This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perlre: Add heading to separate unrelated paragraphs
[perl5.git] / pod / perlre.pod
CommitLineData
a0d0e21e 1=head1 NAME
d74e8afc 2X<regular expression> X<regex> X<regexp>
a0d0e21e
LW
3
4perlre - Perl regular expressions
5
6=head1 DESCRIPTION
7
5d458dd8 8This page describes the syntax of regular expressions in Perl.
91e0c79e 9
cc46d5f2 10If you haven't used regular expressions before, a quick-start
91e0c79e
MJD
11introduction is available in L<perlrequick>, and a longer tutorial
12introduction is available in L<perlretut>.
13
14For reference on how regular expressions are used in matching
15operations, plus various examples of the same, see discussions of
16C<m//>, C<s///>, C<qr//> and C<??> in L<perlop/"Regexp Quote-Like
17Operators">.
cb1a09d0 18
0d017f4d
WL
19
20=head2 Modifiers
21
19799a22 22Matching operations can have various modifiers. Modifiers
5a964f20 23that relate to the interpretation of the regular expression inside
19799a22 24are listed below. Modifiers that alter the way a regular expression
5d458dd8 25is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and
1e66bd83 26L<perlop/"Gory details of parsing quoted constructs">.
a0d0e21e 27
55497cff 28=over 4
29
54310121 30=item m
d74e8afc 31X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline>
55497cff 32
33Treat string as multiple lines. That is, change "^" and "$" from matching
14218588 34the start or end of the string to matching the start or end of any
7f761169 35line anywhere within the string.
55497cff 36
54310121 37=item s
d74e8afc
ITB
38X</s> X<regex, single-line> X<regexp, single-line>
39X<regular expression, single-line>
55497cff 40
41Treat string as single line. That is, change "." to match any character
19799a22 42whatsoever, even a newline, which normally it would not match.
55497cff 43
34d67d80 44Used together, as C</ms>, they let the "." match any character whatsoever,
fb55449c 45while still allowing "^" and "$" to match, respectively, just after
19799a22 46and just before newlines within the string.
7b8d334a 47
87e95b7f
YO
48=item i
49X</i> X<regex, case-insensitive> X<regexp, case-insensitive>
50X<regular expression, case-insensitive>
51
52Do case-insensitive pattern matching.
53
54If C<use locale> is in effect, the case map is taken from the current
55locale. See L<perllocale>.
56
54310121 57=item x
d74e8afc 58X</x>
55497cff 59
60Extend your pattern's legibility by permitting whitespace and comments.
61
87e95b7f
YO
62=item p
63X</p> X<regex, preserve> X<regexp, preserve>
64
632a1772 65Preserve the string matched such that ${^PREMATCH}, ${^MATCH}, and
87e95b7f
YO
66${^POSTMATCH} are available for use after matching.
67
e2e6bec7
DN
68=item g and c
69X</g> X</c>
70
71Global matching, and keep the Current position after failed matching.
72Unlike i, m, s and x, these two flags affect the way the regex is used
73rather than the regex itself. See
74L<perlretut/"Using regular expressions in Perl"> for further explanation
75of the g and c modifiers.
76
55497cff 77=back
a0d0e21e
LW
78
79These are usually written as "the C</x> modifier", even though the delimiter
14218588 80in question might not really be a slash. Any of these
a0d0e21e 81modifiers may also be embedded within the regular expression itself using
14218588 82the C<(?...)> construct. See below.
a0d0e21e 83
4633a7c4 84The C</x> modifier itself needs a little more explanation. It tells
7b059540 85the regular expression parser to ignore most whitespace that is neither
55497cff 86backslashed nor within a character class. You can use this to break up
4633a7c4 87your regular expression into (slightly) more readable parts. The C<#>
54310121 88character is also treated as a metacharacter introducing a comment,
55497cff 89just as in ordinary Perl code. This also means that if you want real
14218588 90whitespace or C<#> characters in the pattern (outside a character
f9a3ff1a 91class, where they are unaffected by C</x>), then you'll either have to
7b059540
KW
92escape them (using backslashes or C<\Q...\E>) or encode them using octal,
93hex, or C<\N{}> escapes. Taken together, these features go a long way towards
8933a740
RGS
94making Perl's regular expressions more readable. Note that you have to
95be careful not to include the pattern delimiter in the comment--perl has
96no way of knowing you did not intend to close the pattern early. See
97the C-comment deletion code in L<perlop>. Also note that anything inside
7651b971
KW
98a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect
99whether space interpretation within a single multi-character construct. For
100example in C<\x{...}>, regardless of the C</x> modifier, there can be no
9bb1f947 101spaces. Same for a L<quantifier|/Quantifiers> such as C<{3}> or
f9e949fd
KW
102C<{5,}>. Similarly, C<(?:...)> can't have a space between the C<?> and C<:>,
103but can between the C<(> and C<?>. Within any delimiters for such a
104construct, allowed spaces are not affected by C</x>, and depend on the
105construct. For example, C<\x{...}> can't have spaces because hexadecimal
106numbers don't have spaces in them. But, Unicode properties can have spaces, so
107in C<\p{...}> there can be spaces that follow the Unicode rules, for which see
9bb1f947 108L<perluniprops/Properties accessible through \p{} and \P{}>.
d74e8afc 109X</x>
a0d0e21e
LW
110
111=head2 Regular Expressions
112
04838cea
RGS
113=head3 Metacharacters
114
384f06ae 115The patterns used in Perl pattern matching evolved from those supplied in
14218588 116the Version 8 regex routines. (The routines are derived
19799a22
GS
117(distantly) from Henry Spencer's freely redistributable reimplementation
118of the V8 routines.) See L<Version 8 Regular Expressions> for
119details.
a0d0e21e
LW
120
121In particular the following metacharacters have their standard I<egrep>-ish
122meanings:
d74e8afc
ITB
123X<metacharacter>
124X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
125
a0d0e21e 126
f793d64a
KW
127 \ Quote the next metacharacter
128 ^ Match the beginning of the line
129 . Match any character (except newline)
130 $ Match the end of the line (or before newline at the end)
131 | Alternation
132 () Grouping
133 [] Bracketed Character class
a0d0e21e 134
14218588
GS
135By default, the "^" character is guaranteed to match only the
136beginning of the string, the "$" character only the end (or before the
137newline at the end), and Perl does certain optimizations with the
a0d0e21e
LW
138assumption that the string contains only one line. Embedded newlines
139will not be matched by "^" or "$". You may, however, wish to treat a
4a6725af 140string as a multi-line buffer, such that the "^" will match after any
0d520e8e
YO
141newline within the string (except if the newline is the last character in
142the string), and "$" will match before any newline. At the
a0d0e21e
LW
143cost of a little more overhead, you can do this by using the /m modifier
144on the pattern match operator. (Older programs did this by setting C<$*>,
f02c194e 145but this practice has been removed in perl 5.9.)
d74e8afc 146X<^> X<$> X</m>
a0d0e21e 147
14218588 148To simplify multi-line substitutions, the "." character never matches a
55497cff 149newline unless you use the C</s> modifier, which in effect tells Perl to pretend
f02c194e 150the string is a single line--even if it isn't.
d74e8afc 151X<.> X</s>
a0d0e21e 152
04838cea
RGS
153=head3 Quantifiers
154
a0d0e21e 155The following standard quantifiers are recognized:
d74e8afc 156X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
a0d0e21e 157
f793d64a
KW
158 * Match 0 or more times
159 + Match 1 or more times
160 ? Match 1 or 0 times
161 {n} Match exactly n times
162 {n,} Match at least n times
163 {n,m} Match at least n but not more than m times
a0d0e21e
LW
164
165(If a curly bracket occurs in any other context, it is treated
b975c076 166as a regular character. In particular, the lower bound
527e91da
BB
167is not optional.) The "*" quantifier is equivalent to C<{0,}>, the "+"
168quantifier to C<{1,}>, and the "?" quantifier to C<{0,1}>. n and m are limited
d0b16107 169to non-negative integral values less than a preset limit defined when perl is built.
9c79236d
GS
170This is usually 32766 on the most common platforms. The actual limit can
171be seen in the error message generated by code such as this:
172
820475bd 173 $_ **= $_ , / {$_} / for 2 .. 42;
a0d0e21e 174
54310121 175By default, a quantified subpattern is "greedy", that is, it will match as
176many times as possible (given a particular starting location) while still
177allowing the rest of the pattern to match. If you want it to match the
178minimum number of times possible, follow the quantifier with a "?". Note
179that the meanings don't change, just the "greediness":
0d017f4d 180X<metacharacter> X<greedy> X<greediness>
d74e8afc 181X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
a0d0e21e 182
f793d64a
KW
183 *? Match 0 or more times, not greedily
184 +? Match 1 or more times, not greedily
185 ?? Match 0 or 1 time, not greedily
186 {n}? Match exactly n times, not greedily
187 {n,}? Match at least n times, not greedily
188 {n,m}? Match at least n but not more than m times, not greedily
a0d0e21e 189
b9b4dddf
YO
190By default, when a quantified subpattern does not allow the rest of the
191overall pattern to match, Perl will backtrack. However, this behaviour is
0d017f4d 192sometimes undesirable. Thus Perl provides the "possessive" quantifier form
b9b4dddf
YO
193as well.
194
f793d64a
KW
195 *+ Match 0 or more times and give nothing back
196 ++ Match 1 or more times and give nothing back
197 ?+ Match 0 or 1 time and give nothing back
198 {n}+ Match exactly n times and give nothing back (redundant)
199 {n,}+ Match at least n times and give nothing back
200 {n,m}+ Match at least n but not more than m times and give nothing back
b9b4dddf
YO
201
202For instance,
203
204 'aaaa' =~ /a++a/
205
206will never match, as the C<a++> will gobble up all the C<a>'s in the
207string and won't leave any for the remaining part of the pattern. This
208feature can be extremely useful to give perl hints about where it
209shouldn't backtrack. For instance, the typical "match a double-quoted
210string" problem can be most efficiently performed when written as:
211
212 /"(?:[^"\\]++|\\.)*+"/
213
0d017f4d 214as we know that if the final quote does not match, backtracking will not
b9b4dddf
YO
215help. See the independent subexpression C<< (?>...) >> for more details;
216possessive quantifiers are just syntactic sugar for that construct. For
217instance the above example could also be written as follows:
218
219 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
220
04838cea
RGS
221=head3 Escape sequences
222
5f05dabc 223Because patterns are processed as double quoted strings, the following
a0d0e21e
LW
224also work:
225
f793d64a
KW
226 \t tab (HT, TAB)
227 \n newline (LF, NL)
228 \r return (CR)
229 \f form feed (FF)
230 \a alarm (bell) (BEL)
231 \e escape (think troff) (ESC)
232 \033 octal char (example: ESC)
233 \x1B hex char (example: ESC)
234 \x{263a} long hex char (example: Unicode SMILEY)
235 \cK control char (example: VT)
236 \N{name} named Unicode character
237 \N{U+263D} Unicode character (example: FIRST QUARTER MOON)
238 \l lowercase next char (think vi)
239 \u uppercase next char (think vi)
240 \L lowercase till \E (think vi)
241 \U uppercase till \E (think vi)
242 \Q quote (disable) pattern metacharacters till \E
243 \E end either case modification or quoted section, think vi
a0d0e21e 244
9bb1f947 245Details are in L<perlop/Quote and Quote-like Operators>.
