Commit | Line | Data |
---|---|---|
a0d0e21e | 1 | =head1 NAME |
d74e8afc | 2 | X<regular expression> X<regex> X<regexp> |
a0d0e21e LW |
3 | |
4 | perlre - Perl regular expressions | |
5 | ||
6 | =head1 DESCRIPTION | |
7 | ||
5d458dd8 | 8 | This page describes the syntax of regular expressions in Perl. |
91e0c79e | 9 | |
cc46d5f2 | 10 | If you haven't used regular expressions before, a quick-start |
91e0c79e MJD |
11 | introduction is available in L<perlrequick>, and a longer tutorial |
12 | introduction is available in L<perlretut>. | |
13 | ||
14 | For reference on how regular expressions are used in matching | |
15 | operations, plus various examples of the same, see discussions of | |
16 | C<m//>, C<s///>, C<qr//> and C<??> in L<perlop/"Regexp Quote-Like | |
17 | Operators">. | |
cb1a09d0 | 18 | |
0d017f4d WL |
19 | |
20 | =head2 Modifiers | |
21 | ||
19799a22 | 22 | Matching operations can have various modifiers. Modifiers |
5a964f20 | 23 | that relate to the interpretation of the regular expression inside |
19799a22 | 24 | are listed below. Modifiers that alter the way a regular expression |
5d458dd8 | 25 | is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and |
1e66bd83 | 26 | L<perlop/"Gory details of parsing quoted constructs">. |
a0d0e21e | 27 | |
55497cff | 28 | =over 4 |
29 | ||
54310121 | 30 | =item m |
d74e8afc | 31 | X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline> |
55497cff | 32 | |
33 | Treat string as multiple lines. That is, change "^" and "$" from matching | |
14218588 | 34 | the start or end of the string to matching the start or end of any |
7f761169 | 35 | line anywhere within the string. |
55497cff | 36 | |
54310121 | 37 | =item s |
d74e8afc ITB |
38 | X</s> X<regex, single-line> X<regexp, single-line> |
39 | X<regular expression, single-line> | |
55497cff | 40 | |
41 | Treat string as single line. That is, change "." to match any character | |
19799a22 | 42 | whatsoever, even a newline, which normally it would not match. |
55497cff | 43 | |
f02c194e | 44 | Used together, as /ms, they let the "." match any character whatsoever, |
fb55449c | 45 | while still allowing "^" and "$" to match, respectively, just after |
19799a22 | 46 | and just before newlines within the string. |
7b8d334a | 47 | |
87e95b7f YO |
48 | =item i |
49 | X</i> X<regex, case-insensitive> X<regexp, case-insensitive> | |
50 | X<regular expression, case-insensitive> | |
51 | ||
52 | Do case-insensitive pattern matching. | |
53 | ||
54 | If C<use locale> is in effect, the case map is taken from the current | |
55 | locale. See L<perllocale>. | |
56 | ||
54310121 | 57 | =item x |
d74e8afc | 58 | X</x> |
55497cff | 59 | |
60 | Extend your pattern's legibility by permitting whitespace and comments. | |
61 | ||
87e95b7f YO |
62 | =item p |
63 | X</p> X<regex, preserve> X<regexp, preserve> | |
64 | ||
65 | Preserve the string matched such that ${^PREMATCH}, {$^MATCH}, and | |
66 | ${^POSTMATCH} are available for use after matching. | |
67 | ||
55497cff | 68 | =back |
a0d0e21e LW |
69 | |
70 | These are usually written as "the C</x> modifier", even though the delimiter | |
14218588 | 71 | in question might not really be a slash. Any of these |
a0d0e21e | 72 | modifiers may also be embedded within the regular expression itself using |
14218588 | 73 | the C<(?...)> construct. See below. |
a0d0e21e | 74 | |
4633a7c4 | 75 | The C</x> modifier itself needs a little more explanation. It tells |
55497cff | 76 | the regular expression parser to ignore whitespace that is neither |
77 | backslashed nor within a character class. You can use this to break up | |
4633a7c4 | 78 | your regular expression into (slightly) more readable parts. The C<#> |
54310121 | 79 | character is also treated as a metacharacter introducing a comment, |
55497cff | 80 | just as in ordinary Perl code. This also means that if you want real |
