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 tutorial introduction
11 is available in L<perlretut>. If you know just a little about them,
12 a quick-start introduction is available in L<perlrequick>.
14 Except for L</The Basics> section, this page assumes you are familiar
15 with regular expression basics, like what is a "pattern", what does it
16 look like, and how it is basically used. For a reference on how they
17 are used, plus various examples of the same, see discussions of C<m//>,
18 C<s///>, C<qr//> and C<"??"> in L<perlop/"Regexp Quote-Like Operators">.
20 New in v5.22, L<C<use re 'strict'>|re/'strict' mode> applies stricter
21 rules than otherwise when compiling regular expression patterns. It can
22 find things that, while legal, may not be what you intended.
25 X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
27 Regular expressions are strings with the very particular syntax and
28 meaning described in this document and auxiliary documents referred to
29 by this one. The strings are called "patterns". Patterns are used to
30 determine if some other string, called the "target", has (or doesn't
31 have) the characteristics specified by the pattern. We call this
32 "matching" the target string against the pattern. Usually the match is
33 done by having the target be the first operand, and the pattern be the
34 second operand, of one of the two binary operators C<=~> and C<!~>,
35 listed in L<perlop/Binding Operators>; and the pattern will have been
36 converted from an ordinary string by one of the operators in
37 L<perlop/"Regexp Quote-Like Operators">, like so:
41 This evaluates to true if and only if the string in the variable C<$foo>
42 contains somewhere in it, the sequence of characters "a", "b", then "c".
43 (The C<=~ m>, or match operator, is described in
44 L<perlop/m/PATTERN/msixpodualngc>.)
46 Patterns that aren't already stored in some variable must be delimited,
47 at both ends, by delimiter characters. These are often, as in the
48 example above, forward slashes, and the typical way a pattern is written
49 in documentation is with those slashes. In most cases, the delimiter
50 is the same character, fore and aft, but there are a few cases where a
51 character looks like it has a mirror-image mate, where the opening
52 version is the beginning delimiter, and the closing one is the ending
57 Most times, the pattern is evaluated in double-quotish context, but it
58 is possible to choose delimiters to force single-quotish, like
62 If the pattern contains its delimiter within it, that delimiter must be
63 escaped. Prefixing it with a backslash (I<e.g.>, C<"/foo\/bar/">)
66 Any single character in a pattern matches that same character in the
67 target string, unless the character is a I<metacharacter> with a special
68 meaning described in this document. A sequence of non-metacharacters
69 matches the same sequence in the target string, as we saw above with
72 Only a few characters (all of them being ASCII punctuation characters)
73 are metacharacters. The most commonly used one is a dot C<".">, which
74 normally matches almost any character (including a dot itself).
76 You can cause characters that normally function as metacharacters to be
77 interpreted literally by prefixing them with a C<"\">, just like the
78 pattern's delimiter must be escaped if it also occurs within the
79 pattern. Thus, C<"\."> matches just a literal dot, C<"."> instead of
80 its normal meaning. This means that the backslash is also a
81 metacharacter, so C<"\\"> matches a single C<"\">. And a sequence that
82 contains an escaped metacharacter matches the same sequence (but without
83 the escape) in the target string. So, the pattern C</blur\\fl/> would
84 match any target string that contains the sequence C<"blur\fl">.
86 The metacharacter C<"|"> is used to match one thing or another. Thus
90 is TRUE if and only if C<$foo> contains either the sequence C<"this"> or
91 the sequence C<"that">. Like all metacharacters, prefixing the C<"|">
92 with a backslash makes it match the plain punctuation character; in its
93 case, the VERTICAL LINE.
97 is TRUE if and only if C<$foo> contains the sequence C<"this|that">.
99 You aren't limited to just a single C<"|">.
101 $foo =~ m/fee|fie|foe|fum/
103 is TRUE if and only if C<$foo> contains any of those 4 sequences from
104 the children's story "Jack and the Beanstalk".
106 As you can see, the C<"|"> binds less tightly than a sequence of
107 ordinary characters. We can override this by using the grouping
108 metacharacters, the parentheses C<"("> and C<")">.
110 $foo =~ m/th(is|at) thing/
112 is TRUE if and only if C<$foo> contains either the sequence S<C<"this
113 thing">> or the sequence S<C<"that thing">>. The portions of the string
114 that match the portions of the pattern enclosed in parentheses are
115 normally made available separately for use later in the pattern,
116 substitution, or program. This is called "capturing", and it can get
117 complicated. See L</Capture groups>.
119 The first alternative includes everything from the last pattern
120 delimiter (C<"(">, C<"(?:"> (described later), I<etc>. or the beginning
121 of the pattern) up to the first C<"|">, and the last alternative
122 contains everything from the last C<"|"> to the next closing pattern
123 delimiter. That's why it's common practice to include alternatives in
124 parentheses: to minimize confusion about where they start and end.
126 Alternatives are tried from left to right, so the first
127 alternative found for which the entire expression matches, is the one that
128 is chosen. This means that alternatives are not necessarily greedy. For
129 example: when matching C<foo|foot> against C<"barefoot">, only the C<"foo">
130 part will match, as that is the first alternative tried, and it successfully
131 matches the target string. (This might not seem important, but it is
132 important when you are capturing matched text using parentheses.)
134 Besides taking away the special meaning of a metacharacter, a prefixed
135 backslash changes some letter and digit characters away from matching
136 just themselves to instead have special meaning. These are called
137 "escape sequences", and all such are described in L<perlrebackslash>. A
138 backslash sequence (of a letter or digit) that doesn't currently have
139 special meaning to Perl will raise a warning if warnings are enabled,
140 as those are reserved for potential future use.
142 One such sequence is C<\b>, which matches a boundary of some sort.
143 C<\b{wb}> and a few others give specialized types of boundaries.
144 (They are all described in detail starting at
145 L<perlrebackslash/\b{}, \b, \B{}, \B>.) Note that these don't match
146 characters, but the zero-width spaces between characters. They are an
147 example of a L<zero-width assertion|/Assertions>. Consider again,
149 $foo =~ m/fee|fie|foe|fum/
151 It evaluates to TRUE if, besides those 4 words, any of the sequences
152 "feed", "field", "Defoe", "fume", and many others are in C<$foo>. By
153 judicious use of C<\b> (or better (because it is designed to handle
154 natural language) C<\b{wb}>), we can make sure that only the Giant's
157 $foo =~ m/\b(fee|fie|foe|fum)\b/
158 $foo =~ m/\b{wb}(fee|fie|foe|fum)\b{wb}/
160 The final example shows that the characters C<"{"> and C<"}"> are
163 Another use for escape sequences is to specify characters that cannot
164 (or which you prefer not to) be written literally. These are described
165 in detail in L<perlrebackslash/Character Escapes>, but the next three
166 paragraphs briefly describe some of them.
168 Various control characters can be written in C language style: C<"\n">
169 matches a newline, C<"\t"> a tab, C<"\r"> a carriage return, C<"\f"> a
172 More generally, C<\I<nnn>>, where I<nnn> is a string of three octal
173 digits, matches the character whose native code point is I<nnn>. You
174 can easily run into trouble if you don't have exactly three digits. So
175 always use three, or since Perl 5.14, you can use C<\o{...}> to specify
176 any number of octal digits.
178 Similarly, C<\xI<nn>>, where I<nn> are hexadecimal digits, matches the
179 character whose native ordinal is I<nn>. Again, not using exactly two
180 digits is a recipe for disaster, but you can use C<\x{...}> to specify
181 any number of hex digits.
183 Besides being a metacharacter, the C<"."> is an example of a "character
184 class", something that can match any single character of a given set of
185 them. In its case, the set is just about all possible characters. Perl
186 predefines several character classes besides the C<".">; there is a
187 separate reference page about just these, L<perlrecharclass>.
189 You can define your own custom character classes, by putting into your
190 pattern in the appropriate place(s), a list of all the characters you
191 want in the set. You do this by enclosing the list within C<[]> bracket
192 characters. These are called "bracketed character classes" when we are
193 being precise, but often the word "bracketed" is dropped. (Dropping it
194 usually doesn't cause confusion.) This means that the C<"["> character
195 is another metacharacter. It doesn't match anything just by itself; it
196 is used only to tell Perl that what follows it is a bracketed character
197 class. If you want to match a literal left square bracket, you must
198 escape it, like C<"\[">. The matching C<"]"> is also a metacharacter;
199 again it doesn't match anything by itself, but just marks the end of
200 your custom class to Perl. It is an example of a "sometimes
201 metacharacter". It isn't a metacharacter if there is no corresponding
202 C<"[">, and matches its literal self:
204 print "]" =~ /]/; # prints 1
206 The list of characters within the character class gives the set of
207 characters matched by the class. C<"[abc]"> matches a single "a" or "b"
208 or "c". But if the first character after the C<"["> is C<"^">, the
209 class instead matches any character not in the list. Within a list, the
210 C<"-"> character specifies a range of characters, so that C<a-z>
211 represents all characters between "a" and "z", inclusive. If you want
212 either C<"-"> or C<"]"> itself to be a member of a class, put it at the
213 start of the list (possibly after a C<"^">), or escape it with a
214 backslash. C<"-"> is also taken literally when it is at the end of the
215 list, just before the closing C<"]">. (The following all specify the
216 same class of three characters: C<[-az]>, C<[az-]>, and C<[a\-z]>. All
217 are different from C<[a-z]>, which specifies a class containing
218 twenty-six characters, even on EBCDIC-based character sets.)
220 There is lots more to bracketed character classes; full details are in
221 L<perlrecharclass/Bracketed Character Classes>.
223 =head3 Metacharacters
225 X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
227 L</The Basics> introduced some of the metacharacters. This section
228 gives them all. Most of them have the same meaning as in the I<egrep>
231 Only the C<"\"> is always a metacharacter. The others are metacharacters
232 just sometimes. The following tables lists all of them, summarizes
233 their use, and gives the contexts where they are metacharacters.
234 Outside those contexts or if prefixed by a C<"\">, they match their
235 corresponding punctuation character. In some cases, their meaning
236 varies depending on various pattern modifiers that alter the default
237 behaviors. See L</Modifiers>.
241 \ Escape the next character Always, except when
243 ^ Match the beginning of the string Not in []
244 (or line, if /m is used)
245 ^ Complement the [] class At the beginning of []
246 . Match any single character except newline Not in []
247 (under /s, includes newline)
248 $ Match the end of the string Not in [], but can
249 (or before newline at the end of the mean interpolate a
250 string; or before any newline if /m is scalar
252 | Alternation Not in []
253 () Grouping Not in []
254 [ Start Bracketed Character class Not in []
255 ] End Bracketed Character class Only in [], and
257 * Matches the preceding element 0 or more Not in []
259 + Matches the preceding element 1 or more Not in []
261 ? Matches the preceding element 0 or 1 Not in []
263 { Starts a sequence that gives number(s) Not in []
264 of times the preceding element can be
266 { when following certain escape sequences
267 starts a modifier to the meaning of the
269 } End sequence started by {
270 - Indicates a range Only in [] interior
271 # Beginning of comment, extends to line end Only with /x modifier
273 Notice that most of the metacharacters lose their special meaning when
274 they occur in a bracketed character class, except C<"^"> has a different
275 meaning when it is at the beginning of such a class. And C<"-"> and C<"]">
276 are metacharacters only at restricted positions within bracketed
277 character classes; while C<"}"> is a metacharacter only when closing a
278 special construct started by C<"{">.
280 In double-quotish context, as is usually the case, you need to be
281 careful about C<"$"> and the non-metacharacter C<"@">. Those could
282 interpolate variables, which may or may not be what you intended.
284 These rules were designed for compactness of expression, rather than
285 legibility and maintainability. The L</E<sol>x and E<sol>xx> pattern
286 modifiers allow you to insert white space to improve readability. And
287 use of S<C<L<re 'strict'|re/'strict' mode>>> adds extra checking to
288 catch some typos that might silently compile into something unintended.
290 By default, the C<"^"> character is guaranteed to match only the
291 beginning of the string, the C<"$"> character only the end (or before the
292 newline at the end), and Perl does certain optimizations with the
293 assumption that the string contains only one line. Embedded newlines
294 will not be matched by C<"^"> or C<"$">. You may, however, wish to treat a
295 string as a multi-line buffer, such that the C<"^"> will match after any
296 newline within the string (except if the newline is the last character in
297 the string), and C<"$"> will match before any newline. At the
298 cost of a little more overhead, you can do this by using the
299 C<L</E<sol>m>> modifier on the pattern match operator. (Older programs
300 did this by setting C<$*>, but this option was removed in perl 5.10.)
303 To simplify multi-line substitutions, the C<"."> character never matches a
304 newline unless you use the L<C<E<sol>s>|/s> modifier, which in effect tells
305 Perl to pretend the string is a single line--even if it isn't.
312 The default behavior for matching can be changed, using various
313 modifiers. Modifiers that relate to the interpretation of the pattern
314 are listed just below. Modifiers that alter the way a pattern is used
315 by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and
316 L<perlop/"Gory details of parsing quoted constructs">. Modifiers can be added
317 dynamically; see L</Extended Patterns> below.
322 X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline>
324 Treat the string being matched against as multiple lines. That is, change C<"^"> and C<"$"> from matching
325 the start of the string's first line and the end of its last line to
326 matching the start and end of each line within the string.
329 X</s> X<regex, single-line> X<regexp, single-line>
330 X<regular expression, single-line>
332 Treat the string as single line. That is, change C<"."> to match any character
333 whatsoever, even a newline, which normally it would not match.
335 Used together, as C</ms>, they let the C<"."> match any character whatsoever,
336 while still allowing C<"^"> and C<"$"> to match, respectively, just after
337 and just before newlines within the string.
340 X</i> X<regex, case-insensitive> X<regexp, case-insensitive>
341 X<regular expression, case-insensitive>
343 Do case-insensitive pattern matching. For example, "A" will match "a"
346 If locale matching rules are in effect, the case map is taken from the
348 locale for code points less than 255, and from Unicode rules for larger
349 code points. However, matches that would cross the Unicode
350 rules/non-Unicode rules boundary (ords 255/256) will not succeed, unless
351 the locale is a UTF-8 one. See L<perllocale>.
353 There are a number of Unicode characters that match a sequence of
354 multiple characters under C</i>. For example,
355 C<LATIN SMALL LIGATURE FI> should match the sequence C<fi>. Perl is not
356 currently able to do this when the multiple characters are in the pattern and
357 are split between groupings, or when one or more are quantified. Thus
359 "\N{LATIN SMALL LIGATURE FI}" =~ /fi/i; # Matches
360 "\N{LATIN SMALL LIGATURE FI}" =~ /[fi][fi]/i; # Doesn't match!
361 "\N{LATIN SMALL LIGATURE FI}" =~ /fi*/i; # Doesn't match!
363 # The below doesn't match, and it isn't clear what $1 and $2 would
364 # be even if it did!!
365 "\N{LATIN SMALL LIGATURE FI}" =~ /(f)(i)/i; # Doesn't match!
