+"/
474 as we know that if the final quote does not match, backtracking will not
475 help. See the independent subexpression
476 L</C<< (?>pattern) >>> for more details;
477 possessive quantifiers are just syntactic sugar for that construct. For
478 instance the above example could also be written as follows:
480 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
482 =head3 Escape sequences
484 Because patterns are processed as double-quoted strings, the following
491 \a alarm (bell) (BEL)
492 \e escape (think troff) (ESC)
493 \cK control char (example: VT)
494 \x{}, \x00 character whose ordinal is the given hexadecimal number
495 \N{name} named Unicode character or character sequence
496 \N{U+263D} Unicode character (example: FIRST QUARTER MOON)
497 \o{}, \000 character whose ordinal is the given octal number
498 \l lowercase next char (think vi)
499 \u uppercase next char (think vi)
500 \L lowercase till \E (think vi)
501 \U uppercase till \E (think vi)
502 \Q quote (disable) pattern metacharacters till \E
503 \E end either case modification or quoted section, think vi
505 Details are in L<perlop/Quote and Quote-like Operators>.
507 =head3 Character Classes and other Special Escapes
509 In addition, Perl defines the following:
510 X<\g> X<\k> X<\K> X<backreference>
512 Sequence Note Description
513 [...] [1] Match a character according to the rules of the
514 bracketed character class defined by the "...".
515 Example: [a-z] matches "a" or "b" or "c" ... or "z"
516 [[:...:]] [2] Match a character according to the rules of the POSIX
517 character class "..." within the outer bracketed
518 character class. Example: [[:upper:]] matches any
520 \w [3] Match a "word" character (alphanumeric plus "_", plus
521 other connector punctuation chars plus Unicode
523 \W [3] Match a non-"word" character
524 \s [3] Match a whitespace character
525 \S [3] Match a non-whitespace character
526 \d [3] Match a decimal digit character
527 \D [3] Match a non-digit character
528 \pP [3] Match P, named property. Use \p{Prop} for longer names
530 \X [4] Match Unicode "eXtended grapheme cluster"
531 \C Match a single C-language char (octet) even if that is
532 part of a larger UTF-8 character. Thus it breaks up
533 characters into their UTF-8 bytes, so you may end up
534 with malformed pieces of UTF-8. Unsupported in
536 \1 [5] Backreference to a specific capture group or buffer.
537 '1' may actually be any positive integer.
538 \g1 [5] Backreference to a specific or previous group,
539 \g{-1} [5] The number may be negative indicating a relative
540 previous group and may optionally be wrapped in
541 curly brackets for safer parsing.
542 \g{name} [5] Named backreference
543 \k<name> [5] Named backreference
544 \K [6] Keep the stuff left of the \K, don't include it in $&
545 \N [7] Any character but \n (experimental). Not affected by
547 \v [3] Vertical whitespace
548 \V [3] Not vertical whitespace
549 \h [3] Horizontal whitespace
550 \H [3] Not horizontal whitespace
557 See L<perlrecharclass/Bracketed Character Classes> for details.
561 See L<perlrecharclass/POSIX Character Classes> for details.
565 See L<perlrecharclass/Backslash sequences> for details.
569 See L<perlrebackslash/Misc> for details.
573 See L</Capture groups> below for details.
577 See L</Extended Patterns> below for details.
581 Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the
582 character or character sequence whose name is C<NAME>; and similarly
583 when of the form C<\N{U+I<hex>}>, it matches the character whose Unicode
584 code point is I<hex>. Otherwise it matches any character but C<\n>.
590 Perl defines the following zero-width assertions:
591 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
592 X<regexp, zero-width assertion>
593 X<regular expression, zero-width assertion>
594 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
596 \b Match a word boundary
597 \B Match except at a word boundary
598 \A Match only at beginning of string
599 \Z Match only at end of string, or before newline at the end
600 \z Match only at end of string
601 \G Match only at pos() (e.g. at the end-of-match position
604 A word boundary (C<\b>) is a spot between two characters
605 that has a C<\w> on one side of it and a C<\W> on the other side
606 of it (in either order), counting the imaginary characters off the
607 beginning and end of the string as matching a C<\W>. (Within
608 character classes C<\b> represents backspace rather than a word
609 boundary, just as it normally does in any double-quoted string.)
610 The C<\A> and C<\Z> are just like "^" and "$", except that they
611 won't match multiple times when the C</m> modifier is used, while
612 "^" and "$" will match at every internal line boundary. To match
613 the actual end of the string and not ignore an optional trailing
615 X<\b> X<\A> X<\Z> X<\z> X</m>
617 The C<\G> assertion can be used to chain global matches (using
618 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
619 It is also useful when writing C<lex>-like scanners, when you have
620 several patterns that you want to match against consequent substrings
621 of your string; see the previous reference. The actual location
622 where C<\G> will match can also be influenced by using C<pos()> as
623 an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
624 matches (see L</"Repeated Patterns Matching a Zero-length Substring">)
625 is modified somewhat, in that contents to the left of C<\G> are
626 not counted when determining the length of the match. Thus the following
627 will not match forever:
632 while ($string =~ /(.\G)/g) {
636 It will print 'A' and then terminate, as it considers the match to
637 be zero-width, and thus will not match at the same position twice in a
640 It is worth noting that C<\G> improperly used can result in an infinite
641 loop. Take care when using patterns that include C<\G> in an alternation.
643 =head3 Capture groups
645 The bracketing construct C<( ... )> creates capture groups (also referred to as
646 capture buffers). To refer to the current contents of a group later on, within
647 the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>)
648 for the second, and so on.
649 This is called a I<backreference>.
650 X<regex, capture buffer> X<regexp, capture buffer>
651 X<regex, capture group> X<regexp, capture group>
652 X<regular expression, capture buffer> X<backreference>
653 X<regular expression, capture group> X<backreference>
654 X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
655 X<named capture buffer> X<regular expression, named capture buffer>
656 X<named capture group> X<regular expression, named capture group>
657 X<%+> X<$+{name}> X<< \k<name> >>
658 There is no limit to the number of captured substrings that you may use.
659 Groups are numbered with the leftmost open parenthesis being number 1, etc. If
660 a group did not match, the associated backreference won't match either. (This
661 can happen if the group is optional, or in a different branch of an
663 You can omit the C<"g">, and write C<"\1">, etc, but there are some issues with
664 this form, described below.
666 You can also refer to capture groups relatively, by using a negative number, so
667 that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture
668 group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For
675 \g{-1} # backref to group 3
676 \g{-3} # backref to group 1
680 would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to
681 interpolate regexes into larger regexes and not have to worry about the
682 capture groups being renumbered.
684 You can dispense with numbers altogether and create named capture groups.
685 The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to
686 reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may
687 also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.)
688 I<name> must not begin with a number, nor contain hyphens.
689 When different groups within the same pattern have the same name, any reference
690 to that name assumes the leftmost defined group. Named groups count in
691 absolute and relative numbering, and so can also be referred to by those
693 (It's possible to do things with named capture groups that would otherwise
696 Capture group contents are dynamically scoped and available to you outside the
697 pattern until the end of the enclosing block or until the next successful
698 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
699 You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">,
700 etc); or by name via the C<%+> hash, using C<"$+{I<name>}">.
702 Braces are required in referring to named capture groups, but are optional for
703 absolute or relative numbered ones. Braces are safer when creating a regex by
704 concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a>
705 contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which
706 is probably not what you intended.
