This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Change 30276 wasn't meant do mess with t/lib/common.pl. Oops.
[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
PP
28=over 4
29
30=item i
d74e8afc
ITB
31X</i> X<regex, case-insensitive> X<regexp, case-insensitive>
32X<regular expression, case-insensitive>
55497cff
PP
33
34Do case-insensitive pattern matching.
35
a034a98d
DD
36If C<use locale> is in effect, the case map is taken from the current
37locale. See L<perllocale>.
38
54310121 39=item m
d74e8afc 40X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline>
55497cff
PP
41
42Treat string as multiple lines. That is, change "^" and "$" from matching
14218588 43the start or end of the string to matching the start or end of any
7f761169 44line anywhere within the string.
55497cff 45
54310121 46=item s
d74e8afc
ITB
47X</s> X<regex, single-line> X<regexp, single-line>
48X<regular expression, single-line>
55497cff
PP
49
50Treat string as single line. That is, change "." to match any character
19799a22 51whatsoever, even a newline, which normally it would not match.
55497cff 52
f02c194e 53Used together, as /ms, they let the "." match any character whatsoever,
fb55449c 54while still allowing "^" and "$" to match, respectively, just after
19799a22 55and just before newlines within the string.
7b8d334a 56
54310121 57=item x
d74e8afc 58X</x>
55497cff
PP
59
60Extend your pattern's legibility by permitting whitespace and comments.
61
62=back
a0d0e21e
LW
63
64These are usually written as "the C</x> modifier", even though the delimiter
14218588 65in question might not really be a slash. Any of these
a0d0e21e 66modifiers may also be embedded within the regular expression itself using
14218588 67the C<(?...)> construct. See below.
a0d0e21e 68
4633a7c4 69The C</x> modifier itself needs a little more explanation. It tells
55497cff
PP
70the regular expression parser to ignore whitespace that is neither
71backslashed nor within a character class. You can use this to break up
4633a7c4 72your regular expression into (slightly) more readable parts. The C<#>
54310121 73character is also treated as a metacharacter introducing a comment,
55497cff 74just as in ordinary Perl code. This also means that if you want real
14218588 75whitespace or C<#> characters in the pattern (outside a character
f9a3ff1a
RGS
76class, where they are unaffected by C</x>), then you'll either have to
77escape them (using backslashes or C<\Q...\E>) or encode them using octal
8933a740
RGS
78or hex escapes. Taken together, these features go a long way towards
79making Perl's regular expressions more readable. Note that you have to
80be careful not to include the pattern delimiter in the comment--perl has
81no way of knowing you did not intend to close the pattern early. See
82the C-comment deletion code in L<perlop>. Also note that anything inside
1031e5db 83a C<\Q...\E> stays unaffected by C</x>.
d74e8afc 84X</x>
a0d0e21e
LW
85
86=head2 Regular Expressions
87
04838cea
RGS
88=head3 Metacharacters
89
0d017f4d 90The patterns used in Perl pattern matching evolved from the ones supplied in
14218588 91the Version 8 regex routines. (The routines are derived
19799a22
GS
92(distantly) from Henry Spencer's freely redistributable reimplementation
93of the V8 routines.) See L<Version 8 Regular Expressions> for
94details.
a0d0e21e
LW
95
96In particular the following metacharacters have their standard I<egrep>-ish
97meanings:
d74e8afc
ITB
98X<metacharacter>
99X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]>
100
a0d0e21e 101
54310121 102 \ Quote the next metacharacter
a0d0e21e
LW
103 ^ Match the beginning of the line
104 . Match any character (except newline)
c07a80fd 105 $ Match the end of the line (or before newline at the end)
a0d0e21e
LW
106 | Alternation
107 () Grouping
108 [] Character class
109
14218588
GS
110By default, the "^" character is guaranteed to match only the
111beginning of the string, the "$" character only the end (or before the
112newline at the end), and Perl does certain optimizations with the
a0d0e21e
LW
113assumption that the string contains only one line. Embedded newlines
114will not be matched by "^" or "$". You may, however, wish to treat a
4a6725af 115string as a multi-line buffer, such that the "^" will match after any
0d520e8e
YO
116newline within the string (except if the newline is the last character in
117the string), and "$" will match before any newline. At the
a0d0e21e
LW
118cost of a little more overhead, you can do this by using the /m modifier
119on the pattern match operator. (Older programs did this by setting C<$*>,
f02c194e 120but this practice has been removed in perl 5.9.)
d74e8afc 121X<^> X<$> X</m>
a0d0e21e 122
14218588 123To simplify multi-line substitutions, the "." character never matches a
55497cff 124newline unless you use the C</s> modifier, which in effect tells Perl to pretend
f02c194e 125the string is a single line--even if it isn't.
d74e8afc 126X<.> X</s>
a0d0e21e 127
04838cea
RGS
128=head3 Quantifiers
129
a0d0e21e 130The following standard quantifiers are recognized:
d74e8afc 131X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}>
a0d0e21e
LW
132
133 * Match 0 or more times
134 + Match 1 or more times
135 ? Match 1 or 0 times
136 {n} Match exactly n times
137 {n,} Match at least n times
138 {n,m} Match at least n but not more than m times
139
140(If a curly bracket occurs in any other context, it is treated
b975c076
JL
141as a regular character. In particular, the lower bound
142is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+"
25f94b33 143modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited
9c79236d
GS
144to integral values less than a preset limit defined when perl is built.
145This is usually 32766 on the most common platforms. The actual limit can
146be seen in the error message generated by code such as this:
147
820475bd 148 $_ **= $_ , / {$_} / for 2 .. 42;
a0d0e21e 149
54310121
PP
150By default, a quantified subpattern is "greedy", that is, it will match as
151many times as possible (given a particular starting location) while still
152allowing the rest of the pattern to match. If you want it to match the
153minimum number of times possible, follow the quantifier with a "?". Note
154that the meanings don't change, just the "greediness":
0d017f4d 155X<metacharacter> X<greedy> X<greediness>
d74e8afc 156X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?>
a0d0e21e 157
0d017f4d
WL
158 *? Match 0 or more times, not greedily
159 +? Match 1 or more times, not greedily
160 ?? Match 0 or 1 time, not greedily
161 {n}? Match exactly n times, not greedily
162 {n,}? Match at least n times, not greedily
163 {n,m}? Match at least n but not more than m times, not greedily
a0d0e21e 164
b9b4dddf
YO
165By default, when a quantified subpattern does not allow the rest of the
166overall pattern to match, Perl will backtrack. However, this behaviour is
0d017f4d 167sometimes undesirable. Thus Perl provides the "possessive" quantifier form
b9b4dddf
YO
168as well.
169
0d017f4d
WL
170 *+ Match 0 or more times and give nothing back
171 ++ Match 1 or more times and give nothing back
172 ?+ Match 0 or 1 time and give nothing back
b9b4dddf 173 {n}+ Match exactly n times and give nothing back (redundant)
04838cea
RGS
174 {n,}+ Match at least n times and give nothing back
175 {n,m}+ Match at least n but not more than m times and give nothing back
b9b4dddf
YO
176
177For instance,
178
179 'aaaa' =~ /a++a/
180
181will never match, as the C<a++> will gobble up all the C<a>'s in the
182string and won't leave any for the remaining part of the pattern. This
183feature can be extremely useful to give perl hints about where it
184shouldn't backtrack. For instance, the typical "match a double-quoted
185string" problem can be most efficiently performed when written as:
186
187 /"(?:[^"\\]++|\\.)*+"/
188
0d017f4d 189as we know that if the final quote does not match, backtracking will not
b9b4dddf
YO
190help. See the independent subexpression C<< (?>...) >> for more details;
191possessive quantifiers are just syntactic sugar for that construct. For
192instance the above example could also be written as follows:
193
194 /"(?>(?:(?>[^"\\]+)|\\.)*)"/
195
04838cea
RGS
196=head3 Escape sequences
197
5f05dabc 198Because patterns are processed as double quoted strings, the following
a0d0e21e 199also work:
0d017f4d 200X<\t> X<\n> X<\r> X<\f> X<\e> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q>
d74e8afc 201X<\0> X<\c> X<\N> X<\x>
a0d0e21e 202
0f36ee90
PP
203 \t tab (HT, TAB)
204 \n newline (LF, NL)
205 \r return (CR)
206 \f form feed (FF)
207 \a alarm (bell) (BEL)
208 \e escape (think troff) (ESC)
0d017f4d
WL
209 \033 octal char (example: ESC)
210 \x1B hex char (example: ESC)
211 \x{263a} wide hex char (example: Unicode SMILEY)
212 \cK control char (example: VT)
4a2d328f 213 \N{name} named char
cb1a09d0
AD
214 \l lowercase next char (think vi)
215 \u uppercase next char (think vi)
216 \L lowercase till \E (think vi)
217 \U uppercase till \E (think vi)
218 \E end case modification (think vi)
5a964f20 219 \Q quote (disable) pattern metacharacters till \E
a0d0e21e 220
a034a98d 221If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u>
423cee85 222and C<\U> is taken from the current locale. See L<perllocale>. For
4a2d328f 223documentation of C<\N{name}>, see L<charnames>.
a034a98d 224
1d2dff63
GS
225You cannot include a literal C<$> or C<@> within a C<\Q> sequence.
226An unescaped C<$> or C<@> interpolates the corresponding variable,
227while escaping will cause the literal string C<\$> to be matched.
228You'll need to write something like C<m/\Quser\E\@\Qhost/>.
229
04838cea
RGS
230=head3 Character classes
231
a0d0e21e 232In addition, Perl defines the following:
d74e8afc 233X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C>
0d017f4d
WL
234X<\g> X<\k> X<\N> X<\K> X<\v> X<\V>
235X<word> X<whitespace> X<character class> X<backreference>
a0d0e21e 236
81714fb9
YO
237 \w Match a "word" character (alphanumeric plus "_")
238 \W Match a non-"word" character
239 \s Match a whitespace character
240 \S Match a non-whitespace character
241 \d Match a digit character
242 \D Match a non-digit character
243 \pP Match P, named property. Use \p{Prop} for longer names.
244 \PP Match non-P
245 \X Match eXtended Unicode "combining character sequence",
246 equivalent to (?:\PM\pM*)
247 \C Match a single C char (octet) even under Unicode.
248 NOTE: breaks up characters into their UTF-8 bytes,
249 so you may end up with malformed pieces of UTF-8.
250 Unsupported in lookbehind.
5d458dd8 251 \1 Backreference to a specific group.
c74340f9 252 '1' may actually be any positive integer.
2bf803e2
YO
253 \g1 Backreference to a specific or previous group,
254 \g{-1} number may be negative indicating a previous buffer and may
255 optionally be wrapped in curly brackets for safer parsing.