367 Perl doesn't match multiple characters in a bracketed
368 character class unless the character that maps to them is explicitly
369 mentioned, and it doesn't match them at all if the character class is
370 inverted, which otherwise could be highly confusing. See
371 L<perlrecharclass/Bracketed Character Classes>, and
372 L<perlrecharclass/Negation>.
374 =item B<C<x>> and B<C<xx>>
377 Extend your pattern's legibility by permitting whitespace and comments.
378 Details in L</E<sol>x and E<sol>xx>
381 X</p> X<regex, preserve> X<regexp, preserve>
383 Preserve the string matched such that C<${^PREMATCH}>, C<${^MATCH}>, and
384 C<${^POSTMATCH}> are available for use after matching.
386 In Perl 5.20 and higher this is ignored. Due to a new copy-on-write
387 mechanism, C<${^PREMATCH}>, C<${^MATCH}>, and C<${^POSTMATCH}> will be available
388 after the match regardless of the modifier.
390 =item B<C<a>>, B<C<d>>, B<C<l>>, and B<C<u>>
391 X</a> X</d> X</l> X</u>
393 These modifiers, all new in 5.14, affect which character-set rules
394 (Unicode, I<etc>.) are used, as described below in
395 L</Character set modifiers>.
398 X</n> X<regex, non-capture> X<regexp, non-capture>
399 X<regular expression, non-capture>
401 Prevent the grouping metacharacters C<()> from capturing. This modifier,
402 new in 5.22, will stop C<$1>, C<$2>, I<etc>... from being filled in.
404 "hello" =~ /(hi|hello)/; # $1 is "hello"
405 "hello" =~ /(hi|hello)/n; # $1 is undef
407 This is equivalent to putting C<?:> at the beginning of every capturing group:
409 "hello" =~ /(?:hi|hello)/; # $1 is undef
411 C</n> can be negated on a per-group basis. Alternatively, named captures
414 "hello" =~ /(?-n:(hi|hello))/n; # $1 is "hello"
415 "hello" =~ /(?<greet>hi|hello)/n; # $1 is "hello", $+{greet} is
418 =item Other Modifiers
420 There are a number of flags that can be found at the end of regular
421 expression constructs that are I<not> generic regular expression flags, but
422 apply to the operation being performed, like matching or substitution (C<m//>
423 or C<s///> respectively).
425 Flags described further in
426 L<perlretut/"Using regular expressions in Perl"> are:
428 c - keep the current position during repeated matching
429 g - globally match the pattern repeatedly in the string
431 Substitution-specific modifiers described in
432 L<perlop/"s/PATTERN/REPLACEMENT/msixpodualngcer"> are:
434 e - evaluate the right-hand side as an expression
435 ee - evaluate the right side as a string then eval the result
436 o - pretend to optimize your code, but actually introduce bugs
437 r - perform non-destructive substitution and return the new value
441 Regular expression modifiers are usually written in documentation
442 as I<e.g.>, "the C</x> modifier", even though the delimiter
443 in question might not really be a slash. The modifiers C</imnsxadlup>
444 may also be embedded within the regular expression itself using
445 the C<(?...)> construct, see L</Extended Patterns> below.
447 =head3 Details on some modifiers
449 Some of the modifiers require more explanation than given in the
452 =head4 C</x> and C</xx>
455 the regular expression parser to ignore most whitespace that is neither
456 backslashed nor within a bracketed character class. You can use this to
457 break up your regular expression into more readable parts.
458 Also, the C<"#"> character is treated as a metacharacter introducing a
459 comment that runs up to the pattern's closing delimiter, or to the end
460 of the current line if the pattern extends onto the next line. Hence,
461 this is very much like an ordinary Perl code comment. (You can include
462 the closing delimiter within the comment only if you precede it with a
463 backslash, so be careful!)
465 Use of C</x> means that if you want real
466 whitespace or C<"#"> characters in the pattern (outside a bracketed character
467 class, which is unaffected by C</x>), then you'll either have to
468 escape them (using backslashes or C<\Q...\E>) or encode them using octal,
469 hex, or C<\N{}> or C<\p{name=...}> escapes.
470 It is ineffective to try to continue a comment onto the next line by
471 escaping the C<\n> with a backslash or C<\Q>.
473 You can use L</(?#text)> to create a comment that ends earlier than the
474 end of the current line, but C<text> also can't contain the closing
475 delimiter unless escaped with a backslash.
477 A common pitfall is to forget that C<"#"> characters begin a comment under
478 C</x> and are not matched literally. Just keep that in mind when trying
479 to puzzle out why a particular C</x> pattern isn't working as expected.
481 Starting in Perl v5.26, if the modifier has a second C<"x"> within it,
482 it does everything that a single C</x> does, but additionally
483 non-backslashed SPACE and TAB characters within bracketed character
484 classes are also generally ignored, and hence can be added to make the
485 classes more readable.
488 /[ ! @ " # $ % ^ & * () = ? <> ' ]/xx
490 may be easier to grasp than the squashed equivalents
495 Taken together, these features go a long way towards
496 making Perl's regular expressions more readable. Here's an example:
498 # Delete (most) C comments.
500 /\* # Match the opening delimiter.
501 .*? # Match a minimal number of characters.
502 \*/ # Match the closing delimiter.
505 Note that anything inside
506 a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect
507 space interpretation within a single multi-character construct. For
508 example C<(?:...)> can't have a space between the C<"(">,
509 C<"?">, and C<":">. Within any delimiters for such a construct, allowed
510 spaces are not affected by C</x>, and depend on the construct. For
511 example, all constructs using curly braces as delimiters, such as
512 C<\x{...}> can have blanks within but adjacent to the braces, but not
513 elsewhere, and no non-blank space characters. An exception are Unicode
514 properties which follow Unicode rules, for which see
515 L<perluniprops/Properties accessible through \p{} and \P{}>.
518 The set of characters that are deemed whitespace are those that Unicode
519 calls "Pattern White Space", namely:
521 U+0009 CHARACTER TABULATION
523 U+000B LINE TABULATION
525 U+000D CARRIAGE RETURN
528 U+200E LEFT-TO-RIGHT MARK
529 U+200F RIGHT-TO-LEFT MARK
530 U+2028 LINE SEPARATOR
531 U+2029 PARAGRAPH SEPARATOR
533 =head4 Character set modifiers
535 C</d>, C</u>, C</a>, and C</l>, available starting in 5.14, are called
536 the character set modifiers; they affect the character set rules
537 used for the regular expression.
539 The C</d>, C</u>, and C</l> modifiers are not likely to be of much use
540 to you, and so you need not worry about them very much. They exist for
541 Perl's internal use, so that complex regular expression data structures
542 can be automatically serialized and later exactly reconstituted,
543 including all their nuances. But, since Perl can't keep a secret, and
544 there may be rare instances where they are useful, they are documented
547 The C</a> modifier, on the other hand, may be useful. Its purpose is to
548 allow code that is to work mostly on ASCII data to not have to concern
551 Briefly, C</l> sets the character set to that of whatever B<L>ocale is in
552 effect at the time of the execution of the pattern match.
554 C</u> sets the character set to B<U>nicode.
556 C</a> also sets the character set to Unicode, BUT adds several
557 restrictions for B<A>SCII-safe matching.
559 C</d> is the old, problematic, pre-5.14 B<D>efault character set
560 behavior. Its only use is to force that old behavior.
562 At any given time, exactly one of these modifiers is in effect. Their
563 existence allows Perl to keep the originally compiled behavior of a
564 regular expression, regardless of what rules are in effect when it is
565 actually executed. And if it is interpolated into a larger regex, the
566 original's rules continue to apply to it, and don't affect the other
569 The C</l> and C</u> modifiers are automatically selected for
570 regular expressions compiled within the scope of various pragmas,
571 and we recommend that in general, you use those pragmas instead of
572 specifying these modifiers explicitly. For one thing, the modifiers
573 affect only pattern matching, and do not extend to even any replacement
574 done, whereas using the pragmas gives consistent results for all
575 appropriate operations within their scopes. For example,
579 will match "foo" using the locale's rules for case-insensitive matching,
580 but the C</l> does not affect how the C<\U> operates. Most likely you
581 want both of them to use locale rules. To do this, instead compile the
582 regular expression within the scope of C<use locale>. This both
583 implicitly adds the C</l>, and applies locale rules to the C<\U>. The
584 lesson is to C<use locale>, and not C</l> explicitly.
586 Similarly, it would be better to use C<use feature 'unicode_strings'>
591 to get Unicode rules, as the C<\L> in the former (but not necessarily
592 the latter) would also use Unicode rules.
594 More detail on each of the modifiers follows. Most likely you don't
595 need to know this detail for C</l>, C</u>, and C</d>, and can skip ahead
596 to L<E<sol>a|/E<sol>a (and E<sol>aa)>.
600 means to use the current locale's rules (see L<perllocale>) when pattern
601 matching. For example, C<\w> will match the "word" characters of that
602 locale, and C<"/i"> case-insensitive matching will match according to
603 the locale's case folding rules. The locale used will be the one in
604 effect at the time of execution of the pattern match. This may not be
605 the same as the compilation-time locale, and can differ from one match
606 to another if there is an intervening call of the
607 L<setlocale() function|perllocale/The setlocale function>.
609 Prior to v5.20, Perl did not support multi-byte locales. Starting then,
610 UTF-8 locales are supported. No other multi byte locales are ever
611 likely to be supported. However, in all locales, one can have code
612 points above 255 and these will always be treated as Unicode no matter
613 what locale is in effect.
615 Under Unicode rules, there are a few case-insensitive matches that cross
616 the 255/256 boundary. Except for UTF-8 locales in Perls v5.20 and
617 later, these are disallowed under C</l>. For example, 0xFF (on ASCII
618 platforms) does not caselessly match the character at 0x178, C<LATIN
619 CAPITAL LETTER Y WITH DIAERESIS>, because 0xFF may not be C<LATIN SMALL
620 LETTER Y WITH DIAERESIS> in the current locale, and Perl has no way of
621 knowing if that character even exists in the locale, much less what code
624 In a UTF-8 locale in v5.20 and later, the only visible difference
625 between locale and non-locale in regular expressions should be tainting
628 This modifier may be specified to be the default by C<use locale>, but
629 see L</Which character set modifier is in effect?>.
634 means to use Unicode rules when pattern matching. On ASCII platforms,
635 this means that the code points between 128 and 255 take on their
636 Latin-1 (ISO-8859-1) meanings (which are the same as Unicode's).
637 (Otherwise Perl considers their meanings to be undefined.) Thus,
638 under this modifier, the ASCII platform effectively becomes a Unicode
639 platform; and hence, for example, C<\w> will match any of the more than
640 100_000 word characters in Unicode.
642 Unlike most locales, which are specific to a language and country pair,
643 Unicode classifies all the characters that are letters I<somewhere> in
645 C<\w>. For example, your locale might not think that C<LATIN SMALL
646 LETTER ETH> is a letter (unless you happen to speak Icelandic), but
647 Unicode does. Similarly, all the characters that are decimal digits
648 somewhere in the world will match C<\d>; this is hundreds, not 10,
649 possible matches. And some of those digits look like some of the 10
650 ASCII digits, but mean a different number, so a human could easily think
651 a number is a different quantity than it really is. For example,
652 C<BENGALI DIGIT FOUR> (U+09EA) looks very much like an
653 C<ASCII DIGIT EIGHT> (U+0038), and C<LEPCHA DIGIT SIX> (U+1C46) looks
654 very much like an C<ASCII DIGIT FIVE> (U+0035). And, C<\d+>, may match
655 strings of digits that are a mixture from different writing systems,
656 creating a security issue. A fraudulent website, for example, could
657 display the price of something using U+1C46, and it would appear to the
658 user that something cost 500 units, but it really costs 600. A browser
659 that enforced script runs (L</Script Runs>) would prevent that
660 fraudulent display. L<Unicode::UCD/num()> can also be used to sort this
661 out. Or the C</a> modifier can be used to force C<\d> to match just the
664 Also, under this modifier, case-insensitive matching works on the full
666 characters. The C<KELVIN SIGN>, for example matches the letters "k" and
667 "K"; and C<LATIN SMALL LIGATURE FF> matches the sequence "ff", which,
668 if you're not prepared, might make it look like a hexadecimal constant,
669 presenting another potential security issue. See
670 L<https://unicode.org/reports/tr36> for a detailed discussion of Unicode
673 This modifier may be specified to be the default by C<use feature
674 'unicode_strings>, C<use locale ':not_characters'>, or
675 C<L<use 5.012|perlfunc/use VERSION>> (or higher),
676 but see L</Which character set modifier is in effect?>.
681 B<IMPORTANT:> Because of the unpredictable behaviors this
682 modifier causes, only use it to maintain weird backward compatibilities.
684 L<< C<unicode_strings>|feature/"The 'unicode_strings' feature" >>
686 in new code to avoid inadvertently enabling this modifier by default.
688 What does this modifier do? It "Depends"!
690 This modifier means to use platform-native matching rules
691 except when there is cause to use Unicode rules instead, as follows:
697 the target string's L<UTF8 flag|perlunifaq/What is "the UTF8 flag"?>
698 (see below) is set; or
702 the pattern's L<UTF8 flag|perlunifaq/What is "the UTF8 flag"?>
703 (see below) is set; or
707 the pattern explicitly mentions a code point that is above 255 (say by
712 the pattern uses a Unicode name (C<\N{...}>); or
716 the pattern uses a Unicode property (C<\p{...}> or C<\P{...}>); or
720 the pattern uses a Unicode break (C<\b{...}> or C<\B{...}>); or
724 the pattern uses C<L</(?[ ])>>
728 the pattern uses L<C<(*script_run: ...)>|/Script Runs>
732 Regarding the "UTF8 flag" references above: normally Perl applications
733 shouldn't think about that flag. It's part of Perl's internals,
734 so it can change whenever Perl wants. C</d> may thus cause unpredictable
735 results. See L<perlunicode/The "Unicode Bug">. This bug
736 has become rather infamous, leading to yet other (without swearing) names
737 for this modifier like "Dicey" and "Dodgy".
739 Here are some examples of how that works on an ASCII platform:
742 utf8::downgrade($str); # $str is not UTF8-flagged.
743 $str =~ /^\w/; # No match, since no UTF8 flag.
745 $str .= "\x{0e0b}"; # Now $str is UTF8-flagged.
746 $str =~ /^\w/; # Match! $str is now UTF8-flagged.
748 $str =~ /^\w/; # Still a match! $str retains its UTF8 flag.
750 Under Perl's default configuration this modifier is automatically
751 selected by default when none of the others are, so yet another name
752 for it (unfortunately) is "Default".
754 Whenever you can, use the
755 L<< C<unicode_strings>|feature/"The 'unicode_strings' feature" >>
756 to cause X</u> to be the default instead.
760 This modifier stands for ASCII-restrict (or ASCII-safe). This modifier
761 may be doubled-up to increase its effect.