708 The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that
709 there were no named nor relative numbered capture groups. Absolute numbered
710 groups were referred to using C<\1>,
711 C<\2>, etc., and this notation is still
712 accepted (and likely always will be). But it leads to some ambiguities if
713 there are more than 9 capture groups, as C<\10> could mean either the tenth
714 capture group, or the character whose ordinal in octal is 010 (a backspace in
715 ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference
716 only if at least 10 left parentheses have opened before it. Likewise C<\11> is
717 a backreference only if at least 11 left parentheses have opened before it.
718 And so on. C<\1> through C<\9> are always interpreted as backreferences.
719 There are several examples below that illustrate these perils. You can avoid
720 the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups;
721 and for octal constants always using C<\o{}>, or for C<\077> and below, using 3
722 digits padded with leading zeros, since a leading zero implies an octal
725 The C<\I<digit>> notation also works in certain circumstances outside
726 the pattern. See L</Warning on \1 Instead of $1> below for details.
730 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
732 /(.)\g1/ # find first doubled char
733 and print "'$1' is the first doubled character\n";
735 /(?<char>.)\k<char>/ # ... a different way
736 and print "'$+{char}' is the first doubled character\n";
738 /(?'char'.)\g1/ # ... mix and match
739 and print "'$1' is the first doubled character\n";
741 if (/Time: (..):(..):(..)/) { # parse out values
747 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference
748 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal
749 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference
750 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal
752 $a = '(.)\1'; # Creates problems when concatenated.
753 $b = '(.)\g{1}'; # Avoids the problems.
754 "aa" =~ /${a}/; # True
755 "aa" =~ /${b}/; # True
756 "aa0" =~ /${a}0/; # False!
757 "aa0" =~ /${b}0/; # True
758 "aa\x08" =~ /${a}0/; # True!
759 "aa\x08" =~ /${b}0/; # False
761 Several special variables also refer back to portions of the previous
762 match. C<$+> returns whatever the last bracket match matched.
763 C<$&> returns the entire matched string. (At one point C<$0> did
764 also, but now it returns the name of the program.) C<$`> returns
765 everything before the matched string. C<$'> returns everything
766 after the matched string. And C<$^N> contains whatever was matched by
767 the most-recently closed group (submatch). C<$^N> can be used in
768 extended patterns (see below), for example to assign a submatch to a
770 X<$+> X<$^N> X<$&> X<$`> X<$'>
772 These special variables, like the C<%+> hash and the numbered match variables
773 (C<$1>, C<$2>, C<$3>, etc.) are dynamically scoped
774 until the end of the enclosing block or until the next successful
775 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
776 X<$+> X<$^N> X<$&> X<$`> X<$'>
777 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
779 B<NOTE>: Failed matches in Perl do not reset the match variables,
780 which makes it easier to write code that tests for a series of more
781 specific cases and remembers the best match.
783 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
784 C<$'> anywhere in the program, it has to provide them for every
785 pattern match. This may substantially slow your program. Perl
786 uses the same mechanism to produce C<$1>, C<$2>, etc, so you also pay a
787 price for each pattern that contains capturing parentheses. (To
788 avoid this cost while retaining the grouping behaviour, use the
789 extended regular expression C<(?: ... )> instead.) But if you never
790 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
791 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
792 if you can, but if you can't (and some algorithms really appreciate
793 them), once you've used them once, use them at will, because you've
794 already paid the price. As of 5.005, C<$&> is not so costly as the
798 As a workaround for this problem, Perl 5.10.0 introduces C<${^PREMATCH}>,
799 C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&>
800 and C<$'>, B<except> that they are only guaranteed to be defined after a
801 successful match that was executed with the C</p> (preserve) modifier.
802 The use of these variables incurs no global performance penalty, unlike
803 their punctuation char equivalents, however at the trade-off that you
804 have to tell perl when you want to use them.
807 =head2 Quoting metacharacters
809 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
810 C<\w>, C<\n>. Unlike some other regular expression languages, there
811 are no backslashed symbols that aren't alphanumeric. So anything
812 that looks like \\, \(, \), \<, \>, \{, or \} is always
813 interpreted as a literal character, not a metacharacter. This was
814 once used in a common idiom to disable or quote the special meanings
815 of regular expression metacharacters in a string that you want to
816 use for a pattern. Simply quote all non-"word" characters:
818 $pattern =~ s/(\W)/\\$1/g;
820 (If C<use locale> is set, then this depends on the current locale.)
821 Today it is more common to use the quotemeta() function or the C<\Q>
822 metaquoting escape sequence to disable all metacharacters' special
825 /$unquoted\Q$quoted\E$unquoted/
827 Beware that if you put literal backslashes (those not inside
828 interpolated variables) between C<\Q> and C<\E>, double-quotish
829 backslash interpolation may lead to confusing results. If you
830 I<need> to use literal backslashes within C<\Q...\E>,
831 consult L<perlop/"Gory details of parsing quoted constructs">.
833 =head2 Extended Patterns
835 Perl also defines a consistent extension syntax for features not
836 found in standard tools like B<awk> and
837 B<lex>. The syntax for most of these is a
838 pair of parentheses with a question mark as the first thing within
839 the parentheses. The character after the question mark indicates
842 The stability of these extensions varies widely. Some have been
843 part of the core language for many years. Others are experimental
844 and may change without warning or be completely removed. Check
845 the documentation on an individual feature to verify its current
848 A question mark was chosen for this and for the minimal-matching
849 construct because 1) question marks are rare in older regular
850 expressions, and 2) whenever you see one, you should stop and
851 "question" exactly what is going on. That's psychology....
858 A comment. The text is ignored. If the C</x> modifier enables
859 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
860 the comment as soon as it sees a C<)>, so there is no way to put a literal
863 =item C<(?adlupimsx-imsx)>
865 =item C<(?^alupimsx)>
868 One or more embedded pattern-match modifiers, to be turned on (or
869 turned off, if preceded by C<->) for the remainder of the pattern or
870 the remainder of the enclosing pattern group (if any).
872 This is particularly useful for dynamic patterns, such as those read in from a
873 configuration file, taken from an argument, or specified in a table
874 somewhere. Consider the case where some patterns want to be
875 case-sensitive and some do not: The case-insensitive ones merely need to
876 include C<(?i)> at the front of the pattern. For example:
879 if ( /$pattern/i ) { }
883 $pattern = "(?i)foobar";
884 if ( /$pattern/ ) { }
886 These modifiers are restored at the end of the enclosing group. For example,
888 ( (?i) blah ) \s+ \g1
890 will match C<blah> in any case, some spaces, and an exact (I<including the case>!)
891 repetition of the previous word, assuming the C</x> modifier, and no C</i>
892 modifier outside this group.
894 These modifiers do not carry over into named subpatterns called in the
895 enclosing group. In other words, a pattern such as C<((?i)(&NAME))> does not
896 change the case-sensitivity of the "NAME" pattern.
898 Any of these modifiers can be set to apply globally to all regular
899 expressions compiled within the scope of a C<use re>. See
900 L<re/"'/flags' mode">.
902 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
903 after the C<"?"> is a shorthand equivalent to C<d-imsx>. Flags (except
904 C<"d">) may follow the caret to override it.
905 But a minus sign is not legal with it.
907 Note that the C<a>, C<d>, C<l>, C<p>, and C<u> modifiers are special in
908 that they can only be enabled, not disabled, and the C<a>, C<d>, C<l>, and
909 C<u> modifiers are mutually exclusive: specifying one de-specifies the
910 others, and a maximum of one (or two C<a>'s) may appear in the
912 example, C<(?-p)> will warn when compiled under C<use warnings>;
913 C<(?-d:...)> and C<(?dl:...)> are fatal errors.
915 Note also that the C<p> modifier is special in that its presence
916 anywhere in a pattern has a global effect.