1d2dff63 246
e1d1eefb 247=head3 Character Classes and other Special Escapes
04838cea 248
a0d0e21e 249In addition, Perl defines the following:
d0b16107 250X<\g> X<\k> X<\K> X<backreference>
a0d0e21e 251
f793d64a
KW
252 Sequence Note Description
253 [...] [1] Match a character according to the rules of the
254 bracketed character class defined by the "...".
255 Example: [a-z] matches "a" or "b" or "c" ... or "z"
256 [[:...:]] [2] Match a character according to the rules of the POSIX
257 character class "..." within the outer bracketed
258 character class. Example: [[:upper:]] matches any
259 uppercase character.
260 \w [3] Match a "word" character (alphanumeric plus "_")
261 \W [3] Match a non-"word" character
262 \s [3] Match a whitespace character
263 \S [3] Match a non-whitespace character
264 \d [3] Match a decimal digit character
265 \D [3] Match a non-digit character
266 \pP [3] Match P, named property. Use \p{Prop} for longer names
267 \PP [3] Match non-P
268 \X [4] Match Unicode "eXtended grapheme cluster"
269 \C Match a single C-language char (octet) even if that is
270 part of a larger UTF-8 character. Thus it breaks up
271 characters into their UTF-8 bytes, so you may end up
272 with malformed pieces of UTF-8. Unsupported in
273 lookbehind.
c27a5cfe 274 \1 [5] Backreference to a specific capture group or buffer.
f793d64a
KW
275 '1' may actually be any positive integer.
276 \g1 [5] Backreference to a specific or previous group,
277 \g{-1} [5] The number may be negative indicating a relative
c27a5cfe 278 previous group and may optionally be wrapped in
f793d64a
KW
279 curly brackets for safer parsing.
280 \g{name} [5] Named backreference
281 \k<name> [5] Named backreference
282 \K [6] Keep the stuff left of the \K, don't include it in $&
283 \N [7] Any character but \n (experimental). Not affected by
284 /s modifier
285 \v [3] Vertical whitespace
286 \V [3] Not vertical whitespace
287 \h [3] Horizontal whitespace
288 \H [3] Not horizontal whitespace
289 \R [4] Linebreak
e1d1eefb 290
9bb1f947
KW
291=over 4
292
293=item [1]
294
295See L<perlrecharclass/Bracketed Character Classes> for details.
df225385 296
9bb1f947 297=item [2]
b8c5462f 298
9bb1f947 299See L<perlrecharclass/POSIX Character Classes> for details.
b8c5462f 300
9bb1f947 301=item [3]
5496314a 302
9bb1f947 303See L<perlrecharclass/Backslash sequences> for details.
5496314a 304
9bb1f947 305=item [4]
5496314a 306
9bb1f947 307See L<perlrebackslash/Misc> for details.
d0b16107 308
9bb1f947 309=item [5]
b8c5462f 310
c27a5cfe 311See L</Capture groups> below for details.
93733859 312
9bb1f947 313=item [6]
b8c5462f 314
9bb1f947
KW
315See L</Extended Patterns> below for details.
316
317=item [7]
318
319Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the
320character whose name is C<NAME>; and similarly when of the form
321C<\N{U+I<wide hex char>}>, it matches the character whose Unicode ordinal is
322I<wide hex char>. Otherwise it matches any character but C<\n>.
323
324=back
d0b16107 325
04838cea
RGS
326=head3 Assertions
327
a0d0e21e 328Perl defines the following zero-width assertions:
d74e8afc
ITB
329X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
330X<regexp, zero-width assertion>
331X<regular expression, zero-width assertion>
332X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
a0d0e21e 333
9bb1f947
KW
334 \b Match a word boundary
335 \B Match except at a word boundary
336 \A Match only at beginning of string
337 \Z Match only at end of string, or before newline at the end
338 \z Match only at end of string
339 \G Match only at pos() (e.g. at the end-of-match position
9da458fc 340 of prior m//g)
a0d0e21e 341
14218588 342A word boundary (C<\b>) is a spot between two characters
19799a22
GS
343that has a C<\w> on one side of it and a C<\W> on the other side
344of it (in either order), counting the imaginary characters off the
345beginning and end of the string as matching a C<\W>. (Within
346character classes C<\b> represents backspace rather than a word
347boundary, just as it normally does in any double-quoted string.)
348The C<\A> and C<\Z> are just like "^" and "$", except that they
349won't match multiple times when the C</m> modifier is used, while
350"^" and "$" will match at every internal line boundary. To match
351the actual end of the string and not ignore an optional trailing
352newline, use C<\z>.
d74e8afc 353X<\b> X<\A> X<\Z> X<\z> X</m>
19799a22
GS
354
355The C<\G> assertion can be used to chain global matches (using
356C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
357It is also useful when writing C<lex>-like scanners, when you have
358several patterns that you want to match against consequent substrings
359of your string, see the previous reference. The actual location
360where C<\G> will match can also be influenced by using C<pos()> as
58e23c8d
YO
361an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
362matches is modified somewhat, in that contents to the left of C<\G> is
363not counted when determining the length of the match. Thus the following
364will not match forever:
d74e8afc 365X<\G>
c47ff5f1 366
58e23c8d
YO
367 $str = 'ABC';
368 pos($str) = 1;
369 while (/.\G/g) {
370 print $&;
371 }
372
373It will print 'A' and then terminate, as it considers the match to
374be zero-width, and thus will not match at the same position twice in a
375row.
376
377It is worth noting that C<\G> improperly used can result in an infinite
378loop. Take care when using patterns that include C<\G> in an alternation.
379
c27a5cfe 380=head3 Capture groups
04838cea 381
c27a5cfe
KW
382The bracketing construct C<( ... )> creates capture groups (also referred to as
383capture buffers). To refer to the current contents of a group later on, within
d8b950dc
KW
384the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>)
385for the second, and so on.
386This is called a I<backreference>.
d74e8afc 387X<regex, capture buffer> X<regexp, capture buffer>
c27a5cfe 388X<regex, capture group> X<regexp, capture group>
d74e8afc 389X<regular expression, capture buffer> X<backreference>
c27a5cfe 390X<regular expression, capture group> X<backreference>
1f1031fe 391X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
d8b950dc
KW
392X<named capture buffer> X<regular expression, named capture buffer>
393X<named capture group> X<regular expression, named capture group>
394X<%+> X<$+{name}> X<< \k<name> >>
395There is no limit to the number of captured substrings that you may use.
396Groups are numbered with the leftmost open parenthesis being number 1, etc. If
397a group did not match, the associated backreference won't match either. (This
398can happen if the group is optional, or in a different branch of an
399alternation.)
400You can omit the C<"g">, and write C<"\1">, etc, but there are some issues with
401this form, described below.
402
403You can also refer to capture groups relatively, by using a negative number, so
404that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture
405group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For
406example:
5624f11d
YO
407
408 /
c27a5cfe
KW
409 (Y) # group 1
410 ( # group 2
411 (X) # group 3
412 \g{-1} # backref to group 3
413 \g{-3} # backref to group 1
5624f11d
YO
414 )
415 /x
416
d8b950dc
KW
417would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to
418interpolate regexes into larger regexes and not have to worry about the
419capture groups being renumbered.
420
421You can dispense with numbers altogether and create named capture groups.
422The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to
423reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may
424also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.)
425I<name> must not begin with a number, nor contain hyphens.
426When different groups within the same pattern have the same name, any reference
427to that name assumes the leftmost defined group. Named groups count in
428absolute and relative numbering, and so can also be referred to by those
429numbers.
430(It's possible to do things with named capture groups that would otherwise
431require C<(??{})>.)
432
433Capture group contents are dynamically scoped and available to you outside the
434pattern until the end of the enclosing block or until the next successful
435match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
436You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">,
437etc); or by name via the C<%+> hash, using C<"$+{I<name>}">.
438
439Braces are required in referring to named capture groups, but are optional for
440absolute or relative numbered ones. Braces are safer when creating a regex by
441concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a>
442contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which
443is probably not what you intended.
444
445The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that
446there were no named nor relative numbered capture groups. Absolute numbered
447groups were referred to using C<\1>, C<\2>, etc, and this notation is still
448accepted (and likely always will be). But it leads to some ambiguities if
449there are more than 9 capture groups, as C<\10> could mean either the tenth
450capture group, or the character whose ordinal in octal is 010 (a backspace in
451ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference
452only if at least 10 left parentheses have opened before it. Likewise C<\11> is
453a backreference only if at least 11 left parentheses have opened before it.
454And so on. C<\1> through C<\9> are always interpreted as backreferences. You
455can minimize the ambiguity by always using C<\g> if you mean capturing groups;
456and always using 3 digits for octal constants, with the first always "0" (which
457works if there are 63 (= \077) or fewer capture groups).
458
459The C<\I<digit>> notation also works in certain circumstances outside
460the pattern. See L</Warning on \1 Instead of $1> below for details.)
81714fb9 461
14218588 462Examples:
a0d0e21e
LW
463
464 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
465
d8b950dc 466 /(.)\g1/ # find first doubled char
81714fb9
YO
467 and print "'$1' is the first doubled character\n";
468
469 /(?<char>.)\k<char>/ # ... a different way
470 and print "'$+{char}' is the first doubled character\n";
471
d8b950dc 472 /(?'char'.)\g1/ # ... mix and match
81714fb9 473 and print "'$1' is the first doubled character\n";
c47ff5f1 474
14218588 475 if (/Time: (..):(..):(..)/) { # parse out values
f793d64a
KW
476 $hours = $1;
477 $minutes = $2;
478 $seconds = $3;
a0d0e21e 479 }
c47ff5f1 480
14218588
GS
481Several special variables also refer back to portions of the previous
482match. C<$+> returns whatever the last bracket match matched.
483C<$&> returns the entire matched string. (At one point C<$0> did
484also, but now it returns the name of the program.) C<$`> returns
77ea4f6d
JV
485everything before the matched string. C<$'> returns everything
486after the matched string. And C<$^N> contains whatever was matched by
487the most-recently closed group (submatch). C<$^N> can be used in
488extended patterns (see below), for example to assign a submatch to a
81714fb9 489variable.
d74e8afc 490X<$+> X<$^N> X<$&> X<$`> X<$'>
14218588 491
d8b950dc
KW
492These special variables, like the C<%+> hash and the numbered match variables
493(C<$1>, C<$2>, C<$3>, etc.) are dynamically scoped
14218588
GS
494until the end of the enclosing block or until the next successful
495match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
d74e8afc
ITB
496X<$+> X<$^N> X<$&> X<$`> X<$'>
497X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
498
0d017f4d 499B<NOTE>: Failed matches in Perl do not reset the match variables,
5146ce24 500which makes it easier to write code that tests for a series of more
665e98b9
JH
501specific cases and remembers the best match.