763 When it appears singly, it causes the sequences C<\d>, C<\s>, C<\w>, and
764 the Posix character classes to match only in the ASCII range. They thus
765 revert to their pre-5.6, pre-Unicode meanings. Under C</a>, C<\d>
766 always means precisely the digits C<"0"> to C<"9">; C<\s> means the five
767 characters C<[ \f\n\r\t]>, and starting in Perl v5.18, the vertical tab;
768 C<\w> means the 63 characters
769 C<[A-Za-z0-9_]>; and likewise, all the Posix classes such as
770 C<[[:print:]]> match only the appropriate ASCII-range characters.
772 This modifier is useful for people who only incidentally use Unicode,
773 and who do not wish to be burdened with its complexities and security
776 With C</a>, one can write C<\d> with confidence that it will only match
777 ASCII characters, and should the need arise to match beyond ASCII, you
778 can instead use C<\p{Digit}> (or C<\p{Word}> for C<\w>). There are
779 similar C<\p{...}> constructs that can match beyond ASCII both white
780 space (see L<perlrecharclass/Whitespace>), and Posix classes (see
781 L<perlrecharclass/POSIX Character Classes>). Thus, this modifier
782 doesn't mean you can't use Unicode, it means that to get Unicode
783 matching you must explicitly use a construct (C<\p{}>, C<\P{}>) that
786 As you would expect, this modifier causes, for example, C<\D> to mean
787 the same thing as C<[^0-9]>; in fact, all non-ASCII characters match
788 C<\D>, C<\S>, and C<\W>. C<\b> still means to match at the boundary
789 between C<\w> and C<\W>, using the C</a> definitions of them (similarly
792 Otherwise, C</a> behaves like the C</u> modifier, in that
793 case-insensitive matching uses Unicode rules; for example, "k" will
794 match the Unicode C<\N{KELVIN SIGN}> under C</i> matching, and code
795 points in the Latin1 range, above ASCII will have Unicode rules when it
796 comes to case-insensitive matching.
798 To forbid ASCII/non-ASCII matches (like "k" with C<\N{KELVIN SIGN}>),
799 specify the C<"a"> twice, for example C</aai> or C</aia>. (The first
800 occurrence of C<"a"> restricts the C<\d>, I<etc>., and the second occurrence
801 adds the C</i> restrictions.) But, note that code points outside the
802 ASCII range will use Unicode rules for C</i> matching, so the modifier
803 doesn't really restrict things to just ASCII; it just forbids the
804 intermixing of ASCII and non-ASCII.
806 To summarize, this modifier provides protection for applications that
807 don't wish to be exposed to all of Unicode. Specifying it twice
808 gives added protection.
810 This modifier may be specified to be the default by C<use re '/a'>
811 or C<use re '/aa'>. If you do so, you may actually have occasion to use
812 the C</u> modifier explicitly if there are a few regular expressions
813 where you do want full Unicode rules (but even here, it's best if
814 everything were under feature C<"unicode_strings">, along with the
815 C<use re '/aa'>). Also see L</Which character set modifier is in
820 =head4 Which character set modifier is in effect?
822 Which of these modifiers is in effect at any given point in a regular
823 expression depends on a fairly complex set of interactions. These have
824 been designed so that in general you don't have to worry about it, but
825 this section gives the gory details. As
826 explained below in L</Extended Patterns> it is possible to explicitly
827 specify modifiers that apply only to portions of a regular expression.
828 The innermost always has priority over any outer ones, and one applying
829 to the whole expression has priority over any of the default settings that are
830 described in the remainder of this section.
832 The C<L<use re 'E<sol>foo'|re/"'/flags' mode">> pragma can be used to set
833 default modifiers (including these) for regular expressions compiled
834 within its scope. This pragma has precedence over the other pragmas
835 listed below that also change the defaults.
837 Otherwise, C<L<use locale|perllocale>> sets the default modifier to C</l>;
838 and C<L<use feature 'unicode_strings|feature>>, or
839 C<L<use 5.012|perlfunc/use VERSION>> (or higher) set the default to
840 C</u> when not in the same scope as either C<L<use locale|perllocale>>
841 or C<L<use bytes|bytes>>.
842 (C<L<use locale ':not_characters'|perllocale/Unicode and UTF-8>> also
843 sets the default to C</u>, overriding any plain C<use locale>.)
844 Unlike the mechanisms mentioned above, these
845 affect operations besides regular expressions pattern matching, and so
846 give more consistent results with other operators, including using
847 C<\U>, C<\l>, I<etc>. in substitution replacements.
849 If none of the above apply, for backwards compatibility reasons, the
850 C</d> modifier is the one in effect by default. As this can lead to
851 unexpected results, it is best to specify which other rule set should be
854 =head4 Character set modifier behavior prior to Perl 5.14
856 Prior to 5.14, there were no explicit modifiers, but C</l> was implied
857 for regexes compiled within the scope of C<use locale>, and C</d> was
858 implied otherwise. However, interpolating a regex into a larger regex
859 would ignore the original compilation in favor of whatever was in effect
860 at the time of the second compilation. There were a number of
861 inconsistencies (bugs) with the C</d> modifier, where Unicode rules
862 would be used when inappropriate, and vice versa. C<\p{}> did not imply
863 Unicode rules, and neither did all occurrences of C<\N{}>, until 5.12.
865 =head2 Regular Expressions
869 Quantifiers are used when a particular portion of a pattern needs to
870 match a certain number (or numbers) of times. If there isn't a
871 quantifier the number of times to match is exactly one. The following
872 standard quantifiers are recognized:
873 X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
875 * Match 0 or more times
876 + Match 1 or more times
878 {n} Match exactly n times
879 {n,} Match at least n times
880 {,n} Match at most n times
881 {n,m} Match at least n but not more than m times
883 (If a non-escaped curly bracket occurs in a context other than one of
884 the quantifiers listed above, where it does not form part of a
885 backslashed sequence like C<\x{...}>, it is either a fatal syntax error,
886 or treated as a regular character, generally with a deprecation warning
887 raised. To escape it, you can precede it with a backslash (C<"\{">) or
888 enclose it within square brackets (C<"[{]">).
889 This change will allow for future syntax extensions (like making the
890 lower bound of a quantifier optional), and better error checking of
893 The C<"*"> quantifier is equivalent to C<{0,}>, the C<"+">
894 quantifier to C<{1,}>, and the C<"?"> quantifier to C<{0,1}>. I<n> and I<m> are limited
895 to non-negative integral values less than a preset limit defined when perl is built.
896 This is usually 65534 on the most common platforms. The actual limit can
897 be seen in the error message generated by code such as this:
899 $_ **= $_ , / {$_} / for 2 .. 42;
901 By default, a quantified subpattern is "greedy", that is, it will match as
902 many times as possible (given a particular starting location) while still
903 allowing the rest of the pattern to match. If you want it to match the
904 minimum number of times possible, follow the quantifier with a C<"?">. Note
905 that the meanings don't change, just the "greediness":
906 X<metacharacter> X<greedy> X<greediness>
907 X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{,n}?> X<{n,m}?>
909 *? Match 0 or more times, not greedily
910 +? Match 1 or more times, not greedily
911 ?? Match 0 or 1 time, not greedily
912 {n}? Match exactly n times, not greedily (redundant)
913 {n,}? Match at least n times, not greedily
914 {,n}? Match at most n times, not greedily
915 {n,m}? Match at least n but not more than m times, not greedily
917 Normally when a quantified subpattern does not allow the rest of the
918 overall pattern to match, Perl will backtrack. However, this behaviour is
919 sometimes undesirable. Thus Perl provides the "possessive" quantifier form
922 *+ Match 0 or more times and give nothing back
923 ++ Match 1 or more times and give nothing back
924 ?+ Match 0 or 1 time and give nothing back
925 {n}+ Match exactly n times and give nothing back (redundant)
926 {n,}+ Match at least n times and give nothing back
927 {,n}+ Match at most n times and give nothing back
928 {n,m}+ Match at least n but not more than m times and give nothing back
934 will never match, as the C<a++> will gobble up all the C<"a">'s in the
935 string and won't leave any for the remaining part of the pattern. This
936 feature can be extremely useful to give perl hints about where it
937 shouldn't backtrack. For instance, the typical "match a double-quoted
938 string" problem can be most efficiently performed when written as:
940 /"(?:[^"\\]++|\\.)*+"/
942 as we know that if the final quote does not match, backtracking will not
943 help. See the independent subexpression
944 C<L</(?E<gt>I<pattern>)>> for more details;
945 possessive quantifiers are just syntactic sugar for that construct. For
946 instance the above example could also be written as follows:
948 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
950 Note that the possessive quantifier modifier can not be combined
951 with the non-greedy modifier. This is because it would make no sense.
952 Consider the follow equivalency table:
960 =head3 Escape sequences
962 Because patterns are processed as double-quoted strings, the following
969 \a alarm (bell) (BEL)
970 \e escape (think troff) (ESC)
971 \cK control char (example: VT)
972 \x{}, \x00 character whose ordinal is the given hexadecimal number
973 \N{name} named Unicode character or character sequence
974 \N{U+263D} Unicode character (example: FIRST QUARTER MOON)
975 \o{}, \000 character whose ordinal is the given octal number
976 \l lowercase next char (think vi)
977 \u uppercase next char (think vi)
978 \L lowercase until \E (think vi)
979 \U uppercase until \E (think vi)
980 \Q quote (disable) pattern metacharacters until \E
981 \E end either case modification or quoted section, think vi
983 Details are in L<perlop/Quote and Quote-like Operators>.
985 =head3 Character Classes and other Special Escapes
987 In addition, Perl defines the following:
988 X<\g> X<\k> X<\K> X<backreference>
990 Sequence Note Description
991 [...] [1] Match a character according to the rules of the
992 bracketed character class defined by the "...".
993 Example: [a-z] matches "a" or "b" or "c" ... or "z"
994 [[:...:]] [2] Match a character according to the rules of the POSIX
995 character class "..." within the outer bracketed
996 character class. Example: [[:upper:]] matches any
998 (?[...]) [8] Extended bracketed character class
999 \w [3] Match a "word" character (alphanumeric plus "_", plus
1000 other connector punctuation chars plus Unicode
1002 \W [3] Match a non-"word" character
1003 \s [3] Match a whitespace character
1004 \S [3] Match a non-whitespace character
1005 \d [3] Match a decimal digit character
1006 \D [3] Match a non-digit character
1007 \pP [3] Match P, named property. Use \p{Prop} for longer names
1009 \X [4] Match Unicode "eXtended grapheme cluster"
1010 \1 [5] Backreference to a specific capture group or buffer.
1011 '1' may actually be any positive integer.
1012 \g1 [5] Backreference to a specific or previous group,
1013 \g{-1} [5] The number may be negative indicating a relative
1014 previous group and may optionally be wrapped in
1015 curly brackets for safer parsing.
1016 \g{name} [5] Named backreference
1017 \k<name> [5] Named backreference
1018 \k'name' [5] Named backreference
1019 \k{name} [5] Named backreference
1020 \K [6] Keep the stuff left of the \K, don't include it in $&
1021 \N [7] Any character but \n. Not affected by /s modifier
1022 \v [3] Vertical whitespace
1023 \V [3] Not vertical whitespace
1024 \h [3] Horizontal whitespace
1025 \H [3] Not horizontal whitespace
1032 See L<perlrecharclass/Bracketed Character Classes> for details.
1036 See L<perlrecharclass/POSIX Character Classes> for details.
1040 See L<perlunicode/Unicode Character Properties> for details
1044 See L<perlrebackslash/Misc> for details.
1048 See L</Capture groups> below for details.
1052 See L</Extended Patterns> below for details.
1056 Note that C<\N> has two meanings. When of the form C<\N{I<NAME>}>, it
1057 matches the character or character sequence whose name is I<NAME>; and
1059 when of the form C<\N{U+I<hex>}>, it matches the character whose Unicode
1060 code point is I<hex>. Otherwise it matches any character but C<\n>.
1064 See L<perlrecharclass/Extended Bracketed Character Classes> for details.
1070 Besides L<C<"^"> and C<"$">|/Metacharacters>, Perl defines the following
1071 zero-width assertions:
1072 X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
1073 X<regexp, zero-width assertion>
1074 X<regular expression, zero-width assertion>
1075 X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
1077 \b{} Match at Unicode boundary of specified type
1078 \B{} Match where corresponding \b{} doesn't match
1079 \b Match a \w\W or \W\w boundary
1080 \B Match except at a \w\W or \W\w boundary
1081 \A Match only at beginning of string
1082 \Z Match only at end of string, or before newline at the end
1083 \z Match only at end of string
1084 \G Match only at pos() (e.g. at the end-of-match position
1087 A Unicode boundary (C<\b{}>), available starting in v5.22, is a spot
1088 between two characters, or before the first character in the string, or
1089 after the final character in the string where certain criteria defined
1090 by Unicode are met. See L<perlrebackslash/\b{}, \b, \B{}, \B> for
1093 A word boundary (C<\b>) is a spot between two characters
1094 that has a C<\w> on one side of it and a C<\W> on the other side
1095 of it (in either order), counting the imaginary characters off the
1096 beginning and end of the string as matching a C<\W>. (Within
1097 character classes C<\b> represents backspace rather than a word
1098 boundary, just as it normally does in any double-quoted string.)
1099 The C<\A> and C<\Z> are just like C<"^"> and C<"$">, except that they
1100 won't match multiple times when the C</m> modifier is used, while
1101 C<"^"> and C<"$"> will match at every internal line boundary. To match
1102 the actual end of the string and not ignore an optional trailing
1104 X<\b> X<\A> X<\Z> X<\z> X</m>
1106 The C<\G> assertion can be used to chain global matches (using
1107 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
1108 It is also useful when writing C<lex>-like scanners, when you have
1109 several patterns that you want to match against consequent substrings
1110 of your string; see the previous reference. The actual location
1111 where C<\G> will match can also be influenced by using C<pos()> as
1112 an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
1113 matches (see L</"Repeated Patterns Matching a Zero-length Substring">)
1114 is modified somewhat, in that contents to the left of C<\G> are
1115 not counted when determining the length of the match. Thus the following
1116 will not match forever:
1121 while ($string =~ /(.\G)/g) {
1125 It will print 'A' and then terminate, as it considers the match to
1126 be zero-width, and thus will not match at the same position twice in a
1129 It is worth noting that C<\G> improperly used can result in an infinite
1130 loop. Take care when using patterns that include C<\G> in an alternation.
1132 Note also that C<s///> will refuse to overwrite part of a substitution
1133 that has already been replaced; so for example this will stop after the
1134 first iteration, rather than iterating its way backwards through the
1140 print; # prints 1234X6789, not XXXXX6789
1143 =head3 Capture groups
1145 The grouping construct C<( ... )> creates capture groups (also referred to as
1146 capture buffers). To refer to the current contents of a group later on, within
1147 the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>)
1148 for the second, and so on.