921 =item C<(?adluimsx-imsx:pattern)>
923 =item C<(?^aluimsx:pattern)>
926 This is for clustering, not capturing; it groups subexpressions like
927 "()", but doesn't make backreferences as "()" does. So
929 @fields = split(/\b(?:a|b|c)\b/)
933 @fields = split(/\b(a|b|c)\b/)
935 but doesn't spit out extra fields. It's also cheaper not to capture
936 characters if you don't need to.
938 Any letters between C<?> and C<:> act as flags modifiers as with
939 C<(?adluimsx-imsx)>. For example,
941 /(?s-i:more.*than).*million/i
943 is equivalent to the more verbose
945 /(?:(?s-i)more.*than).*million/i
947 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
948 after the C<"?"> is a shorthand equivalent to C<d-imsx>. Any positive
949 flags (except C<"d">) may follow the caret, so
957 The caret tells Perl that this cluster doesn't inherit the flags of any
958 surrounding pattern, but uses the system defaults (C<d-imsx>),
959 modified by any flags specified.
961 The caret allows for simpler stringification of compiled regular
962 expressions. These look like
966 with any non-default flags appearing between the caret and the colon.
967 A test that looks at such stringification thus doesn't need to have the
968 system default flags hard-coded in it, just the caret. If new flags are
969 added to Perl, the meaning of the caret's expansion will change to include
970 the default for those flags, so the test will still work, unchanged.
972 Specifying a negative flag after the caret is an error, as the flag is
975 Mnemonic for C<(?^...)>: A fresh beginning since the usual use of a caret is
976 to match at the beginning.
979 X<(?|)> X<Branch reset>
981 This is the "branch reset" pattern, which has the special property
982 that the capture groups are numbered from the same starting point
983 in each alternation branch. It is available starting from perl 5.10.0.
985 Capture groups are numbered from left to right, but inside this
986 construct the numbering is restarted for each branch.
988 The numbering within each branch will be as normal, and any groups
989 following this construct will be numbered as though the construct
990 contained only one branch, that being the one with the most capture
993 This construct is useful when you want to capture one of a
994 number of alternative matches.
996 Consider the following pattern. The numbers underneath show in
997 which group the captured content will be stored.
1000 # before ---------------branch-reset----------- after
1001 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1004 Be careful when using the branch reset pattern in combination with
1005 named captures. Named captures are implemented as being aliases to
1006 numbered groups holding the captures, and that interferes with the
1007 implementation of the branch reset pattern. If you are using named
1008 captures in a branch reset pattern, it's best to use the same names,
1009 in the same order, in each of the alternations:
1011 /(?| (?<a> x ) (?<b> y )
1012 | (?<a> z ) (?<b> w )) /x
1014 Not doing so may lead to surprises:
1016 "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x;
1017 say $+ {a}; # Prints '12'
1018 say $+ {b}; # *Also* prints '12'.
1020 The problem here is that both the group named C<< a >> and the group
1021 named C<< b >> are aliases for the group belonging to C<< $1 >>.
1023 =item Look-Around Assertions
1024 X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
1026 Look-around assertions are zero-width patterns which match a specific
1027 pattern without including it in C<$&>. Positive assertions match when
1028 their subpattern matches, negative assertions match when their subpattern
1029 fails. Look-behind matches text up to the current match position,
1030 look-ahead matches text following the current match position.
1034 =item C<(?=pattern)>
1035 X<(?=)> X<look-ahead, positive> X<lookahead, positive>
1037 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
1038 matches a word followed by a tab, without including the tab in C<$&>.
1040 =item C<(?!pattern)>
1041 X<(?!)> X<look-ahead, negative> X<lookahead, negative>
1043 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
1044 matches any occurrence of "foo" that isn't followed by "bar". Note
1045 however that look-ahead and look-behind are NOT the same thing. You cannot
1046 use this for look-behind.
1048 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
1049 will not do what you want. That's because the C<(?!foo)> is just saying that
1050 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
1051 match. Use look-behind instead (see below).
1053 =item C<(?<=pattern)> C<\K>
1054 X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K>
1056 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
1057 matches a word that follows a tab, without including the tab in C<$&>.
1058 Works only for fixed-width look-behind.
1060 There is a special form of this construct, called C<\K>, which causes the
1061 regex engine to "keep" everything it had matched prior to the C<\K> and
1062 not include it in C<$&>. This effectively provides variable-length
1063 look-behind. The use of C<\K> inside of another look-around assertion
1064 is allowed, but the behaviour is currently not well defined.
1066 For various reasons C<\K> may be significantly more efficient than the
1067 equivalent C<< (?<=...) >> construct, and it is especially useful in
1068 situations where you want to efficiently remove something following
1069 something else in a string. For instance
1073 can be rewritten as the much more efficient
1077 =item C<(?<!pattern)>
1078 X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
1080 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
1081 matches any occurrence of "foo" that does not follow "bar". Works
1082 only for fixed-width look-behind.
1086 =item C<(?'NAME'pattern)>
1088 =item C<< (?<NAME>pattern) >>
1089 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
1091 A named capture group. Identical in every respect to normal capturing
1092 parentheses C<()> but for the additional fact that the group
1093 can be referred to by name in various regular expression
1094 constructs (like C<\g{NAME}>) and can be accessed by name
1095 after a successful match via C<%+> or C<%->. See L<perlvar>
1096 for more details on the C<%+> and C<%-> hashes.
1098 If multiple distinct capture groups have the same name then the
1099 $+{NAME} will refer to the leftmost defined group in the match.
1101 The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent.
1103 B<NOTE:> While the notation of this construct is the same as the similar
1104 function in .NET regexes, the behavior is not. In Perl the groups are
1105 numbered sequentially regardless of being named or not. Thus in the
1110 $+{foo} will be the same as $2, and $3 will contain 'z' instead of
1111 the opposite which is what a .NET regex hacker might expect.
1113 Currently NAME is restricted to simple identifiers only.
1114 In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
1115 its Unicode extension (see L<utf8>),
1116 though it isn't extended by the locale (see L<perllocale>).
1118 B<NOTE:> In order to make things easier for programmers with experience
1119 with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >>
1120 may be used instead of C<< (?<NAME>pattern) >>; however this form does not
1121 support the use of single quotes as a delimiter for the name.
1123 =item C<< \k<NAME> >>
1125 =item C<< \k'NAME' >>
1127 Named backreference. Similar to numeric backreferences, except that
1128 the group is designated by name and not number. If multiple groups
1129 have the same name then it refers to the leftmost defined group in
1132 It is an error to refer to a name not defined by a C<< (?<NAME>) >>
1133 earlier in the pattern.
1135 Both forms are equivalent.
1137 B<NOTE:> In order to make things easier for programmers with experience
1138 with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >>
1139 may be used instead of C<< \k<NAME> >>.
1141 =item C<(?{ code })>
1142 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
1144 B<WARNING>: This extended regular expression feature is considered
1145 experimental, and may be changed without notice. Code executed that
1146 has side effects may not perform identically from version to version
1147 due to the effect of future optimisations in the regex engine.
1149 This zero-width assertion evaluates any embedded Perl code. It
1150 always succeeds, and its C<code> is not interpolated. Currently,
1151 the rules to determine where the C<code> ends are somewhat convoluted.
1153 This feature can be used together with the special variable C<$^N> to
1154 capture the results of submatches in variables without having to keep
1155 track of the number of nested parentheses. For example:
1157 $_ = "The brown fox jumps over the lazy dog";
1158 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
1159 print "color = $color, animal = $animal\n";
1161 Inside the C<(?{...})> block, C<$_> refers to the string the regular
1162 expression is matching against. You can also use C<pos()> to know what is
1163 the current position of matching within this string.