14218588 | 81 | whitespace or C<#> characters in the pattern (outside a character |
f9a3ff1a RGS |
82 | class, where they are unaffected by C</x>), then you'll either have to |
83 | escape them (using backslashes or C<\Q...\E>) or encode them using octal | |
8933a740 RGS |
84 | or hex escapes. Taken together, these features go a long way towards |
85 | making Perl's regular expressions more readable. Note that you have to | |
86 | be careful not to include the pattern delimiter in the comment--perl has | |
87 | no way of knowing you did not intend to close the pattern early. See | |
88 | the C-comment deletion code in L<perlop>. Also note that anything inside | |
1031e5db | 89 | a C<\Q...\E> stays unaffected by C</x>. |
d74e8afc | 90 | X</x> |
a0d0e21e LW |
91 | |
92 | =head2 Regular Expressions | |
93 | ||
04838cea RGS |
94 | =head3 Metacharacters |
95 | ||
0d017f4d | 96 | The patterns used in Perl pattern matching evolved from the ones supplied in |
14218588 | 97 | the Version 8 regex routines. (The routines are derived |
19799a22 GS |
98 | (distantly) from Henry Spencer's freely redistributable reimplementation |
99 | of the V8 routines.) See L<Version 8 Regular Expressions> for | |
100 | details. | |
a0d0e21e LW |
101 | |
102 | In particular the following metacharacters have their standard I<egrep>-ish | |
103 | meanings: | |
d74e8afc ITB |
104 | X<metacharacter> |
105 | X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]> | |
106 | ||
a0d0e21e | 107 | |
54310121 | 108 | \ Quote the next metacharacter |
a0d0e21e LW |
109 | ^ Match the beginning of the line |
110 | . Match any character (except newline) | |
c07a80fd | 111 | $ Match the end of the line (or before newline at the end) |
a0d0e21e LW |
112 | | Alternation |
113 | () Grouping | |
114 | [] Character class | |
115 | ||
14218588 GS |
116 | By default, the "^" character is guaranteed to match only the |
117 | beginning of the string, the "$" character only the end (or before the | |
118 | newline at the end), and Perl does certain optimizations with the | |
a0d0e21e LW |
119 | assumption that the string contains only one line. Embedded newlines |
120 | will not be matched by "^" or "$". You may, however, wish to treat a | |
4a6725af | 121 | string as a multi-line buffer, such that the "^" will match after any |
0d520e8e YO |
122 | newline within the string (except if the newline is the last character in |
123 | the string), and "$" will match before any newline. At the | |
a0d0e21e LW |
124 | cost of a little more overhead, you can do this by using the /m modifier |
125 | on the pattern match operator. (Older programs did this by setting C<$*>, | |
f02c194e | 126 | but this practice has been removed in perl 5.9.) |
d74e8afc | 127 | X<^> X<$> X</m> |
a0d0e21e | 128 | |
14218588 | 129 | To simplify multi-line substitutions, the "." character never matches a |
55497cff | 130 | newline unless you use the C</s> modifier, which in effect tells Perl to pretend |
f02c194e | 131 | the string is a single line--even if it isn't. |
d74e8afc | 132 | X<.> X</s> |
a0d0e21e | 133 | |
04838cea RGS |
134 | =head3 Quantifiers |
135 | ||
a0d0e21e | 136 | The following standard quantifiers are recognized: |
d74e8afc | 137 | X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}> |
a0d0e21e LW |
138 | |
139 | * Match 0 or more times | |
140 | + Match 1 or more times | |
141 | ? Match 1 or 0 times | |
142 | {n} Match exactly n times | |
143 | {n,} Match at least n times | |
144 | {n,m} Match at least n but not more than m times | |
145 | ||
146 | (If a curly bracket occurs in any other context, it is treated | |
b975c076 JL |
147 | as a regular character. In particular, the lower bound |
148 | is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+" | |
25f94b33 | 149 | modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited |
9c79236d GS |
150 | to integral values less than a preset limit defined when perl is built. |
151 | This is usually 32766 on the most common platforms. The actual limit can | |
152 | be seen in the error message generated by code such as this: | |
153 | ||
820475bd | 154 | $_ **= $_ , / {$_} / for 2 .. 42; |
a0d0e21e | 155 | |
54310121 | 156 | By default, a quantified subpattern is "greedy", that is, it will match as |
157 | many times as possible (given a particular starting location) while still | |
158 | allowing the rest of the pattern to match. If you want it to match the | |
159 | minimum number of times possible, follow the quantifier with a "?". Note | |
160 | that the meanings don't change, just the "greediness": | |
0d017f4d | 161 | X<metacharacter> X<greedy> X<greediness> |
d74e8afc | 162 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?> |
a0d0e21e | 163 | |
0d017f4d WL |
164 | *? Match 0 or more times, not greedily |
165 | +? Match 1 or more times, not greedily | |
166 | ?? Match 0 or 1 time, not greedily | |
167 | {n}? Match exactly n times, not greedily | |
168 | {n,}? Match at least n times, not greedily | |
169 | {n,m}? Match at least n but not more than m times, not greedily | |
a0d0e21e | 170 | |
b9b4dddf YO |
171 | By default, when a quantified subpattern does not allow the rest of the |
172 | overall pattern to match, Perl will backtrack. However, this behaviour is | |
0d017f4d | 173 | sometimes undesirable. Thus Perl provides the "possessive" quantifier form |
b9b4dddf YO |
174 | as well. |
175 | ||
0d017f4d WL |
176 | *+ Match 0 or more times and give nothing back |
177 | ++ Match 1 or more times and give nothing back | |
178 | ?+ Match 0 or 1 time and give nothing back | |
b9b4dddf | 179 | {n}+ Match exactly n times and give nothing back (redundant) |
04838cea RGS |
180 | {n,}+ Match at least n times and give nothing back |
181 | {n,m}+ Match at least n but not more than m times and give nothing back | |
b9b4dddf YO |
182 | |
183 | For instance, | |
184 | ||
185 | 'aaaa' =~ /a++a/ | |
186 | ||
187 | will never match, as the C<a++> will gobble up all the C<a>'s in the | |
188 | string and won't leave any for the remaining part of the pattern. This | |
189 | feature can be extremely useful to give perl hints about where it | |
190 | shouldn't backtrack. For instance, the typical "match a double-quoted | |
191 | string" problem can be most efficiently performed when written as: | |
192 | ||
193 | /"(?:[^"\\]++|\\.)*+"/ | |
194 | ||
0d017f4d | 195 | as we know that if the final quote does not match, backtracking will not |
b9b4dddf YO |
196 | help. See the independent subexpression C<< (?>...) >> for more details; |
197 | possessive quantifiers are just syntactic sugar for that construct. For | |
198 | instance the above example could also be written as follows: | |
199 | ||
200 | /"(?>(?:(?>[^"\\]+)|\\.)*)"/ | |
201 | ||
04838cea RGS |
202 | =head3 Escape sequences |
203 | ||
5f05dabc | 204 | Because patterns are processed as double quoted strings, the following |
a0d0e21e | 205 | also work: |
0d017f4d | 206 | X<\t> X<\n> X<\r> X<\f> X<\e> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q> |
d74e8afc | 207 | X<\0> X<\c> X<\N> X<\x> |
a0d0e21e | 208 | |
0f36ee90 | 209 | \t tab (HT, TAB) |
210 | \n newline (LF, NL) | |
211 | \r return (CR) | |
212 | \f form feed (FF) | |
213 | \a alarm (bell) (BEL) | |
214 | \e escape (think troff) (ESC) | |
0d017f4d WL |
215 | \033 octal char (example: ESC) |
216 | \x1B hex char (example: ESC) | |
217 | \x{263a} wide hex char (example: Unicode SMILEY) | |
218 | \cK control char (example: VT) | |
4a2d328f | 219 | \N{name} named char |
cb1a09d0 AD |
220 | \l lowercase next char (think vi) |
221 | \u uppercase next char (think vi) | |
222 | \L lowercase till \E (think vi) | |
223 | \U uppercase till \E (think vi) | |
224 | \E end case modification (think vi) | |