1149 This is called a I<backreference>.
1150 X<regex, capture buffer> X<regexp, capture buffer>
1151 X<regex, capture group> X<regexp, capture group>
1152 X<regular expression, capture buffer> X<backreference>
1153 X<regular expression, capture group> X<backreference>
1154 X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
1155 X<named capture buffer> X<regular expression, named capture buffer>
1156 X<named capture group> X<regular expression, named capture group>
1157 X<%+> X<$+{name}> X<< \k<name> >>
1158 There is no limit to the number of captured substrings that you may use.
1159 Groups are numbered with the leftmost open parenthesis being number 1, I<etc>. If
1160 a group did not match, the associated backreference won't match either. (This
1161 can happen if the group is optional, or in a different branch of an
1163 You can omit the C<"g">, and write C<"\1">, I<etc>, but there are some issues with
1164 this form, described below.
1166 You can also refer to capture groups relatively, by using a negative number, so
1167 that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture
1168 group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For
1175 \g{-1} # backref to group 3
1176 \g{-3} # backref to group 1
1180 would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to
1181 interpolate regexes into larger regexes and not have to worry about the
1182 capture groups being renumbered.
1184 You can dispense with numbers altogether and create named capture groups.
1185 The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to
1186 reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may
1187 also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.)
1188 I<name> must not begin with a number, nor contain hyphens.
1189 When different groups within the same pattern have the same name, any reference
1190 to that name assumes the leftmost defined group. Named groups count in
1191 absolute and relative numbering, and so can also be referred to by those
1193 (It's possible to do things with named capture groups that would otherwise
1196 Capture group contents are dynamically scoped and available to you outside the
1197 pattern until the end of the enclosing block or until the next successful
1198 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
1199 You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">,
1200 I<etc>); or by name via the C<%+> hash, using C<"$+{I<name>}">.
1202 Braces are required in referring to named capture groups, but are optional for
1203 absolute or relative numbered ones. Braces are safer when creating a regex by
1204 concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a>
1205 contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which
1206 is probably not what you intended.
1208 If you use braces, you may also optionally add any number of blank
1209 (space or tab) characters within but adjacent to the braces, like
1210 S<C<\g{ -1 }>>, or S<C<\k{ I<name> }>>.
1212 The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that
1213 there were no named nor relative numbered capture groups. Absolute numbered
1214 groups were referred to using C<\1>,
1215 C<\2>, I<etc>., and this notation is still
1216 accepted (and likely always will be). But it leads to some ambiguities if
1217 there are more than 9 capture groups, as C<\10> could mean either the tenth
1218 capture group, or the character whose ordinal in octal is 010 (a backspace in
1219 ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference
1220 only if at least 10 left parentheses have opened before it. Likewise C<\11> is
1221 a backreference only if at least 11 left parentheses have opened before it.
1222 And so on. C<\1> through C<\9> are always interpreted as backreferences.
1223 There are several examples below that illustrate these perils. You can avoid
1224 the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups;
1225 and for octal constants always using C<\o{}>, or for C<\077> and below, using 3
1226 digits padded with leading zeros, since a leading zero implies an octal
1229 The C<\I<digit>> notation also works in certain circumstances outside
1230 the pattern. See L</Warning on \1 Instead of $1> below for details.
1234 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
1236 /(.)\g1/ # find first doubled char
1237 and print "'$1' is the first doubled character\n";
1239 /(?<char>.)\k<char>/ # ... a different way
1240 and print "'$+{char}' is the first doubled character\n";
1242 /(?'char'.)\g1/ # ... mix and match
1243 and print "'$1' is the first doubled character\n";
1245 if (/Time: (..):(..):(..)/) { # parse out values
1251 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference
1252 /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal
1253 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference
1254 /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal
1256 $a = '(.)\1'; # Creates problems when concatenated.
1257 $b = '(.)\g{1}'; # Avoids the problems.
1258 "aa" =~ /${a}/; # True
1259 "aa" =~ /${b}/; # True
1260 "aa0" =~ /${a}0/; # False!
1261 "aa0" =~ /${b}0/; # True
1262 "aa\x08" =~ /${a}0/; # True!
1263 "aa\x08" =~ /${b}0/; # False
1265 Several special variables also refer back to portions of the previous
1266 match. C<$+> returns whatever the last bracket match matched.
1267 C<$&> returns the entire matched string. (At one point C<$0> did
1268 also, but now it returns the name of the program.) C<$`> returns
1269 everything before the matched string. C<$'> returns everything
1270 after the matched string. And C<$^N> contains whatever was matched by
1271 the most-recently closed group (submatch). C<$^N> can be used in
1272 extended patterns (see below), for example to assign a submatch to a
1274 X<$+> X<$^N> X<$&> X<$`> X<$'>
1276 These special variables, like the C<%+> hash and the numbered match variables
1277 (C<$1>, C<$2>, C<$3>, I<etc>.) are dynamically scoped
1278 until the end of the enclosing block or until the next successful
1279 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
1280 X<$+> X<$^N> X<$&> X<$`> X<$'>
1281 X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
1284 The C<@{^CAPTURE}> array may be used to access ALL of the capture buffers
1285 as an array without needing to know how many there are. For instance
1287 $string=~/$pattern/ and @captured = @{^CAPTURE};
1289 will place a copy of each capture variable, C<$1>, C<$2> etc, into the
1292 Be aware that when interpolating a subscript of the C<@{^CAPTURE}>
1293 array you must use demarcated curly brace notation:
1295 print "@{^CAPTURE[0]}";
1297 See L<perldata/"Demarcated variable names using braces"> for more on
1300 B<NOTE>: Failed matches in Perl do not reset the match variables,
1301 which makes it easier to write code that tests for a series of more
1302 specific cases and remembers the best match.
1304 B<WARNING>: If your code is to run on Perl 5.16 or earlier,
1305 beware that once Perl sees that you need one of C<$&>, C<$`>, or
1306 C<$'> anywhere in the program, it has to provide them for every
1307 pattern match. This may substantially slow your program.
1309 Perl uses the same mechanism to produce C<$1>, C<$2>, I<etc>, so you also
1310 pay a price for each pattern that contains capturing parentheses.
1311 (To avoid this cost while retaining the grouping behaviour, use the
1312 extended regular expression C<(?: ... )> instead.) But if you never
1313 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
1314 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
1315 if you can, but if you can't (and some algorithms really appreciate
1316 them), once you've used them once, use them at will, because you've
1317 already paid the price.
1320 Perl 5.16 introduced a slightly more efficient mechanism that notes
1321 separately whether each of C<$`>, C<$&>, and C<$'> have been seen, and
1322 thus may only need to copy part of the string. Perl 5.20 introduced a
1323 much more efficient copy-on-write mechanism which eliminates any slowdown.
1325 As another workaround for this problem, Perl 5.10.0 introduced C<${^PREMATCH}>,
1326 C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&>
1327 and C<$'>, B<except> that they are only guaranteed to be defined after a
1328 successful match that was executed with the C</p> (preserve) modifier.
1329 The use of these variables incurs no global performance penalty, unlike
1330 their punctuation character equivalents, however at the trade-off that you
1331 have to tell perl when you want to use them. As of Perl 5.20, these three
1332 variables are equivalent to C<$`>, C<$&> and C<$'>, and C</p> is ignored.
1335 =head2 Quoting metacharacters
1337 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
1338 C<\w>, C<\n>. Unlike some other regular expression languages, there
1339 are no backslashed symbols that aren't alphanumeric. So anything
1340 that looks like C<\\>, C<\(>, C<\)>, C<\[>, C<\]>, C<\{>, or C<\}> is
1342 interpreted as a literal character, not a metacharacter. This was
1343 once used in a common idiom to disable or quote the special meanings
1344 of regular expression metacharacters in a string that you want to
1345 use for a pattern. Simply quote all non-"word" characters:
1347 $pattern =~ s/(\W)/\\$1/g;
1349 (If C<use locale> is set, then this depends on the current locale.)
1350 Today it is more common to use the C<L<quotemeta()|perlfunc/quotemeta>>
1351 function or the C<\Q> metaquoting escape sequence to disable all
1352 metacharacters' special meanings like this:
1354 /$unquoted\Q$quoted\E$unquoted/
1356 Beware that if you put literal backslashes (those not inside
1357 interpolated variables) between C<\Q> and C<\E>, double-quotish
1358 backslash interpolation may lead to confusing results. If you
1359 I<need> to use literal backslashes within C<\Q...\E>,
1360 consult L<perlop/"Gory details of parsing quoted constructs">.
1362 C<quotemeta()> and C<\Q> are fully described in L<perlfunc/quotemeta>.
1364 =head2 Extended Patterns
1366 Perl also defines a consistent extension syntax for features not
1367 found in standard tools like B<awk> and
1368 B<lex>. The syntax for most of these is a
1369 pair of parentheses with a question mark as the first thing within
1370 the parentheses. The character after the question mark indicates
1373 A question mark was chosen for this and for the minimal-matching
1374 construct because 1) question marks are rare in older regular
1375 expressions, and 2) whenever you see one, you should stop and
1376 "question" exactly what is going on. That's psychology....
1380 =item C<(?#I<text>)>
1383 A comment. The I<text> is ignored.
1384 Note that Perl closes
1385 the comment as soon as it sees a C<")">, so there is no way to put a literal
1386 C<")"> in the comment. The pattern's closing delimiter must be escaped by
1387 a backslash if it appears in the comment.
1389 See L</E<sol>x> for another way to have comments in patterns.
1391 Note that a comment can go just about anywhere, except in the middle of
1392 an escape sequence. Examples:
1394 qr/foo(?#comment)bar/' # Matches 'foobar'
1396 # The pattern below matches 'abcd', 'abccd', or 'abcccd'
1397 qr/abc(?#comment between literal and its quantifier){1,3}d/
1399 # The pattern below generates a syntax error, because the '\p' must
1400 # be followed immediately by a '{'.
1401 qr/\p(?#comment between \p and its property name){Any}/
1403 # The pattern below generates a syntax error, because the initial
1404 # '\(' is a literal opening parenthesis, and so there is nothing
1405 # for the closing ')' to match
1406 qr/\(?#the backslash means this isn't a comment)p{Any}/
1408 # Comments can be used to fold long patterns into multiple lines
1409 qr/First part of a long regex(?#
1412 =item C<(?adlupimnsx-imnsx)>
1414 =item C<(?^alupimnsx)>
1417 Zero or more embedded pattern-match modifiers, to be turned on (or
1418 turned off if preceded by C<"-">) for the remainder of the pattern or
1419 the remainder of the enclosing pattern group (if any).
1421 This is particularly useful for dynamically-generated patterns,
1422 such as those read in from a
1423 configuration file, taken from an argument, or specified in a table
1424 somewhere. Consider the case where some patterns want to be
1425 case-sensitive and some do not: The case-insensitive ones merely need to
1426 include C<(?i)> at the front of the pattern. For example:
1428 $pattern = "foobar";
1429 if ( /$pattern/i ) { }
1433 $pattern = "(?i)foobar";
1434 if ( /$pattern/ ) { }
1436 These modifiers are restored at the end of the enclosing group. For example,
1438 ( (?i) blah ) \s+ \g1
1440 will match C<blah> in any case, some spaces, and an exact (I<including the case>!)
1441 repetition of the previous word, assuming the C</x> modifier, and no C</i>
1442 modifier outside this group.
1444 These modifiers do not carry over into named subpatterns called in the
1445 enclosing group. In other words, a pattern such as C<((?i)(?&I<NAME>))> does not
1446 change the case-sensitivity of the I<NAME> pattern.
1448 A modifier is overridden by later occurrences of this construct in the
1449 same scope containing the same modifier, so that
1451 /((?im)foo(?-m)bar)/
1453 matches all of C<foobar> case insensitively, but uses C</m> rules for
1454 only the C<foo> portion. The C<"a"> flag overrides C<aa> as well;
1455 likewise C<aa> overrides C<"a">. The same goes for C<"x"> and C<xx>.
1460 both C</x> and C</xx> are turned off during matching C<foo>. And in
1464 C</x> but NOT C</xx> is turned on for matching C<foo>. (One might
1465 mistakenly think that since the inner C<(?x)> is already in the scope of
1466 C</x>, that the result would effectively be the sum of them, yielding
1467 C</xx>. It doesn't work that way.) Similarly, doing something like
1468 C<(?xx-x)foo> turns off all C<"x"> behavior for matching C<foo>, it is not
1469 that you subtract 1 C<"x"> from 2 to get 1 C<"x"> remaining.
1471 Any of these modifiers can be set to apply globally to all regular
1472 expressions compiled within the scope of a C<use re>. See
1473 L<re/"'/flags' mode">.
1475 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
1476 after the C<"?"> is a shorthand equivalent to C<d-imnsx>. Flags (except
1477 C<"d">) may follow the caret to override it.
1478 But a minus sign is not legal with it.
1480 Note that the C<"a">, C<"d">, C<"l">, C<"p">, and C<"u"> modifiers are special in
1481 that they can only be enabled, not disabled, and the C<"a">, C<"d">, C<"l">, and
1482 C<"u"> modifiers are mutually exclusive: specifying one de-specifies the
1483 others, and a maximum of one (or two C<"a">'s) may appear in the
1484 construct. Thus, for
1485 example, C<(?-p)> will warn when compiled under C<use warnings>;
1486 C<(?-d:...)> and C<(?dl:...)> are fatal errors.
1488 Note also that the C<"p"> modifier is special in that its presence
1489 anywhere in a pattern has a global effect.
1491 Having zero modifiers makes this a no-op (so why did you specify it,
1492 unless it's generated code), and starting in v5.30, warns under L<C<use
1493 re 'strict'>|re/'strict' mode>.
1495 =item C<(?:I<pattern>)>
1498 =item C<(?adluimnsx-imnsx:I<pattern>)>
1500 =item C<(?^aluimnsx:I<pattern>)>
1503 This is for clustering, not capturing; it groups subexpressions like
1504 C<"()">, but doesn't make backreferences as C<"()"> does. So
1506 @fields = split(/\b(?:a|b|c)\b/)
1508 matches the same field delimiters as
1510 @fields = split(/\b(a|b|c)\b/)
1512 but doesn't spit out the delimiters themselves as extra fields (even though
1513 that's the behaviour of L<perlfunc/split> when its pattern contains capturing
1514 groups). It's also cheaper not to capture
1515 characters if you don't need to.
1517 Any letters between C<"?"> and C<":"> act as flags modifiers as with
1518 C<(?adluimnsx-imnsx)>. For example,
1520 /(?s-i:more.*than).*million/i
1522 is equivalent to the more verbose
1524 /(?:(?s-i)more.*than).*million/i
1526 Note that any C<()> constructs enclosed within this one will still
1527 capture unless the C</n> modifier is in effect.
1529 Like the L</(?adlupimnsx-imnsx)> construct, C<aa> and C<"a"> override each
1530 other, as do C<xx> and C<"x">. They are not additive. So, doing
1531 something like C<(?xx-x:foo)> turns off all C<"x"> behavior for matching
1534 Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately
1535 after the C<"?"> is a shorthand equivalent to C<d-imnsx>. Any positive
1536 flags (except C<"d">) may follow the caret, so
1544 The caret tells Perl that this cluster doesn't inherit the flags of any
1545 surrounding pattern, but uses the system defaults (C<d-imnsx>),
1546 modified by any flags specified.
