3 perlre - Perl regular expressions
7 This page describes the syntax of regular expressions in Perl. For a
8 description of how to I<use> regular expressions in matching
9 operations, plus various examples of the same, see discussions
10 of C<m//>, C<s///>, C<qr//> and C<??> in L<perlop/"Regexp Quote-Like Operators">.
12 Matching operations can have various modifiers. Modifiers
13 that relate to the interpretation of the regular expression inside
14 are listed below. Modifiers that alter the way a regular expression
15 is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and
16 L<perlop/"Gory details of parsing quoted constructs">.
22 Do case-insensitive pattern matching.
24 If C<use locale> is in effect, the case map is taken from the current
25 locale. See L<perllocale>.
29 Treat string as multiple lines. That is, change "^" and "$" from matching
30 the start or end of the string to matching the start or end of any
31 line anywhere within the string.
35 Treat string as single line. That is, change "." to match any character
36 whatsoever, even a newline, which normally it would not match.
38 The C</s> and C</m> modifiers both override the C<$*> setting. That
39 is, no matter what C<$*> contains, C</s> without C</m> will force
40 "^" to match only at the beginning of the string and "$" to match
41 only at the end (or just before a newline at the end) of the string.
42 Together, as /ms, they let the "." match any character whatsoever,
43 while yet allowing "^" and "$" to match, respectively, just after
44 and just before newlines within the string.
48 Extend your pattern's legibility by permitting whitespace and comments.
52 These are usually written as "the C</x> modifier", even though the delimiter
53 in question might not really be a slash. Any of these
54 modifiers may also be embedded within the regular expression itself using
55 the C<(?...)> construct. See below.
57 The C</x> modifier itself needs a little more explanation. It tells
58 the regular expression parser to ignore whitespace that is neither
59 backslashed nor within a character class. You can use this to break up
60 your regular expression into (slightly) more readable parts. The C<#>
61 character is also treated as a metacharacter introducing a comment,
62 just as in ordinary Perl code. This also means that if you want real
63 whitespace or C<#> characters in the pattern (outside a character
64 class, where they are unaffected by C</x>), that you'll either have to
65 escape them or encode them using octal or hex escapes. Taken together,
66 these features go a long way towards making Perl's regular expressions
67 more readable. Note that you have to be careful not to include the
68 pattern delimiter in the comment--perl has no way of knowing you did
69 not intend to close the pattern early. See the C-comment deletion code
72 =head2 Regular Expressions
74 The patterns used in Perl pattern matching derive from supplied in
75 the Version 8 regex routines. (The routines are derived
76 (distantly) from Henry Spencer's freely redistributable reimplementation
77 of the V8 routines.) See L<Version 8 Regular Expressions> for
80 In particular the following metacharacters have their standard I<egrep>-ish
83 \ Quote the next metacharacter
84 ^ Match the beginning of the line
85 . Match any character (except newline)
86 $ Match the end of the line (or before newline at the end)
91 By default, the "^" character is guaranteed to match only the
92 beginning of the string, the "$" character only the end (or before the
93 newline at the end), and Perl does certain optimizations with the
94 assumption that the string contains only one line. Embedded newlines
95 will not be matched by "^" or "$". You may, however, wish to treat a
96 string as a multi-line buffer, such that the "^" will match after any
97 newline within the string, and "$" will match before any newline. At the
98 cost of a little more overhead, you can do this by using the /m modifier
99 on the pattern match operator. (Older programs did this by setting C<$*>,
100 but this practice is now deprecated.)
102 To simplify multi-line substitutions, the "." character never matches a
103 newline unless you use the C</s> modifier, which in effect tells Perl to pretend
104 the string is a single line--even if it isn't. The C</s> modifier also
105 overrides the setting of C<$*>, in case you have some (badly behaved) older
106 code that sets it in another module.
108 The following standard quantifiers are recognized:
110 * Match 0 or more times
111 + Match 1 or more times
113 {n} Match exactly n times
114 {n,} Match at least n times
115 {n,m} Match at least n but not more than m times
117 (If a curly bracket occurs in any other context, it is treated
118 as a regular character.) The "*" modifier is equivalent to C<{0,}>, the "+"
119 modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited
120 to integral values less than a preset limit defined when perl is built.
121 This is usually 32766 on the most common platforms. The actual limit can
122 be seen in the error message generated by code such as this:
124 $_ **= $_ , / {$_} / for 2 .. 42;
126 By default, a quantified subpattern is "greedy", that is, it will match as
127 many times as possible (given a particular starting location) while still
128 allowing the rest of the pattern to match. If you want it to match the
129 minimum number of times possible, follow the quantifier with a "?". Note
130 that the meanings don't change, just the "greediness":
132 *? Match 0 or more times
133 +? Match 1 or more times
135 {n}? Match exactly n times
136 {n,}? Match at least n times
137 {n,m}? Match at least n but not more than m times
139 Because patterns are processed as double quoted strings, the following
146 \a alarm (bell) (BEL)
147 \e escape (think troff) (ESC)
148 \033 octal char (think of a PDP-11)
150 \x{263a} wide hex char (Unicode SMILEY)
153 \l lowercase next char (think vi)
154 \u uppercase next char (think vi)
155 \L lowercase till \E (think vi)
156 \U uppercase till \E (think vi)
157 \E end case modification (think vi)
158 \Q quote (disable) pattern metacharacters till \E
160 If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u>
161 and C<\U> is taken from the current locale. See L<perllocale>. For
162 documentation of C<\N{name}>, see L<charnames>.