5a964f20 | 225 | \Q quote (disable) pattern metacharacters till \E |
a0d0e21e | 226 | |
a034a98d | 227 | If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u> |
423cee85 | 228 | and C<\U> is taken from the current locale. See L<perllocale>. For |
4a2d328f | 229 | documentation of C<\N{name}>, see L<charnames>. |
a034a98d | 230 | |
1d2dff63 GS |
231 | You cannot include a literal C<$> or C<@> within a C<\Q> sequence. |
232 | An unescaped C<$> or C<@> interpolates the corresponding variable, | |
233 | while escaping will cause the literal string C<\$> to be matched. | |
234 | You'll need to write something like C<m/\Quser\E\@\Qhost/>. | |
235 | ||
04838cea RGS |
236 | =head3 Character classes |
237 | ||
a0d0e21e | 238 | In addition, Perl defines the following: |
d74e8afc | 239 | X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C> |
0d017f4d WL |
240 | X<\g> X<\k> X<\N> X<\K> X<\v> X<\V> |
241 | X<word> X<whitespace> X<character class> X<backreference> | |
a0d0e21e | 242 | |
81714fb9 YO |
243 | \w Match a "word" character (alphanumeric plus "_") |
244 | \W Match a non-"word" character | |
245 | \s Match a whitespace character | |
246 | \S Match a non-whitespace character | |
247 | \d Match a digit character | |
248 | \D Match a non-digit character | |
249 | \pP Match P, named property. Use \p{Prop} for longer names. | |
250 | \PP Match non-P | |
251 | \X Match eXtended Unicode "combining character sequence", | |
252 | equivalent to (?:\PM\pM*) | |
253 | \C Match a single C char (octet) even under Unicode. | |
254 | NOTE: breaks up characters into their UTF-8 bytes, | |
255 | so you may end up with malformed pieces of UTF-8. | |
256 | Unsupported in lookbehind. | |
5d458dd8 | 257 | \1 Backreference to a specific group. |
c74340f9 | 258 | '1' may actually be any positive integer. |
2bf803e2 YO |
259 | \g1 Backreference to a specific or previous group, |
260 | \g{-1} number may be negative indicating a previous buffer and may | |
261 | optionally be wrapped in curly brackets for safer parsing. | |
1f1031fe | 262 | \g{name} Named backreference |
81714fb9 | 263 | \k<name> Named backreference |
5d458dd8 | 264 | \N{name} Named unicode character, or unicode escape |
81714fb9 YO |
265 | \x12 Hexadecimal escape sequence |
266 | \x{1234} Long hexadecimal escape sequence | |
ee9b8eae YO |
267 | \K Keep the stuff left of the \K, don't include it in $& |
268 | \v Shortcut for (*PRUNE) | |
269 | \V Shortcut for (*SKIP) | |
a0d0e21e | 270 | |
08ce8fc6 JH |
271 | A C<\w> matches a single alphanumeric character (an alphabetic |
272 | character, or a decimal digit) or C<_>, not a whole word. Use C<\w+> | |
273 | to match a string of Perl-identifier characters (which isn't the same | |
274 | as matching an English word). If C<use locale> is in effect, the list | |
275 | of alphabetic characters generated by C<\w> is taken from the current | |
276 | locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>, | |
0d017f4d WL |
277 | C<\d>, and C<\D> within character classes, but they aren't usable |
278 | as either end of a range. If any of them precedes or follows a "-", | |
279 | the "-" is understood literally. If Unicode is in effect, C<\s> matches | |
280 | also "\x{85}", "\x{2028}, and "\x{2029}". See L<perlunicode> for more | |
281 | details about C<\pP>, C<\PP>, C<\X> and the possibility of defining | |
282 | your own C<\p> and C<\P> properties, and L<perluniintro> about Unicode | |
283 | in general. | |
d74e8afc | 284 | X<\w> X<\W> X<word> |
a0d0e21e | 285 | |
b8c5462f | 286 | The POSIX character class syntax |
d74e8afc | 287 | X<character class> |
b8c5462f | 288 | |
820475bd | 289 | [:class:] |
b8c5462f | 290 | |
0d017f4d | 291 | is also available. Note that the C<[> and C<]> brackets are I<literal>; |
5496314a SP |
292 | they must always be used within a character class expression. |