1f1031fe 256 \g{name} Named backreference
81714fb9 257 \k<name> Named backreference
5d458dd8 258 \N{name} Named unicode character, or unicode escape
81714fb9
YO
259 \x12 Hexadecimal escape sequence
260 \x{1234} Long hexadecimal escape sequence
ee9b8eae
YO
261 \K Keep the stuff left of the \K, don't include it in $&
262 \v Shortcut for (*PRUNE)
263 \V Shortcut for (*SKIP)
a0d0e21e 264
08ce8fc6
JH
265A C<\w> matches a single alphanumeric character (an alphabetic
266character, or a decimal digit) or C<_>, not a whole word. Use C<\w+>
267to match a string of Perl-identifier characters (which isn't the same
268as matching an English word). If C<use locale> is in effect, the list
269of alphabetic characters generated by C<\w> is taken from the current
270locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>,
0d017f4d
WL
271C<\d>, and C<\D> within character classes, but they aren't usable
272as either end of a range. If any of them precedes or follows a "-",
273the "-" is understood literally. If Unicode is in effect, C<\s> matches
274also "\x{85}", "\x{2028}, and "\x{2029}". See L<perlunicode> for more
275details about C<\pP>, C<\PP>, C<\X> and the possibility of defining
276your own C<\p> and C<\P> properties, and L<perluniintro> about Unicode
277in general.
d74e8afc 278X<\w> X<\W> X<word>
a0d0e21e 279
b8c5462f 280The POSIX character class syntax
d74e8afc 281X<character class>
b8c5462f 282
820475bd 283 [:class:]
b8c5462f 284
0d017f4d 285is also available. Note that the C<[> and C<]> brackets are I<literal>;
5496314a
SP
286they must always be used within a character class expression.
287
288 # this is correct:
289 $string =~ /[[:alpha:]]/;
290
291 # this is not, and will generate a warning:
292 $string =~ /[:alpha:]/;
293
294The available classes and their backslash equivalents (if available) are
295as follows:
d74e8afc
ITB
296X<character class>
297X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph>
298X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit>
b8c5462f
JH
299
300 alpha
301 alnum
302 ascii
aaa51d5e 303 blank [1]
b8c5462f
JH
304 cntrl
305 digit \d
306 graph
307 lower
308 print
309 punct
aaa51d5e 310 space \s [2]
b8c5462f 311 upper
aaa51d5e 312 word \w [3]
b8c5462f
JH
313 xdigit
314
07698885
RGS
315=over
316
317=item [1]
318
b432a672 319A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace".
07698885
RGS
320
321=item [2]
322
323Not exactly equivalent to C<\s> since the C<[[:space:]]> includes
0d017f4d 324also the (very rare) "vertical tabulator", "\cK" or chr(11) in ASCII.
07698885
RGS
325
326=item [3]
327
08ce8fc6 328A Perl extension, see above.
07698885
RGS
329
330=back
aaa51d5e 331
26b44a0a 332For example use C<[:upper:]> to match all the uppercase characters.
aaa51d5e
JF
333Note that the C<[]> are part of the C<[::]> construct, not part of the
334whole character class. For example:
b8c5462f 335
820475bd 336 [01[:alpha:]%]
b8c5462f 337
0d017f4d 338matches zero, one, any alphabetic character, and the percent sign.
b8c5462f 339
72ff2908
JH
340The following equivalences to Unicode \p{} constructs and equivalent
341backslash character classes (if available), will hold:
d74e8afc 342X<character class> X<\p> X<\p{}>
72ff2908 343
5496314a 344 [[:...:]] \p{...} backslash
b8c5462f
JH
345
346 alpha IsAlpha
347 alnum IsAlnum
348 ascii IsASCII
0d017f4d 349 blank
b8c5462f 350 cntrl IsCntrl
3bec3564 351 digit IsDigit \d
b8c5462f
JH
352 graph IsGraph
353 lower IsLower
354 print IsPrint
355 punct IsPunct
356 space IsSpace
3bec3564 357 IsSpacePerl \s
b8c5462f
JH
358 upper IsUpper
359 word IsWord
360 xdigit IsXDigit
361
5496314a 362For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent.
b8c5462f
JH
363
364If the C<utf8> pragma is not used but the C<locale> pragma is, the
aaa51d5e 365classes correlate with the usual isalpha(3) interface (except for
b432a672 366"word" and "blank").
b8c5462f
JH
367
368The assumedly non-obviously named classes are:
369
370=over 4
371
372=item cntrl
d74e8afc 373X<cntrl>
b8c5462f 374
820475bd
GS
375Any control character. Usually characters that don't produce output as
376such but instead control the terminal somehow: for example newline and
377backspace are control characters. All characters with ord() less than
0d017f4d 37832 are usually classified as control characters (assuming ASCII,
7be5a6cf
JF
379the ISO Latin character sets, and Unicode), as is the character with
380the ord() value of 127 (C<DEL>).
b8c5462f
JH
381
382=item graph
d74e8afc 383X<graph>
b8c5462f 384
f1cbbd6e 385Any alphanumeric or punctuation (special) character.
b8c5462f
JH
386
387=item print
d74e8afc 388X<print>
b8c5462f 389
f79b3095 390Any alphanumeric or punctuation (special) character or the space character.
b8c5462f
JH
391
392=item punct
d74e8afc 393X<punct>
b8c5462f 394
f1cbbd6e 395Any punctuation (special) character.
b8c5462f
JH
396
397=item xdigit
d74e8afc 398X<xdigit>
b8c5462f 399
593df60c 400Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would
820475bd 401work just fine) it is included for completeness.
b8c5462f 402
b8c5462f
JH
403=back
404
405You can negate the [::] character classes by prefixing the class name
406with a '^'. This is a Perl extension. For example:
d74e8afc 407X<character class, negation>
b8c5462f 408
5496314a 409 POSIX traditional Unicode
93733859 410
5496314a
SP
411 [[:^digit:]] \D \P{IsDigit}
412 [[:^space:]] \S \P{IsSpace}
413 [[:^word:]] \W \P{IsWord}
b8c5462f 414
54c18d04
MK
415Perl respects the POSIX standard in that POSIX character classes are
416only supported within a character class. The POSIX character classes
417[.cc.] and [=cc=] are recognized but B<not> supported and trying to
418use them will cause an error.
b8c5462f 419
04838cea
RGS
420=head3 Assertions
421
a0d0e21e 422Perl defines the following zero-width assertions:
d74e8afc
ITB
423X<zero-width assertion> X<assertion> X<regex, zero-width assertion>
424X<regexp, zero-width assertion>
425X<regular expression, zero-width assertion>
426X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G>
a0d0e21e
LW
427
428 \b Match a word boundary
0d017f4d 429 \B Match except at a word boundary
b85d18e9
IZ
430 \A Match only at beginning of string
431 \Z Match only at end of string, or before newline at the end
432 \z Match only at end of string
9da458fc
IZ
433 \G Match only at pos() (e.g. at the end-of-match position
434 of prior m//g)
a0d0e21e 435
14218588 436A word boundary (C<\b>) is a spot between two characters
19799a22
GS
437that has a C<\w> on one side of it and a C<\W> on the other side
438of it (in either order), counting the imaginary characters off the
439beginning and end of the string as matching a C<\W>. (Within
440character classes C<\b> represents backspace rather than a word
441boundary, just as it normally does in any double-quoted string.)
442The C<\A> and C<\Z> are just like "^" and "$", except that they
443won't match multiple times when the C</m> modifier is used, while
444"^" and "$" will match at every internal line boundary. To match
445the actual end of the string and not ignore an optional trailing
446newline, use C<\z>.
d74e8afc 447X<\b> X<\A> X<\Z> X<\z> X</m>
19799a22
GS
448
449The C<\G> assertion can be used to chain global matches (using
450C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">.
451It is also useful when writing C<lex>-like scanners, when you have
452several patterns that you want to match against consequent substrings
453of your string, see the previous reference. The actual location
454where C<\G> will match can also be influenced by using C<pos()> as
58e23c8d
YO
455an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length
456matches is modified somewhat, in that contents to the left of C<\G> is
457not counted when determining the length of the match. Thus the following
458will not match forever:
d74e8afc 459X<\G>
c47ff5f1 460
58e23c8d
YO
461 $str = 'ABC';
462 pos($str) = 1;
463 while (/.\G/g) {
464 print $&;
465 }
466
467It will print 'A' and then terminate, as it considers the match to
468be zero-width, and thus will not match at the same position twice in a
469row.
470
471It is worth noting that C<\G> improperly used can result in an infinite
472loop. Take care when using patterns that include C<\G> in an alternation.
473
04838cea
RGS
474=head3 Capture buffers
475
0d017f4d
WL
476The bracketing construct C<( ... )> creates capture buffers. To refer
477to the current contents of a buffer later on, within the same pattern,
478use \1 for the first, \2 for the second, and so on.
479Outside the match use "$" instead of "\". (The
81714fb9 480\<digit> notation works in certain circumstances outside
14218588
GS
481the match. See the warning below about \1 vs $1 for details.)
482Referring back to another part of the match is called a
483I<backreference>.
d74e8afc
ITB
484X<regex, capture buffer> X<regexp, capture buffer>
485X<regular expression, capture buffer> X<backreference>
14218588
GS
486
487There is no limit to the number of captured substrings that you may
488use. However Perl also uses \10, \11, etc. as aliases for \010,
fb55449c
JH
489\011, etc. (Recall that 0 means octal, so \011 is the character at
490number 9 in your coded character set; which would be the 10th character,
81714fb9
YO
491a horizontal tab under ASCII.) Perl resolves this
492ambiguity by interpreting \10 as a backreference only if at least 10
493left parentheses have opened before it. Likewise \11 is a
494backreference only if at least 11 left parentheses have opened
495before it. And so on. \1 through \9 are always interpreted as
5624f11d 496backreferences.
c74340f9 497
1f1031fe 498X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference>
2bf803e2 499In order to provide a safer and easier way to construct patterns using
0d017f4d 500backreferences, Perl 5.10 provides the C<\g{N}> notation. The curly
2bf803e2
YO
501brackets are optional, however omitting them is less safe as the meaning
502of the pattern can be changed by text (such as digits) following it.
503When N is a positive integer the C<\g{N}> notation is exactly equivalent
504to using normal backreferences. When N is a negative integer then it is
505a relative backreference referring to the previous N'th capturing group.
1f1031fe
YO
506When the bracket form is used and N is not an integer, it is treated as a
507reference to a named buffer.
2bf803e2
YO
508
509Thus C<\g{-1}> refers to the last buffer, C<\g{-2}> refers to the
510buffer before that. For example:
5624f11d
YO
511
512 /
513 (Y) # buffer 1
514 ( # buffer 2
515 (X) # buffer 3
2bf803e2
YO
516 \g{-1} # backref to buffer 3
517 \g{-3} # backref to buffer 1
5624f11d
YO
518 )
519 /x
520
2bf803e2 521and would match the same as C</(Y) ( (X) \3 \1 )/x>.
14218588 522
81714fb9 523Additionally, as of Perl 5.10 you may use named capture buffers and named
1f1031fe 524backreferences. The notation is C<< (?<name>...) >> to declare and C<< \k<name> >>
0d017f4d
WL
525to reference. You may also use apostrophes instead of angle brackets to delimit the
526name; and you may use the bracketed C<< \g{name} >> backreference syntax.
527It's possible to refer to a named capture buffer by absolute and relative number as well.
528Outside the pattern, a named capture buffer is available via the C<%+> hash.
529When different buffers within the same pattern have the same name, C<$+{name}>
530and C<< \k<name> >> refer to the leftmost defined group. (Thus it's possible
531to do things with named capture buffers that would otherwise require C<(??{})>
532code to accomplish.)