502
14218588
GS
503B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
504C<$'> anywhere in the program, it has to provide them for every
505pattern match. This may substantially slow your program. Perl
d8b950dc 506uses the same mechanism to produce C<$1>, C<$2>, etc, so you also pay a
14218588
GS
507price for each pattern that contains capturing parentheses. (To
508avoid this cost while retaining the grouping behaviour, use the
509extended regular expression C<(?: ... )> instead.) But if you never
510use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
511parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
512if you can, but if you can't (and some algorithms really appreciate
513them), once you've used them once, use them at will, because you've
514already paid the price. As of 5.005, C<$&> is not so costly as the
515other two.
d74e8afc 516X<$&> X<$`> X<$'>
68dc0745 517
99d59c4d 518As a workaround for this problem, Perl 5.10.0 introduces C<${^PREMATCH}>,
cde0cee5
YO
519C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&>
520and C<$'>, B<except> that they are only guaranteed to be defined after a
87e95b7f 521successful match that was executed with the C</p> (preserve) modifier.
cde0cee5
YO
522The use of these variables incurs no global performance penalty, unlike
523their punctuation char equivalents, however at the trade-off that you
524have to tell perl when you want to use them.
87e95b7f 525X</p> X<p modifier>
cde0cee5 526
9d727203
KW
527=head2 Quoting metacharacters
528
19799a22
GS
529Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
530C<\w>, C<\n>. Unlike some other regular expression languages, there
531are no backslashed symbols that aren't alphanumeric. So anything
c47ff5f1 532that looks like \\, \(, \), \<, \>, \{, or \} is always
19799a22
GS
533interpreted as a literal character, not a metacharacter. This was
534once used in a common idiom to disable or quote the special meanings
535of regular expression metacharacters in a string that you want to
36bbe248 536use for a pattern. Simply quote all non-"word" characters:
a0d0e21e
LW
537
538 $pattern =~ s/(\W)/\\$1/g;
539
f1cbbd6e 540(If C<use locale> is set, then this depends on the current locale.)
14218588
GS
541Today it is more common to use the quotemeta() function or the C<\Q>
542metaquoting escape sequence to disable all metacharacters' special
543meanings like this:
a0d0e21e
LW
544
545 /$unquoted\Q$quoted\E$unquoted/
546
9da458fc
IZ
547Beware that if you put literal backslashes (those not inside
548interpolated variables) between C<\Q> and C<\E>, double-quotish
549backslash interpolation may lead to confusing results. If you
550I<need> to use literal backslashes within C<\Q...\E>,
551consult L<perlop/"Gory details of parsing quoted constructs">.
552
19799a22
GS
553=head2 Extended Patterns
554
14218588
GS
555Perl also defines a consistent extension syntax for features not
556found in standard tools like B<awk> and B<lex>. The syntax is a
557pair of parentheses with a question mark as the first thing within
558the parentheses. The character after the question mark indicates
559the extension.
19799a22 560
14218588
GS
561The stability of these extensions varies widely. Some have been
562part of the core language for many years. Others are experimental
563and may change without warning or be completely removed. Check
564the documentation on an individual feature to verify its current
565status.
19799a22 566
14218588
GS
567A question mark was chosen for this and for the minimal-matching
568construct because 1) question marks are rare in older regular
569expressions, and 2) whenever you see one, you should stop and
570"question" exactly what is going on. That's psychology...
a0d0e21e
LW
571
572=over 10
573
cc6b7395 574=item C<(?#text)>
d74e8afc 575X<(?#)>
a0d0e21e 576
14218588 577A comment. The text is ignored. If the C</x> modifier enables
19799a22 578whitespace formatting, a simple C<#> will suffice. Note that Perl closes
259138e3
GS
579the comment as soon as it sees a C<)>, so there is no way to put a literal
580C<)> in the comment.
a0d0e21e 581
f7819f85 582=item C<(?pimsx-imsx)>
d74e8afc 583X<(?)>
19799a22 584
0b6d1084
JH
585One or more embedded pattern-match modifiers, to be turned on (or
586turned off, if preceded by C<->) for the remainder of the pattern or
587the remainder of the enclosing pattern group (if any). This is
588particularly useful for dynamic patterns, such as those read in from a
0d017f4d
WL
589configuration file, taken from an argument, or specified in a table
590somewhere. Consider the case where some patterns want to be case
591sensitive and some do not: The case insensitive ones merely need to
592include C<(?i)> at the front of the pattern. For example:
19799a22
GS
593
594 $pattern = "foobar";
5d458dd8 595 if ( /$pattern/i ) { }
19799a22
GS
596
597 # more flexible:
598
599 $pattern = "(?i)foobar";
5d458dd8 600 if ( /$pattern/ ) { }
19799a22 601
0b6d1084 602These modifiers are restored at the end of the enclosing group. For example,
19799a22 603
d8b950dc 604 ( (?i) blah ) \s+ \g1
19799a22 605
0d017f4d
WL
606will match C<blah> in any case, some spaces, and an exact (I<including the case>!)
607repetition of the previous word, assuming the C</x> modifier, and no C</i>
608modifier outside this group.
19799a22 609
8eb5594e
DR
610These modifiers do not carry over into named subpatterns called in the
611enclosing group. In other words, a pattern such as C<((?i)(&NAME))> does not
612change the case-sensitivity of the "NAME" pattern.
613
5530442b 614Note that the C<p> modifier is special in that it can only be enabled,
cde0cee5 615not disabled, and that its presence anywhere in a pattern has a global
5530442b 616effect. Thus C<(?-p)> and C<(?-p:...)> are meaningless and will warn
cde0cee5
YO
617when executed under C<use warnings>.
618
5a964f20 619=item C<(?:pattern)>
d74e8afc 620X<(?:)>
a0d0e21e 621
ca9dfc88
IZ
622=item C<(?imsx-imsx:pattern)>
623
5a964f20
TC
624This is for clustering, not capturing; it groups subexpressions like
625"()", but doesn't make backreferences as "()" does. So
a0d0e21e 626
5a964f20 627 @fields = split(/\b(?:a|b|c)\b/)
a0d0e21e
LW
628
629is like
630
5a964f20 631 @fields = split(/\b(a|b|c)\b/)
a0d0e21e 632
19799a22
GS
633but doesn't spit out extra fields. It's also cheaper not to capture
634characters if you don't need to.
a0d0e21e 635
19799a22 636Any letters between C<?> and C<:> act as flags modifiers as with
5d458dd8 637C<(?imsx-imsx)>. For example,
ca9dfc88
IZ
638
639 /(?s-i:more.*than).*million/i
640
14218588 641is equivalent to the more verbose
ca9dfc88
IZ
642
643 /(?:(?s-i)more.*than).*million/i
644
594d7033
YO
645=item C<(?|pattern)>
646X<(?|)> X<Branch reset>
647
648This is the "branch reset" pattern, which has the special property
c27a5cfe 649that the capture groups are numbered from the same starting point
99d59c4d 650in each alternation branch. It is available starting from perl 5.10.0.
4deaaa80 651
c27a5cfe 652Capture groups are numbered from left to right, but inside this
693596a8 653construct the numbering is restarted for each branch.
4deaaa80 654
c27a5cfe 655The numbering within each branch will be as normal, and any groups
4deaaa80
PJ
656following this construct will be numbered as though the construct
657contained only one branch, that being the one with the most capture
c27a5cfe 658groups in it.
4deaaa80
PJ
659
660This construct will be useful when you want to capture one of a
661number of alternative matches.
662
663Consider the following pattern. The numbers underneath show in
c27a5cfe 664which group the captured content will be stored.
594d7033
YO
665
666
667 # before ---------------branch-reset----------- after
668 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
669 # 1 2 2 3 2 3 4
670
ab106183
A
671Be careful when using the branch reset pattern in combination with
672named captures. Named captures are implemented as being aliases to
c27a5cfe 673numbered groups holding the captures, and that interferes with the
ab106183
A
674implementation of the branch reset pattern. If you are using named
675captures in a branch reset pattern, it's best to use the same names,
676in the same order, in each of the alternations:
677
678 /(?| (?<a> x ) (?<b> y )
679 | (?<a> z ) (?<b> w )) /x
680
681Not doing so may lead to surprises:
682
683 "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x;
684 say $+ {a}; # Prints '12'
685 say $+ {b}; # *Also* prints '12'.
686
c27a5cfe
KW
687The problem here is that both the group named C<< a >> and the group
688named C<< b >> are aliases for the group belonging to C<< $1 >>.
90a18110 689
ee9b8eae
YO
690=item Look-Around Assertions
691X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
692
693Look-around assertions are zero width patterns which match a specific
694pattern without including it in C<$&>. Positive assertions match when
695their subpattern matches, negative assertions match when their subpattern
696fails. Look-behind matches text up to the current match position,
697look-ahead matches text following the current match position.
698
699=over 4
700
5a964f20 701=item C<(?=pattern)>
d74e8afc 702X<(?=)> X<look-ahead, positive> X<lookahead, positive>
a0d0e21e 703
19799a22 704A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
a0d0e21e
LW
705matches a word followed by a tab, without including the tab in C<$&>.
706
5a964f20 707=item C<(?!pattern)>
d74e8afc 708X<(?!)> X<look-ahead, negative> X<lookahead, negative>
a0d0e21e 709
19799a22 710A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
a0d0e21e 711matches any occurrence of "foo" that isn't followed by "bar". Note
19799a22
GS
712however that look-ahead and look-behind are NOT the same thing. You cannot
713use this for look-behind.
7b8d334a 714
5a964f20 715If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
7b8d334a
GS
716will not do what you want. That's because the C<(?!foo)> is just saying that
717the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
718match. You would have to do something like C</(?!foo)...bar/> for that. We
719say "like" because there's the case of your "bar" not having three characters
720before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
721Sometimes it's still easier just to say:
a0d0e21e 722
a3cb178b 723 if (/bar/ && $` !~ /foo$/)
a0d0e21e 724
19799a22 725For look-behind see below.
c277df42 726
ee9b8eae
YO
727=item C<(?<=pattern)> C<\K>
728X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K>
c277df42 729
c47ff5f1 730A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
19799a22
GS
731matches a word that follows a tab, without including the tab in C<$&>.
732Works only for fixed-width look-behind.
c277df42 733
ee9b8eae
YO
734There is a special form of this construct, called C<\K>, which causes the
735regex engine to "keep" everything it had matched prior to the C<\K> and
736not include it in C<$&>. This effectively provides variable length
737look-behind. The use of C<\K> inside of another look-around assertion
738is allowed, but the behaviour is currently not well defined.
739
c62285ac 740For various reasons C<\K> may be significantly more efficient than the
ee9b8eae
YO
741equivalent C<< (?<=...) >> construct, and it is especially useful in
742situations where you want to efficiently remove something following
743something else in a string. For instance
744
745 s/(foo)bar/$1/g;
746
747can be rewritten as the much more efficient
748
749 s/foo\Kbar//g;
750
5a964f20 751=item C<(?<!pattern)>
d74e8afc 752X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
c277df42 753
19799a22
GS
754A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
755matches any occurrence of "foo" that does not follow "bar". Works
756only for fixed-width look-behind.
c277df42 757
ee9b8eae
YO
758=back
759
81714fb9
YO
760=item C<(?'NAME'pattern)>
761
762=item C<< (?<NAME>pattern) >>
763X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
764
c27a5cfe 765A named capture group. Identical in every respect to normal capturing
90a18110 766parentheses C<()> but for the additional fact that C<%+> or C<%-> may be
c27a5cfe 767used after a successful match to refer to a named group. See C<perlvar>
90a18110 768for more details on the C<%+> and C<%-> hashes.