1165 The C<code> is properly scoped in the following sense: If the assertion
1166 is backtracked (compare L<"Backtracking">), all changes introduced after
1167 C<local>ization are undone, so that
1171 (?{ $cnt = 0 }) # Initialize $cnt.
1175 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
1179 (?{ $res = $cnt }) # On success copy to
1180 # non-localized location.
1183 will set C<$res = 4>. Note that after the match, C<$cnt> returns to the globally
1184 introduced value, because the scopes that restrict C<local> operators
1187 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
1188 switch. If I<not> used in this way, the result of evaluation of
1189 C<code> is put into the special variable C<$^R>. This happens
1190 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
1191 inside the same regular expression.
1193 The assignment to C<$^R> above is properly localized, so the old
1194 value of C<$^R> is restored if the assertion is backtracked; compare
1197 For reasons of security, this construct is forbidden if the regular
1198 expression involves run-time interpolation of variables, unless the
1199 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
1200 variables contain results of the C<qr//> operator (see
1201 L<perlop/"qr/STRINGE<sol>msixpodual">).
1203 This restriction is due to the wide-spread and remarkably convenient
1204 custom of using run-time determined strings as patterns. For example:
1210 Before Perl knew how to execute interpolated code within a pattern,
1211 this operation was completely safe from a security point of view,
1212 although it could raise an exception from an illegal pattern. If
1213 you turn on the C<use re 'eval'>, though, it is no longer secure,
1214 so you should only do so if you are also using taint checking.
1215 Better yet, use the carefully constrained evaluation within a Safe
1216 compartment. See L<perlsec> for details about both these mechanisms.
1218 B<WARNING>: Use of lexical (C<my>) variables in these blocks is
1219 broken. The result is unpredictable and will make perl unstable. The
1220 workaround is to use global (C<our>) variables.
1222 B<WARNING>: In perl 5.12.x and earlier, the regex engine
1223 was not re-entrant, so interpolated code could not
1224 safely invoke the regex engine either directly with
1225 C<m//> or C<s///>), or indirectly with functions such as
1226 C<split>. Invoking the regex engine in these blocks would make perl
1229 =item C<(??{ code })>
1231 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
1233 B<WARNING>: This extended regular expression feature is considered
1234 experimental, and may be changed without notice. Code executed that
1235 has side effects may not perform identically from version to version
1236 due to the effect of future optimisations in the regex engine.
1238 This is a "postponed" regular subexpression. The C<code> is evaluated
1239 at run time, at the moment this subexpression may match. The result
1240 of evaluation is considered a regular expression and matched as
1241 if it were inserted instead of this construct. Note that this means
1242 that the contents of capture groups defined inside an eval'ed pattern
1243 are not available outside of the pattern, and vice versa, there is no
1244 way for the inner pattern to refer to a capture group defined outside.
1247 ('a' x 100)=~/(??{'(.)' x 100})/
1249 B<will> match, it will B<not> set $1.
1251 The C<code> is not interpolated. As before, the rules to determine
1252 where the C<code> ends are currently somewhat convoluted.
1254 The following pattern matches a parenthesized group:
1259 (?> [^()]+ ) # Non-parens without backtracking
1261 (??{ $re }) # Group with matching parens
1266 See also C<(?PARNO)> for a different, more efficient way to accomplish
1269 For reasons of security, this construct is forbidden if the regular
1270 expression involves run-time interpolation of variables, unless the
1271 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
1272 variables contain results of the C<qr//> operator (see
1273 L<perlop/"qrE<sol>STRINGE<sol>msixpodual">).
1275 In perl 5.12.x and earlier, because the regex engine was not re-entrant,
1276 delayed code could not safely invoke the regex engine either directly with
1277 C<m//> or C<s///>), or indirectly with functions such as C<split>.
1279 Recursing deeper than 50 times without consuming any input string will
1280 result in a fatal error. The maximum depth is compiled into perl, so
1281 changing it requires a custom build.
1283 =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)>
1284 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
1285 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
1286 X<regex, relative recursion>
1288 Similar to C<(??{ code })> except it does not involve compiling any code,
1289 instead it treats the contents of a capture group as an independent
1290 pattern that must match at the current position. Capture groups
1291 contained by the pattern will have the value as determined by the
1292 outermost recursion.
1294 PARNO is a sequence of digits (not starting with 0) whose value reflects
1295 the paren-number of the capture group to recurse to. C<(?R)> recurses to
1296 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
1297 C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed
1298 to be relative, with negative numbers indicating preceding capture groups
1299 and positive ones following. Thus C<(?-1)> refers to the most recently
1300 declared group, and C<(?+1)> indicates the next group to be declared.
1301 Note that the counting for relative recursion differs from that of
1302 relative backreferences, in that with recursion unclosed groups B<are>
1305 The following pattern matches a function foo() which may contain
1306 balanced parentheses as the argument.
1308 $re = qr{ ( # paren group 1 (full function)
1310 ( # paren group 2 (parens)
1312 ( # paren group 3 (contents of parens)
1314 (?> [^()]+ ) # Non-parens without backtracking
1316 (?2) # Recurse to start of paren group 2
1324 If the pattern was used as follows
1326 'foo(bar(baz)+baz(bop))'=~/$re/
1327 and print "\$1 = $1\n",
1331 the output produced should be the following:
1333 $1 = foo(bar(baz)+baz(bop))
1334 $2 = (bar(baz)+baz(bop))
1335 $3 = bar(baz)+baz(bop)
1337 If there is no corresponding capture group defined, then it is a
1338 fatal error. Recursing deeper than 50 times without consuming any input
1339 string will also result in a fatal error. The maximum depth is compiled
1340 into perl, so changing it requires a custom build.
1342 The following shows how using negative indexing can make it
1343 easier to embed recursive patterns inside of a C<qr//> construct
1346 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
1347 if (/foo $parens \s+ + \s+ bar $parens/x) {
1348 # do something here...
1351 B<Note> that this pattern does not behave the same way as the equivalent
1352 PCRE or Python construct of the same form. In Perl you can backtrack into
1353 a recursed group, in PCRE and Python the recursed into group is treated
1354 as atomic. Also, modifiers are resolved at compile time, so constructs
1355 like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
1361 Recurse to a named subpattern. Identical to C<(?PARNO)> except that the
1362 parenthesis to recurse to is determined by name. If multiple parentheses have
1363 the same name, then it recurses to the leftmost.
1365 It is an error to refer to a name that is not declared somewhere in the
1368 B<NOTE:> In order to make things easier for programmers with experience
1369 with the Python or PCRE regex engines the pattern C<< (?P>NAME) >>
1370 may be used instead of C<< (?&NAME) >>.
1372 =item C<(?(condition)yes-pattern|no-pattern)>
1375 =item C<(?(condition)yes-pattern)>
1377 Conditional expression. Matches C<yes-pattern> if C<condition> yields
1378 a true value, matches C<no-pattern> otherwise. A missing pattern always
1381 C<(condition)> should be either an integer in
1382 parentheses (which is valid if the corresponding pair of parentheses
1383 matched), a look-ahead/look-behind/evaluate zero-width assertion, a
1384 name in angle brackets or single quotes (which is valid if a group
1385 with the given name matched), or the special symbol (R) (true when
1386 evaluated inside of recursion or eval). Additionally the R may be
1387 followed by a number, (which will be true when evaluated when recursing
1388 inside of the appropriate group), or by C<&NAME>, in which case it will
1389 be true only when evaluated during recursion in the named group.
1391 Here's a summary of the possible predicates:
1397 Checks if the numbered capturing group has matched something.
1399 =item (<NAME>) ('NAME')
1401 Checks if a group with the given name has matched something.
1403 =item (?=...) (?!...) (?<=...) (?<!...)