533X<named capture buffer> X<regular expression, named capture buffer>
534X<%+> X<$+{name}> X<\k{name}>
81714fb9 535
14218588 536Examples:
a0d0e21e
LW
537
538 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
539
81714fb9
YO
540 /(.)\1/ # find first doubled char
541 and print "'$1' is the first doubled character\n";
542
543 /(?<char>.)\k<char>/ # ... a different way
544 and print "'$+{char}' is the first doubled character\n";
545
0d017f4d 546 /(?'char'.)\1/ # ... mix and match
81714fb9 547 and print "'$1' is the first doubled character\n";
c47ff5f1 548
14218588 549 if (/Time: (..):(..):(..)/) { # parse out values
a0d0e21e
LW
550 $hours = $1;
551 $minutes = $2;
552 $seconds = $3;
553 }
c47ff5f1 554
14218588
GS
555Several special variables also refer back to portions of the previous
556match. C<$+> returns whatever the last bracket match matched.
557C<$&> returns the entire matched string. (At one point C<$0> did
558also, but now it returns the name of the program.) C<$`> returns
77ea4f6d
JV
559everything before the matched string. C<$'> returns everything
560after the matched string. And C<$^N> contains whatever was matched by
561the most-recently closed group (submatch). C<$^N> can be used in
562extended patterns (see below), for example to assign a submatch to a
81714fb9 563variable.
d74e8afc 564X<$+> X<$^N> X<$&> X<$`> X<$'>
14218588 565
665e98b9 566The numbered match variables ($1, $2, $3, etc.) and the related punctuation
77ea4f6d 567set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped
14218588
GS
568until the end of the enclosing block or until the next successful
569match, whichever comes first. (See L<perlsyn/"Compound Statements">.)
d74e8afc
ITB
570X<$+> X<$^N> X<$&> X<$`> X<$'>
571X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9>
572
14218588 573
0d017f4d 574B<NOTE>: Failed matches in Perl do not reset the match variables,
5146ce24 575which makes it easier to write code that tests for a series of more
665e98b9
JH
576specific cases and remembers the best match.
577
14218588
GS
578B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or
579C<$'> anywhere in the program, it has to provide them for every
580pattern match. This may substantially slow your program. Perl
581uses the same mechanism to produce $1, $2, etc, so you also pay a
582price for each pattern that contains capturing parentheses. (To
583avoid this cost while retaining the grouping behaviour, use the
584extended regular expression C<(?: ... )> instead.) But if you never
585use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing
586parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`>
587if you can, but if you can't (and some algorithms really appreciate
588them), once you've used them once, use them at will, because you've
589already paid the price. As of 5.005, C<$&> is not so costly as the
590other two.
d74e8afc 591X<$&> X<$`> X<$'>
68dc0745 592
cde0cee5
YO
593As a workaround for this problem, Perl 5.10 introduces C<${^PREMATCH}>,
594C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&>
595and C<$'>, B<except> that they are only guaranteed to be defined after a
596successful match that was executed with the C</k> (keep-copy) modifier.
597The use of these variables incurs no global performance penalty, unlike
598their punctuation char equivalents, however at the trade-off that you
599have to tell perl when you want to use them.
600X</k> X<k modifier>
601
19799a22
GS
602Backslashed metacharacters in Perl are alphanumeric, such as C<\b>,
603C<\w>, C<\n>. Unlike some other regular expression languages, there
604are no backslashed symbols that aren't alphanumeric. So anything
c47ff5f1 605that looks like \\, \(, \), \<, \>, \{, or \} is always
19799a22
GS
606interpreted as a literal character, not a metacharacter. This was
607once used in a common idiom to disable or quote the special meanings
608of regular expression metacharacters in a string that you want to
36bbe248 609use for a pattern. Simply quote all non-"word" characters:
a0d0e21e
LW
610
611 $pattern =~ s/(\W)/\\$1/g;
612
f1cbbd6e 613(If C<use locale> is set, then this depends on the current locale.)
14218588
GS
614Today it is more common to use the quotemeta() function or the C<\Q>
615metaquoting escape sequence to disable all metacharacters' special
616meanings like this:
a0d0e21e
LW
617
618 /$unquoted\Q$quoted\E$unquoted/
619
9da458fc
IZ
620Beware that if you put literal backslashes (those not inside
621interpolated variables) between C<\Q> and C<\E>, double-quotish
622backslash interpolation may lead to confusing results. If you
623I<need> to use literal backslashes within C<\Q...\E>,
624consult L<perlop/"Gory details of parsing quoted constructs">.
625
19799a22
GS
626=head2 Extended Patterns
627
14218588
GS
628Perl also defines a consistent extension syntax for features not
629found in standard tools like B<awk> and B<lex>. The syntax is a
630pair of parentheses with a question mark as the first thing within
631the parentheses. The character after the question mark indicates
632the extension.
19799a22 633
14218588
GS
634The stability of these extensions varies widely. Some have been
635part of the core language for many years. Others are experimental
636and may change without warning or be completely removed. Check
637the documentation on an individual feature to verify its current
638status.
19799a22 639
14218588
GS
640A question mark was chosen for this and for the minimal-matching
641construct because 1) question marks are rare in older regular
642expressions, and 2) whenever you see one, you should stop and
643"question" exactly what is going on. That's psychology...
a0d0e21e
LW
644
645=over 10
646
cc6b7395 647=item C<(?#text)>
d74e8afc 648X<(?#)>
a0d0e21e 649
14218588 650A comment. The text is ignored. If the C</x> modifier enables
19799a22 651whitespace formatting, a simple C<#> will suffice. Note that Perl closes
259138e3
GS
652the comment as soon as it sees a C<)>, so there is no way to put a literal
653C<)> in the comment.
a0d0e21e 654
cde0cee5 655=item C<(?kimsx-imsx)>
d74e8afc 656X<(?)>
19799a22 657
0b6d1084
JH
658One or more embedded pattern-match modifiers, to be turned on (or
659turned off, if preceded by C<->) for the remainder of the pattern or
660the remainder of the enclosing pattern group (if any). This is
661particularly useful for dynamic patterns, such as those read in from a
0d017f4d
WL
662configuration file, taken from an argument, or specified in a table
663somewhere. Consider the case where some patterns want to be case
664sensitive and some do not: The case insensitive ones merely need to
665include C<(?i)> at the front of the pattern. For example:
19799a22
GS
666
667 $pattern = "foobar";
5d458dd8 668 if ( /$pattern/i ) { }
19799a22
GS
669
670 # more flexible:
671
672 $pattern = "(?i)foobar";
5d458dd8 673 if ( /$pattern/ ) { }
19799a22 674
0b6d1084 675These modifiers are restored at the end of the enclosing group. For example,
19799a22
GS
676
677 ( (?i) blah ) \s+ \1
678
0d017f4d
WL
679will match C<blah> in any case, some spaces, and an exact (I<including the case>!)
680repetition of the previous word, assuming the C</x> modifier, and no C</i>
681modifier outside this group.
19799a22 682
cde0cee5
YO
683Note that the C<k> modifier is special in that it can only be enabled,
684not disabled, and that its presence anywhere in a pattern has a global
685effect. Thus C<(?-k)> and C<(?-k:...)> are meaningless and will warn
686when executed under C<use warnings>.
687
5a964f20 688=item C<(?:pattern)>
d74e8afc 689X<(?:)>
a0d0e21e 690
ca9dfc88
IZ
691=item C<(?imsx-imsx:pattern)>
692
5a964f20
TC
693This is for clustering, not capturing; it groups subexpressions like
694"()", but doesn't make backreferences as "()" does. So
a0d0e21e 695
5a964f20 696 @fields = split(/\b(?:a|b|c)\b/)
a0d0e21e
LW
697
698is like
699
5a964f20 700 @fields = split(/\b(a|b|c)\b/)
a0d0e21e 701
19799a22
GS
702but doesn't spit out extra fields. It's also cheaper not to capture
703characters if you don't need to.
a0d0e21e 704
19799a22 705Any letters between C<?> and C<:> act as flags modifiers as with
5d458dd8 706C<(?imsx-imsx)>. For example,
ca9dfc88
IZ
707
708 /(?s-i:more.*than).*million/i
709
14218588 710is equivalent to the more verbose
ca9dfc88
IZ
711
712 /(?:(?s-i)more.*than).*million/i
713
594d7033
YO
714=item C<(?|pattern)>
715X<(?|)> X<Branch reset>
716
717This is the "branch reset" pattern, which has the special property
718that the capture buffers are numbered from the same starting point
4deaaa80
PJ
719in each branch. It is available starting from perl 5.10.
720
721Normally capture buffers in a pattern are numbered sequentially,
722from left to right. Inside this construct that behaviour is
723overridden so that the capture buffers are shared between all the
724branches and take their values from the branch that matched.
725
726The numbering within each branch will be as normal, and any buffers
727following this construct will be numbered as though the construct
728contained only one branch, that being the one with the most capture
729buffers in it.
730
731This construct will be useful when you want to capture one of a
732number of alternative matches.
733
734Consider the following pattern. The numbers underneath show in
735which buffer the captured content will be stored.
594d7033
YO
736
737
738 # before ---------------branch-reset----------- after
739 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
740 # 1 2 2 3 2 3 4
741
ee9b8eae
YO
742=item Look-Around Assertions
743X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
744
745Look-around assertions are zero width patterns which match a specific
746pattern without including it in C<$&>. Positive assertions match when
747their subpattern matches, negative assertions match when their subpattern
748fails. Look-behind matches text up to the current match position,
749look-ahead matches text following the current match position.
750
751=over 4
752
5a964f20 753=item C<(?=pattern)>
d74e8afc 754X<(?=)> X<look-ahead, positive> X<lookahead, positive>
a0d0e21e 755
19799a22 756A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/>
a0d0e21e
LW
757matches a word followed by a tab, without including the tab in C<$&>.
758
5a964f20 759=item C<(?!pattern)>
d74e8afc 760X<(?!)> X<look-ahead, negative> X<lookahead, negative>
a0d0e21e 761
19799a22 762A zero-width negative look-ahead assertion. For example C</foo(?!bar)/>
a0d0e21e 763matches any occurrence of "foo" that isn't followed by "bar". Note
19799a22
GS
764however that look-ahead and look-behind are NOT the same thing. You cannot
765use this for look-behind.
7b8d334a 766
5a964f20 767If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/>
7b8d334a
GS
768will not do what you want. That's because the C<(?!foo)> is just saying that
769the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will
770match. You would have to do something like C</(?!foo)...bar/> for that. We
771say "like" because there's the case of your "bar" not having three characters
772before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>.
773Sometimes it's still easier just to say:
a0d0e21e 774
a3cb178b 775 if (/bar/ && $` !~ /foo$/)
a0d0e21e 776
19799a22 777For look-behind see below.
c277df42 778
ee9b8eae
YO
779=item C<(?<=pattern)> C<\K>
780X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K>
c277df42 781
c47ff5f1 782A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/>
19799a22
GS
783matches a word that follows a tab, without including the tab in C<$&>.
784Works only for fixed-width look-behind.
c277df42 785
ee9b8eae
YO
786There is a special form of this construct, called C<\K>, which causes the
787regex engine to "keep" everything it had matched prior to the C<\K> and
788not include it in C<$&>. This effectively provides variable length
789look-behind. The use of C<\K> inside of another look-around assertion
790is allowed, but the behaviour is currently not well defined.