81714fb9 769
c27a5cfe
KW
770If multiple distinct capture groups have the same name then the
771$+{NAME} will refer to the leftmost defined group in the match.
81714fb9 772
0d017f4d 773The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent.
81714fb9
YO
774
775B<NOTE:> While the notation of this construct is the same as the similar
c27a5cfe 776function in .NET regexes, the behavior is not. In Perl the groups are
81714fb9
YO
777numbered sequentially regardless of being named or not. Thus in the
778pattern
779
780 /(x)(?<foo>y)(z)/
781
782$+{foo} will be the same as $2, and $3 will contain 'z' instead of
783the opposite which is what a .NET regex hacker might expect.
784
1f1031fe
YO
785Currently NAME is restricted to simple identifiers only.
786In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
787its Unicode extension (see L<utf8>),
788though it isn't extended by the locale (see L<perllocale>).
81714fb9 789
1f1031fe 790B<NOTE:> In order to make things easier for programmers with experience
ae5648b3 791with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >>
0d017f4d 792may be used instead of C<< (?<NAME>pattern) >>; however this form does not
64c5a566 793support the use of single quotes as a delimiter for the name.
81714fb9 794
1f1031fe
YO
795=item C<< \k<NAME> >>
796
797=item C<< \k'NAME' >>
81714fb9
YO
798
799Named backreference. Similar to numeric backreferences, except that
800the group is designated by name and not number. If multiple groups
801have the same name then it refers to the leftmost defined group in
802the current match.
803
0d017f4d 804It is an error to refer to a name not defined by a C<< (?<NAME>) >>
81714fb9
YO
805earlier in the pattern.
806
807Both forms are equivalent.
808
1f1031fe 809B<NOTE:> In order to make things easier for programmers with experience
0d017f4d 810with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >>
64c5a566 811may be used instead of C<< \k<NAME> >>.
1f1031fe 812
cc6b7395 813=item C<(?{ code })>
d74e8afc 814X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
c277df42 815
19799a22 816B<WARNING>: This extended regular expression feature is considered
b9b4dddf
YO
817experimental, and may be changed without notice. Code executed that
818has side effects may not perform identically from version to version
819due to the effect of future optimisations in the regex engine.
c277df42 820
cc46d5f2 821This zero-width assertion evaluates any embedded Perl code. It
19799a22
GS
822always succeeds, and its C<code> is not interpolated. Currently,
823the rules to determine where the C<code> ends are somewhat convoluted.
824
77ea4f6d
JV
825This feature can be used together with the special variable C<$^N> to
826capture the results of submatches in variables without having to keep
827track of the number of nested parentheses. For example:
828
829 $_ = "The brown fox jumps over the lazy dog";
830 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
831 print "color = $color, animal = $animal\n";
832
754091cb
RGS
833Inside the C<(?{...})> block, C<$_> refers to the string the regular
834expression is matching against. You can also use C<pos()> to know what is
fa11829f 835the current position of matching within this string.
754091cb 836
19799a22
GS
837The C<code> is properly scoped in the following sense: If the assertion
838is backtracked (compare L<"Backtracking">), all changes introduced after
839C<local>ization are undone, so that
b9ac3b5b
GS
840
841 $_ = 'a' x 8;
5d458dd8 842 m<
f793d64a 843 (?{ $cnt = 0 }) # Initialize $cnt.
b9ac3b5b 844 (
5d458dd8 845 a
b9ac3b5b 846 (?{
f793d64a 847 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
b9ac3b5b 848 })
5d458dd8 849 )*
b9ac3b5b 850 aaaa
f793d64a
KW
851 (?{ $res = $cnt }) # On success copy to
852 # non-localized location.
b9ac3b5b
GS
853 >x;
854
0d017f4d 855will set C<$res = 4>. Note that after the match, C<$cnt> returns to the globally
14218588 856introduced value, because the scopes that restrict C<local> operators
b9ac3b5b
GS
857are unwound.
858
19799a22
GS
859This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
860switch. If I<not> used in this way, the result of evaluation of
861C<code> is put into the special variable C<$^R>. This happens
862immediately, so C<$^R> can be used from other C<(?{ code })> assertions
863inside the same regular expression.
b9ac3b5b 864
19799a22
GS
865The assignment to C<$^R> above is properly localized, so the old
866value of C<$^R> is restored if the assertion is backtracked; compare
867L<"Backtracking">.
b9ac3b5b 868
19799a22
GS
869For reasons of security, this construct is forbidden if the regular
870expression involves run-time interpolation of variables, unless the
871perilous C<use re 'eval'> pragma has been used (see L<re>), or the
872variables contain results of C<qr//> operator (see
9bb1f947 873L<perlop/"qr/STRINGE<sol>msixpo">).
871b0233 874
0d017f4d 875This restriction is due to the wide-spread and remarkably convenient
19799a22 876custom of using run-time determined strings as patterns. For example:
871b0233
IZ
877
878 $re = <>;
879 chomp $re;
880 $string =~ /$re/;
881
14218588
GS
882Before Perl knew how to execute interpolated code within a pattern,
883this operation was completely safe from a security point of view,
884although it could raise an exception from an illegal pattern. If
885you turn on the C<use re 'eval'>, though, it is no longer secure,
886so you should only do so if you are also using taint checking.
887Better yet, use the carefully constrained evaluation within a Safe
cc46d5f2 888compartment. See L<perlsec> for details about both these mechanisms.
871b0233 889
e95d7314
GG
890B<WARNING>: Use of lexical (C<my>) variables in these blocks is
891broken. The result is unpredictable and will make perl unstable. The
892workaround is to use global (C<our>) variables.
893
894B<WARNING>: Because Perl's regex engine is currently not re-entrant,
895interpolated code may not invoke the regex engine either directly with
896C<m//> or C<s///>), or indirectly with functions such as
897C<split>. Invoking the regex engine in these blocks will make perl
898unstable.
8988a1bb 899
14455d6c 900=item C<(??{ code })>
d74e8afc
ITB
901X<(??{})>
902X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
0f5d15d6 903
19799a22 904B<WARNING>: This extended regular expression feature is considered
b9b4dddf
YO
905experimental, and may be changed without notice. Code executed that
906has side effects may not perform identically from version to version
907due to the effect of future optimisations in the regex engine.
0f5d15d6 908
19799a22
GS
909This is a "postponed" regular subexpression. The C<code> is evaluated
910at run time, at the moment this subexpression may match. The result
911of evaluation is considered as a regular expression and matched as
61528107 912if it were inserted instead of this construct. Note that this means
c27a5cfe 913that the contents of capture groups defined inside an eval'ed pattern
6bda09f9 914are not available outside of the pattern, and vice versa, there is no
c27a5cfe 915way for the inner pattern to refer to a capture group defined outside.
6bda09f9
YO
916Thus,
917
918 ('a' x 100)=~/(??{'(.)' x 100})/
919
81714fb9 920B<will> match, it will B<not> set $1.
0f5d15d6 921
428594d9 922The C<code> is not interpolated. As before, the rules to determine
19799a22
GS
923where the C<code> ends are currently somewhat convoluted.
924
925The following pattern matches a parenthesized group:
0f5d15d6
IZ
926
927 $re = qr{
f793d64a
KW
928 \(
929 (?:
930 (?> [^()]+ ) # Non-parens without backtracking
931 |
932 (??{ $re }) # Group with matching parens
933 )*
934 \)
935 }x;
0f5d15d6 936
6bda09f9
YO
937See also C<(?PARNO)> for a different, more efficient way to accomplish
938the same task.
939
0b370c0a
A
940For reasons of security, this construct is forbidden if the regular
941expression involves run-time interpolation of variables, unless the
942perilous C<use re 'eval'> pragma has been used (see L<re>), or the
943variables contain results of C<qr//> operator (see
9bb1f947 944L<perlop/"qrE<sol>STRINGE<sol>msixpo">).
0b370c0a 945
5d458dd8 946Because perl's regex engine is not currently re-entrant, delayed
8988a1bb
DD
947code may not invoke the regex engine either directly with C<m//> or C<s///>),
948or indirectly with functions such as C<split>.
949
5d458dd8
YO
950Recursing deeper than 50 times without consuming any input string will
951result in a fatal error. The maximum depth is compiled into perl, so
6bda09f9
YO
952changing it requires a custom build.
953
542fa716
YO
954=item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)>
955X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
6bda09f9 956X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
542fa716 957X<regex, relative recursion>
6bda09f9 958
81714fb9 959Similar to C<(??{ code })> except it does not involve compiling any code,
c27a5cfe
KW
960instead it treats the contents of a capture group as an independent
961pattern that must match at the current position. Capture groups
81714fb9 962contained by the pattern will have the value as determined by the
6bda09f9
YO
963outermost recursion.
964
894be9b7 965PARNO is a sequence of digits (not starting with 0) whose value reflects
c27a5cfe 966the paren-number of the capture group to recurse to. C<(?R)> recurses to
894be9b7 967the beginning of the whole pattern. C<(?0)> is an alternate syntax for
542fa716 968C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed
c27a5cfe 969to be relative, with negative numbers indicating preceding capture groups
542fa716 970and positive ones following. Thus C<(?-1)> refers to the most recently
c27a5cfe 971declared group, and C<(?+1)> indicates the next group to be declared.
c74340f9 972Note that the counting for relative recursion differs from that of
c27a5cfe 973relative backreferences, in that with recursion unclosed groups B<are>
c74340f9 974included.
6bda09f9 975
81714fb9 976The following pattern matches a function foo() which may contain
f145b7e9 977balanced parentheses as the argument.
6bda09f9
YO
978
979 $re = qr{ ( # paren group 1 (full function)
81714fb9 980 foo
6bda09f9
YO
981 ( # paren group 2 (parens)
982 \(
983 ( # paren group 3 (contents of parens)
984 (?:
985 (?> [^()]+ ) # Non-parens without backtracking
986 |
987 (?2) # Recurse to start of paren group 2
988 )*
989 )
990 \)
991 )
992 )
993 }x;
994
995If the pattern was used as follows
996
997 'foo(bar(baz)+baz(bop))'=~/$re/
998 and print "\$1 = $1\n",
999 "\$2 = $2\n",
1000 "\$3 = $3\n";
1001
1002the output produced should be the following:
1003
1004 $1 = foo(bar(baz)+baz(bop))
1005 $2 = (bar(baz)+baz(bop))
81714fb9 1006 $3 = bar(baz)+baz(bop)
6bda09f9 1007
c27a5cfe 1008If there is no corresponding capture group defined, then it is a
61528107 1009fatal error. Recursing deeper than 50 times without consuming any input
81714fb9 1010string will also result in a fatal error. The maximum depth is compiled
6bda09f9
YO
1011into perl, so changing it requires a custom build.