1405 Checks whether the pattern matches (or does not match, for the '!'
1410 Treats the return value of the code block as the condition.
1414 Checks if the expression has been evaluated inside of recursion.
1418 Checks if the expression has been evaluated while executing directly
1419 inside of the n-th capture group. This check is the regex equivalent of
1421 if ((caller(0))[3] eq 'subname') { ... }
1423 In other words, it does not check the full recursion stack.
1427 Similar to C<(R1)>, this predicate checks to see if we're executing
1428 directly inside of the leftmost group with a given name (this is the same
1429 logic used by C<(?&NAME)> to disambiguate). It does not check the full
1430 stack, but only the name of the innermost active recursion.
1434 In this case, the yes-pattern is never directly executed, and no
1435 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1436 See below for details.
1447 matches a chunk of non-parentheses, possibly included in parentheses
1450 A special form is the C<(DEFINE)> predicate, which never executes its
1451 yes-pattern directly, and does not allow a no-pattern. This allows one to
1452 define subpatterns which will be executed only by the recursion mechanism.
1453 This way, you can define a set of regular expression rules that can be
1454 bundled into any pattern you choose.
1456 It is recommended that for this usage you put the DEFINE block at the
1457 end of the pattern, and that you name any subpatterns defined within it.
1459 Also, it's worth noting that patterns defined this way probably will
1460 not be as efficient, as the optimiser is not very clever about
1463 An example of how this might be used is as follows:
1465 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
1471 Note that capture groups matched inside of recursion are not accessible
1472 after the recursion returns, so the extra layer of capturing groups is
1473 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1474 C<$+{NAME}> would be.
1476 =item C<< (?>pattern) >>
1477 X<backtrack> X<backtracking> X<atomic> X<possessive>
1479 An "independent" subexpression, one which matches the substring
1480 that a I<standalone> C<pattern> would match if anchored at the given
1481 position, and it matches I<nothing other than this substring>. This
1482 construct is useful for optimizations of what would otherwise be
1483 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
1484 It may also be useful in places where the "grab all you can, and do not
1485 give anything back" semantic is desirable.
1487 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
1488 (anchored at the beginning of string, as above) will match I<all>
1489 characters C<a> at the beginning of string, leaving no C<a> for
1490 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1491 since the match of the subgroup C<a*> is influenced by the following
1492 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1493 C<a*ab> will match fewer characters than a standalone C<a*>, since
1494 this makes the tail match.
1496 C<< (?>pattern) >> does not disable backtracking altogether once it has
1497 matched. It is still possible to backtrack past the construct, but not
1498 into it. So C<< ((?>a*)|(?>b*))ar >> will still match "bar".
1500 An effect similar to C<< (?>pattern) >> may be achieved by writing
1501 C<(?=(pattern))\g{-1}>. This matches the same substring as a standalone
1502 C<a+>, and the following C<\g{-1}> eats the matched string; it therefore
1503 makes a zero-length assertion into an analogue of C<< (?>...) >>.
1504 (The difference between these two constructs is that the second one
1505 uses a capturing group, thus shifting ordinals of backreferences
1506 in the rest of a regular expression.)
1508 Consider this pattern:
1519 That will efficiently match a nonempty group with matching parentheses
1520 two levels deep or less. However, if there is no such group, it
1521 will take virtually forever on a long string. That's because there
1522 are so many different ways to split a long string into several
1523 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1524 to a subpattern of the above pattern. Consider how the pattern
1525 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1526 seconds, but that each extra letter doubles this time. This
1527 exponential performance will make it appear that your program has
1528 hung. However, a tiny change to this pattern
1532 (?> [^()]+ ) # change x+ above to (?> x+ )
1539 which uses C<< (?>...) >> matches exactly when the one above does (verifying
1540 this yourself would be a productive exercise), but finishes in a fourth
1541 the time when used on a similar string with 1000000 C<a>s. Be aware,
1542 however, that, when this construct is followed by a
1543 quantifier, it currently triggers a warning message under
1544 the C<use warnings> pragma or B<-w> switch saying it
1545 C<"matches null string many times in regex">.
1547 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
1548 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
1549 This was only 4 times slower on a string with 1000000 C<a>s.
1551 The "grab all you can, and do not give anything back" semantic is desirable
1552 in many situations where on the first sight a simple C<()*> looks like
1553 the correct solution. Suppose we parse text with comments being delimited
1554 by C<#> followed by some optional (horizontal) whitespace. Contrary to
1555 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
1556 the comment delimiter, because it may "give up" some whitespace if
1557 the remainder of the pattern can be made to match that way. The correct
1558 answer is either one of these:
1563 For example, to grab non-empty comments into $1, one should use either
1566 / (?> \# [ \t]* ) ( .+ ) /x;
1567 / \# [ \t]* ( [^ \t] .* ) /x;
1569 Which one you pick depends on which of these expressions better reflects
1570 the above specification of comments.
1572 In some literature this construct is called "atomic matching" or
1573 "possessive matching".
1575 Possessive quantifiers are equivalent to putting the item they are applied
1576 to inside of one of these constructs. The following equivalences apply:
1578 Quantifier Form Bracketing Form
1579 --------------- ---------------
1583 PAT{min,max}+ (?>PAT{min,max})
1587 =head2 Special Backtracking Control Verbs
1589 B<WARNING:> These patterns are experimental and subject to change or
1590 removal in a future version of Perl. Their usage in production code should
1591 be noted to avoid problems during upgrades.
1593 These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1594 otherwise stated the ARG argument is optional; in some cases, it is
1597 Any pattern containing a special backtracking verb that allows an argument
1598 has the special behaviour that when executed it sets the current package's
1599 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1602 On failure, the C<$REGERROR> variable will be set to the ARG value of the
1603 verb pattern, if the verb was involved in the failure of the match. If the
1604 ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1605 name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1606 none. Also, the C<$REGMARK> variable will be set to FALSE.
1608 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1609 the C<$REGMARK> variable will be set to the name of the last
1610 C<(*MARK:NAME)> pattern executed. See the explanation for the
1611 C<(*MARK:NAME)> verb below for more details.
1613 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1614 and most other regex-related variables. They are not local to a scope, nor
1615 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1616 Use C<local> to localize changes to them to a specific scope if necessary.
1618 If a pattern does not contain a special backtracking verb that allows an
1619 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
1623 =item Verbs that take an argument
1627 =item C<(*PRUNE)> C<(*PRUNE:NAME)>
1628 X<(*PRUNE)> X<(*PRUNE:NAME)>
1630 This zero-width pattern prunes the backtracking tree at the current point
1631 when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1632 where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1633 A may backtrack as necessary to match. Once it is reached, matching
1634 continues in B, which may also backtrack as necessary; however, should B
1635 not match, then no further backtracking will take place, and the pattern
1636 will fail outright at the current starting position.
1638 The following example counts all the possible matching strings in a
1639 pattern (without actually matching any of them).
1641 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
1642 print "Count=$count\n";
1657 If we add a C<(*PRUNE)> before the count like the following
1659 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
1660 print "Count=$count\n";
1662 we prevent backtracking and find the count of the longest matching string
1663 at each matching starting point like so:
1670 Any number of C<(*PRUNE)> assertions may be used in a pattern.
1672 See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1673 control backtracking. In some cases, the use of C<(*PRUNE)> can be
1674 replaced with a C<< (?>pattern) >> with no functional difference; however,
1675 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1676 C<< (?>pattern) >> alone.
1679 =item C<(*SKIP)> C<(*SKIP:NAME)>
1682 This zero-width pattern is similar to C<(*PRUNE)>, except that on
1683 failure it also signifies that whatever text that was matched leading up
1684 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1685 of this pattern. This effectively means that the regex engine "skips" forward
1686 to this position on failure and tries to match again, (assuming that
1687 there is sufficient room to match).