164 You cannot include a literal C<$> or C<@> within a C<\Q> sequence.
165 An unescaped C<$> or C<@> interpolates the corresponding variable,
166 while escaping will cause the literal string C<\$> to be matched.
167 You'll need to write something like C<m/\Quser\E\@\Qhost/>.
169 In addition, Perl defines the following:
171 \w Match a "word" character (alphanumeric plus "_")
172 \W Match a non-"word" character
173 \s Match a whitespace character
174 \S Match a non-whitespace character
175 \d Match a digit character
176 \D Match a non-digit character
177 \pP Match P, named property. Use \p{Prop} for longer names.
179 \X Match eXtended Unicode "combining character sequence",
180 equivalent to C<(?:\PM\pM*)>
181 \C Match a single C char (octet) even under utf8.
183 A C<\w> matches a single alphanumeric character or C<_>, not a whole word.
184 Use C<\w+> to match a string of Perl-identifier characters (which isn't
185 the same as matching an English word). If C<use locale> is in effect, the
186 list of alphabetic characters generated by C<\w> is taken from the
187 current locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
188 C<\d>, and C<\D> within character classes, but if you try to use them
189 as endpoints of a range, that's not a range, the "-" is understood literally.
190 See L<utf8> for details about C<\pP>, C<\PP>, and C<\X>.
192 The POSIX character class syntax
196 is also available. The available classes and their backslash
197 equivalents (if available) are as follows:
213 For example use C<[:upper:]> to match all the uppercase characters.
214 Note that the C<[]> are part of the C<[::]> construct, not part of the whole
215 character class. For example:
219 matches zero, one, any alphabetic character, and the percentage sign.
221 If the C<utf8> pragma is used, the following equivalences to Unicode
222 \p{} constructs hold:
238 For example C<[:lower:]> and C<\p{IsLower}> are equivalent.
240 If the C<utf8> pragma is not used but the C<locale> pragma is, the
241 classes correlate with the isalpha(3) interface (except for `word',
242 which is a Perl extension, mirroring C<\w>).
244 The assumedly non-obviously named classes are:
250 Any control character. Usually characters that don't produce output as
251 such but instead control the terminal somehow: for example newline and
252 backspace are control characters. All characters with ord() less than
253 32 are most often classified as control characters (assuming ASCII,
254 the ISO Latin character sets, and Unicode).
258 Any alphanumeric or punctuation (special) character.
262 Any alphanumeric or punctuation (special) character or space.
266 Any punctuation (special) character.
270 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
271 work just fine) it is included for completeness.
275 You can negate the [::] character classes by prefixing the class name
276 with a '^'. This is a Perl extension. For example:
278 POSIX trad. Perl utf8 Perl
280 [:^digit:] \D \P{IsDigit}
281 [:^space:] \S \P{IsSpace}
282 [:^word:] \W \P{IsWord}
284 The POSIX character classes [.cc.] and [=cc=] are recognized but
285 B<not> supported and trying to use them will cause an error.
287 Perl defines the following zero-width assertions:
289 \b Match a word boundary
290 \B Match a non-(word boundary)
291 \A Match only at beginning of string
292 \Z Match only at end of string, or before newline at the end
293 \z Match only at end of string
294 \G Match only at pos() (e.g. at the end-of-match position
297 A word boundary (C<\b>) is a spot between two characters
298 that has a C<\w> on one side of it and a C<\W> on the other side
299 of it (in either order), counting the imaginary characters off the
300 beginning and end of the string as matching a C<\W>. (Within
301 character classes C<\b> represents backspace rather than a word
302 boundary, just as it normally does in any double-quoted string.)
303 The C<\A> and C<\Z> are just like "^" and "$", except that they
304 won't match multiple times when the C</m> modifier is used, while
305 "^" and "$" will match at every internal line boundary. To match
306 the actual end of the string and not ignore an optional trailing
309 The C<\G> assertion can be used to chain global matches (using
310 C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
311 It is also useful when writing C<lex>-like scanners, when you have
312 several patterns that you want to match against consequent substrings
313 of your string, see the previous reference. The actual location
314 where C<\G> will match can also be influenced by using C<pos()> as
315 an lvalue. See L<perlfunc/pos>.
317 The bracketing construct C<( ... )> creates capture buffers. To
318 refer to the digit'th buffer use \<digit> within the
319 match. Outside the match use "$" instead of "\". (The
320 \<digit> notation works in certain circumstances outside
321 the match. See the warning below about \1 vs $1 for details.)