1548 The caret allows for simpler stringification of compiled regular
1549 expressions. These look like
1553 with any non-default flags appearing between the caret and the colon.
1554 A test that looks at such stringification thus doesn't need to have the
1555 system default flags hard-coded in it, just the caret. If new flags are
1556 added to Perl, the meaning of the caret's expansion will change to include
1557 the default for those flags, so the test will still work, unchanged.
1559 Specifying a negative flag after the caret is an error, as the flag is
1562 Mnemonic for C<(?^...)>: A fresh beginning since the usual use of a caret is
1563 to match at the beginning.
1565 =item C<(?|I<pattern>)>
1566 X<(?|)> X<Branch reset>
1568 This is the "branch reset" pattern, which has the special property
1569 that the capture groups are numbered from the same starting point
1570 in each alternation branch. It is available starting from perl 5.10.0.
1572 Capture groups are numbered from left to right, but inside this
1573 construct the numbering is restarted for each branch.
1575 The numbering within each branch will be as normal, and any groups
1576 following this construct will be numbered as though the construct
1577 contained only one branch, that being the one with the most capture
1580 This construct is useful when you want to capture one of a
1581 number of alternative matches.
1583 Consider the following pattern. The numbers underneath show in
1584 which group the captured content will be stored.
1587 # before ---------------branch-reset----------- after
1588 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1591 Be careful when using the branch reset pattern in combination with
1592 named captures. Named captures are implemented as being aliases to
1593 numbered groups holding the captures, and that interferes with the
1594 implementation of the branch reset pattern. If you are using named
1595 captures in a branch reset pattern, it's best to use the same names,
1596 in the same order, in each of the alternations:
1598 /(?| (?<a> x ) (?<b> y )
1599 | (?<a> z ) (?<b> w )) /x
1601 Not doing so may lead to surprises:
1603 "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x;
1604 say $+{a}; # Prints '12'
1605 say $+{b}; # *Also* prints '12'.
1607 The problem here is that both the group named C<< a >> and the group
1608 named C<< b >> are aliases for the group belonging to C<< $1 >>.
1610 =item Lookaround Assertions
1611 X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
1613 Lookaround assertions are zero-width patterns which match a specific
1614 pattern without including it in C<$&>. Positive assertions match when
1615 their subpattern matches, negative assertions match when their subpattern
1616 fails. Lookbehind matches text up to the current match position,
1617 lookahead matches text following the current match position.
1621 =item C<(?=I<pattern>)>
1623 =item C<(*pla:I<pattern>)>
1625 =item C<(*positive_lookahead:I<pattern>)>
1628 X<(*positive_lookahead>
1629 X<look-ahead, positive> X<lookahead, positive>
1631 A zero-width positive lookahead assertion. For example, C</\w+(?=\t)/>
1632 matches a word followed by a tab, without including the tab in C<$&>.
1634 =item C<(?!I<pattern>)>
1636 =item C<(*nla:I<pattern>)>
1638 =item C<(*negative_lookahead:I<pattern>)>
1641 X<(*negative_lookahead>
1642 X<look-ahead, negative> X<lookahead, negative>
1644 A zero-width negative lookahead assertion. For example C</foo(?!bar)/>
1645 matches any occurrence of "foo" that isn't followed by "bar". Note
1646 however that lookahead and lookbehind are NOT the same thing. You cannot
1647 use this for lookbehind.
1649 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
1650 will not do what you want. That's because the C<(?!foo)> is just saying that
1651 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
1652 match. Use lookbehind instead (see below).
1654 =item C<(?<=I<pattern>)>
1658 =item C<(*plb:I<pattern>)>
1660 =item C<(*positive_lookbehind:I<pattern>)>
1663 X<(*positive_lookbehind>
1664 X<look-behind, positive> X<lookbehind, positive> X<\K>
1666 A zero-width positive lookbehind assertion. For example, C</(?<=\t)\w+/>
1667 matches a word that follows a tab, without including the tab in C<$&>.
1669 Prior to Perl 5.30, it worked only for fixed-width lookbehind, but
1670 starting in that release, it can handle variable lengths from 1 to 255
1671 characters as an experimental feature. The feature is enabled
1672 automatically if you use a variable length positive lookbehind assertion.
1674 In Perl 5.35.10 the scope of the experimental nature of this construct
1675 has been reduced, and experimental warnings will only be produced when
1676 the construct contains capturing parenthesis. The warnings will be
1677 raised at pattern compilation time, unless turned off, in the
1678 C<experimental::vlb> category. This is to warn you that the exact
1679 contents of capturing buffers in a variable length positive lookbehind
1680 is not well defined and is subject to change in a future release of perl.
1682 Currently if you use capture buffers inside of a positive variable length
1683 lookbehind the result will be the longest and thus leftmost match possible.
1686 "aax" =~ /(?=x)(?<=(a|aa))/
1687 "aax" =~ /(?=x)(?<=(aa|a))/
1688 "aax" =~ /(?=x)(?<=(a{1,2}?)/
1689 "aax" =~ /(?=x)(?<=(a{1,2})/
1691 will all result in C<$1> containing C<"aa">. It is possible in a future
1692 release of perl we will change this behavior.
1694 There is a special form of this construct, called C<\K>
1695 (available since Perl 5.10.0), which causes the
1696 regex engine to "keep" everything it had matched prior to the C<\K> and
1697 not include it in C<$&>. This effectively provides non-experimental
1698 variable-length lookbehind of any length.
1700 And, there is a technique that can be used to handle variable length
1701 lookbehinds on earlier releases, and longer than 255 characters. It is
1703 L<http://www.drregex.com/2019/02/variable-length-lookbehinds-actually.html>.
1705 Note that under C</i>, a few single characters match two or three other
1706 characters. This makes them variable length, and the 255 length applies
1707 to the maximum number of characters in the match. For
1708 example C<qr/\N{LATIN SMALL LETTER SHARP S}/i> matches the sequence
1709 C<"ss">. Your lookbehind assertion could contain 127 Sharp S
1710 characters under C</i>, but adding a 128th would generate a compilation
1711 error, as that could match 256 C<"s"> characters in a row.
1713 The use of C<\K> inside of another lookaround assertion
1714 is allowed, but the behaviour is currently not well defined.
1716 For various reasons C<\K> may be significantly more efficient than the
1717 equivalent C<< (?<=...) >> construct, and it is especially useful in
1718 situations where you want to efficiently remove something following
1719 something else in a string. For instance
1723 can be rewritten as the much more efficient
1727 Use of the non-greedy modifier C<"?"> may not give you the expected
1728 results if it is within a capturing group within the construct.
1730 =item C<(?<!I<pattern>)>
1732 =item C<(*nlb:I<pattern>)>
1734 =item C<(*negative_lookbehind:I<pattern>)>
1737 X<(*negative_lookbehind>
1738 X<look-behind, negative> X<lookbehind, negative>
1740 A zero-width negative lookbehind assertion. For example C</(?<!bar)foo/>
1741 matches any occurrence of "foo" that does not follow "bar".
1743 Prior to Perl 5.30, it worked only for fixed-width lookbehind, but
1744 starting in that release, it can handle variable lengths from 1 to 255
1745 characters as an experimental feature. The feature is enabled
1746 automatically if you use a variable length negative lookbehind assertion.
1748 In Perl 5.35.10 the scope of the experimental nature of this construct
1749 has been reduced, and experimental warnings will only be produced when
1750 the construct contains capturing parentheses. The warnings will be
1751 raised at pattern compilation time, unless turned off, in the
1752 C<experimental::vlb> category. This is to warn you that the exact
1753 contents of capturing buffers in a variable length negative lookbehind
1754 is not well defined and is subject to change in a future release of perl.
1756 Currently if you use capture buffers inside of a negative variable length
1757 lookbehind the result may not be what you expect, for instance:
1759 say "axfoo"=~/(?=foo)(?<!(a|ax)(?{ say $1 }))/ ? "y" : "n";
1761 will output the following:
1766 which does not make sense as this should print out "ax" as the "a" does
1767 not line up at the correct place. Another example would be:
1769 say "yes: '$1-$2'" if "aayfoo"=~/(?=foo)(?<!(a|aa)(a|aa)x)/;
1771 will output the following:
1775 It is possible in a future release of perl we will change this behavior
1776 so both of these examples produced more reasonable output.
1778 Note that we are confident that the construct will match and reject
1779 patterns appropriately, the undefined behavior strictly relates to the
1780 value of the capture buffer during or after matching.
1782 There is a technique that can be used to handle variable length
1783 lookbehind on earlier releases, and longer than 255 characters. It is
1785 L<http://www.drregex.com/2019/02/variable-length-lookbehinds-actually.html>.
1787 Note that under C</i>, a few single characters match two or three other
1788 characters. This makes them variable length, and the 255 length applies
1789 to the maximum number of characters in the match. For
1790 example C<qr/\N{LATIN SMALL LETTER SHARP S}/i> matches the sequence
1791 C<"ss">. Your lookbehind assertion could contain 127 Sharp S
1792 characters under C</i>, but adding a 128th would generate a compilation
1793 error, as that could match 256 C<"s"> characters in a row.
1795 Use of the non-greedy modifier C<"?"> may not give you the expected
1796 results if it is within a capturing group within the construct.
1800 =item C<< (?<I<NAME>>I<pattern>) >>
1802 =item C<(?'I<NAME>'I<pattern>)>
1803 X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
1805 A named capture group. Identical in every respect to normal capturing
1806 parentheses C<()> but for the additional fact that the group
1807 can be referred to by name in various regular expression
1808 constructs (like C<\g{I<NAME>}>) and can be accessed by name
1809 after a successful match via C<%+> or C<%->. See L<perlvar>
1810 for more details on the C<%+> and C<%-> hashes.
1812 If multiple distinct capture groups have the same name, then
1813 C<$+{I<NAME>}> will refer to the leftmost defined group in the match.
1815 The forms C<(?'I<NAME>'I<pattern>)> and C<< (?<I<NAME>>I<pattern>) >>
1818 B<NOTE:> While the notation of this construct is the same as the similar
1819 function in .NET regexes, the behavior is not. In Perl the groups are
1820 numbered sequentially regardless of being named or not. Thus in the
1825 C<$+{foo}> will be the same as C<$2>, and C<$3> will contain 'z' instead of
1826 the opposite which is what a .NET regex hacker might expect.
1828 Currently I<NAME> is restricted to simple identifiers only.
1829 In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
1830 its Unicode extension (see L<utf8>),
1831 though it isn't extended by the locale (see L<perllocale>).
1833 B<NOTE:> In order to make things easier for programmers with experience
1834 with the Python or PCRE regex engines, the pattern C<<
1835 (?PE<lt>I<NAME>E<gt>I<pattern>) >>
1836 may be used instead of C<< (?<I<NAME>>I<pattern>) >>; however this form does not
1837 support the use of single quotes as a delimiter for the name.
1839 =item C<< \k<I<NAME>> >>
1841 =item C<< \k'I<NAME>' >>
1843 =item C<< \k{I<NAME>} >>
1845 Named backreference. Similar to numeric backreferences, except that
1846 the group is designated by name and not number. If multiple groups
1847 have the same name then it refers to the leftmost defined group in
1850 It is an error to refer to a name not defined by a C<< (?<I<NAME>>) >>
1851 earlier in the pattern.
1853 All three forms are equivalent, although with C<< \k{ I<NAME> } >>,
1854 you may optionally have blanks within but adjacent to the braces, as
1857 B<NOTE:> In order to make things easier for programmers with experience
1858 with the Python or PCRE regex engines, the pattern C<< (?P=I<NAME>) >>
1859 may be used instead of C<< \k<I<NAME>> >>.
1861 =item C<(?{ I<code> })>
1862 X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
1864 B<WARNING>: Using this feature safely requires that you understand its
1865 limitations. Code executed that has side effects may not perform identically
1866 from version to version due to the effect of future optimisations in the regex
1867 engine. For more information on this, see L</Embedded Code Execution
1870 This zero-width assertion executes any embedded Perl code. It always
1871 succeeds, and its return value is set as C<$^R>.
1873 In literal patterns, the code is parsed at the same time as the
1874 surrounding code. While within the pattern, control is passed temporarily
1875 back to the perl parser, until the logically-balancing closing brace is
1876 encountered. This is similar to the way that an array index expression in
1877 a literal string is handled, for example
1879 "abc$array[ 1 + f('[') + g()]def"
1881 In particular, braces do not need to be balanced:
1883 s/abc(?{ f('{'); })/def/
1885 Even in a pattern that is interpolated and compiled at run-time, literal
1886 code blocks will be compiled once, at perl compile time; the following
1890 my $qr = qr/(?{ BEGIN { print "A" } })/;
1892 /$foo$qr(?{ BEGIN { print "B" } })/;
1895 In patterns where the text of the code is derived from run-time
1896 information rather than appearing literally in a source code /pattern/,
1897 the code is compiled at the same time that the pattern is compiled, and
1898 for reasons of security, C<use re 'eval'> must be in scope. This is to
1899 stop user-supplied patterns containing code snippets from being
1902 In situations where you need to enable this with C<use re 'eval'>, you should
1903 also have taint checking enabled. Better yet, use the carefully
1904 constrained evaluation within a Safe compartment. See L<perlsec> for
1905 details about both these mechanisms.
1907 From the viewpoint of parsing, lexical variable scope and closures,
1911 behaves approximately like
1913 /AAA/ && do { BBB } && /CCC/
1917 qr/AAA(?{ BBB })CCC/
1919 behaves approximately like
1921 sub { /AAA/ && do { BBB } && /CCC/ }
1925 { my $i = 1; $r = qr/(?{ print $i })/ }
1929 Inside a C<(?{...})> block, C<$_> refers to the string the regular
1930 expression is matching against. You can also use C<pos()> to know what is
1931 the current position of matching within this string.
1933 The code block introduces a new scope from the perspective of lexical
1934 variable declarations, but B<not> from the perspective of C<local> and
1935 similar localizing behaviours. So later code blocks within the same
1936 pattern will still see the values which were localized in earlier blocks.
1937 These accumulated localizations are undone either at the end of a
1938 successful match, or if the assertion is backtracked (compare
1939 L</"Backtracking">). For example,
1943 (?{ $cnt = 0 }) # Initialize $cnt.
1947 local $cnt = $cnt + 1; # Update $cnt,
1948 # backtracking-safe.
1952 (?{ $res = $cnt }) # On success copy to
1953 # non-localized location.
1956 will initially increment C<$cnt> up to 8; then during backtracking, its
1957 value will be unwound back to 4, which is the value assigned to C<$res>.