1012
542fa716
YO
1013The following shows how using negative indexing can make it
1014easier to embed recursive patterns inside of a C<qr//> construct
1015for later use:
1016
1017 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
1018 if (/foo $parens \s+ + \s+ bar $parens/x) {
1019 # do something here...
1020 }
1021
81714fb9 1022B<Note> that this pattern does not behave the same way as the equivalent
0d017f4d 1023PCRE or Python construct of the same form. In Perl you can backtrack into
6bda09f9 1024a recursed group, in PCRE and Python the recursed into group is treated
542fa716
YO
1025as atomic. Also, modifiers are resolved at compile time, so constructs
1026like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
1027be processed.
6bda09f9 1028
894be9b7
YO
1029=item C<(?&NAME)>
1030X<(?&NAME)>
1031
0d017f4d
WL
1032Recurse to a named subpattern. Identical to C<(?PARNO)> except that the
1033parenthesis to recurse to is determined by name. If multiple parentheses have
894be9b7
YO
1034the same name, then it recurses to the leftmost.
1035
1036It is an error to refer to a name that is not declared somewhere in the
1037pattern.
1038
1f1031fe
YO
1039B<NOTE:> In order to make things easier for programmers with experience
1040with the Python or PCRE regex engines the pattern C<< (?P>NAME) >>
64c5a566 1041may be used instead of C<< (?&NAME) >>.
1f1031fe 1042
e2e6a0f1
YO
1043=item C<(?(condition)yes-pattern|no-pattern)>
1044X<(?()>
286f584a 1045
e2e6a0f1 1046=item C<(?(condition)yes-pattern)>
286f584a 1047
e2e6a0f1
YO
1048Conditional expression. C<(condition)> should be either an integer in
1049parentheses (which is valid if the corresponding pair of parentheses
1050matched), a look-ahead/look-behind/evaluate zero-width assertion, a
c27a5cfe 1051name in angle brackets or single quotes (which is valid if a group
e2e6a0f1
YO
1052with the given name matched), or the special symbol (R) (true when
1053evaluated inside of recursion or eval). Additionally the R may be
1054followed by a number, (which will be true when evaluated when recursing
1055inside of the appropriate group), or by C<&NAME>, in which case it will
1056be true only when evaluated during recursion in the named group.
1057
1058Here's a summary of the possible predicates:
1059
1060=over 4
1061
1062=item (1) (2) ...
1063
c27a5cfe 1064Checks if the numbered capturing group has matched something.
e2e6a0f1
YO
1065
1066=item (<NAME>) ('NAME')
1067
c27a5cfe 1068Checks if a group with the given name has matched something.
e2e6a0f1
YO
1069
1070=item (?{ CODE })
1071
1072Treats the code block as the condition.
1073
1074=item (R)
1075
1076Checks if the expression has been evaluated inside of recursion.
1077
1078=item (R1) (R2) ...
1079
1080Checks if the expression has been evaluated while executing directly
1081inside of the n-th capture group. This check is the regex equivalent of
1082
1083 if ((caller(0))[3] eq 'subname') { ... }
1084
1085In other words, it does not check the full recursion stack.
1086
1087=item (R&NAME)
1088
1089Similar to C<(R1)>, this predicate checks to see if we're executing
1090directly inside of the leftmost group with a given name (this is the same
1091logic used by C<(?&NAME)> to disambiguate). It does not check the full
1092stack, but only the name of the innermost active recursion.
1093
1094=item (DEFINE)
1095
1096In this case, the yes-pattern is never directly executed, and no
1097no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1098See below for details.
1099
1100=back
1101
1102For example:
1103
1104 m{ ( \( )?
1105 [^()]+
1106 (?(1) \) )
1107 }x
1108
1109matches a chunk of non-parentheses, possibly included in parentheses
1110themselves.
1111
1112A special form is the C<(DEFINE)> predicate, which never executes directly
1113its yes-pattern, and does not allow a no-pattern. This allows to define
1114subpatterns which will be executed only by using the recursion mechanism.
1115This way, you can define a set of regular expression rules that can be
1116bundled into any pattern you choose.
1117
1118It is recommended that for this usage you put the DEFINE block at the
1119end of the pattern, and that you name any subpatterns defined within it.
1120
1121Also, it's worth noting that patterns defined this way probably will
1122not be as efficient, as the optimiser is not very clever about
1123handling them.
1124
1125An example of how this might be used is as follows:
1126
2bf803e2 1127 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
e2e6a0f1 1128 (?(DEFINE)
2bf803e2
YO
1129 (?<NAME_PAT>....)
1130 (?<ADRESS_PAT>....)
e2e6a0f1
YO
1131 )/x
1132
c27a5cfe
KW
1133Note that capture groups matched inside of recursion are not accessible
1134after the recursion returns, so the extra layer of capturing groups is
e2e6a0f1
YO
1135necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1136C<$+{NAME}> would be.
286f584a 1137
c47ff5f1 1138=item C<< (?>pattern) >>
6bda09f9 1139X<backtrack> X<backtracking> X<atomic> X<possessive>
5a964f20 1140
19799a22
GS
1141An "independent" subexpression, one which matches the substring
1142that a I<standalone> C<pattern> would match if anchored at the given
9da458fc 1143position, and it matches I<nothing other than this substring>. This
19799a22
GS
1144construct is useful for optimizations of what would otherwise be
1145"eternal" matches, because it will not backtrack (see L<"Backtracking">).
9da458fc
IZ
1146It may also be useful in places where the "grab all you can, and do not
1147give anything back" semantic is desirable.
19799a22 1148
c47ff5f1 1149For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
19799a22
GS
1150(anchored at the beginning of string, as above) will match I<all>
1151characters C<a> at the beginning of string, leaving no C<a> for
1152C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1153since the match of the subgroup C<a*> is influenced by the following
1154group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1155C<a*ab> will match fewer characters than a standalone C<a*>, since
1156this makes the tail match.
1157
c47ff5f1 1158An effect similar to C<< (?>pattern) >> may be achieved by writing
d8b950dc
KW
1159C<(?=(pattern))\g1>. This matches the same substring as a standalone
1160C<a+>, and the following C<\g1> eats the matched string; it therefore
c47ff5f1 1161makes a zero-length assertion into an analogue of C<< (?>...) >>.
19799a22
GS
1162(The difference between these two constructs is that the second one
1163uses a capturing group, thus shifting ordinals of backreferences
1164in the rest of a regular expression.)
1165
1166Consider this pattern:
c277df42 1167
871b0233 1168 m{ \(
e2e6a0f1 1169 (
f793d64a 1170 [^()]+ # x+
e2e6a0f1 1171 |
871b0233
IZ
1172 \( [^()]* \)
1173 )+
e2e6a0f1 1174 \)
871b0233 1175 }x
5a964f20 1176
19799a22
GS
1177That will efficiently match a nonempty group with matching parentheses
1178two levels deep or less. However, if there is no such group, it
1179will take virtually forever on a long string. That's because there
1180are so many different ways to split a long string into several
1181substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1182to a subpattern of the above pattern. Consider how the pattern
1183above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1184seconds, but that each extra letter doubles this time. This
1185exponential performance will make it appear that your program has
14218588 1186hung. However, a tiny change to this pattern
5a964f20 1187
e2e6a0f1
YO
1188 m{ \(
1189 (
f793d64a 1190 (?> [^()]+ ) # change x+ above to (?> x+ )
e2e6a0f1 1191 |
871b0233
IZ
1192 \( [^()]* \)
1193 )+
e2e6a0f1 1194 \)
871b0233 1195 }x
c277df42 1196
c47ff5f1 1197which uses C<< (?>...) >> matches exactly when the one above does (verifying
5a964f20
TC
1198this yourself would be a productive exercise), but finishes in a fourth
1199the time when used on a similar string with 1000000 C<a>s. Be aware,
1200however, that this pattern currently triggers a warning message under
9f1b1f2d 1201the C<use warnings> pragma or B<-w> switch saying it
6bab786b 1202C<"matches null string many times in regex">.
c277df42 1203
c47ff5f1 1204On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
19799a22 1205effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
c277df42
IZ
1206This was only 4 times slower on a string with 1000000 C<a>s.
1207
9da458fc
IZ
1208The "grab all you can, and do not give anything back" semantic is desirable
1209in many situations where on the first sight a simple C<()*> looks like
1210the correct solution. Suppose we parse text with comments being delimited
1211by C<#> followed by some optional (horizontal) whitespace. Contrary to
4375e838 1212its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
9da458fc
IZ
1213the comment delimiter, because it may "give up" some whitespace if
1214the remainder of the pattern can be made to match that way. The correct
1215answer is either one of these:
1216
1217 (?>#[ \t]*)
1218 #[ \t]*(?![ \t])
1219
1220For example, to grab non-empty comments into $1, one should use either
1221one of these:
1222
1223 / (?> \# [ \t]* ) ( .+ ) /x;
1224 / \# [ \t]* ( [^ \t] .* ) /x;
1225
1226Which one you pick depends on which of these expressions better reflects
1227the above specification of comments.
1228
6bda09f9
YO
1229In some literature this construct is called "atomic matching" or
1230"possessive matching".
1231
b9b4dddf
YO
1232Possessive quantifiers are equivalent to putting the item they are applied
1233to inside of one of these constructs. The following equivalences apply:
1234
1235 Quantifier Form Bracketing Form
1236 --------------- ---------------
1237 PAT*+ (?>PAT*)
1238 PAT++ (?>PAT+)
1239 PAT?+ (?>PAT?)
1240 PAT{min,max}+ (?>PAT{min,max})
1241
e2e6a0f1
YO
1242=back
1243
1244=head2 Special Backtracking Control Verbs
1245
1246B<WARNING:> These patterns are experimental and subject to change or
0d017f4d 1247removal in a future version of Perl. Their usage in production code should
e2e6a0f1
YO
1248be noted to avoid problems during upgrades.
1249
1250These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1251otherwise stated the ARG argument is optional; in some cases, it is
1252forbidden.
1253
1254Any pattern containing a special backtracking verb that allows an argument
e1020413 1255has the special behaviour that when executed it sets the current package's
5d458dd8
YO
1256C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1257rules apply:
e2e6a0f1 1258
5d458dd8
YO
1259On failure, the C<$REGERROR> variable will be set to the ARG value of the
1260verb pattern, if the verb was involved in the failure of the match. If the
1261ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1262name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1263none. Also, the C<$REGMARK> variable will be set to FALSE.
e2e6a0f1 1264
5d458dd8
YO
1265On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1266the C<$REGMARK> variable will be set to the name of the last
1267C<(*MARK:NAME)> pattern executed. See the explanation for the
1268C<(*MARK:NAME)> verb below for more details.
e2e6a0f1 1269
5d458dd8
YO
1270B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1271and most other regex related variables. They are not local to a scope, nor
1272readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1273Use C<local> to localize changes to them to a specific scope if necessary.
e2e6a0f1
YO
1274
1275If a pattern does not contain a special backtracking verb that allows an
5d458dd8 1276argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
e2e6a0f1
YO
1277
1278=over 4
1279
1280=item Verbs that take an argument
1281
1282=over 4
1283
5d458dd8 1284=item C<(*PRUNE)> C<(*PRUNE:NAME)>
f7819f85 1285X<(*PRUNE)> X<(*PRUNE:NAME)>
54612592 1286
5d458dd8
YO
1287This zero-width pattern prunes the backtracking tree at the current point
1288when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1289where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1290A may backtrack as necessary to match. Once it is reached, matching
1291continues in B, which may also backtrack as necessary; however, should B
1292not match, then no further backtracking will take place, and the pattern
1293will fail outright at the current starting position.