322 Referring back to another part of the match is called a
325 There is no limit to the number of captured substrings that you may
326 use. However Perl also uses \10, \11, etc. as aliases for \010,
327 \011, etc. (Recall that 0 means octal, so \011 is the 9'th ASCII
328 character, a tab.) Perl resolves this ambiguity by interpreting
329 \10 as a backreference only if at least 10 left parentheses have
330 opened before it. Likewise \11 is a backreference only if at least
331 11 left parentheses have opened before it. And so on. \1 through
332 \9 are always interpreted as backreferences."
336 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
338 if (/(.)\1/) { # find first doubled char
339 print "'$1' is the first doubled character\n";
342 if (/Time: (..):(..):(..)/) { # parse out values
348 Several special variables also refer back to portions of the previous
349 match. C<$+> returns whatever the last bracket match matched.
350 C<$&> returns the entire matched string. (At one point C<$0> did
351 also, but now it returns the name of the program.) C<$`> returns
352 everything before the matched string. And C<$'> returns everything
353 after the matched string.
355 The numbered variables ($1, $2, $3, etc.) and the related punctuation
356 set (C<<$+>, C<$&>, C<$`>, and C<$'>) are all dynamically scoped
357 until the end of the enclosing block or until the next successful
358 match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
360 B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
361 C<$'> anywhere in the program, it has to provide them for every
362 pattern match. This may substantially slow your program. Perl
363 uses the same mechanism to produce $1, $2, etc, so you also pay a
364 price for each pattern that contains capturing parentheses. (To
365 avoid this cost while retaining the grouping behaviour, use the
366 extended regular expression C<(?: ... )> instead.) But if you never
367 use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
368 parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
369 if you can, but if you can't (and some algorithms really appreciate
370 them), once you've used them once, use them at will, because you've
371 already paid the price. As of 5.005, C<$&> is not so costly as the
374 Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
375 C<\w>, C<\n>. Unlike some other regular expression languages, there
376 are no backslashed symbols that aren't alphanumeric. So anything
377 that looks like \\, \(, \), \<, \>, \{, or \} is always
378 interpreted as a literal character, not a metacharacter. This was
379 once used in a common idiom to disable or quote the special meanings
380 of regular expression metacharacters in a string that you want to
381 use for a pattern. Simply quote all non-"word" characters:
383 $pattern =~ s/(\W)/\\$1/g;
385 (If C<use locale> is set, then this depends on the current locale.)
386 Today it is more common to use the quotemeta() function or the C<\Q>
387 metaquoting escape sequence to disable all metacharacters' special
390 /$unquoted\Q$quoted\E$unquoted/
392 Beware that if you put literal backslashes (those not inside
393 interpolated variables) between C<\Q> and C<\E>, double-quotish
394 backslash interpolation may lead to confusing results. If you
395 I<need> to use literal backslashes within C<\Q...\E>,
396 consult L<perlop/"Gory details of parsing quoted constructs">.
398 =head2 Extended Patterns
400 Perl also defines a consistent extension syntax for features not
401 found in standard tools like B<awk> and B<lex>. The syntax is a
402 pair of parentheses with a question mark as the first thing within
403 the parentheses. The character after the question mark indicates
406 The stability of these extensions varies widely. Some have been
407 part of the core language for many years. Others are experimental
408 and may change without warning or be completely removed. Check
409 the documentation on an individual feature to verify its current
412 A question mark was chosen for this and for the minimal-matching
413 construct because 1) question marks are rare in older regular
414 expressions, and 2) whenever you see one, you should stop and
415 "question" exactly what is going on. That's psychology...
421 A comment. The text is ignored. If the C</x> modifier enables
422 whitespace formatting, a simple C<#> will suffice. Note that Perl closes
423 the comment as soon as it sees a C<)>, so there is no way to put a literal
426 =item C<(?imsx-imsx)>
428 One or more embedded pattern-match modifiers. This is particularly
429 useful for dynamic patterns, such as those read in from a configuration
430 file, read in as an argument, are specified in a table somewhere,
431 etc. Consider the case that some of which want to be case sensitive
432 and some do not. The case insensitive ones need to include merely
433 C<(?i)> at the front of the pattern. For example:
436 if ( /$pattern/i ) { }
440 $pattern = "(?i)foobar";
441 if ( /$pattern/ ) { }
443 Letters after a C<-> turn those modifiers off. These modifiers are
444 localized inside an enclosing group (if any). For example,
448 will match a repeated (I<including the case>!) word C<blah> in any
449 case, assuming C<x> modifier, and no C<i> modifier outside this
454 =item C<(?imsx-imsx:pattern)>
456 This is for clustering, not capturing; it groups subexpressions like
457 "()", but doesn't make backreferences as "()" does. So
459 @fields = split(/\b(?:a|b|c)\b/)
463 @fields = split(/\b(a|b|c)\b/)
465 but doesn't spit out extra fields. It's also cheaper not to capture
466 characters if you don't need to.