791
792For various reasons C<\K> may be signifigantly more efficient than the
793equivalent C<< (?<=...) >> construct, and it is especially useful in
794situations where you want to efficiently remove something following
795something else in a string. For instance
796
797 s/(foo)bar/$1/g;
798
799can be rewritten as the much more efficient
800
801 s/foo\Kbar//g;
802
5a964f20 803=item C<(?<!pattern)>
d74e8afc 804X<(?<!)> X<look-behind, negative> X<lookbehind, negative>
c277df42 805
19799a22
GS
806A zero-width negative look-behind assertion. For example C</(?<!bar)foo/>
807matches any occurrence of "foo" that does not follow "bar". Works
808only for fixed-width look-behind.
c277df42 809
ee9b8eae
YO
810=back
811
81714fb9
YO
812=item C<(?'NAME'pattern)>
813
814=item C<< (?<NAME>pattern) >>
815X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture>
816
817A named capture buffer. Identical in every respect to normal capturing
0d017f4d 818parentheses C<()> but for the additional fact that C<%+> may be used after
81714fb9
YO
819a succesful match to refer to a named buffer. See C<perlvar> for more
820details on the C<%+> hash.
821
822If multiple distinct capture buffers have the same name then the
823$+{NAME} will refer to the leftmost defined buffer in the match.
824
0d017f4d 825The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent.
81714fb9
YO
826
827B<NOTE:> While the notation of this construct is the same as the similar
0d017f4d 828function in .NET regexes, the behavior is not. In Perl the buffers are
81714fb9
YO
829numbered sequentially regardless of being named or not. Thus in the
830pattern
831
832 /(x)(?<foo>y)(z)/
833
834$+{foo} will be the same as $2, and $3 will contain 'z' instead of
835the opposite which is what a .NET regex hacker might expect.
836
1f1031fe
YO
837Currently NAME is restricted to simple identifiers only.
838In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or
839its Unicode extension (see L<utf8>),
840though it isn't extended by the locale (see L<perllocale>).
81714fb9 841
1f1031fe 842B<NOTE:> In order to make things easier for programmers with experience
0d017f4d
WL
843with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >>
844may be used instead of C<< (?<NAME>pattern) >>; however this form does not
1f1031fe
YO
845support the use of single quotes as a delimiter for the name. This is
846only available in Perl 5.10 or later.
81714fb9 847
1f1031fe
YO
848=item C<< \k<NAME> >>
849
850=item C<< \k'NAME' >>
81714fb9
YO
851
852Named backreference. Similar to numeric backreferences, except that
853the group is designated by name and not number. If multiple groups
854have the same name then it refers to the leftmost defined group in
855the current match.
856
0d017f4d 857It is an error to refer to a name not defined by a C<< (?<NAME>) >>
81714fb9
YO
858earlier in the pattern.
859
860Both forms are equivalent.
861
1f1031fe 862B<NOTE:> In order to make things easier for programmers with experience
0d017f4d
WL
863with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >>
864may be used instead of C<< \k<NAME> >> in Perl 5.10 or later.
1f1031fe 865
cc6b7395 866=item C<(?{ code })>
d74e8afc 867X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in>
c277df42 868
19799a22 869B<WARNING>: This extended regular expression feature is considered
b9b4dddf
YO
870experimental, and may be changed without notice. Code executed that
871has side effects may not perform identically from version to version
872due to the effect of future optimisations in the regex engine.
c277df42 873
cc46d5f2 874This zero-width assertion evaluates any embedded Perl code. It
19799a22
GS
875always succeeds, and its C<code> is not interpolated. Currently,
876the rules to determine where the C<code> ends are somewhat convoluted.
877
77ea4f6d
JV
878This feature can be used together with the special variable C<$^N> to
879capture the results of submatches in variables without having to keep
880track of the number of nested parentheses. For example:
881
882 $_ = "The brown fox jumps over the lazy dog";
883 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
884 print "color = $color, animal = $animal\n";
885
754091cb
RGS
886Inside the C<(?{...})> block, C<$_> refers to the string the regular
887expression is matching against. You can also use C<pos()> to know what is
fa11829f 888the current position of matching within this string.
754091cb 889
19799a22
GS
890The C<code> is properly scoped in the following sense: If the assertion
891is backtracked (compare L<"Backtracking">), all changes introduced after
892C<local>ization are undone, so that
b9ac3b5b
GS
893
894 $_ = 'a' x 8;
5d458dd8 895 m<
b9ac3b5b
GS
896 (?{ $cnt = 0 }) # Initialize $cnt.
897 (
5d458dd8 898 a
b9ac3b5b
GS
899 (?{
900 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
901 })
5d458dd8 902 )*
b9ac3b5b
GS
903 aaaa
904 (?{ $res = $cnt }) # On success copy to non-localized
905 # location.
906 >x;
907
0d017f4d 908will set C<$res = 4>. Note that after the match, C<$cnt> returns to the globally
14218588 909introduced value, because the scopes that restrict C<local> operators
b9ac3b5b
GS
910are unwound.
911
19799a22
GS
912This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)>
913switch. If I<not> used in this way, the result of evaluation of
914C<code> is put into the special variable C<$^R>. This happens
915immediately, so C<$^R> can be used from other C<(?{ code })> assertions
916inside the same regular expression.
b9ac3b5b 917
19799a22
GS
918The assignment to C<$^R> above is properly localized, so the old
919value of C<$^R> is restored if the assertion is backtracked; compare
920L<"Backtracking">.
b9ac3b5b 921
61528107
SP
922Due to an unfortunate implementation issue, the Perl code contained in these
923blocks is treated as a compile time closure that can have seemingly bizarre
6bda09f9 924consequences when used with lexically scoped variables inside of subroutines
61528107
SP
925or loops. There are various workarounds for this, including simply using
926global variables instead. If you are using this construct and strange results
6bda09f9
YO
927occur then check for the use of lexically scoped variables.
928
19799a22
GS
929For reasons of security, this construct is forbidden if the regular
930expression involves run-time interpolation of variables, unless the
931perilous C<use re 'eval'> pragma has been used (see L<re>), or the
932variables contain results of C<qr//> operator (see
5d458dd8 933L<perlop/"qr/STRING/imosx">).
871b0233 934
0d017f4d 935This restriction is due to the wide-spread and remarkably convenient
19799a22 936custom of using run-time determined strings as patterns. For example:
871b0233
IZ
937
938 $re = <>;
939 chomp $re;
940 $string =~ /$re/;
941
14218588
GS
942Before Perl knew how to execute interpolated code within a pattern,
943this operation was completely safe from a security point of view,
944although it could raise an exception from an illegal pattern. If
945you turn on the C<use re 'eval'>, though, it is no longer secure,
946so you should only do so if you are also using taint checking.
947Better yet, use the carefully constrained evaluation within a Safe
cc46d5f2 948compartment. See L<perlsec> for details about both these mechanisms.
871b0233 949
0d017f4d 950Because Perl's regex engine is currently not re-entrant, interpolated
8988a1bb
DD
951code may not invoke the regex engine either directly with C<m//> or C<s///>),
952or indirectly with functions such as C<split>.
953
14455d6c 954=item C<(??{ code })>
d74e8afc
ITB
955X<(??{})>
956X<regex, postponed> X<regexp, postponed> X<regular expression, postponed>
0f5d15d6 957
19799a22 958B<WARNING>: This extended regular expression feature is considered
b9b4dddf
YO
959experimental, and may be changed without notice. Code executed that
960has side effects may not perform identically from version to version
961due to the effect of future optimisations in the regex engine.
0f5d15d6 962
19799a22
GS
963This is a "postponed" regular subexpression. The C<code> is evaluated
964at run time, at the moment this subexpression may match. The result
965of evaluation is considered as a regular expression and matched as
61528107 966if it were inserted instead of this construct. Note that this means
6bda09f9
YO
967that the contents of capture buffers defined inside an eval'ed pattern
968are not available outside of the pattern, and vice versa, there is no
969way for the inner pattern to refer to a capture buffer defined outside.
970Thus,
971
972 ('a' x 100)=~/(??{'(.)' x 100})/
973
81714fb9 974B<will> match, it will B<not> set $1.
0f5d15d6 975
428594d9 976The C<code> is not interpolated. As before, the rules to determine
19799a22
GS
977where the C<code> ends are currently somewhat convoluted.
978
979The following pattern matches a parenthesized group:
0f5d15d6
IZ
980
981 $re = qr{
982 \(
983 (?:
984 (?> [^()]+ ) # Non-parens without backtracking
985 |
14455d6c 986 (??{ $re }) # Group with matching parens
0f5d15d6
IZ
987 )*
988 \)
989 }x;
990
6bda09f9
YO
991See also C<(?PARNO)> for a different, more efficient way to accomplish
992the same task.
993
5d458dd8 994Because perl's regex engine is not currently re-entrant, delayed
8988a1bb
DD
995code may not invoke the regex engine either directly with C<m//> or C<s///>),
996or indirectly with functions such as C<split>.
997
5d458dd8
YO
998Recursing deeper than 50 times without consuming any input string will
999result in a fatal error. The maximum depth is compiled into perl, so
6bda09f9
YO
1000changing it requires a custom build.
1001
542fa716
YO
1002=item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)>
1003X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)>
6bda09f9 1004X<regex, recursive> X<regexp, recursive> X<regular expression, recursive>
542fa716 1005X<regex, relative recursion>
6bda09f9 1006
81714fb9
YO
1007Similar to C<(??{ code })> except it does not involve compiling any code,
1008instead it treats the contents of a capture buffer as an independent
61528107 1009pattern that must match at the current position. Capture buffers
81714fb9 1010contained by the pattern will have the value as determined by the
6bda09f9
YO
1011outermost recursion.
1012
894be9b7
YO
1013PARNO is a sequence of digits (not starting with 0) whose value reflects
1014the paren-number of the capture buffer to recurse to. C<(?R)> recurses to
1015the beginning of the whole pattern. C<(?0)> is an alternate syntax for
542fa716
YO
1016C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed
1017to be relative, with negative numbers indicating preceding capture buffers
1018and positive ones following. Thus C<(?-1)> refers to the most recently
1019declared buffer, and C<(?+1)> indicates the next buffer to be declared.
c74340f9
YO
1020Note that the counting for relative recursion differs from that of
1021relative backreferences, in that with recursion unclosed buffers B<are>
1022included.
6bda09f9 1023
81714fb9 1024The following pattern matches a function foo() which may contain
f145b7e9 1025balanced parentheses as the argument.
6bda09f9
YO
1026
1027 $re = qr{ ( # paren group 1 (full function)
81714fb9 1028 foo
6bda09f9
YO
1029 ( # paren group 2 (parens)
1030 \(
1031 ( # paren group 3 (contents of parens)
1032 (?:
1033 (?> [^()]+ ) # Non-parens without backtracking
1034 |
1035 (?2) # Recurse to start of paren group 2
1036 )*
1037 )
1038 \)
1039 )
1040 )
1041 }x;
1042
1043If the pattern was used as follows
1044
1045 'foo(bar(baz)+baz(bop))'=~/$re/
1046 and print "\$1 = $1\n",
1047 "\$2 = $2\n",
1048 "\$3 = $3\n";
1049
1050the output produced should be the following:
1051
1052 $1 = foo(bar(baz)+baz(bop))
1053 $2 = (bar(baz)+baz(bop))
81714fb9 1054 $3 = bar(baz)+baz(bop)
6bda09f9 1055
81714fb9 1056If there is no corresponding capture buffer defined, then it is a
61528107 1057fatal error. Recursing deeper than 50 times without consuming any input
81714fb9 1058string will also result in a fatal error. The maximum depth is compiled
6bda09f9
YO
1059into perl, so changing it requires a custom build.