54612592
YO
1294
1295The following example counts all the possible matching strings in a
1296pattern (without actually matching any of them).
1297
e2e6a0f1 1298 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
54612592
YO
1299 print "Count=$count\n";
1300
1301which produces:
1302
1303 aaab
1304 aaa
1305 aa
1306 a
1307 aab
1308 aa
1309 a
1310 ab
1311 a
1312 Count=9
1313
5d458dd8 1314If we add a C<(*PRUNE)> before the count like the following
54612592 1315
5d458dd8 1316 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
54612592
YO
1317 print "Count=$count\n";
1318
1319we prevent backtracking and find the count of the longest matching
353c6505 1320at each matching starting point like so:
54612592
YO
1321
1322 aaab
1323 aab
1324 ab
1325 Count=3
1326
5d458dd8 1327Any number of C<(*PRUNE)> assertions may be used in a pattern.
54612592 1328
5d458dd8
YO
1329See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1330control backtracking. In some cases, the use of C<(*PRUNE)> can be
1331replaced with a C<< (?>pattern) >> with no functional difference; however,
1332C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1333C<< (?>pattern) >> alone.
54612592 1334
e2e6a0f1 1335
5d458dd8
YO
1336=item C<(*SKIP)> C<(*SKIP:NAME)>
1337X<(*SKIP)>
e2e6a0f1 1338
5d458dd8 1339This zero-width pattern is similar to C<(*PRUNE)>, except that on
e2e6a0f1 1340failure it also signifies that whatever text that was matched leading up
5d458dd8
YO
1341to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1342of this pattern. This effectively means that the regex engine "skips" forward
1343to this position on failure and tries to match again, (assuming that
1344there is sufficient room to match).
1345
1346The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1347C<(*MARK:NAME)> was encountered while matching, then it is that position
1348which is used as the "skip point". If no C<(*MARK)> of that name was
1349encountered, then the C<(*SKIP)> operator has no effect. When used
1350without a name the "skip point" is where the match point was when
1351executing the (*SKIP) pattern.
1352
1353Compare the following to the examples in C<(*PRUNE)>, note the string
24b23f37
YO
1354is twice as long:
1355
5d458dd8 1356 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
24b23f37
YO
1357 print "Count=$count\n";
1358
1359outputs
1360
1361 aaab
1362 aaab
1363 Count=2
1364
5d458dd8 1365Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
353c6505 1366executed, the next starting point will be where the cursor was when the
5d458dd8
YO
1367C<(*SKIP)> was executed.
1368
5d458dd8
YO
1369=item C<(*MARK:NAME)> C<(*:NAME)>
1370X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)>
1371
1372This zero-width pattern can be used to mark the point reached in a string
1373when a certain part of the pattern has been successfully matched. This
1374mark may be given a name. A later C<(*SKIP)> pattern will then skip
1375forward to that point if backtracked into on failure. Any number of
b4222fa9 1376C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated.
5d458dd8
YO
1377
1378In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1379can be used to "label" a pattern branch, so that after matching, the
1380program can determine which branches of the pattern were involved in the
1381match.
1382
1383When a match is successful, the C<$REGMARK> variable will be set to the
1384name of the most recently executed C<(*MARK:NAME)> that was involved
1385in the match.
1386
1387This can be used to determine which branch of a pattern was matched
c27a5cfe 1388without using a separate capture group for each branch, which in turn
5d458dd8
YO
1389can result in a performance improvement, as perl cannot optimize
1390C</(?:(x)|(y)|(z))/> as efficiently as something like
1391C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1392
1393When a match has failed, and unless another verb has been involved in
1394failing the match and has provided its own name to use, the C<$REGERROR>
1395variable will be set to the name of the most recently executed
1396C<(*MARK:NAME)>.
1397
1398See C<(*SKIP)> for more details.
1399
b62d2d15
YO
1400As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
1401
5d458dd8
YO
1402=item C<(*THEN)> C<(*THEN:NAME)>
1403
241e7389 1404This is similar to the "cut group" operator C<::> from Perl 6. Like
5d458dd8
YO
1405C<(*PRUNE)>, this verb always matches, and when backtracked into on
1406failure, it causes the regex engine to try the next alternation in the
1407innermost enclosing group (capturing or otherwise).
1408
1409Its name comes from the observation that this operation combined with the
1410alternation operator (C<|>) can be used to create what is essentially a
1411pattern-based if/then/else block:
1412
1413 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
1414
1415Note that if this operator is used and NOT inside of an alternation then
1416it acts exactly like the C<(*PRUNE)> operator.
1417
1418 / A (*PRUNE) B /
1419
1420is the same as
1421
1422 / A (*THEN) B /
1423
1424but
1425
1426 / ( A (*THEN) B | C (*THEN) D ) /
1427
1428is not the same as
1429
1430 / ( A (*PRUNE) B | C (*PRUNE) D ) /
1431
1432as after matching the A but failing on the B the C<(*THEN)> verb will
1433backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
24b23f37 1434
e2e6a0f1
YO
1435=item C<(*COMMIT)>
1436X<(*COMMIT)>
24b23f37 1437
241e7389 1438This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a
5d458dd8
YO
1439zero-width pattern similar to C<(*SKIP)>, except that when backtracked
1440into on failure it causes the match to fail outright. No further attempts
1441to find a valid match by advancing the start pointer will occur again.
1442For example,
24b23f37 1443
e2e6a0f1 1444 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
24b23f37
YO
1445 print "Count=$count\n";
1446
1447outputs
1448
1449 aaab
1450 Count=1
1451
e2e6a0f1
YO
1452In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1453does not match, the regex engine will not try any further matching on the
1454rest of the string.
c277df42 1455
e2e6a0f1 1456=back
9af228c6 1457
e2e6a0f1 1458=item Verbs without an argument
9af228c6
YO
1459
1460=over 4
1461
e2e6a0f1
YO
1462=item C<(*FAIL)> C<(*F)>
1463X<(*FAIL)> X<(*F)>
9af228c6 1464
e2e6a0f1
YO
1465This pattern matches nothing and always fails. It can be used to force the
1466engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1467fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
9af228c6 1468
e2e6a0f1 1469It is probably useful only when combined with C<(?{})> or C<(??{})>.
9af228c6 1470
e2e6a0f1
YO
1471=item C<(*ACCEPT)>
1472X<(*ACCEPT)>
9af228c6 1473
e2e6a0f1
YO
1474B<WARNING:> This feature is highly experimental. It is not recommended
1475for production code.
9af228c6 1476
e2e6a0f1
YO
1477This pattern matches nothing and causes the end of successful matching at
1478the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1479whether there is actually more to match in the string. When inside of a
0d017f4d 1480nested pattern, such as recursion, or in a subpattern dynamically generated
e2e6a0f1 1481via C<(??{})>, only the innermost pattern is ended immediately.
9af228c6 1482
c27a5cfe 1483If the C<(*ACCEPT)> is inside of capturing groups then the groups are
e2e6a0f1
YO
1484marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1485For instance:
9af228c6 1486
e2e6a0f1 1487 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
9af228c6 1488
e2e6a0f1 1489will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
0d017f4d 1490be set. If another branch in the inner parentheses were matched, such as in the
e2e6a0f1 1491string 'ACDE', then the C<D> and C<E> would have to be matched as well.
9af228c6
YO
1492
1493=back
c277df42 1494
a0d0e21e
LW
1495=back
1496
c07a80fd 1497=head2 Backtracking
d74e8afc 1498X<backtrack> X<backtracking>
c07a80fd 1499
35a734be
IZ
1500NOTE: This section presents an abstract approximation of regular
1501expression behavior. For a more rigorous (and complicated) view of
1502the rules involved in selecting a match among possible alternatives,
0d017f4d 1503see L<Combining RE Pieces>.
35a734be 1504
c277df42 1505A fundamental feature of regular expression matching involves the
5a964f20 1506notion called I<backtracking>, which is currently used (when needed)
0d017f4d 1507by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>,
9da458fc
IZ
1508C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1509internally, but the general principle outlined here is valid.
c07a80fd 1510
1511For a regular expression to match, the I<entire> regular expression must
1512match, not just part of it. So if the beginning of a pattern containing a
1513quantifier succeeds in a way that causes later parts in the pattern to
1514fail, the matching engine backs up and recalculates the beginning
1515part--that's why it's called backtracking.
1516
1517Here is an example of backtracking: Let's say you want to find the
1518word following "foo" in the string "Food is on the foo table.":
1519
1520 $_ = "Food is on the foo table.";
1521 if ( /\b(foo)\s+(\w+)/i ) {
f793d64a 1522 print "$2 follows $1.\n";
c07a80fd 1523 }
1524
1525When the match runs, the first part of the regular expression (C<\b(foo)>)
1526finds a possible match right at the beginning of the string, and loads up
1527$1 with "Foo". However, as soon as the matching engine sees that there's
1528no whitespace following the "Foo" that it had saved in $1, it realizes its
68dc0745 1529mistake and starts over again one character after where it had the
c07a80fd 1530tentative match. This time it goes all the way until the next occurrence
1531of "foo". The complete regular expression matches this time, and you get
1532the expected output of "table follows foo."
1533
1534Sometimes minimal matching can help a lot. Imagine you'd like to match
1535everything between "foo" and "bar". Initially, you write something
1536like this:
1537
1538 $_ = "The food is under the bar in the barn.";
1539 if ( /foo(.*)bar/ ) {
f793d64a 1540 print "got <$1>\n";
c07a80fd 1541 }
1542
1543Which perhaps unexpectedly yields:
1544
1545 got <d is under the bar in the >
1546
1547That's because C<.*> was greedy, so you get everything between the
14218588 1548I<first> "foo" and the I<last> "bar". Here it's more effective
c07a80fd 1549to use minimal matching to make sure you get the text between a "foo"
1550and the first "bar" thereafter.
1551
1552 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1553 got <d is under the >
1554
0d017f4d 1555Here's another example. Let's say you'd like to match a number at the end
b6e13d97 1556of a string, and you also want to keep the preceding part of the match.
c07a80fd 1557So you write this:
1558
1559 $_ = "I have 2 numbers: 53147";
f793d64a
KW
1560 if ( /(.*)(\d*)/ ) { # Wrong!
1561 print "Beginning is <$1>, number is <$2>.\n";
c07a80fd 1562 }
1563
1564That won't work at all, because C<.*> was greedy and gobbled up the
1565whole string. As C<\d*> can match on an empty string the complete
1566regular expression matched successfully.