1689 The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1690 C<(*MARK:NAME)> was encountered while matching, then it is that position
1691 which is used as the "skip point". If no C<(*MARK)> of that name was
1692 encountered, then the C<(*SKIP)> operator has no effect. When used
1693 without a name the "skip point" is where the match point was when
1694 executing the (*SKIP) pattern.
1696 Compare the following to the examples in C<(*PRUNE)>; note the string
1699 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
1700 print "Count=$count\n";
1708 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
1709 executed, the next starting point will be where the cursor was when the
1710 C<(*SKIP)> was executed.
1712 =item C<(*MARK:NAME)> C<(*:NAME)>
1713 X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)>
1715 This zero-width pattern can be used to mark the point reached in a string
1716 when a certain part of the pattern has been successfully matched. This
1717 mark may be given a name. A later C<(*SKIP)> pattern will then skip
1718 forward to that point if backtracked into on failure. Any number of
1719 C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated.
1721 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1722 can be used to "label" a pattern branch, so that after matching, the
1723 program can determine which branches of the pattern were involved in the
1726 When a match is successful, the C<$REGMARK> variable will be set to the
1727 name of the most recently executed C<(*MARK:NAME)> that was involved
1730 This can be used to determine which branch of a pattern was matched
1731 without using a separate capture group for each branch, which in turn
1732 can result in a performance improvement, as perl cannot optimize
1733 C</(?:(x)|(y)|(z))/> as efficiently as something like
1734 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1736 When a match has failed, and unless another verb has been involved in
1737 failing the match and has provided its own name to use, the C<$REGERROR>
1738 variable will be set to the name of the most recently executed
1741 See C<(*SKIP)> for more details.
1743 As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
1745 =item C<(*THEN)> C<(*THEN:NAME)>
1747 This is similar to the "cut group" operator C<::> from Perl 6. Like
1748 C<(*PRUNE)>, this verb always matches, and when backtracked into on
1749 failure, it causes the regex engine to try the next alternation in the
1750 innermost enclosing group (capturing or otherwise).
1752 Its name comes from the observation that this operation combined with the
1753 alternation operator (C<|>) can be used to create what is essentially a
1754 pattern-based if/then/else block:
1756 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
1758 Note that if this operator is used and NOT inside of an alternation then
1759 it acts exactly like the C<(*PRUNE)> operator.
1769 / ( A (*THEN) B | C (*THEN) D ) /
1773 / ( A (*PRUNE) B | C (*PRUNE) D ) /
1775 as after matching the A but failing on the B the C<(*THEN)> verb will
1776 backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
1780 =item Verbs without an argument
1787 This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a
1788 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
1789 into on failure it causes the match to fail outright. No further attempts
1790 to find a valid match by advancing the start pointer will occur again.
1793 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
1794 print "Count=$count\n";
1801 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1802 does not match, the regex engine will not try any further matching on the
1805 =item C<(*FAIL)> C<(*F)>
1808 This pattern matches nothing and always fails. It can be used to force the
1809 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1810 fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
1812 It is probably useful only when combined with C<(?{})> or C<(??{})>.
1817 B<WARNING:> This feature is highly experimental. It is not recommended
1818 for production code.
1820 This pattern matches nothing and causes the end of successful matching at
1821 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1822 whether there is actually more to match in the string. When inside of a
1823 nested pattern, such as recursion, or in a subpattern dynamically generated
1824 via C<(??{})>, only the innermost pattern is ended immediately.
1826 If the C<(*ACCEPT)> is inside of capturing groups then the groups are
1827 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1830 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
1832 will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
1833 be set. If another branch in the inner parentheses was matched, such as in the
1834 string 'ACDE', then the C<D> and C<E> would have to be matched as well.
1841 X<backtrack> X<backtracking>
1843 NOTE: This section presents an abstract approximation of regular
1844 expression behavior. For a more rigorous (and complicated) view of
1845 the rules involved in selecting a match among possible alternatives,
1846 see L<Combining RE Pieces>.
1848 A fundamental feature of regular expression matching involves the
1849 notion called I<backtracking>, which is currently used (when needed)
1850 by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>,
1851 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1852 internally, but the general principle outlined here is valid.
1854 For a regular expression to match, the I<entire> regular expression must
1855 match, not just part of it. So if the beginning of a pattern containing a
1856 quantifier succeeds in a way that causes later parts in the pattern to
1857 fail, the matching engine backs up and recalculates the beginning
1858 part--that's why it's called backtracking.
1860 Here is an example of backtracking: Let's say you want to find the
1861 word following "foo" in the string "Food is on the foo table.":
1863 $_ = "Food is on the foo table.";
1864 if ( /\b(foo)\s+(\w+)/i ) {
1865 print "$2 follows $1.\n";
1868 When the match runs, the first part of the regular expression (C<\b(foo)>)
1869 finds a possible match right at the beginning of the string, and loads up
1870 $1 with "Foo". However, as soon as the matching engine sees that there's
1871 no whitespace following the "Foo" that it had saved in $1, it realizes its
1872 mistake and starts over again one character after where it had the
1873 tentative match. This time it goes all the way until the next occurrence
1874 of "foo". The complete regular expression matches this time, and you get
1875 the expected output of "table follows foo."
1877 Sometimes minimal matching can help a lot. Imagine you'd like to match
1878 everything between "foo" and "bar". Initially, you write something
1881 $_ = "The food is under the bar in the barn.";
1882 if ( /foo(.*)bar/ ) {
1886 Which perhaps unexpectedly yields:
1888 got <d is under the bar in the >
1890 That's because C<.*> was greedy, so you get everything between the
1891 I<first> "foo" and the I<last> "bar". Here it's more effective
1892 to use minimal matching to make sure you get the text between a "foo"
1893 and the first "bar" thereafter.
1895 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1896 got <d is under the >
1898 Here's another example. Let's say you'd like to match a number at the end
1899 of a string, and you also want to keep the preceding part of the match.
1902 $_ = "I have 2 numbers: 53147";
1903 if ( /(.*)(\d*)/ ) { # Wrong!
1904 print "Beginning is <$1>, number is <$2>.\n";
1907 That won't work at all, because C<.*> was greedy and gobbled up the
1908 whole string. As C<\d*> can match on an empty string the complete
1909 regular expression matched successfully.
1911 Beginning is <I have 2 numbers: 53147>, number is <>.
1913 Here are some variants, most of which don't work:
1915 $_ = "I have 2 numbers: 53147";
1928 printf "%-12s ", $pat;
1930 print "<$1> <$2>\n";
1936 That will print out:
1938 (.*)(\d*) <I have 2 numbers: 53147> <>
1939 (.*)(\d+) <I have 2 numbers: 5314> <7>
1941 (.*?)(\d+) <I have > <2>
1942 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1943 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1944 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1945 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1947 As you see, this can be a bit tricky. It's important to realize that a
1948 regular expression is merely a set of assertions that gives a definition
1949 of success. There may be 0, 1, or several different ways that the
1950 definition might succeed against a particular string. And if there are
1951 multiple ways it might succeed, you need to understand backtracking to
1952 know which variety of success you will achieve.
1954 When using look-ahead assertions and negations, this can all get even
1955 trickier. Imagine you'd like to find a sequence of non-digits not
1956 followed by "123". You might try to write that as
1959 if ( /^\D*(?!123)/ ) { # Wrong!