1060
542fa716
YO
1061The following shows how using negative indexing can make it
1062easier to embed recursive patterns inside of a C<qr//> construct
1063for later use:
1064
1065 my $parens = qr/(\((?:[^()]++|(?-1))*+\))/;
1066 if (/foo $parens \s+ + \s+ bar $parens/x) {
1067 # do something here...
1068 }
1069
81714fb9 1070B<Note> that this pattern does not behave the same way as the equivalent
0d017f4d 1071PCRE or Python construct of the same form. In Perl you can backtrack into
6bda09f9 1072a recursed group, in PCRE and Python the recursed into group is treated
542fa716
YO
1073as atomic. Also, modifiers are resolved at compile time, so constructs
1074like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will
1075be processed.
6bda09f9 1076
894be9b7
YO
1077=item C<(?&NAME)>
1078X<(?&NAME)>
1079
0d017f4d
WL
1080Recurse to a named subpattern. Identical to C<(?PARNO)> except that the
1081parenthesis to recurse to is determined by name. If multiple parentheses have
894be9b7
YO
1082the same name, then it recurses to the leftmost.
1083
1084It is an error to refer to a name that is not declared somewhere in the
1085pattern.
1086
1f1031fe
YO
1087B<NOTE:> In order to make things easier for programmers with experience
1088with the Python or PCRE regex engines the pattern C<< (?P>NAME) >>
0d017f4d 1089may be used instead of C<< (?&NAME) >> in Perl 5.10 or later.
1f1031fe 1090
e2e6a0f1
YO
1091=item C<(?(condition)yes-pattern|no-pattern)>
1092X<(?()>
286f584a 1093
e2e6a0f1 1094=item C<(?(condition)yes-pattern)>
286f584a 1095
e2e6a0f1
YO
1096Conditional expression. C<(condition)> should be either an integer in
1097parentheses (which is valid if the corresponding pair of parentheses
1098matched), a look-ahead/look-behind/evaluate zero-width assertion, a
1099name in angle brackets or single quotes (which is valid if a buffer
1100with the given name matched), or the special symbol (R) (true when
1101evaluated inside of recursion or eval). Additionally the R may be
1102followed by a number, (which will be true when evaluated when recursing
1103inside of the appropriate group), or by C<&NAME>, in which case it will
1104be true only when evaluated during recursion in the named group.
1105
1106Here's a summary of the possible predicates:
1107
1108=over 4
1109
1110=item (1) (2) ...
1111
1112Checks if the numbered capturing buffer has matched something.
1113
1114=item (<NAME>) ('NAME')
1115
1116Checks if a buffer with the given name has matched something.
1117
1118=item (?{ CODE })
1119
1120Treats the code block as the condition.
1121
1122=item (R)
1123
1124Checks if the expression has been evaluated inside of recursion.
1125
1126=item (R1) (R2) ...
1127
1128Checks if the expression has been evaluated while executing directly
1129inside of the n-th capture group. This check is the regex equivalent of
1130
1131 if ((caller(0))[3] eq 'subname') { ... }
1132
1133In other words, it does not check the full recursion stack.
1134
1135=item (R&NAME)
1136
1137Similar to C<(R1)>, this predicate checks to see if we're executing
1138directly inside of the leftmost group with a given name (this is the same
1139logic used by C<(?&NAME)> to disambiguate). It does not check the full
1140stack, but only the name of the innermost active recursion.
1141
1142=item (DEFINE)
1143
1144In this case, the yes-pattern is never directly executed, and no
1145no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient.
1146See below for details.
1147
1148=back
1149
1150For example:
1151
1152 m{ ( \( )?
1153 [^()]+
1154 (?(1) \) )
1155 }x
1156
1157matches a chunk of non-parentheses, possibly included in parentheses
1158themselves.
1159
1160A special form is the C<(DEFINE)> predicate, which never executes directly
1161its yes-pattern, and does not allow a no-pattern. This allows to define
1162subpatterns which will be executed only by using the recursion mechanism.
1163This way, you can define a set of regular expression rules that can be
1164bundled into any pattern you choose.
1165
1166It is recommended that for this usage you put the DEFINE block at the
1167end of the pattern, and that you name any subpatterns defined within it.
1168
1169Also, it's worth noting that patterns defined this way probably will
1170not be as efficient, as the optimiser is not very clever about
1171handling them.
1172
1173An example of how this might be used is as follows:
1174
2bf803e2 1175 /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT))
e2e6a0f1 1176 (?(DEFINE)
2bf803e2
YO
1177 (?<NAME_PAT>....)
1178 (?<ADRESS_PAT>....)
e2e6a0f1
YO
1179 )/x
1180
1181Note that capture buffers matched inside of recursion are not accessible
0d017f4d 1182after the recursion returns, so the extra layer of capturing buffers is
e2e6a0f1
YO
1183necessary. Thus C<$+{NAME_PAT}> would not be defined even though
1184C<$+{NAME}> would be.
286f584a 1185
c47ff5f1 1186=item C<< (?>pattern) >>
6bda09f9 1187X<backtrack> X<backtracking> X<atomic> X<possessive>
5a964f20 1188
19799a22
GS
1189An "independent" subexpression, one which matches the substring
1190that a I<standalone> C<pattern> would match if anchored at the given
9da458fc 1191position, and it matches I<nothing other than this substring>. This
19799a22
GS
1192construct is useful for optimizations of what would otherwise be
1193"eternal" matches, because it will not backtrack (see L<"Backtracking">).
9da458fc
IZ
1194It may also be useful in places where the "grab all you can, and do not
1195give anything back" semantic is desirable.
19799a22 1196
c47ff5f1 1197For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >>
19799a22
GS
1198(anchored at the beginning of string, as above) will match I<all>
1199characters C<a> at the beginning of string, leaving no C<a> for
1200C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>,
1201since the match of the subgroup C<a*> is influenced by the following
1202group C<ab> (see L<"Backtracking">). In particular, C<a*> inside
1203C<a*ab> will match fewer characters than a standalone C<a*>, since
1204this makes the tail match.
1205
c47ff5f1 1206An effect similar to C<< (?>pattern) >> may be achieved by writing
19799a22
GS
1207C<(?=(pattern))\1>. This matches the same substring as a standalone
1208C<a+>, and the following C<\1> eats the matched string; it therefore
c47ff5f1 1209makes a zero-length assertion into an analogue of C<< (?>...) >>.
19799a22
GS
1210(The difference between these two constructs is that the second one
1211uses a capturing group, thus shifting ordinals of backreferences
1212in the rest of a regular expression.)
1213
1214Consider this pattern:
c277df42 1215
871b0233 1216 m{ \(
e2e6a0f1
YO
1217 (
1218 [^()]+ # x+
1219 |
871b0233
IZ
1220 \( [^()]* \)
1221 )+
e2e6a0f1 1222 \)
871b0233 1223 }x
5a964f20 1224
19799a22
GS
1225That will efficiently match a nonempty group with matching parentheses
1226two levels deep or less. However, if there is no such group, it
1227will take virtually forever on a long string. That's because there
1228are so many different ways to split a long string into several
1229substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar
1230to a subpattern of the above pattern. Consider how the pattern
1231above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several
1232seconds, but that each extra letter doubles this time. This
1233exponential performance will make it appear that your program has
14218588 1234hung. However, a tiny change to this pattern
5a964f20 1235
e2e6a0f1
YO
1236 m{ \(
1237 (
1238 (?> [^()]+ ) # change x+ above to (?> x+ )
1239 |
871b0233
IZ
1240 \( [^()]* \)
1241 )+
e2e6a0f1 1242 \)
871b0233 1243 }x
c277df42 1244
c47ff5f1 1245which uses C<< (?>...) >> matches exactly when the one above does (verifying
5a964f20
TC
1246this yourself would be a productive exercise), but finishes in a fourth
1247the time when used on a similar string with 1000000 C<a>s. Be aware,
1248however, that this pattern currently triggers a warning message under
9f1b1f2d 1249the C<use warnings> pragma or B<-w> switch saying it
6bab786b 1250C<"matches null string many times in regex">.
c277df42 1251
c47ff5f1 1252On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable
19799a22 1253effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>.
c277df42
IZ
1254This was only 4 times slower on a string with 1000000 C<a>s.
1255
9da458fc
IZ
1256The "grab all you can, and do not give anything back" semantic is desirable
1257in many situations where on the first sight a simple C<()*> looks like
1258the correct solution. Suppose we parse text with comments being delimited
1259by C<#> followed by some optional (horizontal) whitespace. Contrary to
4375e838 1260its appearance, C<#[ \t]*> I<is not> the correct subexpression to match
9da458fc
IZ
1261the comment delimiter, because it may "give up" some whitespace if
1262the remainder of the pattern can be made to match that way. The correct
1263answer is either one of these:
1264
1265 (?>#[ \t]*)
1266 #[ \t]*(?![ \t])
1267
1268For example, to grab non-empty comments into $1, one should use either
1269one of these:
1270
1271 / (?> \# [ \t]* ) ( .+ ) /x;
1272 / \# [ \t]* ( [^ \t] .* ) /x;
1273
1274Which one you pick depends on which of these expressions better reflects
1275the above specification of comments.
1276
6bda09f9
YO
1277In some literature this construct is called "atomic matching" or
1278"possessive matching".
1279
b9b4dddf
YO
1280Possessive quantifiers are equivalent to putting the item they are applied
1281to inside of one of these constructs. The following equivalences apply:
1282
1283 Quantifier Form Bracketing Form
1284 --------------- ---------------
1285 PAT*+ (?>PAT*)
1286 PAT++ (?>PAT+)
1287 PAT?+ (?>PAT?)
1288 PAT{min,max}+ (?>PAT{min,max})
1289
e2e6a0f1
YO
1290=back
1291
1292=head2 Special Backtracking Control Verbs
1293
1294B<WARNING:> These patterns are experimental and subject to change or
0d017f4d 1295removal in a future version of Perl. Their usage in production code should
e2e6a0f1
YO
1296be noted to avoid problems during upgrades.
1297
1298These special patterns are generally of the form C<(*VERB:ARG)>. Unless
1299otherwise stated the ARG argument is optional; in some cases, it is
1300forbidden.
1301
1302Any pattern containing a special backtracking verb that allows an argument
1303has the special behaviour that when executed it sets the current packages'
5d458dd8
YO
1304C<$REGERROR> and C<$REGMARK> variables. When doing so the following
1305rules apply:
e2e6a0f1 1306
5d458dd8
YO
1307On failure, the C<$REGERROR> variable will be set to the ARG value of the
1308verb pattern, if the verb was involved in the failure of the match. If the
1309ARG part of the pattern was omitted, then C<$REGERROR> will be set to the
1310name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was
1311none. Also, the C<$REGMARK> variable will be set to FALSE.
e2e6a0f1 1312
5d458dd8
YO
1313On a successful match, the C<$REGERROR> variable will be set to FALSE, and
1314the C<$REGMARK> variable will be set to the name of the last
1315C<(*MARK:NAME)> pattern executed. See the explanation for the
1316C<(*MARK:NAME)> verb below for more details.
e2e6a0f1 1317
5d458dd8
YO
1318B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1>
1319and most other regex related variables. They are not local to a scope, nor
1320readonly, but instead are volatile package variables similar to C<$AUTOLOAD>.