1567
8e1088bc 1568 Beginning is <I have 2 numbers: 53147>, number is <>.
c07a80fd 1569
1570Here are some variants, most of which don't work:
1571
1572 $_ = "I have 2 numbers: 53147";
1573 @pats = qw{
f793d64a
KW
1574 (.*)(\d*)
1575 (.*)(\d+)
1576 (.*?)(\d*)
1577 (.*?)(\d+)
1578 (.*)(\d+)$
1579 (.*?)(\d+)$
1580 (.*)\b(\d+)$
1581 (.*\D)(\d+)$
c07a80fd 1582 };
1583
1584 for $pat (@pats) {
f793d64a
KW
1585 printf "%-12s ", $pat;
1586 if ( /$pat/ ) {
1587 print "<$1> <$2>\n";
1588 } else {
1589 print "FAIL\n";
1590 }
c07a80fd 1591 }
1592
1593That will print out:
1594
1595 (.*)(\d*) <I have 2 numbers: 53147> <>
1596 (.*)(\d+) <I have 2 numbers: 5314> <7>
1597 (.*?)(\d*) <> <>
1598 (.*?)(\d+) <I have > <2>
1599 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1600 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1601 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1602 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1603
1604As you see, this can be a bit tricky. It's important to realize that a
1605regular expression is merely a set of assertions that gives a definition
1606of success. There may be 0, 1, or several different ways that the
1607definition might succeed against a particular string. And if there are
5a964f20
TC
1608multiple ways it might succeed, you need to understand backtracking to
1609know which variety of success you will achieve.
c07a80fd 1610
19799a22 1611When using look-ahead assertions and negations, this can all get even
8b19b778 1612trickier. Imagine you'd like to find a sequence of non-digits not
c07a80fd 1613followed by "123". You might try to write that as
1614
871b0233 1615 $_ = "ABC123";
f793d64a
KW
1616 if ( /^\D*(?!123)/ ) { # Wrong!
1617 print "Yup, no 123 in $_\n";
871b0233 1618 }
c07a80fd 1619
1620But that isn't going to match; at least, not the way you're hoping. It
1621claims that there is no 123 in the string. Here's a clearer picture of
9b9391b2 1622why that pattern matches, contrary to popular expectations:
c07a80fd 1623
4358a253
SS
1624 $x = 'ABC123';
1625 $y = 'ABC445';
c07a80fd 1626
4358a253
SS
1627 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1628 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
c07a80fd 1629
4358a253
SS
1630 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1631 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
c07a80fd 1632
1633This prints
1634
1635 2: got ABC
1636 3: got AB
1637 4: got ABC
1638
5f05dabc 1639You might have expected test 3 to fail because it seems to a more
c07a80fd 1640general purpose version of test 1. The important difference between
1641them is that test 3 contains a quantifier (C<\D*>) and so can use
1642backtracking, whereas test 1 will not. What's happening is
1643that you've asked "Is it true that at the start of $x, following 0 or more
5f05dabc 1644non-digits, you have something that's not 123?" If the pattern matcher had
c07a80fd 1645let C<\D*> expand to "ABC", this would have caused the whole pattern to
54310121 1646fail.
14218588 1647
c07a80fd 1648The search engine will initially match C<\D*> with "ABC". Then it will
14218588 1649try to match C<(?!123> with "123", which fails. But because
c07a80fd 1650a quantifier (C<\D*>) has been used in the regular expression, the
1651search engine can backtrack and retry the match differently
54310121 1652in the hope of matching the complete regular expression.
c07a80fd 1653
5a964f20
TC
1654The pattern really, I<really> wants to succeed, so it uses the
1655standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
c07a80fd 1656time. Now there's indeed something following "AB" that is not
14218588 1657"123". It's "C123", which suffices.
c07a80fd 1658
14218588
GS
1659We can deal with this by using both an assertion and a negation.
1660We'll say that the first part in $1 must be followed both by a digit
1661and by something that's not "123". Remember that the look-aheads
1662are zero-width expressions--they only look, but don't consume any
1663of the string in their match. So rewriting this way produces what
c07a80fd 1664you'd expect; that is, case 5 will fail, but case 6 succeeds:
1665
4358a253
SS
1666 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1667 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
c07a80fd 1668
1669 6: got ABC
1670
5a964f20 1671In other words, the two zero-width assertions next to each other work as though
19799a22 1672they're ANDed together, just as you'd use any built-in assertions: C</^$/>
c07a80fd 1673matches only if you're at the beginning of the line AND the end of the
1674line simultaneously. The deeper underlying truth is that juxtaposition in
1675regular expressions always means AND, except when you write an explicit OR
1676using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1677although the attempted matches are made at different positions because "a"
1678is not a zero-width assertion, but a one-width assertion.
1679
0d017f4d 1680B<WARNING>: Particularly complicated regular expressions can take
14218588 1681exponential time to solve because of the immense number of possible
0d017f4d 1682ways they can use backtracking to try for a match. For example, without
9da458fc
IZ
1683internal optimizations done by the regular expression engine, this will
1684take a painfully long time to run:
c07a80fd 1685
e1901655
IZ
1686 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1687
1688And if you used C<*>'s in the internal groups instead of limiting them
1689to 0 through 5 matches, then it would take forever--or until you ran
1690out of stack space. Moreover, these internal optimizations are not
1691always applicable. For example, if you put C<{0,5}> instead of C<*>
1692on the external group, no current optimization is applicable, and the
1693match takes a long time to finish.
c07a80fd 1694
9da458fc
IZ
1695A powerful tool for optimizing such beasts is what is known as an
1696"independent group",
c47ff5f1 1697which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
9da458fc 1698zero-length look-ahead/look-behind assertions will not backtrack to make
5d458dd8 1699the tail match, since they are in "logical" context: only
14218588 1700whether they match is considered relevant. For an example
9da458fc 1701where side-effects of look-ahead I<might> have influenced the
c47ff5f1 1702following match, see L<C<< (?>pattern) >>>.
c277df42 1703
a0d0e21e 1704=head2 Version 8 Regular Expressions
d74e8afc 1705X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
a0d0e21e 1706
5a964f20 1707In case you're not familiar with the "regular" Version 8 regex
a0d0e21e
LW
1708routines, here are the pattern-matching rules not described above.
1709
54310121 1710Any single character matches itself, unless it is a I<metacharacter>
a0d0e21e 1711with a special meaning described here or above. You can cause
5a964f20 1712characters that normally function as metacharacters to be interpreted
5f05dabc 1713literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
0d017f4d
WL
1714character; "\\" matches a "\"). This escape mechanism is also required
1715for the character used as the pattern delimiter.
1716
1717A series of characters matches that series of characters in the target
1718string, so the pattern C<blurfl> would match "blurfl" in the target
1719string.
a0d0e21e
LW
1720
1721You can specify a character class, by enclosing a list of characters
5d458dd8 1722in C<[]>, which will match any character from the list. If the
a0d0e21e 1723first character after the "[" is "^", the class matches any character not
14218588 1724in the list. Within a list, the "-" character specifies a
5a964f20 1725range, so that C<a-z> represents all characters between "a" and "z",
8a4f6ac2
GS
1726inclusive. If you want either "-" or "]" itself to be a member of a
1727class, put it at the start of the list (possibly after a "^"), or
1728escape it with a backslash. "-" is also taken literally when it is
1729at the end of the list, just before the closing "]". (The
84850974
DD
1730following all specify the same class of three characters: C<[-az]>,
1731C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
5d458dd8
YO
1732specifies a class containing twenty-six characters, even on EBCDIC-based
1733character sets.) Also, if you try to use the character
1734classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1735a range, the "-" is understood literally.
a0d0e21e 1736
8ada0baa
JH
1737Note also that the whole range idea is rather unportable between
1738character sets--and even within character sets they may cause results
1739you probably didn't expect. A sound principle is to use only ranges
0d017f4d 1740that begin from and end at either alphabetics of equal case ([a-e],
8ada0baa
JH
1741[A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1742spell out the character sets in full.
1743
54310121 1744Characters may be specified using a metacharacter syntax much like that
a0d0e21e
LW
1745used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1746"\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
5d458dd8
YO
1747of octal digits, matches the character whose coded character set value
1748is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1749matches the character whose numeric value is I<nn>. The expression \cI<x>
1750matches the character control-I<x>. Finally, the "." metacharacter
fb55449c 1751matches any character except "\n" (unless you use C</s>).
a0d0e21e
LW
1752
1753You can specify a series of alternatives for a pattern using "|" to
1754separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
5a964f20 1755or "foe" in the target string (as would C<f(e|i|o)e>). The
a0d0e21e
LW
1756first alternative includes everything from the last pattern delimiter
1757("(", "[", or the beginning of the pattern) up to the first "|", and
1758the last alternative contains everything from the last "|" to the next
14218588
GS
1759pattern delimiter. That's why it's common practice to include
1760alternatives in parentheses: to minimize confusion about where they
a3cb178b
GS
1761start and end.
1762
5a964f20 1763Alternatives are tried from left to right, so the first
a3cb178b
GS
1764alternative found for which the entire expression matches, is the one that
1765is chosen. This means that alternatives are not necessarily greedy. For
628afcb5 1766example: when matching C<foo|foot> against "barefoot", only the "foo"
a3cb178b
GS
1767part will match, as that is the first alternative tried, and it successfully
1768matches the target string. (This might not seem important, but it is
1769important when you are capturing matched text using parentheses.)
1770
5a964f20 1771Also remember that "|" is interpreted as a literal within square brackets,
a3cb178b 1772so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
a0d0e21e 1773
14218588
GS
1774Within a pattern, you may designate subpatterns for later reference
1775by enclosing them in parentheses, and you may refer back to the
1776I<n>th subpattern later in the pattern using the metacharacter
1777\I<n>. Subpatterns are numbered based on the left to right order
1778of their opening parenthesis. A backreference matches whatever
1779actually matched the subpattern in the string being examined, not
d8b950dc 1780the rules for that subpattern. Therefore, C<(0|0x)\d*\s\g1\d*> will
14218588
GS
1781match "0x1234 0x4321", but not "0x1234 01234", because subpattern
17821 matched "0x", even though the rule C<0|0x> could potentially match
1783the leading 0 in the second number.
cb1a09d0 1784
0d017f4d 1785=head2 Warning on \1 Instead of $1
cb1a09d0 1786
5a964f20 1787Some people get too used to writing things like:
cb1a09d0
AD
1788
1789 $pattern =~ s/(\W)/\\\1/g;
1790
3ff1c45a
KW
1791This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid
1792shocking the
cb1a09d0 1793B<sed> addicts, but it's a dirty habit to get into. That's because in
d1be9408 1794PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
cb1a09d0
AD
1795the usual double-quoted string means a control-A. The customary Unix
1796meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1797of doing that, you get yourself into trouble if you then add an C</e>
1798modifier.
1799
f793d64a 1800 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
cb1a09d0
AD
1801
1802Or if you try to do
1803
1804 s/(\d+)/\1000/;
1805
1806You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
14218588 1807C<${1}000>. The operation of interpolation should not be confused
cb1a09d0
AD
1808with the operation of matching a backreference. Certainly they mean two
1809different things on the I<left> side of the C<s///>.
9fa51da4 1810
0d017f4d 1811=head2 Repeated Patterns Matching a Zero-length Substring
c84d73f1 1812
19799a22 1813B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
c84d73f1
IZ
1814
1815Regular expressions provide a terse and powerful programming language. As
1816with most other power tools, power comes together with the ability
1817to wreak havoc.