468 Any letters between C<?> and C<:> act as flags modifiers as with
469 C<(?imsx-imsx)>. For example,
471 /(?s-i:more.*than).*million/i
473 is equivalent to the more verbose
475 /(?:(?s-i)more.*than).*million/i
479 A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
480 matches a word followed by a tab, without including the tab in C<$&>.
484 A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
485 matches any occurrence of "foo" that isn't followed by "bar". Note
486 however that look-ahead and look-behind are NOT the same thing. You cannot
487 use this for look-behind.
489 If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
490 will not do what you want. That's because the C<(?!foo)> is just saying that
491 the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
492 match. You would have to do something like C</(?!foo)...bar/> for that. We
493 say "like" because there's the case of your "bar" not having three characters
494 before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
495 Sometimes it's still easier just to say:
497 if (/bar/ && $` !~ /foo$/)
499 For look-behind see below.
501 =item C<(?<=pattern)>
503 A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
504 matches a word that follows a tab, without including the tab in C<$&>.
505 Works only for fixed-width look-behind.
507 =item C<(?<!pattern)>
509 A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
510 matches any occurrence of "foo" that does not follow "bar". Works
511 only for fixed-width look-behind.
515 B<WARNING>: This extended regular expression feature is considered
516 highly experimental, and may be changed or deleted without notice.
518 This zero-width assertion evaluate any embedded Perl code. It
519 always succeeds, and its C<code> is not interpolated. Currently,
520 the rules to determine where the C<code> ends are somewhat convoluted.
522 The C<code> is properly scoped in the following sense: If the assertion
523 is backtracked (compare L<"Backtracking">), all changes introduced after
524 C<local>ization are undone, so that
528 (?{ $cnt = 0 }) # Initialize $cnt.
532 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
536 (?{ $res = $cnt }) # On success copy to non-localized
540 will set C<$res = 4>. Note that after the match, $cnt returns to the globally
541 introduced value, because the scopes that restrict C<local> operators
544 This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
545 switch. If I<not> used in this way, the result of evaluation of
546 C<code> is put into the special variable C<$^R>. This happens
547 immediately, so C<$^R> can be used from other C<(?{ code })> assertions
548 inside the same regular expression.
550 The assignment to C<$^R> above is properly localized, so the old
551 value of C<$^R> is restored if the assertion is backtracked; compare
554 For reasons of security, this construct is forbidden if the regular
555 expression involves run-time interpolation of variables, unless the
556 perilous C<use re 'eval'> pragma has been used (see L<re>), or the
557 variables contain results of C<qr//> operator (see
558 L<perlop/"qr/STRING/imosx">).
560 This restriction is because of the wide-spread and remarkably convenient
561 custom of using run-time determined strings as patterns. For example:
567 Before Perl knew how to execute interpolated code within a pattern,
568 this operation was completely safe from a security point of view,
569 although it could raise an exception from an illegal pattern. If
570 you turn on the C<use re 'eval'>, though, it is no longer secure,
571 so you should only do so if you are also using taint checking.
572 Better yet, use the carefully constrained evaluation within a Safe
573 module. See L<perlsec> for details about both these mechanisms.
575 =item C<(??{ code })>
577 B<WARNING>: This extended regular expression feature is considered
578 highly experimental, and may be changed or deleted without notice.
579 A simplified version of the syntax may be introduced for commonly
582 This is a "postponed" regular subexpression. The C<code> is evaluated
583 at run time, at the moment this subexpression may match. The result
584 of evaluation is considered as a regular expression and matched as
585 if it were inserted instead of this construct.
587 The C<code> is not interpolated. As before, the rules to determine
588 where the C<code> ends are currently somewhat convoluted.
590 The following pattern matches a parenthesized group:
595 (?> [^()]+ ) # Non-parens without backtracking
597 (??{ $re }) # Group with matching parens
602 =item C<< (?>pattern) >>
604 B<WARNING>: This extended regular expression feature is considered
605 highly experimental, and may be changed or deleted without notice.
607 An "independent" subexpression, one which matches the substring
608 that a I<standalone> C<pattern> would match if anchored at the given
609 position, and it matches I<nothing other than this substring>. This
610 construct is useful for optimizations of what would otherwise be
611 "eternal" matches, because it will not backtrack (see L<"Backtracking">).
612 It may also be useful in places where the "grab all you can, and do not
613 give anything back" semantic is desirable.
615 For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
616 (anchored at the beginning of string, as above) will match I<all>
617 characters C<a> at the beginning of string, leaving no C<a> for
618 C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
619 since the match of the subgroup C<a*> is influenced by the following
620 group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
621 C<a*ab> will match fewer characters than a standalone C<a*>, since
622 this makes the tail match.
624 An effect similar to C<< (?>pattern) >> may be achieved by writing
625 C<(?=(pattern))\1>. This matches the same substring as a standalone
626 C<a+>, and the following C<\1> eats the matched string; it therefore
627 makes a zero-length assertion into an analogue of C<< (?>...) >>.