1321Use C<local> to localize changes to them to a specific scope if necessary.
e2e6a0f1
YO
1322
1323If a pattern does not contain a special backtracking verb that allows an
5d458dd8 1324argument, then C<$REGERROR> and C<$REGMARK> are not touched at all.
e2e6a0f1
YO
1325
1326=over 4
1327
1328=item Verbs that take an argument
1329
1330=over 4
1331
5d458dd8 1332=item C<(*PRUNE)> C<(*PRUNE:NAME)>
ee9b8eae 1333X<(*PRUNE)> X<(*PRUNE:NAME)> X<\v>
54612592 1334
5d458dd8
YO
1335This zero-width pattern prunes the backtracking tree at the current point
1336when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>,
1337where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached,
1338A may backtrack as necessary to match. Once it is reached, matching
1339continues in B, which may also backtrack as necessary; however, should B
1340not match, then no further backtracking will take place, and the pattern
1341will fail outright at the current starting position.
54612592 1342
0d017f4d 1343As a shortcut, C<\v> is exactly equivalent to C<(*PRUNE)>.
ee9b8eae 1344
54612592
YO
1345The following example counts all the possible matching strings in a
1346pattern (without actually matching any of them).
1347
e2e6a0f1 1348 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/;
54612592
YO
1349 print "Count=$count\n";
1350
1351which produces:
1352
1353 aaab
1354 aaa
1355 aa
1356 a
1357 aab
1358 aa
1359 a
1360 ab
1361 a
1362 Count=9
1363
5d458dd8 1364If we add a C<(*PRUNE)> before the count like the following
54612592 1365
5d458dd8 1366 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/;
54612592
YO
1367 print "Count=$count\n";
1368
1369we prevent backtracking and find the count of the longest matching
1370at each matching startpoint like so:
1371
1372 aaab
1373 aab
1374 ab
1375 Count=3
1376
5d458dd8 1377Any number of C<(*PRUNE)> assertions may be used in a pattern.
54612592 1378
5d458dd8
YO
1379See also C<< (?>pattern) >> and possessive quantifiers for other ways to
1380control backtracking. In some cases, the use of C<(*PRUNE)> can be
1381replaced with a C<< (?>pattern) >> with no functional difference; however,
1382C<(*PRUNE)> can be used to handle cases that cannot be expressed using a
1383C<< (?>pattern) >> alone.
54612592 1384
e2e6a0f1 1385
5d458dd8
YO
1386=item C<(*SKIP)> C<(*SKIP:NAME)>
1387X<(*SKIP)>
e2e6a0f1 1388
5d458dd8 1389This zero-width pattern is similar to C<(*PRUNE)>, except that on
e2e6a0f1 1390failure it also signifies that whatever text that was matched leading up
5d458dd8
YO
1391to the C<(*SKIP)> pattern being executed cannot be part of I<any> match
1392of this pattern. This effectively means that the regex engine "skips" forward
1393to this position on failure and tries to match again, (assuming that
1394there is sufficient room to match).
1395
0d017f4d 1396As a shortcut C<\V> is exactly equivalent to C<(*SKIP)>.
ee9b8eae 1397
5d458dd8
YO
1398The name of the C<(*SKIP:NAME)> pattern has special significance. If a
1399C<(*MARK:NAME)> was encountered while matching, then it is that position
1400which is used as the "skip point". If no C<(*MARK)> of that name was
1401encountered, then the C<(*SKIP)> operator has no effect. When used
1402without a name the "skip point" is where the match point was when
1403executing the (*SKIP) pattern.
1404
1405Compare the following to the examples in C<(*PRUNE)>, note the string
24b23f37
YO
1406is twice as long:
1407
5d458dd8 1408 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/;
24b23f37
YO
1409 print "Count=$count\n";
1410
1411outputs
1412
1413 aaab
1414 aaab
1415 Count=2
1416
5d458dd8 1417Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)>
e2e6a0f1 1418executed, the next startpoint will be where the cursor was when the
5d458dd8
YO
1419C<(*SKIP)> was executed.
1420
5d458dd8
YO
1421=item C<(*MARK:NAME)> C<(*:NAME)>
1422X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)>
1423
1424This zero-width pattern can be used to mark the point reached in a string
1425when a certain part of the pattern has been successfully matched. This
1426mark may be given a name. A later C<(*SKIP)> pattern will then skip
1427forward to that point if backtracked into on failure. Any number of
1428C<(*MARK)> patterns are allowed, and the NAME portion is optional and may
1429be duplicated.
1430
1431In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)>
1432can be used to "label" a pattern branch, so that after matching, the
1433program can determine which branches of the pattern were involved in the
1434match.
1435
1436When a match is successful, the C<$REGMARK> variable will be set to the
1437name of the most recently executed C<(*MARK:NAME)> that was involved
1438in the match.
1439
1440This can be used to determine which branch of a pattern was matched
1441without using a seperate capture buffer for each branch, which in turn
1442can result in a performance improvement, as perl cannot optimize
1443C</(?:(x)|(y)|(z))/> as efficiently as something like
1444C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>.
1445
1446When a match has failed, and unless another verb has been involved in
1447failing the match and has provided its own name to use, the C<$REGERROR>
1448variable will be set to the name of the most recently executed
1449C<(*MARK:NAME)>.
1450
1451See C<(*SKIP)> for more details.
1452
b62d2d15
YO
1453As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>.
1454
5d458dd8
YO
1455=item C<(*THEN)> C<(*THEN:NAME)>
1456
1457This is similar to the "cut group" operator C<::> from Perl6. Like
1458C<(*PRUNE)>, this verb always matches, and when backtracked into on
1459failure, it causes the regex engine to try the next alternation in the
1460innermost enclosing group (capturing or otherwise).
1461
1462Its name comes from the observation that this operation combined with the
1463alternation operator (C<|>) can be used to create what is essentially a
1464pattern-based if/then/else block:
1465
1466 ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ )
1467
1468Note that if this operator is used and NOT inside of an alternation then
1469it acts exactly like the C<(*PRUNE)> operator.
1470
1471 / A (*PRUNE) B /
1472
1473is the same as
1474
1475 / A (*THEN) B /
1476
1477but
1478
1479 / ( A (*THEN) B | C (*THEN) D ) /
1480
1481is not the same as
1482
1483 / ( A (*PRUNE) B | C (*PRUNE) D ) /
1484
1485as after matching the A but failing on the B the C<(*THEN)> verb will
1486backtrack and try C; but the C<(*PRUNE)> verb will simply fail.
24b23f37 1487
e2e6a0f1
YO
1488=item C<(*COMMIT)>
1489X<(*COMMIT)>
24b23f37 1490
5d458dd8
YO
1491This is the Perl6 "commit pattern" C<< <commit> >> or C<:::>. It's a
1492zero-width pattern similar to C<(*SKIP)>, except that when backtracked
1493into on failure it causes the match to fail outright. No further attempts
1494to find a valid match by advancing the start pointer will occur again.
1495For example,
24b23f37 1496
e2e6a0f1 1497 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/;
24b23f37
YO
1498 print "Count=$count\n";
1499
1500outputs
1501
1502 aaab
1503 Count=1
1504
e2e6a0f1
YO
1505In other words, once the C<(*COMMIT)> has been entered, and if the pattern
1506does not match, the regex engine will not try any further matching on the
1507rest of the string.
c277df42 1508
e2e6a0f1 1509=back
9af228c6 1510
e2e6a0f1 1511=item Verbs without an argument
9af228c6
YO
1512
1513=over 4
1514
e2e6a0f1
YO
1515=item C<(*FAIL)> C<(*F)>
1516X<(*FAIL)> X<(*F)>
9af228c6 1517
e2e6a0f1
YO
1518This pattern matches nothing and always fails. It can be used to force the
1519engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In
1520fact, C<(?!)> gets optimised into C<(*FAIL)> internally.
9af228c6 1521
e2e6a0f1 1522It is probably useful only when combined with C<(?{})> or C<(??{})>.
9af228c6 1523
e2e6a0f1
YO
1524=item C<(*ACCEPT)>
1525X<(*ACCEPT)>
9af228c6 1526
e2e6a0f1
YO
1527B<WARNING:> This feature is highly experimental. It is not recommended
1528for production code.
9af228c6 1529
e2e6a0f1
YO
1530This pattern matches nothing and causes the end of successful matching at
1531the point at which the C<(*ACCEPT)> pattern was encountered, regardless of
1532whether there is actually more to match in the string. When inside of a
0d017f4d 1533nested pattern, such as recursion, or in a subpattern dynamically generated
e2e6a0f1 1534via C<(??{})>, only the innermost pattern is ended immediately.
9af228c6 1535
e2e6a0f1
YO
1536If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are
1537marked as ended at the point at which the C<(*ACCEPT)> was encountered.
1538For instance:
9af228c6 1539
e2e6a0f1 1540 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x;
9af228c6 1541
e2e6a0f1 1542will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not
0d017f4d 1543be set. If another branch in the inner parentheses were matched, such as in the
e2e6a0f1 1544string 'ACDE', then the C<D> and C<E> would have to be matched as well.
9af228c6
YO
1545
1546=back
c277df42 1547
a0d0e21e
LW
1548=back
1549
c07a80fd 1550=head2 Backtracking
d74e8afc 1551X<backtrack> X<backtracking>
c07a80fd 1552
35a734be
IZ
1553NOTE: This section presents an abstract approximation of regular
1554expression behavior. For a more rigorous (and complicated) view of
1555the rules involved in selecting a match among possible alternatives,
0d017f4d 1556see L<Combining RE Pieces>.
35a734be 1557
c277df42 1558A fundamental feature of regular expression matching involves the
5a964f20 1559notion called I<backtracking>, which is currently used (when needed)
0d017f4d 1560by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>,
9da458fc
IZ
1561C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized
1562internally, but the general principle outlined here is valid.
c07a80fd
PP
1563
1564For a regular expression to match, the I<entire> regular expression must
1565match, not just part of it. So if the beginning of a pattern containing a
1566quantifier succeeds in a way that causes later parts in the pattern to
1567fail, the matching engine backs up and recalculates the beginning
1568part--that's why it's called backtracking.
1569
1570Here is an example of backtracking: Let's say you want to find the
1571word following "foo" in the string "Food is on the foo table.":
1572
1573 $_ = "Food is on the foo table.";
1574 if ( /\b(foo)\s+(\w+)/i ) {
1575 print "$2 follows $1.\n";
1576 }
1577
1578When the match runs, the first part of the regular expression (C<\b(foo)>)
1579finds a possible match right at the beginning of the string, and loads up
1580$1 with "Foo". However, as soon as the matching engine sees that there's
1581no whitespace following the "Foo" that it had saved in $1, it realizes its
68dc0745 1582mistake and starts over again one character after where it had the
c07a80fd
PP
1583tentative match. This time it goes all the way until the next occurrence
1584of "foo". The complete regular expression matches this time, and you get
1585the expected output of "table follows foo."
1586
1587Sometimes minimal matching can help a lot. Imagine you'd like to match
1588everything between "foo" and "bar". Initially, you write something
1589like this:
1590
1591 $_ = "The food is under the bar in the barn.";
1592 if ( /foo(.*)bar/ ) {
1593 print "got <$1>\n";
1594 }
1595
1596Which perhaps unexpectedly yields:
1597
1598 got <d is under the bar in the >
1599
1600That's because C<.*> was greedy, so you get everything between the
14218588 1601I<first> "foo" and the I<last> "bar". Here it's more effective
c07a80fd
PP
1602to use minimal matching to make sure you get the text between a "foo"
1603and the first "bar" thereafter.