1958 At the end of the regex execution, C<$cnt> will be wound back to its initial
1961 This assertion may be used as the condition in a
1963 (?(condition)yes-pattern|no-pattern)
1965 switch. If I<not> used in this way, the result of evaluation of I<code>
1966 is put into the special variable C<$^R>. This happens immediately, so
1967 C<$^R> can be used from other C<(?{ I<code> })> assertions inside the same
1970 The assignment to C<$^R> above is properly localized, so the old
1971 value of C<$^R> is restored if the assertion is backtracked; compare
1974 Note that the special variable C<$^N> is particularly useful with code
1975 blocks to capture the results of submatches in variables without having to
1976 keep track of the number of nested parentheses. For example:
1978 $_ = "The brown fox jumps over the lazy dog";
1979 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
1980 print "color = $color, animal = $animal\n";
1983 =item C<(??{ I<code> })>
1985 X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
1987 B<WARNING>: Using this feature safely requires that you understand its
1988 limitations. Code executed that has side effects may not perform
1989 identically from version to version due to the effect of future
1990 optimisations in the regex engine. For more information on this, see
1991 L</Embedded Code Execution Frequency>.
1993 This is a "postponed" regular subexpression. It behaves in I<exactly> the
1994 same way as a C<(?{ I<code> })> code block as described above, except that
1995 its return value, rather than being assigned to C<$^R>, is treated as a
1996 pattern, compiled if it's a string (or used as-is if its a qr// object),
1997 then matched as if it were inserted instead of this construct.
1999 During the matching of this sub-pattern, it has its own set of
2000 captures which are valid during the sub-match, but are discarded once
2001 control returns to the main pattern. For example, the following matches,
2002 with the inner pattern capturing "B" and matching "BB", while the outer
2003 pattern captures "A";
2005 my $inner = '(.)\1';
2006 "ABBA" =~ /^(.)(??{ $inner })\1/;
2007 print $1; # prints "A";
2009 Note that this means that there is no way for the inner pattern to refer
2010 to a capture group defined outside. (The code block itself can use C<$1>,
2011 I<etc>., to refer to the enclosing pattern's capture groups.) Thus, although
2013 ('a' x 100)=~/(??{'(.)' x 100})/
2015 I<will> match, it will I<not> set C<$1> on exit.
2017 The following pattern matches a parenthesized group:
2022 (?> [^()]+ ) # Non-parens without backtracking
2024 (??{ $re }) # Group with matching parens
2030 L<C<(?I<PARNO>)>|/(?I<PARNO>) (?-I<PARNO>) (?+I<PARNO>) (?R) (?0)>
2031 for a different, more efficient way to accomplish
2034 Executing a postponed regular expression too many times without
2035 consuming any input string will also result in a fatal error. The depth
2036 at which that happens is compiled into perl, so it can be changed with a
2039 =item C<(?I<PARNO>)> C<(?-I<PARNO>)> C<(?+I<PARNO>)> C<(?R)> C<(?0)>
2040 X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
2041 X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
2042 X<regex, relative recursion> X<GOSUB> X<GOSTART>
2044 Recursive subpattern. Treat the contents of a given capture buffer in the
2045 current pattern as an independent subpattern and attempt to match it at
2046 the current position in the string. Information about capture state from
2047 the caller for things like backreferences is available to the subpattern,
2048 but capture buffers set by the subpattern are not visible to the caller.
2050 Similar to C<(??{ I<code> })> except that it does not involve executing any
2051 code or potentially compiling a returned pattern string; instead it treats
2052 the part of the current pattern contained within a specified capture group
2053 as an independent pattern that must match at the current position. Also
2054 different is the treatment of capture buffers, unlike C<(??{ I<code> })>
2055 recursive patterns have access to their caller's match state, so one can
2056 use backreferences safely.
2058 I<PARNO> is a sequence of digits (not starting with 0) whose value reflects
2059 the paren-number of the capture group to recurse to. C<(?R)> recurses to
2060 the beginning of the whole pattern. C<(?0)> is an alternate syntax for
2061 C<(?R)>. If I<PARNO> is preceded by a plus or minus sign then it is assumed
2062 to be relative, with negative numbers indicating preceding capture groups
2063 and positive ones following. Thus C<(?-1)> refers to the most recently
2064 declared group, and C<(?+1)> indicates the next group to be declared.
2065 Note that the counting for relative recursion differs from that of
2066 relative backreferences, in that with recursion unclosed groups B<are>
2069 The following pattern matches a function C<foo()> which may contain
2070 balanced parentheses as the argument.
2072 $re = qr{ ( # paren group 1 (full function)
2074 ( # paren group 2 (parens)
2076 ( # paren group 3 (contents of parens)
2078 (?> [^()]+ ) # Non-parens without backtracking
2080 (?2) # Recurse to start of paren group 2
2088 If the pattern was used as follows
2090 'foo(bar(baz)+baz(bop))'=~/$re/
2091 and print "\$1 = $1\n",
2095 the output produced should be the following:
2097 $1 = foo(bar(baz)+baz(bop))
2098 $2 = (bar(baz)+baz(bop))
2099 $3 = bar(baz)+baz(bop)
2101 If there is no corresponding capture group defined, then it is a
2102 fatal error. Recursing deeply without consuming any input string will
2103 also result in a fatal error. The depth at which that happens is
2104 compiled into perl, so it can be changed with a custom build.
2106 The following shows how using negative indexing can make it
2107 easier to embed recursive patterns inside of a C<qr//> construct
2110 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
2111 if (/foo $parens \s+ \+ \s+ bar $parens/x) {
2112 # do something here...
2115 B<Note> that this pattern does not behave the same way as the equivalent
2116 PCRE or Python construct of the same form. In Perl you can backtrack into
2117 a recursed group, in PCRE and Python the recursed into group is treated
2118 as atomic. Also, modifiers are resolved at compile time, so constructs
2119 like C<(?i:(?1))> or C<(?:(?i)(?1))> do not affect how the sub-pattern will
2122 =item C<(?&I<NAME>)>
2125 Recurse to a named subpattern. Identical to C<(?I<PARNO>)> except that the
2126 parenthesis to recurse to is determined by name. If multiple parentheses have
2127 the same name, then it recurses to the leftmost.
2129 It is an error to refer to a name that is not declared somewhere in the
2132 B<NOTE:> In order to make things easier for programmers with experience
2133 with the Python or PCRE regex engines the pattern C<< (?P>I<NAME>) >>
2134 may be used instead of C<< (?&I<NAME>) >>.
2136 =item C<(?(I<condition>)I<yes-pattern>|I<no-pattern>)>
2139 =item C<(?(I<condition>)I<yes-pattern>)>
2141 Conditional expression. Matches I<yes-pattern> if I<condition> yields
2142 a true value, matches I<no-pattern> otherwise. A missing pattern always
2145 C<(I<condition>)> should be one of:
2149 =item an integer in parentheses
2151 (which is valid if the corresponding pair of parentheses
2154 =item a lookahead/lookbehind/evaluate zero-width assertion;
2156 =item a name in angle brackets or single quotes
2158 (which is valid if a group with the given name matched);
2160 =item the special symbol C<(R)>
2162 (true when evaluated inside of recursion or eval). Additionally the
2164 followed by a number, (which will be true when evaluated when recursing
2165 inside of the appropriate group), or by C<&I<NAME>>, in which case it will
2166 be true only when evaluated during recursion in the named group.
2170 Here's a summary of the possible predicates:
2174 =item C<(1)> C<(2)> ...
2176 Checks if the numbered capturing group has matched something.
2177 Full syntax: C<< (?(1)then|else) >>
2179 =item C<(E<lt>I<NAME>E<gt>)> C<('I<NAME>')>
2181 Checks if a group with the given name has matched something.
2182 Full syntax: C<< (?(<name>)then|else) >>
2184 =item C<(?=...)> C<(?!...)> C<(?<=...)> C<(?<!...)>
2186 Checks whether the pattern matches (or does not match, for the C<"!">
2188 Full syntax: C<< (?(?=I<lookahead>)I<then>|I<else>) >>
2190 =item C<(?{ I<CODE> })>
2192 Treats the return value of the code block as the condition.
2193 Full syntax: C<< (?(?{ I<code> })I<then>|I<else>) >>
2197 Checks if the expression has been evaluated inside of recursion.
2198 Full syntax: C<< (?(R)I<then>|I<else>) >>
2200 =item C<(R1)> C<(R2)> ...
2202 Checks if the expression has been evaluated while executing directly
2203 inside of the n-th capture group. This check is the regex equivalent of
2205 if ((caller(0))[3] eq 'subname') { ... }
2207 In other words, it does not check the full recursion stack.
2209 Full syntax: C<< (?(R1)I<then>|I<else>) >>
2211 =item C<(R&I<NAME>)>
2213 Similar to C<(R1)>, this predicate checks to see if we're executing
2214 directly inside of the leftmost group with a given name (this is the same
2215 logic used by C<(?&I<NAME>)> to disambiguate). It does not check the full
2216 stack, but only the name of the innermost active recursion.
2217 Full syntax: C<< (?(R&I<name>)I<then>|I<else>) >>
2221 In this case, the yes-pattern is never directly executed, and no
2222 no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
2223 See below for details.
2224 Full syntax: C<< (?(DEFINE)I<definitions>...) >>
2235 matches a chunk of non-parentheses, possibly included in parentheses
2238 A special form is the C<(DEFINE)> predicate, which never executes its
2239 yes-pattern directly, and does not allow a no-pattern. This allows one to
2240 define subpatterns which will be executed only by the recursion mechanism.
2241 This way, you can define a set of regular expression rules that can be
2242 bundled into any pattern you choose.
2244 It is recommended that for this usage you put the DEFINE block at the
2245 end of the pattern, and that you name any subpatterns defined within it.
2247 Also, it's worth noting that patterns defined this way probably will
2248 not be as efficient, as the optimizer is not very clever about
2251 An example of how this might be used is as follows:
2253 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
2256 (?<ADDRESS_PAT>....)
2259 Note that capture groups matched inside of recursion are not accessible
2260 after the recursion returns, so the extra layer of capturing groups is
2261 necessary. Thus C<$+{NAME_PAT}> would not be defined even though
2262 C<$+{NAME}> would be.
2264 Finally, keep in mind that subpatterns created inside a DEFINE block
2265 count towards the absolute and relative number of captures, so this:
2267 my @captures = "a" =~ /(.) # First capture
2269 (?<EXAMPLE> 1 ) # Second capture
2271 say scalar @captures;
2273 Will output 2, not 1. This is particularly important if you intend to
2274 compile the definitions with the C<qr//> operator, and later
2275 interpolate them in another pattern.
2277 =item C<< (?>I<pattern>) >>
2279 =item C<< (*atomic:I<pattern>) >>
2282 X<backtrack> X<backtracking> X<atomic> X<possessive>
2284 An "independent" subexpression, one which matches the substring
2285 that a standalone I<pattern> would match if anchored at the given
2286 position, and it matches I<nothing other than this substring>. This
2287 construct is useful for optimizations of what would otherwise be
2288 "eternal" matches, because it will not backtrack (see L</"Backtracking">).
2289 It may also be useful in places where the "grab all you can, and do not
2290 give anything back" semantic is desirable.
2292 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
2293 (anchored at the beginning of string, as above) will match I<all>
2294 characters C<"a"> at the beginning of string, leaving no C<"a"> for
2295 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
2296 since the match of the subgroup C<a*> is influenced by the following
2297 group C<ab> (see L</"Backtracking">). In particular, C<a*> inside
2298 C<a*ab> will match fewer characters than a standalone C<a*>, since
2299 this makes the tail match.
2301 C<< (?>I<pattern>) >> does not disable backtracking altogether once it has
2302 matched. It is still possible to backtrack past the construct, but not
2303 into it. So C<< ((?>a*)|(?>b*))ar >> will still match "bar".
2305 An effect similar to C<< (?>I<pattern>) >> may be achieved by writing
2306 C<(?=(I<pattern>))\g{-1}>. This matches the same substring as a standalone
2307 C<a+>, and the following C<\g{-1}> eats the matched string; it therefore
2308 makes a zero-length assertion into an analogue of C<< (?>...) >>.
2309 (The difference between these two constructs is that the second one
2310 uses a capturing group, thus shifting ordinals of backreferences
2311 in the rest of a regular expression.)
2313 Consider this pattern:
2324 That will efficiently match a nonempty group with matching parentheses
2325 two levels deep or less. However, if there is no such group, it
2326 will take virtually forever on a long string. That's because there
2327 are so many different ways to split a long string into several
2328 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
2329 to a subpattern of the above pattern. Consider how the pattern
2330 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
2331 seconds, but that each extra letter doubles this time. This
2332 exponential performance will make it appear that your program has
2333 hung. However, a tiny change to this pattern
2337 (?> [^()]+ ) # change x+ above to (?> x+ )
2344 which uses C<< (?>...) >> matches exactly when the one above does (verifying
2345 this yourself would be a productive exercise), but finishes in a fourth
2346 the time when used on a similar string with 1000000 C<"a">s. Be aware,
2347 however, that, when this construct is followed by a
2348 quantifier, it currently triggers a warning message under
2349 the C<use warnings> pragma or B<-w> switch saying it
2350 C<"matches null string many times in regex">.
2352 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
2353 effect may be achieved by negative lookahead, as in C<[^()]+ (?! [^()] )>.
2354 This was only 4 times slower on a string with 1000000 C<"a">s.
2356 The "grab all you can, and do not give anything back" semantic is desirable
2357 in many situations where on the first sight a simple C<()*> looks like
2358 the correct solution. Suppose we parse text with comments being delimited
2359 by C<"#"> followed by some optional (horizontal) whitespace. Contrary to
2360 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
2361 the comment delimiter, because it may "give up" some whitespace if
2362 the remainder of the pattern can be made to match that way. The correct
2363 answer is either one of these:
2368 For example, to grab non-empty comments into C<$1>, one should use either
2371 / (?> \# [ \t]* ) ( .+ ) /x;
2372 / \# [ \t]* ( [^ \t] .* ) /x;
2374 Which one you pick depends on which of these expressions better reflects
2375 the above specification of comments.
2377 In some literature this construct is called "atomic matching" or
2378 "possessive matching".
2380 Possessive quantifiers are equivalent to putting the item they are applied
2381 to inside of one of these constructs. The following equivalences apply:
2383 Quantifier Form Bracketing Form
2384 --------------- ---------------
2388 PAT{min,max}+ (?>PAT{min,max})
2390 Nested C<(?E<gt>...)> constructs are not no-ops, even if at first glance
2391 they might seem to be. This is because the nested C<(?E<gt>...)> can
2392 restrict internal backtracking that otherwise might occur. For example,
2394 "abc" =~ /(?>a[bc]*c)/
2398 "abc" =~ /(?>a(?>[bc]*)c)/
2404 See L<perlrecharclass/Extended Bracketed Character Classes>.
2409 X<backtrack> X<backtracking>
2411 NOTE: This section presents an abstract approximation of regular
2412 expression behavior. For a more rigorous (and complicated) view of
2413 the rules involved in selecting a match among possible alternatives,
2414 see L</Combining RE Pieces>.