628 (The difference between these two constructs is that the second one
629 uses a capturing group, thus shifting ordinals of backreferences
630 in the rest of a regular expression.)
632 Consider this pattern:
643 That will efficiently match a nonempty group with matching parentheses
644 two levels deep or less. However, if there is no such group, it
645 will take virtually forever on a long string. That's because there
646 are so many different ways to split a long string into several
647 substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
648 to a subpattern of the above pattern. Consider how the pattern
649 above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
650 seconds, but that each extra letter doubles this time. This
651 exponential performance will make it appear that your program has
652 hung. However, a tiny change to this pattern
656 (?> [^()]+ ) # change x+ above to (?> x+ )
663 which uses C<< (?>...) >> matches exactly when the one above does (verifying
664 this yourself would be a productive exercise), but finishes in a fourth
665 the time when used on a similar string with 1000000 C<a>s. Be aware,
666 however, that this pattern currently triggers a warning message under
667 the C<use warnings> pragma or B<-w> switch saying it
668 C<"matches the null string many times">):
670 On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
671 effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
672 This was only 4 times slower on a string with 1000000 C<a>s.
674 The "grab all you can, and do not give anything back" semantic is desirable
675 in many situations where on the first sight a simple C<()*> looks like
676 the correct solution. Suppose we parse text with comments being delimited
677 by C<#> followed by some optional (horizontal) whitespace. Contrary to
678 its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
679 the comment delimiter, because it may "give up" some whitespace if
680 the remainder of the pattern can be made to match that way. The correct
681 answer is either one of these:
686 For example, to grab non-empty comments into $1, one should use either
689 / (?> \# [ \t]* ) ( .+ ) /x;
690 / \# [ \t]* ( [^ \t] .* ) /x;
692 Which one you pick depends on which of these expressions better reflects
693 the above specification of comments.
695 =item C<(?(condition)yes-pattern|no-pattern)>
697 =item C<(?(condition)yes-pattern)>
699 B<WARNING>: This extended regular expression feature is considered
700 highly experimental, and may be changed or deleted without notice.
702 Conditional expression. C<(condition)> should be either an integer in
703 parentheses (which is valid if the corresponding pair of parentheses
704 matched), or look-ahead/look-behind/evaluate zero-width assertion.
713 matches a chunk of non-parentheses, possibly included in parentheses
720 NOTE: This section presents an abstract approximation of regular
721 expression behavior. For a more rigorous (and complicated) view of
722 the rules involved in selecting a match among possible alternatives,
723 see L<Combining pieces together>.
725 A fundamental feature of regular expression matching involves the
726 notion called I<backtracking>, which is currently used (when needed)
727 by all regular expression quantifiers, namely C<*>, C<*?>, C<+>,
728 C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
729 internally, but the general principle outlined here is valid.
731 For a regular expression to match, the I<entire> regular expression must
732 match, not just part of it. So if the beginning of a pattern containing a
733 quantifier succeeds in a way that causes later parts in the pattern to
734 fail, the matching engine backs up and recalculates the beginning
735 part--that's why it's called backtracking.
737 Here is an example of backtracking: Let's say you want to find the
738 word following "foo" in the string "Food is on the foo table.":
740 $_ = "Food is on the foo table.";
741 if ( /\b(foo)\s+(\w+)/i ) {
742 print "$2 follows $1.\n";
745 When the match runs, the first part of the regular expression (C<\b(foo)>)
746 finds a possible match right at the beginning of the string, and loads up
747 $1 with "Foo". However, as soon as the matching engine sees that there's
748 no whitespace following the "Foo" that it had saved in $1, it realizes its
749 mistake and starts over again one character after where it had the
750 tentative match. This time it goes all the way until the next occurrence
751 of "foo". The complete regular expression matches this time, and you get
752 the expected output of "table follows foo."
754 Sometimes minimal matching can help a lot. Imagine you'd like to match
755 everything between "foo" and "bar". Initially, you write something
758 $_ = "The food is under the bar in the barn.";
759 if ( /foo(.*)bar/ ) {
763 Which perhaps unexpectedly yields:
765 got <d is under the bar in the >
767 That's because C<.*> was greedy, so you get everything between the
768 I<first> "foo" and the I<last> "bar". Here it's more effective
769 to use minimal matching to make sure you get the text between a "foo"
770 and the first "bar" thereafter.
772 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
773 got <d is under the >
775 Here's another example: let's say you'd like to match a number at the end
776 of a string, and you also want to keep the preceding part the match.
779 $_ = "I have 2 numbers: 53147";
780 if ( /(.*)(\d*)/ ) { # Wrong!
781 print "Beginning is <$1>, number is <$2>.\n";
784 That won't work at all, because C<.*> was greedy and gobbled up the
785 whole string. As C<\d*> can match on an empty string the complete
786 regular expression matched successfully.