1604
1605 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
1606 got <d is under the >
1607
0d017f4d 1608Here's another example. Let's say you'd like to match a number at the end
b6e13d97 1609of a string, and you also want to keep the preceding part of the match.
c07a80fd
PP
1610So you write this:
1611
1612 $_ = "I have 2 numbers: 53147";
1613 if ( /(.*)(\d*)/ ) { # Wrong!
1614 print "Beginning is <$1>, number is <$2>.\n";
1615 }
1616
1617That won't work at all, because C<.*> was greedy and gobbled up the
1618whole string. As C<\d*> can match on an empty string the complete
1619regular expression matched successfully.
1620
8e1088bc 1621 Beginning is <I have 2 numbers: 53147>, number is <>.
c07a80fd
PP
1622
1623Here are some variants, most of which don't work:
1624
1625 $_ = "I have 2 numbers: 53147";
1626 @pats = qw{
1627 (.*)(\d*)
1628 (.*)(\d+)
1629 (.*?)(\d*)
1630 (.*?)(\d+)
1631 (.*)(\d+)$
1632 (.*?)(\d+)$
1633 (.*)\b(\d+)$
1634 (.*\D)(\d+)$
1635 };
1636
1637 for $pat (@pats) {
1638 printf "%-12s ", $pat;
1639 if ( /$pat/ ) {
1640 print "<$1> <$2>\n";
1641 } else {
1642 print "FAIL\n";
1643 }
1644 }
1645
1646That will print out:
1647
1648 (.*)(\d*) <I have 2 numbers: 53147> <>
1649 (.*)(\d+) <I have 2 numbers: 5314> <7>
1650 (.*?)(\d*) <> <>
1651 (.*?)(\d+) <I have > <2>
1652 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
1653 (.*?)(\d+)$ <I have 2 numbers: > <53147>
1654 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
1655 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
1656
1657As you see, this can be a bit tricky. It's important to realize that a
1658regular expression is merely a set of assertions that gives a definition
1659of success. There may be 0, 1, or several different ways that the
1660definition might succeed against a particular string. And if there are
5a964f20
TC
1661multiple ways it might succeed, you need to understand backtracking to
1662know which variety of success you will achieve.
c07a80fd 1663
19799a22 1664When using look-ahead assertions and negations, this can all get even
8b19b778 1665trickier. Imagine you'd like to find a sequence of non-digits not
c07a80fd
PP
1666followed by "123". You might try to write that as
1667
871b0233
IZ
1668 $_ = "ABC123";
1669 if ( /^\D*(?!123)/ ) { # Wrong!
1670 print "Yup, no 123 in $_\n";
1671 }
c07a80fd
PP
1672
1673But that isn't going to match; at least, not the way you're hoping. It
1674claims that there is no 123 in the string. Here's a clearer picture of
9b9391b2 1675why that pattern matches, contrary to popular expectations:
c07a80fd 1676
4358a253
SS
1677 $x = 'ABC123';
1678 $y = 'ABC445';
c07a80fd 1679
4358a253
SS
1680 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
1681 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
c07a80fd 1682
4358a253
SS
1683 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
1684 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
c07a80fd
PP
1685
1686This prints
1687
1688 2: got ABC
1689 3: got AB
1690 4: got ABC
1691
5f05dabc 1692You might have expected test 3 to fail because it seems to a more
c07a80fd
PP
1693general purpose version of test 1. The important difference between
1694them is that test 3 contains a quantifier (C<\D*>) and so can use
1695backtracking, whereas test 1 will not. What's happening is
1696that you've asked "Is it true that at the start of $x, following 0 or more
5f05dabc 1697non-digits, you have something that's not 123?" If the pattern matcher had
c07a80fd 1698let C<\D*> expand to "ABC", this would have caused the whole pattern to
54310121 1699fail.
14218588 1700
c07a80fd 1701The search engine will initially match C<\D*> with "ABC". Then it will
14218588 1702try to match C<(?!123> with "123", which fails. But because
c07a80fd
PP
1703a quantifier (C<\D*>) has been used in the regular expression, the
1704search engine can backtrack and retry the match differently
54310121 1705in the hope of matching the complete regular expression.
c07a80fd 1706
5a964f20
TC
1707The pattern really, I<really> wants to succeed, so it uses the
1708standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this
c07a80fd 1709time. Now there's indeed something following "AB" that is not
14218588 1710"123". It's "C123", which suffices.
c07a80fd 1711
14218588
GS
1712We can deal with this by using both an assertion and a negation.
1713We'll say that the first part in $1 must be followed both by a digit
1714and by something that's not "123". Remember that the look-aheads
1715are zero-width expressions--they only look, but don't consume any
1716of the string in their match. So rewriting this way produces what
c07a80fd
PP
1717you'd expect; that is, case 5 will fail, but case 6 succeeds:
1718
4358a253
SS
1719 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
1720 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
c07a80fd
PP
1721
1722 6: got ABC
1723
5a964f20 1724In other words, the two zero-width assertions next to each other work as though
19799a22 1725they're ANDed together, just as you'd use any built-in assertions: C</^$/>
c07a80fd
PP
1726matches only if you're at the beginning of the line AND the end of the
1727line simultaneously. The deeper underlying truth is that juxtaposition in
1728regular expressions always means AND, except when you write an explicit OR
1729using the vertical bar. C</ab/> means match "a" AND (then) match "b",
1730although the attempted matches are made at different positions because "a"
1731is not a zero-width assertion, but a one-width assertion.
1732
0d017f4d 1733B<WARNING>: Particularly complicated regular expressions can take
14218588 1734exponential time to solve because of the immense number of possible
0d017f4d 1735ways they can use backtracking to try for a match. For example, without
9da458fc
IZ
1736internal optimizations done by the regular expression engine, this will
1737take a painfully long time to run:
c07a80fd 1738
e1901655
IZ
1739 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
1740
1741And if you used C<*>'s in the internal groups instead of limiting them
1742to 0 through 5 matches, then it would take forever--or until you ran
1743out of stack space. Moreover, these internal optimizations are not
1744always applicable. For example, if you put C<{0,5}> instead of C<*>
1745on the external group, no current optimization is applicable, and the
1746match takes a long time to finish.
c07a80fd 1747
9da458fc
IZ
1748A powerful tool for optimizing such beasts is what is known as an
1749"independent group",
c47ff5f1 1750which does not backtrack (see L<C<< (?>pattern) >>>). Note also that
9da458fc 1751zero-length look-ahead/look-behind assertions will not backtrack to make
5d458dd8 1752the tail match, since they are in "logical" context: only
14218588 1753whether they match is considered relevant. For an example
9da458fc 1754where side-effects of look-ahead I<might> have influenced the
c47ff5f1 1755following match, see L<C<< (?>pattern) >>>.
c277df42 1756
a0d0e21e 1757=head2 Version 8 Regular Expressions
d74e8afc 1758X<regular expression, version 8> X<regex, version 8> X<regexp, version 8>
a0d0e21e 1759
5a964f20 1760In case you're not familiar with the "regular" Version 8 regex
a0d0e21e
LW
1761routines, here are the pattern-matching rules not described above.
1762
54310121 1763Any single character matches itself, unless it is a I<metacharacter>
a0d0e21e 1764with a special meaning described here or above. You can cause
5a964f20 1765characters that normally function as metacharacters to be interpreted
5f05dabc 1766literally by prefixing them with a "\" (e.g., "\." matches a ".", not any
0d017f4d
WL
1767character; "\\" matches a "\"). This escape mechanism is also required
1768for the character used as the pattern delimiter.
1769
1770A series of characters matches that series of characters in the target
1771string, so the pattern C<blurfl> would match "blurfl" in the target
1772string.
a0d0e21e
LW
1773
1774You can specify a character class, by enclosing a list of characters
5d458dd8 1775in C<[]>, which will match any character from the list. If the
a0d0e21e 1776first character after the "[" is "^", the class matches any character not
14218588 1777in the list. Within a list, the "-" character specifies a
5a964f20 1778range, so that C<a-z> represents all characters between "a" and "z",
8a4f6ac2
GS
1779inclusive. If you want either "-" or "]" itself to be a member of a
1780class, put it at the start of the list (possibly after a "^"), or
1781escape it with a backslash. "-" is also taken literally when it is
1782at the end of the list, just before the closing "]". (The
84850974
DD
1783following all specify the same class of three characters: C<[-az]>,
1784C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which
5d458dd8
YO
1785specifies a class containing twenty-six characters, even on EBCDIC-based
1786character sets.) Also, if you try to use the character
1787classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of
1788a range, the "-" is understood literally.
a0d0e21e 1789
8ada0baa
JH
1790Note also that the whole range idea is rather unportable between
1791character sets--and even within character sets they may cause results
1792you probably didn't expect. A sound principle is to use only ranges
0d017f4d 1793that begin from and end at either alphabetics of equal case ([a-e],
8ada0baa
JH
1794[A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt,
1795spell out the character sets in full.
1796
54310121 1797Characters may be specified using a metacharacter syntax much like that
a0d0e21e
LW
1798used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
1799"\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string
5d458dd8
YO
1800of octal digits, matches the character whose coded character set value
1801is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits,
1802matches the character whose numeric value is I<nn>. The expression \cI<x>
1803matches the character control-I<x>. Finally, the "." metacharacter
fb55449c 1804matches any character except "\n" (unless you use C</s>).
a0d0e21e
LW
1805
1806You can specify a series of alternatives for a pattern using "|" to
1807separate them, so that C<fee|fie|foe> will match any of "fee", "fie",
5a964f20 1808or "foe" in the target string (as would C<f(e|i|o)e>). The
a0d0e21e
LW
1809first alternative includes everything from the last pattern delimiter
1810("(", "[", or the beginning of the pattern) up to the first "|", and
1811the last alternative contains everything from the last "|" to the next
14218588
GS
1812pattern delimiter. That's why it's common practice to include
1813alternatives in parentheses: to minimize confusion about where they
a3cb178b
GS
1814start and end.
1815
5a964f20 1816Alternatives are tried from left to right, so the first
a3cb178b
GS
1817alternative found for which the entire expression matches, is the one that
1818is chosen. This means that alternatives are not necessarily greedy. For
628afcb5 1819example: when matching C<foo|foot> against "barefoot", only the "foo"
a3cb178b
GS
1820part will match, as that is the first alternative tried, and it successfully
1821matches the target string. (This might not seem important, but it is
1822important when you are capturing matched text using parentheses.)
1823
5a964f20 1824Also remember that "|" is interpreted as a literal within square brackets,
a3cb178b 1825so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>.
a0d0e21e 1826
14218588
GS
1827Within a pattern, you may designate subpatterns for later reference
1828by enclosing them in parentheses, and you may refer back to the
1829I<n>th subpattern later in the pattern using the metacharacter
1830\I<n>. Subpatterns are numbered based on the left to right order
1831of their opening parenthesis. A backreference matches whatever
1832actually matched the subpattern in the string being examined, not
1833the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will
1834match "0x1234 0x4321", but not "0x1234 01234", because subpattern
18351 matched "0x", even though the rule C<0|0x> could potentially match
1836the leading 0 in the second number.
cb1a09d0 1837
0d017f4d 1838=head2 Warning on \1 Instead of $1
cb1a09d0 1839
5a964f20 1840Some people get too used to writing things like:
cb1a09d0
AD
1841
1842 $pattern =~ s/(\W)/\\\1/g;
1843
1844This is grandfathered for the RHS of a substitute to avoid shocking the
1845B<sed> addicts, but it's a dirty habit to get into. That's because in
d1be9408 1846PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in
cb1a09d0
AD
1847the usual double-quoted string means a control-A. The customary Unix
1848meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit
1849of doing that, you get yourself into trouble if you then add an C</e>
1850modifier.