1960 print "Yup, no 123 in $_\n";
1963 But that isn't going to match; at least, not the way you're hoping. It
1964 claims that there is no 123 in the string. Here's a clearer picture of
1965 why that pattern matches, contrary to popular expectations:
1970 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1971 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
1973 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1974 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
1982 You might have expected test 3 to fail because it seems to a more
1983 general purpose version of test 1. The important difference between
1984 them is that test 3 contains a quantifier (C<\D*>) and so can use
1985 backtracking, whereas test 1 will not. What's happening is
1986 that you've asked "Is it true that at the start of $x, following 0 or more
1987 non-digits, you have something that's not 123?" If the pattern matcher had
1988 let C<\D*> expand to "ABC", this would have caused the whole pattern to
1991 The search engine will initially match C<\D*> with "ABC". Then it will
1992 try to match C<(?!123)> with "123", which fails. But because
1993 a quantifier (C<\D*>) has been used in the regular expression, the
1994 search engine can backtrack and retry the match differently
1995 in the hope of matching the complete regular expression.
1997 The pattern really, I<really> wants to succeed, so it uses the
1998 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
1999 time. Now there's indeed something following "AB" that is not
2000 "123". It's "C123", which suffices.
2002 We can deal with this by using both an assertion and a negation.
2003 We'll say that the first part in $1 must be followed both by a digit
2004 and by something that's not "123". Remember that the look-aheads
2005 are zero-width expressions--they only look, but don't consume any
2006 of the string in their match. So rewriting this way produces what
2007 you'd expect; that is, case 5 will fail, but case 6 succeeds:
2009 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
2010 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
2014 In other words, the two zero-width assertions next to each other work as though
2015 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
2016 matches only if you're at the beginning of the line AND the end of the
2017 line simultaneously. The deeper underlying truth is that juxtaposition in
2018 regular expressions always means AND, except when you write an explicit OR
2019 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
2020 although the attempted matches are made at different positions because "a"
2021 is not a zero-width assertion, but a one-width assertion.
2023 B<WARNING>: Particularly complicated regular expressions can take
2024 exponential time to solve because of the immense number of possible
2025 ways they can use backtracking to try for a match. For example, without
2026 internal optimizations done by the regular expression engine, this will
2027 take a painfully long time to run:
2029 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
2031 And if you used C<*>'s in the internal groups instead of limiting them
2032 to 0 through 5 matches, then it would take forever--or until you ran
2033 out of stack space. Moreover, these internal optimizations are not
2034 always applicable. For example, if you put C<{0,5}> instead of C<*>
2035 on the external group, no current optimization is applicable, and the
2036 match takes a long time to finish.
2038 A powerful tool for optimizing such beasts is what is known as an
2039 "independent group",
2040 which does not backtrack (see L</C<< (?>pattern) >>>). Note also that
2041 zero-length look-ahead/look-behind assertions will not backtrack to make
2042 the tail match, since they are in "logical" context: only
2043 whether they match is considered relevant. For an example
2044 where side-effects of look-ahead I<might> have influenced the
2045 following match, see L</C<< (?>pattern) >>>.
2047 =head2 Version 8 Regular Expressions
2048 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
2050 In case you're not familiar with the "regular" Version 8 regex
2051 routines, here are the pattern-matching rules not described above.
2053 Any single character matches itself, unless it is a I<metacharacter>
2054 with a special meaning described here or above. You can cause
2055 characters that normally function as metacharacters to be interpreted
2056 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
2057 character; "\\" matches a "\"). This escape mechanism is also required
2058 for the character used as the pattern delimiter.
2060 A series of characters matches that series of characters in the target
2061 string, so the pattern C<blurfl> would match "blurfl" in the target
2064 You can specify a character class, by enclosing a list of characters
2065 in C<[]>, which will match any character from the list. If the
2066 first character after the "[" is "^", the class matches any character not
2067 in the list. Within a list, the "-" character specifies a
2068 range, so that C<a-z> represents all characters between "a" and "z",
2069 inclusive. If you want either "-" or "]" itself to be a member of a
2070 class, put it at the start of the list (possibly after a "^"), or
2071 escape it with a backslash. "-" is also taken literally when it is
2072 at the end of the list, just before the closing "]". (The
2073 following all specify the same class of three characters: C<[-az]>,
2074 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
2075 specifies a class containing twenty-six characters, even on EBCDIC-based
2076 character sets.) Also, if you try to use the character
2077 classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
2078 a range, the "-" is understood literally.
2080 Note also that the whole range idea is rather unportable between
2081 character sets--and even within character sets they may cause results
2082 you probably didn't expect. A sound principle is to use only ranges
2083 that begin from and end at either alphabetics of equal case ([a-e],
2084 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
2085 spell out the character sets in full.
2087 Characters may be specified using a metacharacter syntax much like that
2088 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
2089 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
2090 of three octal digits, matches the character whose coded character set value
2091 is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
2092 matches the character whose ordinal is I<nn>. The expression \cI<x>
2093 matches the character control-I<x>. Finally, the "." metacharacter
2094 matches any character except "\n" (unless you use C</s>).
2096 You can specify a series of alternatives for a pattern using "|" to
2097 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
2098 or "foe" in the target string (as would C<f(e|i|o)e>). The
2099 first alternative includes everything from the last pattern delimiter
2100 ("(", "(?:", etc. or the beginning of the pattern) up to the first "|", and
2101 the last alternative contains everything from the last "|" to the next
2102 closing pattern delimiter. That's why it's common practice to include
2103 alternatives in parentheses: to minimize confusion about where they
2106 Alternatives are tried from left to right, so the first
2107 alternative found for which the entire expression matches, is the one that
2108 is chosen. This means that alternatives are not necessarily greedy. For
2109 example: when matching C<foo|foot> against "barefoot", only the "foo"
2110 part will match, as that is the first alternative tried, and it successfully
2111 matches the target string. (This might not seem important, but it is
2112 important when you are capturing matched text using parentheses.)
2114 Also remember that "|" is interpreted as a literal within square brackets,
2115 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
2117 Within a pattern, you may designate subpatterns for later reference
2118 by enclosing them in parentheses, and you may refer back to the
2119 I<n>th subpattern later in the pattern using the metacharacter
2120 \I<n> or \gI<n>. Subpatterns are numbered based on the left to right order
2121 of their opening parenthesis. A backreference matches whatever
2122 actually matched the subpattern in the string being examined, not
2123 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\g1\d*> will
2124 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
2125 1 matched "0x", even though the rule C<0|0x> could potentially match
2126 the leading 0 in the second number.
2128 =head2 Warning on \1 Instead of $1
2130 Some people get too used to writing things like:
2132 $pattern =~ s/(\W)/\\\1/g;
2134 This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid
2136 B<sed> addicts, but it's a dirty habit to get into. That's because in
2137 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
2138 the usual double-quoted string means a control-A. The customary Unix
2139 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
2140 of doing that, you get yourself into trouble if you then add an C</e>
2143 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
2149 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
2150 C<${1}000>. The operation of interpolation should not be confused
2151 with the operation of matching a backreference. Certainly they mean two
2152 different things on the I<left> side of the C<s///>.
2154 =head2 Repeated Patterns Matching a Zero-length Substring
2156 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
2158 Regular expressions provide a terse and powerful programming language. As
2159 with most other power tools, power comes together with the ability
2162 A common abuse of this power stems from the ability to make infinite
2163 loops using regular expressions, with something as innocuous as:
2165 'foo' =~ m{ ( o? )* }x;
2167 The C<o?> matches at the beginning of C<'foo'>, and since the position
2168 in the string is not moved by the match, C<o?> would match again and again
2169 because of the C<*> quantifier. Another common way to create a similar cycle
2170 is with the looping modifier C<//g>:
2172 @matches = ( 'foo' =~ m{ o? }xg );
2176 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
2178 or the loop implied by split().