788 Beginning is <I have 2 numbers: 53147>, number is <>.
790 Here are some variants, most of which don't work:
792 $_ = "I have 2 numbers: 53147";
805 printf "%-12s ", $pat;
815 (.*)(\d*) <I have 2 numbers: 53147> <>
816 (.*)(\d+) <I have 2 numbers: 5314> <7>
818 (.*?)(\d+) <I have > <2>
819 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
820 (.*?)(\d+)$ <I have 2 numbers: > <53147>
821 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
822 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
824 As you see, this can be a bit tricky. It's important to realize that a
825 regular expression is merely a set of assertions that gives a definition
826 of success. There may be 0, 1, or several different ways that the
827 definition might succeed against a particular string. And if there are
828 multiple ways it might succeed, you need to understand backtracking to
829 know which variety of success you will achieve.
831 When using look-ahead assertions and negations, this can all get even
832 tricker. Imagine you'd like to find a sequence of non-digits not
833 followed by "123". You might try to write that as
836 if ( /^\D*(?!123)/ ) { # Wrong!
837 print "Yup, no 123 in $_\n";
840 But that isn't going to match; at least, not the way you're hoping. It
841 claims that there is no 123 in the string. Here's a clearer picture of
842 why it that pattern matches, contrary to popular expectations:
847 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/ ;
848 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/ ;
850 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/ ;
851 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/ ;
859 You might have expected test 3 to fail because it seems to a more
860 general purpose version of test 1. The important difference between
861 them is that test 3 contains a quantifier (C<\D*>) and so can use
862 backtracking, whereas test 1 will not. What's happening is
863 that you've asked "Is it true that at the start of $x, following 0 or more
864 non-digits, you have something that's not 123?" If the pattern matcher had
865 let C<\D*> expand to "ABC", this would have caused the whole pattern to
868 The search engine will initially match C<\D*> with "ABC". Then it will
869 try to match C<(?!123> with "123", which fails. But because
870 a quantifier (C<\D*>) has been used in the regular expression, the
871 search engine can backtrack and retry the match differently
872 in the hope of matching the complete regular expression.
874 The pattern really, I<really> wants to succeed, so it uses the
875 standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
876 time. Now there's indeed something following "AB" that is not
877 "123". It's "C123", which suffices.
879 We can deal with this by using both an assertion and a negation.
880 We'll say that the first part in $1 must be followed both by a digit
881 and by something that's not "123". Remember that the look-aheads
882 are zero-width expressions--they only look, but don't consume any
883 of the string in their match. So rewriting this way produces what
884 you'd expect; that is, case 5 will fail, but case 6 succeeds:
886 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/ ;
887 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/ ;
891 In other words, the two zero-width assertions next to each other work as though
892 they're ANDed together, just as you'd use any built-in assertions: C</^$/>
893 matches only if you're at the beginning of the line AND the end of the
894 line simultaneously. The deeper underlying truth is that juxtaposition in
895 regular expressions always means AND, except when you write an explicit OR
896 using the vertical bar. C</ab/> means match "a" AND (then) match "b",
897 although the attempted matches are made at different positions because "a"
898 is not a zero-width assertion, but a one-width assertion.
900 B<WARNING>: particularly complicated regular expressions can take
901 exponential time to solve because of the immense number of possible
902 ways they can use backtracking to try match. For example, without
903 internal optimizations done by the regular expression engine, this will
904 take a painfully long time to run:
906 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5}){0,5}[c]/
908 And if you used C<*>'s instead of limiting it to 0 through 5 matches,
909 then it would take forever--or until you ran out of stack space.
911 A powerful tool for optimizing such beasts is what is known as an
913 which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
914 zero-length look-ahead/look-behind assertions will not backtrack to make
915 the tail match, since they are in "logical" context: only
916 whether they match is considered relevant. For an example
917 where side-effects of look-ahead I<might> have influenced the
918 following match, see L<C<< (?>pattern) >>>.
920 =head2 Version 8 Regular Expressions
922 In case you're not familiar with the "regular" Version 8 regex
923 routines, here are the pattern-matching rules not described above.
925 Any single character matches itself, unless it is a I<metacharacter>
926 with a special meaning described here or above. You can cause
927 characters that normally function as metacharacters to be interpreted
928 literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
929 character; "\\" matches a "\"). A series of characters matches that
930 series of characters in the target string, so the pattern C<blurfl>
931 would match "blurfl" in the target string.
933 You can specify a character class, by enclosing a list of characters
934 in C<[]>, which will match any one character from the list. If the
935 first character after the "[" is "^", the class matches any character not
936 in the list. Within a list, the "-" character specifies a
937 range, so that C<a-z> represents all characters between "a" and "z",
938 inclusive. If you want either "-" or "]" itself to be a member of a
939 class, put it at the start of the list (possibly after a "^"), or
940 escape it with a backslash. "-" is also taken literally when it is
941 at the end of the list, just before the closing "]". (The
942 following all specify the same class of three characters: C<[-az]>,
943 C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
944 specifies a class containing twenty-six characters.)