2416 A fundamental feature of regular expression matching involves the
2417 notion called I<backtracking>, which is currently used (when needed)
2418 by all regular non-possessive expression quantifiers, namely C<"*">,
2419 C<*?>, C<"+">, C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often
2420 optimized internally, but the general principle outlined here is valid.
2422 For a regular expression to match, the I<entire> regular expression must
2423 match, not just part of it. So if the beginning of a pattern containing a
2424 quantifier succeeds in a way that causes later parts in the pattern to
2425 fail, the matching engine backs up and recalculates the beginning
2426 part--that's why it's called backtracking.
2428 Here is an example of backtracking: Let's say you want to find the
2429 word following "foo" in the string "Food is on the foo table.":
2431 $_ = "Food is on the foo table.";
2432 if ( /\b(foo)\s+(\w+)/i ) {
2433 print "$2 follows $1.\n";
2436 When the match runs, the first part of the regular expression (C<\b(foo)>)
2437 finds a possible match right at the beginning of the string, and loads up
2438 C<$1> with "Foo". However, as soon as the matching engine sees that there's
2439 no whitespace following the "Foo" that it had saved in C<$1>, it realizes its
2440 mistake and starts over again one character after where it had the
2441 tentative match. This time it goes all the way until the next occurrence
2442 of "foo". The complete regular expression matches this time, and you get
2443 the expected output of "table follows foo."
2445 Sometimes minimal matching can help a lot. Imagine you'd like to match
2446 everything between "foo" and "bar". Initially, you write something
2449 $_ = "The food is under the bar in the barn.";
2450 if ( /foo(.*)bar/ ) {
2454 Which perhaps unexpectedly yields:
2456 got <d is under the bar in the >
2458 That's because C<.*> was greedy, so you get everything between the
2459 I<first> "foo" and the I<last> "bar". Here it's more effective
2460 to use minimal matching to make sure you get the text between a "foo"
2461 and the first "bar" thereafter.
2463 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
2464 got <d is under the >
2466 Here's another example. Let's say you'd like to match a number at the end
2467 of a string, and you also want to keep the preceding part of the match.
2470 $_ = "I have 2 numbers: 53147";
2471 if ( /(.*)(\d*)/ ) { # Wrong!
2472 print "Beginning is <$1>, number is <$2>.\n";
2475 That won't work at all, because C<.*> was greedy and gobbled up the
2476 whole string. As C<\d*> can match on an empty string the complete
2477 regular expression matched successfully.
2479 Beginning is <I have 2 numbers: 53147>, number is <>.
2481 Here are some variants, most of which don't work:
2483 $_ = "I have 2 numbers: 53147";
2496 printf "%-12s ", $pat;
2498 print "<$1> <$2>\n";
2504 That will print out:
2506 (.*)(\d*) <I have 2 numbers: 53147> <>
2507 (.*)(\d+) <I have 2 numbers: 5314> <7>
2509 (.*?)(\d+) <I have > <2>
2510 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
2511 (.*?)(\d+)$ <I have 2 numbers: > <53147>
2512 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
2513 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
2515 As you see, this can be a bit tricky. It's important to realize that a
2516 regular expression is merely a set of assertions that gives a definition
2517 of success. There may be 0, 1, or several different ways that the
2518 definition might succeed against a particular string. And if there are
2519 multiple ways it might succeed, you need to understand backtracking to
2520 know which variety of success you will achieve.
2522 When using lookahead assertions and negations, this can all get even
2523 trickier. Imagine you'd like to find a sequence of non-digits not
2524 followed by "123". You might try to write that as
2527 if ( /^\D*(?!123)/ ) { # Wrong!
2528 print "Yup, no 123 in $_\n";
2531 But that isn't going to match; at least, not the way you're hoping. It
2532 claims that there is no 123 in the string. Here's a clearer picture of
2533 why that pattern matches, contrary to popular expectations:
2538 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
2539 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
2541 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
2542 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
2550 You might have expected test 3 to fail because it seems to a more
2551 general purpose version of test 1. The important difference between
2552 them is that test 3 contains a quantifier (C<\D*>) and so can use
2553 backtracking, whereas test 1 will not. What's happening is
2554 that you've asked "Is it true that at the start of C<$x>, following 0 or more
2555 non-digits, you have something that's not 123?" If the pattern matcher had
2556 let C<\D*> expand to "ABC", this would have caused the whole pattern to
2559 The search engine will initially match C<\D*> with "ABC". Then it will
2560 try to match C<(?!123)> with "123", which fails. But because
2561 a quantifier (C<\D*>) has been used in the regular expression, the
2562 search engine can backtrack and retry the match differently
2563 in the hope of matching the complete regular expression.
2565 The pattern really, I<really> wants to succeed, so it uses the
2566 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
2567 time. Now there's indeed something following "AB" that is not
2568 "123". It's "C123", which suffices.
2570 We can deal with this by using both an assertion and a negation.
2571 We'll say that the first part in C<$1> must be followed both by a digit
2572 and by something that's not "123". Remember that the lookaheads
2573 are zero-width expressions--they only look, but don't consume any
2574 of the string in their match. So rewriting this way produces what
2575 you'd expect; that is, case 5 will fail, but case 6 succeeds:
2577 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
2578 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
2582 In other words, the two zero-width assertions next to each other work as though
2583 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
2584 matches only if you're at the beginning of the line AND the end of the
2585 line simultaneously. The deeper underlying truth is that juxtaposition in
2586 regular expressions always means AND, except when you write an explicit OR
2587 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
2588 although the attempted matches are made at different positions because "a"
2589 is not a zero-width assertion, but a one-width assertion.
2591 B<WARNING>: Particularly complicated regular expressions can take
2592 exponential time to solve because of the immense number of possible
2593 ways they can use backtracking to try for a match. For example, without
2594 internal optimizations done by the regular expression engine, this will
2595 take a painfully long time to run:
2597 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
2599 And if you used C<"*">'s in the internal groups instead of limiting them
2600 to 0 through 5 matches, then it would take forever--or until you ran
2601 out of stack space. Moreover, these internal optimizations are not
2602 always applicable. For example, if you put C<{0,5}> instead of C<"*">
2603 on the external group, no current optimization is applicable, and the
2604 match takes a long time to finish.
2606 A powerful tool for optimizing such beasts is what is known as an
2607 "independent group",
2608 which does not backtrack (see C<L</(?E<gt>pattern)>>). Note also that
2609 zero-length lookahead/lookbehind assertions will not backtrack to make
2610 the tail match, since they are in "logical" context: only
2611 whether they match is considered relevant. For an example
2612 where side-effects of lookahead I<might> have influenced the
2613 following match, see C<L</(?E<gt>pattern)>>.
2616 X<(*script_run:...)> X<(sr:...)>
2617 X<(*atomic_script_run:...)> X<(asr:...)>
2619 A script run is basically a sequence of characters, all from the same
2620 Unicode script (see L<perlunicode/Scripts>), such as Latin or Greek. In
2621 most places a single word would never be written in multiple scripts,
2622 unless it is a spoofing attack. An infamous example, is
2626 Those letters could all be Latin (as in the example just above), or they
2627 could be all Cyrillic (except for the dot), or they could be a mixture
2628 of the two. In the case of an internet address the C<.com> would be in
2629 Latin, And any Cyrillic ones would cause it to be a mixture, not a
2630 script run. Someone clicking on such a link would not be directed to
2631 the real Paypal website, but an attacker would craft a look-alike one to
2632 attempt to gather sensitive information from the person.
2634 Starting in Perl 5.28, it is now easy to detect strings that aren't
2635 script runs. Simply enclose just about any pattern like either of
2638 (*script_run:pattern)
2641 What happens is that after I<pattern> succeeds in matching, it is
2642 subjected to the additional criterion that every character in it must be
2643 from the same script (see exceptions below). If this isn't true,
2644 backtracking occurs until something all in the same script is found that
2645 matches, or all possibilities are exhausted. This can cause a lot of
2646 backtracking, but generally, only malicious input will result in this,
2647 though the slow down could cause a denial of service attack. If your
2648 needs permit, it is best to make the pattern atomic to cut down on the
2649 amount of backtracking. This is so likely to be what you want, that
2650 instead of writing this:
2652 (*script_run:(?>pattern))
2654 you can write either of these:
2656 (*atomic_script_run:pattern)
2659 (See C<L</(?E<gt>I<pattern>)>>.)
2661 In Taiwan, Japan, and Korea, it is common for text to have a mixture of
2662 characters from their native scripts and base Chinese. Perl follows
2663 Unicode's UTS 39 (L<https://unicode.org/reports/tr39/>) Unicode Security
2664 Mechanisms in allowing such mixtures. For example, the Japanese scripts
2665 Katakana and Hiragana are commonly mixed together in practice, along
2666 with some Chinese characters, and hence are treated as being in a single
2669 The rules used for matching decimal digits are slightly stricter. Many
2670 scripts have their own sets of digits equivalent to the Western C<0>
2671 through C<9> ones. A few, such as Arabic, have more than one set. For
2672 a string to be considered a script run, all digits in it must come from
2673 the same set of ten, as determined by the first digit encountered.
2676 qr/(*script_run: \d+ \b )/x
2678 guarantees that the digits matched will all be from the same set of 10.
2679 You won't get a look-alike digit from a different script that has a
2680 different value than what it appears to be.
2682 Unicode has three pseudo scripts that are handled specially.
2684 "Unknown" is applied to code points whose meaning has yet to be
2685 determined. Perl currently will match as a script run, any single
2686 character string consisting of one of these code points. But any string
2687 longer than one code point containing one of these will not be
2688 considered a script run.
2690 "Inherited" is applied to characters that modify another, such as an
2691 accent of some type. These are considered to be in the script of the
2692 master character, and so never cause a script run to not match.
2694 The other one is "Common". This consists of mostly punctuation, emoji,
2695 and characters used in mathematics and music, the ASCII digits C<0>
2696 through C<9>, and full-width forms of these digits. These characters
2697 can appear intermixed in text in many of the world's scripts. These
2698 also don't cause a script run to not match. But like other scripts, all
2699 digits in a run must come from the same set of 10.
2701 This construct is non-capturing. You can add parentheses to I<pattern>
2702 to capture, if desired. You will have to do this if you plan to use
2703 L</(*ACCEPT) (*ACCEPT:arg)> and not have it bypass the script run
2706 The C<Script_Extensions> property as modified by UTS 39
2707 (L<https://unicode.org/reports/tr39/>) is used as the basis for this
2716 All length 0 or length 1 sequences are script runs.
2720 A longer sequence is a script run if and only if B<all> of the following
2729 No code point in the sequence has the C<Script_Extension> property of
2732 This currently means that all code points in the sequence have been
2733 assigned by Unicode to be characters that aren't private use nor
2734 surrogate code points.
2738 All characters in the sequence come from the Common script and/or the
2739 Inherited script and/or a single other script.
2741 The script of a character is determined by the C<Script_Extensions>
2742 property as modified by UTS 39 (L<https://unicode.org/reports/tr39/>), as
2747 All decimal digits in the sequence come from the same block of 10
2754 =head2 Special Backtracking Control Verbs
2756 These special patterns are generally of the form C<(*I<VERB>:I<arg>)>. Unless
2757 otherwise stated the I<arg> argument is optional; in some cases, it is
2760 Any pattern containing a special backtracking verb that allows an argument
2761 has the special behaviour that when executed it sets the current package's
2762 C<$REGERROR> and C<$REGMARK> variables. When doing so the following
2765 On failure, the C<$REGERROR> variable will be set to the I<arg> value of the
2766 verb pattern, if the verb was involved in the failure of the match. If the
2767 I<arg> part of the pattern was omitted, then C<$REGERROR> will be set to the
2768 name of the last C<(*MARK:I<NAME>)> pattern executed, or to TRUE if there was
2769 none. Also, the C<$REGMARK> variable will be set to FALSE.
2771 On a successful match, the C<$REGERROR> variable will be set to FALSE, and
2772 the C<$REGMARK> variable will be set to the name of the last
2773 C<(*MARK:I<NAME>)> pattern executed. See the explanation for the
2774 C<(*MARK:I<NAME>)> verb below for more details.
2776 B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
2777 and most other regex-related variables. They are not local to a scope, nor
2778 readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
2779 They are set in the package containing the code that I<executed> the regex
2780 (rather than the one that compiled it, where those differ). If necessary, you
2781 can use C<local> to localize changes to these variables to a specific scope
2782 before executing a regex.
2784 If a pattern does not contain a special backtracking verb that allows an
2785 argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
2793 =item C<(*PRUNE)> C<(*PRUNE:I<NAME>)>
2794 X<(*PRUNE)> X<(*PRUNE:NAME)>
2796 This zero-width pattern prunes the backtracking tree at the current point
2797 when backtracked into on failure. Consider the pattern C</I<A> (*PRUNE) I<B>/>,
2798 where I<A> and I<B> are complex patterns. Until the C<(*PRUNE)> verb is reached,
2799 I<A> may backtrack as necessary to match. Once it is reached, matching
2800 continues in I<B>, which may also backtrack as necessary; however, should B
2801 not match, then no further backtracking will take place, and the pattern
2802 will fail outright at the current starting position.
2804 The following example counts all the possible matching strings in a
2805 pattern (without actually matching any of them).
2807 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
2808 print "Count=$count\n";
2823 If we add a C<(*PRUNE)> before the count like the following
2825 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
2826 print "Count=$count\n";
2828 we prevent backtracking and find the count of the longest matching string
2829 at each matching starting point like so:
2836 Any number of C<(*PRUNE)> assertions may be used in a pattern.
2838 See also C<<< L<< /(?>I<pattern>) >> >>> and possessive quantifiers for
2840 control backtracking. In some cases, the use of C<(*PRUNE)> can be
2841 replaced with a C<< (?>pattern) >> with no functional difference; however,
2842 C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
2843 C<< (?>pattern) >> alone.
2845 =item C<(*SKIP)> C<(*SKIP:I<NAME>)>
2848 This zero-width pattern is similar to C<(*PRUNE)>, except that on
2849 failure it also signifies that whatever text that was matched leading up
2850 to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
2851 of this pattern. This effectively means that the regex engine "skips" forward
2852 to this position on failure and tries to match again, (assuming that
2853 there is sufficient room to match).
2855 The name of the C<(*SKIP:I<NAME>)> pattern has special significance. If a
2856 C<(*MARK:I<NAME>)> was encountered while matching, then it is that position
2857 which is used as the "skip point". If no C<(*MARK)> of that name was
2858 encountered, then the C<(*SKIP)> operator has no effect. When used
2859 without a name the "skip point" is where the match point was when
2860 executing the C<(*SKIP)> pattern.