1851
5a964f20 1852 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
cb1a09d0
AD
1853
1854Or if you try to do
1855
1856 s/(\d+)/\1000/;
1857
1858You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with
14218588 1859C<${1}000>. The operation of interpolation should not be confused
cb1a09d0
AD
1860with the operation of matching a backreference. Certainly they mean two
1861different things on the I<left> side of the C<s///>.
9fa51da4 1862
0d017f4d 1863=head2 Repeated Patterns Matching a Zero-length Substring
c84d73f1 1864
19799a22 1865B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite.
c84d73f1
IZ
1866
1867Regular expressions provide a terse and powerful programming language. As
1868with most other power tools, power comes together with the ability
1869to wreak havoc.
1870
1871A common abuse of this power stems from the ability to make infinite
628afcb5 1872loops using regular expressions, with something as innocuous as:
c84d73f1
IZ
1873
1874 'foo' =~ m{ ( o? )* }x;
1875
0d017f4d 1876The C<o?> matches at the beginning of C<'foo'>, and since the position
c84d73f1 1877in the string is not moved by the match, C<o?> would match again and again
14218588 1878because of the C<*> modifier. Another common way to create a similar cycle
c84d73f1
IZ
1879is with the looping modifier C<//g>:
1880
1881 @matches = ( 'foo' =~ m{ o? }xg );
1882
1883or
1884
1885 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1886
1887or the loop implied by split().
1888
1889However, long experience has shown that many programming tasks may
14218588
GS
1890be significantly simplified by using repeated subexpressions that
1891may match zero-length substrings. Here's a simple example being:
c84d73f1
IZ
1892
1893 @chars = split //, $string; # // is not magic in split
1894 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1895
9da458fc 1896Thus Perl allows such constructs, by I<forcefully breaking
c84d73f1
IZ
1897the infinite loop>. The rules for this are different for lower-level
1898loops given by the greedy modifiers C<*+{}>, and for higher-level
1899ones like the C</g> modifier or split() operator.
1900
19799a22
GS
1901The lower-level loops are I<interrupted> (that is, the loop is
1902broken) when Perl detects that a repeated expression matched a
1903zero-length substring. Thus
c84d73f1
IZ
1904
1905 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;
1906
5d458dd8 1907is made equivalent to
c84d73f1 1908
5d458dd8
YO
1909 m{ (?: NON_ZERO_LENGTH )*
1910 |
1911 (?: ZERO_LENGTH )?
c84d73f1
IZ
1912 }x;
1913
1914The higher level-loops preserve an additional state between iterations:
5d458dd8 1915whether the last match was zero-length. To break the loop, the following
c84d73f1 1916match after a zero-length match is prohibited to have a length of zero.
5d458dd8 1917This prohibition interacts with backtracking (see L<"Backtracking">),
c84d73f1
IZ
1918and so the I<second best> match is chosen if the I<best> match is of
1919zero length.
1920
19799a22 1921For example:
c84d73f1
IZ
1922
1923 $_ = 'bar';
1924 s/\w??/<$&>/g;
1925
20fb949f 1926results in C<< <><b><><a><><r><> >>. At each position of the string the best
5d458dd8 1927match given by non-greedy C<??> is the zero-length match, and the I<second
c84d73f1
IZ
1928best> match is what is matched by C<\w>. Thus zero-length matches
1929alternate with one-character-long matches.
1930
5d458dd8 1931Similarly, for repeated C<m/()/g> the second-best match is the match at the
c84d73f1
IZ
1932position one notch further in the string.
1933
19799a22 1934The additional state of being I<matched with zero-length> is associated with
c84d73f1 1935the matched string, and is reset by each assignment to pos().
9da458fc
IZ
1936Zero-length matches at the end of the previous match are ignored
1937during C<split>.
c84d73f1 1938
0d017f4d 1939=head2 Combining RE Pieces
35a734be
IZ
1940
1941Each of the elementary pieces of regular expressions which were described
1942before (such as C<ab> or C<\Z>) could match at most one substring
1943at the given position of the input string. However, in a typical regular
1944expression these elementary pieces are combined into more complicated
1945patterns using combining operators C<ST>, C<S|T>, C<S*> etc
1946(in these examples C<S> and C<T> are regular subexpressions).
1947
1948Such combinations can include alternatives, leading to a problem of choice:
1949if we match a regular expression C<a|ab> against C<"abc">, will it match
1950substring C<"a"> or C<"ab">? One way to describe which substring is
1951actually matched is the concept of backtracking (see L<"Backtracking">).
1952However, this description is too low-level and makes you think
1953in terms of a particular implementation.
1954
1955Another description starts with notions of "better"/"worse". All the
1956substrings which may be matched by the given regular expression can be
1957sorted from the "best" match to the "worst" match, and it is the "best"
1958match which is chosen. This substitutes the question of "what is chosen?"
1959by the question of "which matches are better, and which are worse?".
1960
1961Again, for elementary pieces there is no such question, since at most
1962one match at a given position is possible. This section describes the
1963notion of better/worse for combining operators. In the description
1964below C<S> and C<T> are regular subexpressions.
1965
13a2d996 1966=over 4
35a734be
IZ
1967
1968=item C<ST>
1969
1970Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are
1971substrings which can be matched by C<S>, C<B> and C<B'> are substrings
5d458dd8 1972which can be matched by C<T>.
35a734be
IZ
1973
1974If C<A> is better match for C<S> than C<A'>, C<AB> is a better
1975match than C<A'B'>.
1976
1977If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if
1978C<B> is better match for C<T> than C<B'>.
1979
1980=item C<S|T>
1981
1982When C<S> can match, it is a better match than when only C<T> can match.
1983
1984Ordering of two matches for C<S> is the same as for C<S>. Similar for
1985two matches for C<T>.
1986
1987=item C<S{REPEAT_COUNT}>
1988
1989Matches as C<SSS...S> (repeated as many times as necessary).
1990
1991=item C<S{min,max}>
1992
1993Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>.
1994
1995=item C<S{min,max}?>
1996
1997Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>.
1998
1999=item C<S?>, C<S*>, C<S+>
2000
2001Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively.
2002
2003=item C<S??>, C<S*?>, C<S+?>
2004
2005Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively.
2006
c47ff5f1 2007=item C<< (?>S) >>
35a734be
IZ
2008
2009Matches the best match for C<S> and only that.
2010
2011=item C<(?=S)>, C<(?<=S)>
2012
2013Only the best match for C<S> is considered. (This is important only if
2014C<S> has capturing parentheses, and backreferences are used somewhere
2015else in the whole regular expression.)
2016
2017=item C<(?!S)>, C<(?<!S)>
2018
2019For this grouping operator there is no need to describe the ordering, since
2020only whether or not C<S> can match is important.
2021
6bda09f9 2022=item C<(??{ EXPR })>, C<(?PARNO)>
35a734be
IZ
2023
2024The ordering is the same as for the regular expression which is
6bda09f9 2025the result of EXPR, or the pattern contained by capture buffer PARNO.
35a734be
IZ
2026
2027=item C<(?(condition)yes-pattern|no-pattern)>
2028
2029Recall that which of C<yes-pattern> or C<no-pattern> actually matches is
2030already determined. The ordering of the matches is the same as for the
2031chosen subexpression.
2032
2033=back
2034
2035The above recipes describe the ordering of matches I<at a given position>.
2036One more rule is needed to understand how a match is determined for the
2037whole regular expression: a match at an earlier position is always better
2038than a match at a later position.
2039
0d017f4d 2040=head2 Creating Custom RE Engines
c84d73f1
IZ
2041
2042Overloaded constants (see L<overload>) provide a simple way to extend
2043the functionality of the RE engine.
2044
2045Suppose that we want to enable a new RE escape-sequence C<\Y|> which
0d017f4d 2046matches at a boundary between whitespace characters and non-whitespace
c84d73f1
IZ
2047characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly
2048at these positions, so we want to have each C<\Y|> in the place of the
2049more complicated version. We can create a module C<customre> to do
2050this:
2051
2052 package customre;
2053 use overload;
2054
2055 sub import {
2056 shift;
2057 die "No argument to customre::import allowed" if @_;
2058 overload::constant 'qr' => \&convert;
2059 }
2060
2061 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
2062
580a9fe1
RGS
2063 # We must also take care of not escaping the legitimate \\Y|
2064 # sequence, hence the presence of '\\' in the conversion rules.
5d458dd8 2065 my %rules = ( '\\' => '\\\\',
c84d73f1
IZ
2066 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ );
2067 sub convert {
2068 my $re = shift;
5d458dd8 2069 $re =~ s{
c84d73f1
IZ
2070 \\ ( \\ | Y . )
2071 }
5d458dd8 2072 { $rules{$1} or invalid($re,$1) }sgex;
c84d73f1
IZ
2073 return $re;
2074 }
2075
2076Now C<use customre> enables the new escape in constant regular
2077expressions, i.e., those without any runtime variable interpolations.
2078As documented in L<overload>, this conversion will work only over
2079literal parts of regular expressions. For C<\Y|$re\Y|> the variable
2080part of this regular expression needs to be converted explicitly
2081(but only if the special meaning of C<\Y|> should be enabled inside $re):
2082
2083 use customre;
2084 $re = <>;
2085 chomp $re;
2086 $re = customre::convert $re;
2087 /\Y|$re\Y|/;
2088
1f1031fe
YO
2089=head1 PCRE/Python Support
2090
2091As of Perl 5.10 Perl supports several Python/PCRE specific extensions
2092to the regex syntax. While Perl programmers are encouraged to use the
2093Perl specific syntax, the following are legal in Perl 5.10:
2094
2095=over 4
2096
ee9b8eae 2097=item C<< (?PE<lt>NAMEE<gt>pattern) >>
1f1031fe
YO
2098
2099Define a named capture buffer. Equivalent to C<< (?<NAME>pattern) >>.
2100
2101=item C<< (?P=NAME) >>
2102
2103Backreference to a named capture buffer. Equivalent to C<< \g{NAME} >>.
2104
2105=item C<< (?P>NAME) >>
2106
2107Subroutine call to a named capture buffer. Equivalent to C<< (?&NAME) >>.
2108
ee9b8eae 2109=back
1f1031fe 2110
19799a22
GS
2111=head1 BUGS
2112
9da458fc
IZ
2113This document varies from difficult to understand to completely
2114and utterly opaque. The wandering prose riddled with jargon is
2115hard to fathom in several places.
2116
2117This document needs a rewrite that separates the tutorial content
2118from the reference content.
19799a22
GS
2119
2120=head1 SEE ALSO
9fa51da4 2121
91e0c79e
MJD
2122L<perlrequick>.
2123
2124L<perlretut>.
2125
9b599b2a
GS
2126L<perlop/"Regexp Quote-Like Operators">.
2127
1e66bd83
PP
2128L<perlop/"Gory details of parsing quoted constructs">.
2129
14218588
GS
2130L<perlfaq6>.
2131
9b599b2a
GS
2132L<perlfunc/pos>.
2133
2134L<perllocale>.
2135
fb55449c
JH
2136L<perlebcdic>.
2137
14218588
GS
2138I<Mastering Regular Expressions> by Jeffrey Friedl, published
2139by O'Reilly and Associates.