945 Also, if you try to use the character classes C<\w>, C<\W>, C<\s>,
946 C<\S>, C<\d>, or C<\D> as endpoints of a range, that's not a range,
947 the "-" is understood literally.
949 Note also that the whole range idea is rather unportable between
950 character sets--and even within character sets they may cause results
951 you probably didn't expect. A sound principle is to use only ranges
952 that begin from and end at either alphabets of equal case ([a-e],
953 [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
954 spell out the character sets in full.
956 Characters may be specified using a metacharacter syntax much like that
957 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
958 "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
959 of octal digits, matches the character whose ASCII value is I<nnn>.
960 Similarly, \xI<nn>, where I<nn> are hexadecimal digits, matches the
961 character whose ASCII value is I<nn>. The expression \cI<x> matches the
962 ASCII character control-I<x>. Finally, the "." metacharacter matches any
963 character except "\n" (unless you use C</s>).
965 You can specify a series of alternatives for a pattern using "|" to
966 separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
967 or "foe" in the target string (as would C<f(e|i|o)e>). The
968 first alternative includes everything from the last pattern delimiter
969 ("(", "[", or the beginning of the pattern) up to the first "|", and
970 the last alternative contains everything from the last "|" to the next
971 pattern delimiter. That's why it's common practice to include
972 alternatives in parentheses: to minimize confusion about where they
975 Alternatives are tried from left to right, so the first
976 alternative found for which the entire expression matches, is the one that
977 is chosen. This means that alternatives are not necessarily greedy. For
978 example: when matching C<foo|foot> against "barefoot", only the "foo"
979 part will match, as that is the first alternative tried, and it successfully
980 matches the target string. (This might not seem important, but it is
981 important when you are capturing matched text using parentheses.)
983 Also remember that "|" is interpreted as a literal within square brackets,
984 so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
986 Within a pattern, you may designate subpatterns for later reference
987 by enclosing them in parentheses, and you may refer back to the
988 I<n>th subpattern later in the pattern using the metacharacter
989 \I<n>. Subpatterns are numbered based on the left to right order
990 of their opening parenthesis. A backreference matches whatever
991 actually matched the subpattern in the string being examined, not
992 the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
993 match "0x1234 0x4321", but not "0x1234 01234", because subpattern
994 1 matched "0x", even though the rule C<0|0x> could potentially match
995 the leading 0 in the second number.
997 =head2 Warning on \1 vs $1
999 Some people get too used to writing things like:
1001 $pattern =~ s/(\W)/\\\1/g;
1003 This is grandfathered for the RHS of a substitute to avoid shocking the
1004 B<sed> addicts, but it's a dirty habit to get into. That's because in
1005 PerlThink, the righthand side of a C<s///> is a double-quoted string. C<\1> in
1006 the usual double-quoted string means a control-A. The customary Unix
1007 meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1008 of doing that, you get yourself into trouble if you then add an C</e>
1011 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1017 You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
1018 C<${1}000>. The operation of interpolation should not be confused
1019 with the operation of matching a backreference. Certainly they mean two
1020 different things on the I<left> side of the C<s///>.
1022 =head2 Repeated patterns matching zero-length substring
1024 B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
1026 Regular expressions provide a terse and powerful programming language. As
1027 with most other power tools, power comes together with the ability
1030 A common abuse of this power stems from the ability to make infinite
1031 loops using regular expressions, with something as innocuous as:
1033 'foo' =~ m{ ( o? )* }x;
1035 The C<o?> can match at the beginning of C<'foo'>, and since the position
1036 in the string is not moved by the match, C<o?> would match again and again
1037 because of the C<*> modifier. Another common way to create a similar cycle
1038 is with the looping modifier C<//g>:
1040 @matches = ( 'foo' =~ m{ o? }xg );
1044 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1046 or the loop implied by split().
1048 However, long experience has shown that many programming tasks may
1049 be significantly simplified by using repeated subexpressions that
1050 may match zero-length substrings. Here's a simple example being:
1052 @chars = split //, $string; # // is not magic in split
1053 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1055 Thus Perl allows such constructs, by I<forcefully breaking
1056 the infinite loop>. The rules for this are different for lower-level
1057 loops given by the greedy modifiers C<*+{}>, and for higher-level
1058 ones like the C</g> modifier or split() operator.
1060 The lower-level loops are I<interrupted> (that is, the loop is
1061 broken) when Perl detects that a repeated expression matched a
1062 zero-length substring. Thus
1064 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1066 is made equivalent to
1068 m{ (?: NON_ZERO_LENGTH )*
1073 The higher level-loops preserve an additional state between iterations:
1074 whether the last match was zero-length. To break the loop, the following
1075 match after a zero-length match is prohibited to have a length of zero.