2180 However, long experience has shown that many programming tasks may
2181 be significantly simplified by using repeated subexpressions that
2182 may match zero-length substrings. Here's a simple example being:
2184 @chars = split //, $string; # // is not magic in split
2185 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
2187 Thus Perl allows such constructs, by I<forcefully breaking
2188 the infinite loop>. The rules for this are different for lower-level
2189 loops given by the greedy quantifiers C<*+{}>, and for higher-level
2190 ones like the C</g> modifier or split() operator.
2192 The lower-level loops are I<interrupted> (that is, the loop is
2193 broken) when Perl detects that a repeated expression matched a
2194 zero-length substring. Thus
2196 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
2198 is made equivalent to
2200 m{ (?: NON_ZERO_LENGTH )* (?: ZERO_LENGTH )? }x;
2202 For example, this program
2209 (?{print "hello"}) # print hello whenever this
2211 (?=(b)) # zero-width assertion
2212 )* # any number of times
2223 Notice that "hello" is only printed once, as when Perl sees that the sixth
2224 iteration of the outermost C<(?:)*> matches a zero-length string, it stops
2227 The higher-level loops preserve an additional state between iterations:
2228 whether the last match was zero-length. To break the loop, the following
2229 match after a zero-length match is prohibited to have a length of zero.
2230 This prohibition interacts with backtracking (see L<"Backtracking">),
2231 and so the I<second best> match is chosen if the I<best> match is of
2239 results in C<< <><b><><a><><r><> >>. At each position of the string the best
2240 match given by non-greedy C<??> is the zero-length match, and the I<second
2241 best> match is what is matched by C<\w>. Thus zero-length matches
2242 alternate with one-character-long matches.
2244 Similarly, for repeated C<m/()/g> the second-best match is the match at the
2245 position one notch further in the string.
2247 The additional state of being I<matched with zero-length> is associated with
2248 the matched string, and is reset by each assignment to pos().
2249 Zero-length matches at the end of the previous match are ignored
2252 =head2 Combining RE Pieces
2254 Each of the elementary pieces of regular expressions which were described
2255 before (such as C<ab> or C<\Z>) could match at most one substring
2256 at the given position of the input string. However, in a typical regular
2257 expression these elementary pieces are combined into more complicated
2258 patterns using combining operators C<ST>, C<S|T>, C<S*> etc.
2259 (in these examples C<S> and C<T> are regular subexpressions).
2261 Such combinations can include alternatives, leading to a problem of choice:
2262 if we match a regular expression C<a|ab> against C<"abc">, will it match
2263 substring C<"a"> or C<"ab">? One way to describe which substring is
2264 actually matched is the concept of backtracking (see L<"Backtracking">).
2265 However, this description is too low-level and makes you think
2266 in terms of a particular implementation.
2268 Another description starts with notions of "better"/"worse". All the
2269 substrings which may be matched by the given regular expression can be
2270 sorted from the "best" match to the "worst" match, and it is the "best"
2271 match which is chosen. This substitutes the question of "what is chosen?"
2272 by the question of "which matches are better, and which are worse?".
2274 Again, for elementary pieces there is no such question, since at most
2275 one match at a given position is possible. This section describes the
2276 notion of better/worse for combining operators. In the description
2277 below C<S> and C<T> are regular subexpressions.
2283 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
2284 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
2285 which can be matched by C<T>.
2287 If C<A> is a better match for C<S> than C<A'>, C<AB> is a better
2290 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
2291 C<B> is a better match for C<T> than C<B'>.
2295 When C<S> can match, it is a better match than when only C<T> can match.
2297 Ordering of two matches for C<S> is the same as for C<S>. Similar for
2298 two matches for C<T>.
2300 =item C<S{REPEAT_COUNT}>
2302 Matches as C<SSS...S> (repeated as many times as necessary).
2306 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
2308 =item C<S{min,max}?>
2310 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
2312 =item C<S?>, C<S*>, C<S+>
2314 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
2316 =item C<S??>, C<S*?>, C<S+?>
2318 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
2322 Matches the best match for C<S> and only that.
2324 =item C<(?=S)>, C<(?<=S)>
2326 Only the best match for C<S> is considered. (This is important only if
2327 C<S> has capturing parentheses, and backreferences are used somewhere
2328 else in the whole regular expression.)
2330 =item C<(?!S)>, C<(?<!S)>
2332 For this grouping operator there is no need to describe the ordering, since
2333 only whether or not C<S> can match is important.
2335 =item C<(??{ EXPR })>, C<(?PARNO)>
2337 The ordering is the same as for the regular expression which is
2338 the result of EXPR, or the pattern contained by capture group PARNO.
2340 =item C<(?(condition)yes-pattern|no-pattern)>
2342 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
2343 already determined. The ordering of the matches is the same as for the
2344 chosen subexpression.
2348 The above recipes describe the ordering of matches I<at a given position>.
2349 One more rule is needed to understand how a match is determined for the
2350 whole regular expression: a match at an earlier position is always better
2351 than a match at a later position.
2353 =head2 Creating Custom RE Engines
2355 As of Perl 5.10.0, one can create custom regular expression engines. This
2356 is not for the faint of heart, as they have to plug in at the C level. See
2357 L<perlreapi> for more details.
2359 As an alternative, overloaded constants (see L<overload>) provide a simple
2360 way to extend the functionality of the RE engine, by substituting one
2361 pattern for another.
2363 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
2364 matches at a boundary between whitespace characters and non-whitespace
2365 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
2366 at these positions, so we want to have each C<\Y|> in the place of the
2367 more complicated version. We can create a module C<customre> to do
2375 die "No argument to customre::import allowed" if @_;
2376 overload::constant 'qr' => \&convert;
2379 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
2381 # We must also take care of not escaping the legitimate \\Y|
2382 # sequence, hence the presence of '\\' in the conversion rules.
2383 my %rules = ( '\\' => '\\\\',
2384 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
2390 { $rules{$1} or invalid($re,$1) }sgex;
2394 Now C<use customre> enables the new escape in constant regular
2395 expressions, i.e., those without any runtime variable interpolations.
2396 As documented in L<overload>, this conversion will work only over
2397 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
2398 part of this regular expression needs to be converted explicitly
2399 (but only if the special meaning of C<\Y|> should be enabled inside $re):
2404 $re = customre::convert $re;
2407 =head2 PCRE/Python Support
2409 As of Perl 5.10.0, Perl supports several Python/PCRE-specific extensions
2410 to the regex syntax. While Perl programmers are encouraged to use the
2411 Perl-specific syntax, the following are also accepted:
2415 =item C<< (?PE<lt>NAMEE<gt>pattern) >>
2417 Define a named capture group. Equivalent to C<< (?<NAME>pattern) >>.
2419 =item C<< (?P=NAME) >>
2421 Backreference to a named capture group. Equivalent to C<< \g{NAME} >>.
2423 =item C<< (?P>NAME) >>
2425 Subroutine call to a named capture group. Equivalent to C<< (?&NAME) >>.
2431 Many regular expression constructs don't work on EBCDIC platforms.
2433 There are a number of issues with regard to case-insensitive matching
2434 in Unicode rules. See C<i> under L</Modifiers> above.
2436 This document varies from difficult to understand to completely
2437 and utterly opaque. The wandering prose riddled with jargon is
2438 hard to fathom in several places.
2440 This document needs a rewrite that separates the tutorial content
2441 from the reference content.
2449 L<perlop/"Regexp Quote-Like Operators">.
2451 L<perlop/"Gory details of parsing quoted constructs">.
2461 I<Mastering Regular Expressions> by Jeffrey Friedl, published
2462 by O'Reilly and Associates.