1818
1819A common abuse of this power stems from the ability to make infinite
628afcb5 1820loops using regular expressions, with something as innocuous as:
c84d73f1
IZ
1821
1822 'foo' =~ m{ ( o? )* }x;
1823
0d017f4d 1824The C<o?> matches at the beginning of C<'foo'>, and since the position
c84d73f1 1825in the string is not moved by the match, C<o?> would match again and again
527e91da 1826because of the C<*> quantifier. Another common way to create a similar cycle
c84d73f1
IZ
1827is with the looping modifier C<//g>:
1828
1829 @matches = ( 'foo' =~ m{ o? }xg );
1830
1831or
1832
1833 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1834
1835or the loop implied by split().
1836
1837However, long experience has shown that many programming tasks may
14218588
GS
1838be significantly simplified by using repeated subexpressions that
1839may match zero-length substrings. Here's a simple example being:
c84d73f1 1840
f793d64a 1841 @chars = split //, $string; # // is not magic in split
c84d73f1
IZ
1842 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1843
9da458fc 1844Thus Perl allows such constructs, by I<forcefully breaking
c84d73f1 1845the infinite loop>. The rules for this are different for lower-level
527e91da 1846loops given by the greedy quantifiers C<*+{}>, and for higher-level
c84d73f1
IZ
1847ones like the C</g> modifier or split() operator.
1848
19799a22
GS
1849The lower-level loops are I<interrupted> (that is, the loop is
1850broken) when Perl detects that a repeated expression matched a
1851zero-length substring. Thus
c84d73f1
IZ
1852
1853 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1854
5d458dd8 1855is made equivalent to
c84d73f1 1856
5d458dd8
YO
1857 m{ (?: NON_ZERO_LENGTH )*
1858 |
1859 (?: ZERO_LENGTH )?
c84d73f1
IZ
1860 }x;
1861
1862The higher level-loops preserve an additional state between iterations:
5d458dd8 1863whether the last match was zero-length. To break the loop, the following
c84d73f1 1864match after a zero-length match is prohibited to have a length of zero.
5d458dd8 1865This prohibition interacts with backtracking (see L<"Backtracking">),
c84d73f1
IZ
1866and so the I<second best> match is chosen if the I<best> match is of
1867zero length.
1868
19799a22 1869For example:
c84d73f1
IZ
1870
1871 $_ = 'bar';
1872 s/\w??/<$&>/g;
1873
20fb949f 1874results in C<< <><b><><a><><r><> >>. At each position of the string the best
5d458dd8 1875match given by non-greedy C<??> is the zero-length match, and the I<second
c84d73f1
IZ
1876best> match is what is matched by C<\w>. Thus zero-length matches
1877alternate with one-character-long matches.
1878
5d458dd8 1879Similarly, for repeated C<m/()/g> the second-best match is the match at the
c84d73f1
IZ
1880position one notch further in the string.
1881
19799a22 1882The additional state of being I<matched with zero-length> is associated with
c84d73f1 1883the matched string, and is reset by each assignment to pos().
9da458fc
IZ
1884Zero-length matches at the end of the previous match are ignored
1885during C<split>.
c84d73f1 1886
0d017f4d 1887=head2 Combining RE Pieces
35a734be
IZ
1888
1889Each of the elementary pieces of regular expressions which were described
1890before (such as C<ab> or C<\Z>) could match at most one substring
1891at the given position of the input string. However, in a typical regular
1892expression these elementary pieces are combined into more complicated
1893patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1894(in these examples C<S> and C<T> are regular subexpressions).
1895
1896Such combinations can include alternatives, leading to a problem of choice:
1897if we match a regular expression C<a|ab> against C<"abc">, will it match
1898substring C<"a"> or C<"ab">? One way to describe which substring is
1899actually matched is the concept of backtracking (see L<"Backtracking">).
1900However, this description is too low-level and makes you think
1901in terms of a particular implementation.
1902
1903Another description starts with notions of "better"/"worse". All the
1904substrings which may be matched by the given regular expression can be
1905sorted from the "best" match to the "worst" match, and it is the "best"
1906match which is chosen. This substitutes the question of "what is chosen?"
1907by the question of "which matches are better, and which are worse?".
1908
1909Again, for elementary pieces there is no such question, since at most
1910one match at a given position is possible. This section describes the
1911notion of better/worse for combining operators. In the description
1912below C<S> and C<T> are regular subexpressions.
1913
13a2d996 1914=over 4
35a734be
IZ
1915
1916=item C<ST>
1917
1918Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1919substrings which can be matched by C<S>, C<B> and C<B'> are substrings
5d458dd8 1920which can be matched by C<T>.
35a734be
IZ
1921
1922If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1923match than C<A'B'>.
1924
1925If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1926C<B> is better match for C<T> than C<B'>.
1927
1928=item C<S|T>
1929
1930When C<S> can match, it is a better match than when only C<T> can match.
1931
1932Ordering of two matches for C<S> is the same as for C<S>. Similar for
1933two matches for C<T>.
1934
1935=item C<S{REPEAT_COUNT}>
1936
1937Matches as C<SSS...S> (repeated as many times as necessary).
1938
1939=item C<S{min,max}>
1940
1941Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1942
1943=item C<S{min,max}?>
1944
1945Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1946
1947=item C<S?>, C<S*>, C<S+>
1948
1949Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
1950
1951=item C<S??>, C<S*?>, C<S+?>
1952
1953Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
1954
c47ff5f1 1955=item C<< (?>S) >>
35a734be
IZ
1956
1957Matches the best match for C<S> and only that.
1958
1959=item C<(?=S)>, C<(?<=S)>
1960
1961Only the best match for C<S> is considered. (This is important only if
1962C<S> has capturing parentheses, and backreferences are used somewhere
1963else in the whole regular expression.)
1964
1965=item C<(?!S)>, C<(?<!S)>
1966
1967For this grouping operator there is no need to describe the ordering, since
1968only whether or not C<S> can match is important.
1969
6bda09f9 1970=item C<(??{ EXPR })>, C<(?PARNO)>
35a734be
IZ
1971
1972The ordering is the same as for the regular expression which is
c27a5cfe 1973the result of EXPR, or the pattern contained by capture group PARNO.
35a734be
IZ
1974
1975=item C<(?(condition)yes-pattern|no-pattern)>
1976
1977Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
1978already determined. The ordering of the matches is the same as for the
1979chosen subexpression.
1980
1981=back
1982
1983The above recipes describe the ordering of matches I<at a given position>.
1984One more rule is needed to understand how a match is determined for the
1985whole regular expression: a match at an earlier position is always better
1986than a match at a later position.
1987
0d017f4d 1988=head2 Creating Custom RE Engines
c84d73f1
IZ
1989
1990Overloaded constants (see L<overload>) provide a simple way to extend
1991the functionality of the RE engine.
1992
1993Suppose that we want to enable a new RE escape-sequence C<\Y|> which
0d017f4d 1994matches at a boundary between whitespace characters and non-whitespace
c84d73f1
IZ
1995characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
1996at these positions, so we want to have each C<\Y|> in the place of the
1997more complicated version. We can create a module C<customre> to do
1998this:
1999
2000 package customre;
2001 use overload;
2002
2003 sub import {
2004 shift;
2005 die "No argument to customre::import allowed" if @_;
2006 overload::constant 'qr' => \&convert;
2007 }
2008
2009 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
2010
580a9fe1
RGS
2011 # We must also take care of not escaping the legitimate \\Y|
2012 # sequence, hence the presence of '\\' in the conversion rules.
5d458dd8 2013 my %rules = ( '\\' => '\\\\',
f793d64a 2014 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
c84d73f1
IZ
2015 sub convert {
2016 my $re = shift;
5d458dd8 2017 $re =~ s{
c84d73f1
IZ
2018 \\ ( \\ | Y . )
2019 }
5d458dd8 2020 { $rules{$1} or invalid($re,$1) }sgex;
c84d73f1
IZ
2021 return $re;
2022 }
2023
2024Now C<use customre> enables the new escape in constant regular
2025expressions, i.e., those without any runtime variable interpolations.
2026As documented in L<overload>, this conversion will work only over
2027literal parts of regular expressions. For C<\Y|$re\Y|> the variable
2028part of this regular expression needs to be converted explicitly
2029(but only if the special meaning of C<\Y|> should be enabled inside $re):
2030
2031 use customre;
2032 $re = <>;
2033 chomp $re;
2034 $re = customre::convert $re;
2035 /\Y|$re\Y|/;
2036
1f1031fe
YO
2037=head1 PCRE/Python Support
2038
99d59c4d 2039As of Perl 5.10.0, Perl supports several Python/PCRE specific extensions
1f1031fe 2040to the regex syntax. While Perl programmers are encouraged to use the
99d59c4d 2041Perl specific syntax, the following are also accepted:
1f1031fe
YO
2042
2043=over 4
2044
ae5648b3 2045=item C<< (?PE<lt>NAMEE<gt>pattern) >>
1f1031fe 2046
c27a5cfe 2047Define a named capture group. Equivalent to C<< (?<NAME>pattern) >>.
1f1031fe
YO
2048
2049=item C<< (?P=NAME) >>
2050
c27a5cfe 2051Backreference to a named capture group. Equivalent to C<< \g{NAME} >>.
1f1031fe
YO
2052
2053=item C<< (?P>NAME) >>
2054
c27a5cfe 2055Subroutine call to a named capture group. Equivalent to C<< (?&NAME) >>.
1f1031fe 2056
ee9b8eae 2057=back
1f1031fe 2058
19799a22
GS
2059=head1 BUGS
2060
78288b8e
KW
2061There are numerous problems with case insensitive matching of characters
2062outside the ASCII range, especially with those whose folds are multiple
2063characters, such as ligatures like C<LATIN SMALL LIGATURE FF>.
2064
f253210b
KW
2065In a bracketed character class with case insensitive matching, ranges only work
2066for ASCII characters. For example,
2067C<m/[\N{CYRILLIC CAPITAL LETTER A}-\N{CYRILLIC CAPITAL LETTER YA}]/i>
2068doesn't match all the Russian upper and lower case letters.
2069
88c9975e
KW
2070Many regular expression constructs don't work on EBCDIC platforms.
2071
9da458fc
IZ
2072This document varies from difficult to understand to completely
2073and utterly opaque. The wandering prose riddled with jargon is
2074hard to fathom in several places.
2075
2076This document needs a rewrite that separates the tutorial content
2077from the reference content.
19799a22
GS
2078
2079=head1 SEE ALSO
9fa51da4 2080
91e0c79e
MJD
2081L<perlrequick>.
2082
2083L<perlretut>.
2084
9b599b2a
GS
2085L<perlop/"Regexp Quote-Like Operators">.
2086
1e66bd83
PP
2087L<perlop/"Gory details of parsing quoted constructs">.
2088
14218588
GS
2089L<perlfaq6>.
2090
9b599b2a
GS
2091L<perlfunc/pos>.
2092
2093L<perllocale>.
2094
fb55449c
JH
2095L<perlebcdic>.
2096
14218588
GS
2097I<Mastering Regular Expressions> by Jeffrey Friedl, published
2098by O'Reilly and Associates.