2862 Compare the following to the examples in C<(*PRUNE)>; note the string
2865 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
2866 print "Count=$count\n";
2874 Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
2875 executed, the next starting point will be where the cursor was when the
2876 C<(*SKIP)> was executed.
2878 =item C<(*MARK:I<NAME>)> C<(*:I<NAME>)>
2879 X<(*MARK)> X<(*MARK:NAME)> X<(*:NAME)>
2881 This zero-width pattern can be used to mark the point reached in a string
2882 when a certain part of the pattern has been successfully matched. This
2883 mark may be given a name. A later C<(*SKIP)> pattern will then skip
2884 forward to that point if backtracked into on failure. Any number of
2885 C<(*MARK)> patterns are allowed, and the I<NAME> portion may be duplicated.
2887 In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:I<NAME>)>
2888 can be used to "label" a pattern branch, so that after matching, the
2889 program can determine which branches of the pattern were involved in the
2892 When a match is successful, the C<$REGMARK> variable will be set to the
2893 name of the most recently executed C<(*MARK:I<NAME>)> that was involved
2896 This can be used to determine which branch of a pattern was matched
2897 without using a separate capture group for each branch, which in turn
2898 can result in a performance improvement, as perl cannot optimize
2899 C</(?:(x)|(y)|(z))/> as efficiently as something like
2900 C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
2902 When a match has failed, and unless another verb has been involved in
2903 failing the match and has provided its own name to use, the C<$REGERROR>
2904 variable will be set to the name of the most recently executed
2907 See L</(*SKIP)> for more details.
2909 As a shortcut C<(*MARK:I<NAME>)> can be written C<(*:I<NAME>)>.
2911 =item C<(*THEN)> C<(*THEN:I<NAME>)>
2913 This is similar to the "cut group" operator C<::> from Raku. Like
2914 C<(*PRUNE)>, this verb always matches, and when backtracked into on
2915 failure, it causes the regex engine to try the next alternation in the
2916 innermost enclosing group (capturing or otherwise) that has alternations.
2917 The two branches of a C<(?(I<condition>)I<yes-pattern>|I<no-pattern>)> do not
2918 count as an alternation, as far as C<(*THEN)> is concerned.
2920 Its name comes from the observation that this operation combined with the
2921 alternation operator (C<"|">) can be used to create what is essentially a
2922 pattern-based if/then/else block:
2924 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
2926 Note that if this operator is used and NOT inside of an alternation then
2927 it acts exactly like the C<(*PRUNE)> operator.
2937 / ( A (*THEN) B | C ) /
2941 / ( A (*PRUNE) B | C ) /
2943 as after matching the I<A> but failing on the I<B> the C<(*THEN)> verb will
2944 backtrack and try I<C>; but the C<(*PRUNE)> verb will simply fail.
2946 =item C<(*COMMIT)> C<(*COMMIT:I<arg>)>
2949 This is the Raku "commit pattern" C<< <commit> >> or C<:::>. It's a
2950 zero-width pattern similar to C<(*SKIP)>, except that when backtracked
2951 into on failure it causes the match to fail outright. No further attempts
2952 to find a valid match by advancing the start pointer will occur again.
2955 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
2956 print "Count=$count\n";
2963 In other words, once the C<(*COMMIT)> has been entered, and if the pattern
2964 does not match, the regex engine will not try any further matching on the
2967 =item C<(*FAIL)> C<(*F)> C<(*FAIL:I<arg>)>
2970 This pattern matches nothing and always fails. It can be used to force the
2971 engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
2972 fact, C<(?!)> gets optimised into C<(*FAIL)> internally. You can provide
2973 an argument so that if the match fails because of this C<FAIL> directive
2974 the argument can be obtained from C<$REGERROR>.
2976 It is probably useful only when combined with C<(?{})> or C<(??{})>.
2978 =item C<(*ACCEPT)> C<(*ACCEPT:I<arg>)>
2981 This pattern matches nothing and causes the end of successful matching at
2982 the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
2983 whether there is actually more to match in the string. When inside of a
2984 nested pattern, such as recursion, or in a subpattern dynamically generated
2985 via C<(??{})>, only the innermost pattern is ended immediately.
2987 If the C<(*ACCEPT)> is inside of capturing groups then the groups are
2988 marked as ended at the point at which the C<(*ACCEPT)> was encountered.
2991 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
2993 will match, and C<$1> will be C<AB> and C<$2> will be C<"B">, C<$3> will not
2994 be set. If another branch in the inner parentheses was matched, such as in the
2995 string 'ACDE', then the C<"D"> and C<"E"> would have to be matched as well.
2997 You can provide an argument, which will be available in the var
2998 C<$REGMARK> after the match completes.
3004 =head2 Warning on C<\1> Instead of C<$1>
3006 Some people get too used to writing things like:
3008 $pattern =~ s/(\W)/\\\1/g;
3010 This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid
3012 B<sed> addicts, but it's a dirty habit to get into. That's because in
3013 PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
3014 the usual double-quoted string means a control-A. The customary Unix
3015 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
3016 of doing that, you get yourself into trouble if you then add an C</e>
3019 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
3025 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
3026 C<${1}000>. The operation of interpolation should not be confused
3027 with the operation of matching a backreference. Certainly they mean two
3028 different things on the I<left> side of the C<s///>.
3030 =head2 Repeated Patterns Matching a Zero-length Substring
3032 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
3034 Regular expressions provide a terse and powerful programming language. As
3035 with most other power tools, power comes together with the ability
3038 A common abuse of this power stems from the ability to make infinite
3039 loops using regular expressions, with something as innocuous as:
3041 'foo' =~ m{ ( o? )* }x;
3043 The C<o?> matches at the beginning of "C<foo>", and since the position
3044 in the string is not moved by the match, C<o?> would match again and again
3045 because of the C<"*"> quantifier. Another common way to create a similar cycle
3046 is with the looping modifier C</g>:
3048 @matches = ( 'foo' =~ m{ o? }xg );
3052 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
3054 or the loop implied by C<split()>.
3056 However, long experience has shown that many programming tasks may
3057 be significantly simplified by using repeated subexpressions that
3058 may match zero-length substrings. Here's a simple example being:
3060 @chars = split //, $string; # // is not magic in split
3061 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
3063 Thus Perl allows such constructs, by I<forcefully breaking
3064 the infinite loop>. The rules for this are different for lower-level
3065 loops given by the greedy quantifiers C<*+{}>, and for higher-level
3066 ones like the C</g> modifier or C<split()> operator.
3068 The lower-level loops are I<interrupted> (that is, the loop is
3069 broken) when Perl detects that a repeated expression matched a
3070 zero-length substring. Thus
3072 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
3074 is made equivalent to
3076 m{ (?: NON_ZERO_LENGTH )* (?: ZERO_LENGTH )? }x;
3078 For example, this program
3085 (?{print "hello"}) # print hello whenever this
3087 (?=(b)) # zero-width assertion
3088 )* # any number of times
3099 Notice that "hello" is only printed once, as when Perl sees that the sixth
3100 iteration of the outermost C<(?:)*> matches a zero-length string, it stops
3103 The higher-level loops preserve an additional state between iterations:
3104 whether the last match was zero-length. To break the loop, the following
3105 match after a zero-length match is prohibited to have a length of zero.
3106 This prohibition interacts with backtracking (see L</"Backtracking">),
3107 and so the I<second best> match is chosen if the I<best> match is of
3115 results in C<< <><b><><a><><r><> >>. At each position of the string the best
3116 match given by non-greedy C<??> is the zero-length match, and the I<second
3117 best> match is what is matched by C<\w>. Thus zero-length matches
3118 alternate with one-character-long matches.
3120 Similarly, for repeated C<m/()/g> the second-best match is the match at the
3121 position one notch further in the string.
3123 The additional state of being I<matched with zero-length> is associated with
3124 the matched string, and is reset by each assignment to C<pos()>.
3125 Zero-length matches at the end of the previous match are ignored
3128 =head2 Combining RE Pieces
3130 Each of the elementary pieces of regular expressions which were described
3131 before (such as C<ab> or C<\Z>) could match at most one substring
3132 at the given position of the input string. However, in a typical regular
3133 expression these elementary pieces are combined into more complicated
3134 patterns using combining operators C<ST>, C<S|T>, C<S*> I<etc>.
3135 (in these examples C<"S"> and C<"T"> are regular subexpressions).
3137 Such combinations can include alternatives, leading to a problem of choice:
3138 if we match a regular expression C<a|ab> against C<"abc">, will it match
3139 substring C<"a"> or C<"ab">? One way to describe which substring is
3140 actually matched is the concept of backtracking (see L</"Backtracking">).
3141 However, this description is too low-level and makes you think
3142 in terms of a particular implementation.
3144 Another description starts with notions of "better"/"worse". All the
3145 substrings which may be matched by the given regular expression can be
3146 sorted from the "best" match to the "worst" match, and it is the "best"
3147 match which is chosen. This substitutes the question of "what is chosen?"
3148 by the question of "which matches are better, and which are worse?".
3150 Again, for elementary pieces there is no such question, since at most
3151 one match at a given position is possible. This section describes the
3152 notion of better/worse for combining operators. In the description
3153 below C<"S"> and C<"T"> are regular subexpressions.
3159 Consider two possible matches, C<AB> and C<A'B'>, C<"A"> and C<A'> are
3160 substrings which can be matched by C<"S">, C<"B"> and C<B'> are substrings
3161 which can be matched by C<"T">.
3163 If C<"A"> is a better match for C<"S"> than C<A'>, C<AB> is a better
3166 If C<"A"> and C<A'> coincide: C<AB> is a better match than C<AB'> if
3167 C<"B"> is a better match for C<"T"> than C<B'>.
3171 When C<"S"> can match, it is a better match than when only C<"T"> can match.
3173 Ordering of two matches for C<"S"> is the same as for C<"S">. Similar for
3174 two matches for C<"T">.
3176 =item C<S{REPEAT_COUNT}>
3178 Matches as C<SSS...S> (repeated as many times as necessary).
3182 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
3184 =item C<S{min,max}?>
3186 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
3188 =item C<S?>, C<S*>, C<S+>
3190 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
3192 =item C<S??>, C<S*?>, C<S+?>
3194 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
3198 Matches the best match for C<"S"> and only that.
3200 =item C<(?=S)>, C<(?<=S)>
3202 Only the best match for C<"S"> is considered. (This is important only if
3203 C<"S"> has capturing parentheses, and backreferences are used somewhere
3204 else in the whole regular expression.)
3206 =item C<(?!S)>, C<(?<!S)>
3208 For this grouping operator there is no need to describe the ordering, since
3209 only whether or not C<"S"> can match is important.
3211 =item C<(??{ I<EXPR> })>, C<(?I<PARNO>)>
3213 The ordering is the same as for the regular expression which is
3214 the result of I<EXPR>, or the pattern contained by capture group I<PARNO>.
3216 =item C<(?(I<condition>)I<yes-pattern>|I<no-pattern>)>
3218 Recall that which of I<yes-pattern> or I<no-pattern> actually matches is
3219 already determined. The ordering of the matches is the same as for the
3220 chosen subexpression.
3224 The above recipes describe the ordering of matches I<at a given position>.
3225 One more rule is needed to understand how a match is determined for the
3226 whole regular expression: a match at an earlier position is always better
3227 than a match at a later position.
3229 =head2 Creating Custom RE Engines
3231 As of Perl 5.10.0, one can create custom regular expression engines. This
3232 is not for the faint of heart, as they have to plug in at the C level. See
3233 L<perlreapi> for more details.
3235 As an alternative, overloaded constants (see L<overload>) provide a simple
3236 way to extend the functionality of the RE engine, by substituting one
3237 pattern for another.
3239 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
3240 matches at a boundary between whitespace characters and non-whitespace
3241 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
3242 at these positions, so we want to have each C<\Y|> in the place of the
3243 more complicated version. We can create a module C<customre> to do
3251 die "No argument to customre::import allowed" if @_;
3252 overload::constant 'qr' => \&convert;
3255 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
3257 # We must also take care of not escaping the legitimate \\Y|
3258 # sequence, hence the presence of '\\' in the conversion rules.
3259 my %rules = ( '\\' => '\\\\',
3260 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
3266 { $rules{$1} or invalid($re,$1) }sgex;
3270 Now C<use customre> enables the new escape in constant regular
3271 expressions, I<i.e.>, those without any runtime variable interpolations.
3272 As documented in L<overload>, this conversion will work only over
3273 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
3274 part of this regular expression needs to be converted explicitly
3275 (but only if the special meaning of C<\Y|> should be enabled inside C<$re>):
3280 $re = customre::convert $re;
3283 =head2 Embedded Code Execution Frequency
3285 The exact rules for how often C<(??{})> and C<(?{})> are executed in a pattern
3286 are unspecified. In the case of a successful match you can assume that
3287 they DWIM and will be executed in left to right order the appropriate
3288 number of times in the accepting path of the pattern as would any other
3289 meta-pattern. How non-accepting pathways and match failures affect the
3290 number of times a pattern is executed is specifically unspecified and
3291 may vary depending on what optimizations can be applied to the pattern
3292 and is likely to change from version to version.
3296 "aaabcdeeeee"=~/a(?{print "a"})b(?{print "b"})cde/;
3298 the exact number of times "a" or "b" are printed out is unspecified for
3299 failure, but you may assume they will be printed at least once during
3300 a successful match, additionally you may assume that if "b" is printed,
3301 it will be preceded by at least one "a".
3303 In the case of branching constructs like the following:
3305 /a(b|(?{ print "a" }))c(?{ print "c" })/;
3307 you can assume that the input "ac" will output "ac", and that "abc"
3308 will output only "c".
3310 When embedded code is quantified, successful matches will call the
3311 code once for each matched iteration of the quantifier. For
3314 "good" =~ /g(?:o(?{print "o"}))*d/;
3316 will output "o" twice.
3318 =head2 PCRE/Python Support
3320 As of Perl 5.10.0, Perl supports several Python/PCRE-specific extensions
3321 to the regex syntax. While Perl programmers are encouraged to use the
3322 Perl-specific syntax, the following are also accepted:
3326 =item C<< (?PE<lt>I<NAME>E<gt>I<pattern>) >>
3328 Define a named capture group. Equivalent to C<< (?<I<NAME>>I<pattern>) >>.
3330 =item C<< (?P=I<NAME>) >>
3332 Backreference to a named capture group. Equivalent to C<< \g{I<NAME>} >>.
3334 =item C<< (?P>I<NAME>) >>
3336 Subroutine call to a named capture group. Equivalent to C<< (?&I<NAME>) >>.
3342 There are a number of issues with regard to case-insensitive matching
3343 in Unicode rules. See C<"i"> under L</Modifiers> above.
3345 This document varies from difficult to understand to completely
3346 and utterly opaque. The wandering prose riddled with jargon is
3347 hard to fathom in several places.
3349 This document needs a rewrite that separates the tutorial content
3350 from the reference content.
3354 The syntax of patterns used in Perl pattern matching evolved from those
3355 supplied in the Bell Labs Research Unix 8th Edition (Version 8) regex
3356 routines. (The code is actually derived (distantly) from Henry
3357 Spencer's freely redistributable reimplementation of those V8 routines.)
3363 L<perlop/"Regexp Quote-Like Operators">.
3365 L<perlop/"Gory details of parsing quoted constructs">.
3375 I<Mastering Regular Expressions> by Jeffrey Friedl, published
3376 by O'Reilly and Associates.