1076 This prohibition interacts with backtracking (see L<"Backtracking">),
1077 and so the I<second best> match is chosen if the I<best> match is of
1085 results in C<"<><b><><a><><r><>">. At each position of the string the best
1086 match given by non-greedy C<??> is the zero-length match, and the I<second
1087 best> match is what is matched by C<\w>. Thus zero-length matches
1088 alternate with one-character-long matches.
1090 Similarly, for repeated C<m/()/g> the second-best match is the match at the
1091 position one notch further in the string.
1093 The additional state of being I<matched with zero-length> is associated with
1094 the matched string, and is reset by each assignment to pos().
1095 Zero-length matches at the end of the previous match are ignored
1098 =head2 Combining pieces together
1100 Each of the elementary pieces of regular expressions which were described
1101 before (such as C<ab> or C<\Z>) could match at most one substring
1102 at the given position of the input string. However, in a typical regular
1103 expression these elementary pieces are combined into more complicated
1104 patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1105 (in these examples C<S> and C<T> are regular subexpressions).
1107 Such combinations can include alternatives, leading to a problem of choice:
1108 if we match a regular expression C<a|ab> against C<"abc">, will it match
1109 substring C<"a"> or C<"ab">? One way to describe which substring is
1110 actually matched is the concept of backtracking (see L<"Backtracking">).
1111 However, this description is too low-level and makes you think
1112 in terms of a particular implementation.
1114 Another description starts with notions of "better"/"worse". All the
1115 substrings which may be matched by the given regular expression can be
1116 sorted from the "best" match to the "worst" match, and it is the "best"
1117 match which is chosen. This substitutes the question of "what is chosen?"
1118 by the question of "which matches are better, and which are worse?".
1120 Again, for elementary pieces there is no such question, since at most
1121 one match at a given position is possible. This section describes the
1122 notion of better/worse for combining operators. In the description
1123 below C<S> and C<T> are regular subexpressions.
1129 Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1130 substrings which can be matched by C<S>, C<B> and C<B'> are substrings
1131 which can be matched by C<T>.
1133 If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1136 If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1137 C<B> is better match for C<T> than C<B'>.
1141 When C<S> can match, it is a better match than when only C<T> can match.
1143 Ordering of two matches for C<S> is the same as for C<S>. Similar for
1144 two matches for C<T>.
1146 =item C<S{REPEAT_COUNT}>
1148 Matches as C<SSS...S> (repeated as many times as necessary).
1152 Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1154 =item C<S{min,max}?>
1156 Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1158 =item C<S?>, C<S*>, C<S+>
1160 Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1162 =item C<S??>, C<S*?>, C<S+?>
1164 Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1168 Matches the best match for C<S> and only that.
1170 =item C<(?=S)>, C<(?<=S)>
1172 Only the best match for C<S> is considered. (This is important only if
1173 C<S> has capturing parentheses, and backreferences are used somewhere
1174 else in the whole regular expression.)
1176 =item C<(?!S)>, C<(?<!S)>
1178 For this grouping operator there is no need to describe the ordering, since
1179 only whether or not C<S> can match is important.
1181 =item C<(??{ EXPR })>
1183 The ordering is the same as for the regular expression which is
1186 =item C<(?(condition)yes-pattern|no-pattern)>
1188 Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1189 already determined. The ordering of the matches is the same as for the
1190 chosen subexpression.
1194 The above recipes describe the ordering of matches I<at a given position>.
1195 One more rule is needed to understand how a match is determined for the
1196 whole regular expression: a match at an earlier position is always better
1197 than a match at a later position.
1199 =head2 Creating custom RE engines
1201 Overloaded constants (see L<overload>) provide a simple way to extend
1202 the functionality of the RE engine.
1204 Suppose that we want to enable a new RE escape-sequence C<\Y|> which
1205 matches at boundary between white-space characters and non-whitespace
1206 characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1207 at these positions, so we want to have each C<\Y|> in the place of the
1208 more complicated version. We can create a module C<customre> to do
1216 die "No argument to customre::import allowed" if @_;
1217 overload::constant 'qr' => \&convert;
1220 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1222 my %rules = ( '\\' => '\\',
1223 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
1229 { $rules{$1} or invalid($re,$1) }sgex;
1233 Now C<use customre> enables the new escape in constant regular
1234 expressions, i.e., those without any runtime variable interpolations.
1235 As documented in L<overload>, this conversion will work only over
1236 literal parts of regular expressions. For C<\Y|$re\Y|> the variable
1237 part of this regular expression needs to be converted explicitly
1238 (but only if the special meaning of C<\Y|> should be enabled inside $re):
1243 $re = customre::convert $re;
1248 This document varies from difficult to understand to completely
1249 and utterly opaque. The wandering prose riddled with jargon is
1250 hard to fathom in several places.
1252 This document needs a rewrite that separates the tutorial content
1253 from the reference content.
1257 L<perlop/"Regexp Quote-Like Operators">.
1259 L<perlop/"Gory details of parsing quoted constructs">.
1267 I<Mastering Regular Expressions> by Jeffrey Friedl, published
1268 by O'Reilly and Associates.