293 | ||
294 | # this is correct: | |
295 | $string =~ /[[:alpha:]]/; | |
296 | ||
297 | # this is not, and will generate a warning: | |
298 | $string =~ /[:alpha:]/; | |
299 | ||
300 | The available classes and their backslash equivalents (if available) are | |
301 | as follows: | |
d74e8afc ITB |
302 | X<character class> |
303 | X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph> | |
304 | X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit> | |
b8c5462f JH |
305 | |
306 | alpha | |
307 | alnum | |
308 | ascii | |
aaa51d5e | 309 | blank [1] |
b8c5462f JH |
310 | cntrl |
311 | digit \d | |
312 | graph | |
313 | lower | |
314 | ||
315 | punct | |
aaa51d5e | 316 | space \s [2] |
b8c5462f | 317 | upper |
aaa51d5e | 318 | word \w [3] |
b8c5462f JH |
319 | xdigit |
320 | ||
07698885 RGS |
321 | =over |
322 | ||
323 | =item [1] | |
324 | ||
b432a672 | 325 | A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace". |
07698885 RGS |
326 | |
327 | =item [2] | |
328 | ||
329 | Not exactly equivalent to C<\s> since the C<[[:space:]]> includes | |
0d017f4d | 330 | also the (very rare) "vertical tabulator", "\cK" or chr(11) in ASCII. |
07698885 RGS |
331 | |
332 | =item [3] | |
333 | ||
08ce8fc6 | 334 | A Perl extension, see above. |
07698885 RGS |
335 | |
336 | =back | |
aaa51d5e | 337 | |
26b44a0a | 338 | For example use C<[:upper:]> to match all the uppercase characters. |
aaa51d5e JF |
339 | Note that the C<[]> are part of the C<[::]> construct, not part of the |
340 | whole character class. For example: | |
b8c5462f | 341 | |
820475bd | 342 | [01[:alpha:]%] |
b8c5462f | 343 | |
0d017f4d | 344 | matches zero, one, any alphabetic character, and the percent sign. |
b8c5462f | 345 | |
72ff2908 JH |
346 | The following equivalences to Unicode \p{} constructs and equivalent |
347 | backslash character classes (if available), will hold: | |
d74e8afc | 348 | X<character class> X<\p> X<\p{}> |
72ff2908 | 349 | |
5496314a | 350 | [[:...:]] \p{...} backslash |
b8c5462f JH |
351 | |
352 | alpha IsAlpha | |
353 | alnum IsAlnum | |
354 | ascii IsASCII | |
0d017f4d | 355 | blank |
b8c5462f | 356 | cntrl IsCntrl |
3bec3564 | 357 | digit IsDigit \d |
b8c5462f JH |
358 | graph IsGraph |
359 | lower IsLower | |
360 | print IsPrint | |
361 | punct IsPunct | |
362 | space IsSpace | |
3bec3564 | 363 | IsSpacePerl \s |
b8c5462f JH |
364 | upper IsUpper |
365 | word IsWord | |
366 | xdigit IsXDigit | |
367 | ||
5496314a | 368 | For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent. |
b8c5462f JH |
369 | |
370 | If the C<utf8> pragma is not used but the C<locale> pragma is, the | |
aaa51d5e | 371 | classes correlate with the usual isalpha(3) interface (except for |
b432a672 | 372 | "word" and "blank"). |
b8c5462f JH |
373 | |
374 | The assumedly non-obviously named classes are: | |
375 | ||
376 | =over 4 | |
377 | ||
378 | =item cntrl | |
d74e8afc | 379 | X<cntrl> |
b8c5462f | 380 | |
820475bd GS |
381 | Any control character. Usually characters that don't produce output as |
382 | such but instead control the terminal somehow: for example newline and | |
383 | backspace are control characters. All characters with ord() less than | |
0d017f4d | 384 | 32 are usually classified as control characters (assuming ASCII, |
7be5a6cf JF |
385 | the ISO Latin character sets, and Unicode), as is the character with |
386 | the ord() value of 127 (C<DEL>). | |
b8c5462f JH |
387 | |
388 | =item graph | |
d74e8afc | 389 | X<graph> |
b8c5462f | 390 | |
f1cbbd6e | 391 | Any alphanumeric or punctuation (special) character. |
b8c5462f JH |
392 | |
393 | =item print | |
d74e8afc | 394 | X<print> |
b8c5462f | 395 | |
f79b3095 | 396 | Any alphanumeric or punctuation (special) character or the space character. |
b8c5462f JH |
397 | |
398 | =item punct | |
d74e8afc | 399 | X<punct> |
b8c5462f | 400 | |
f1cbbd6e | 401 | Any punctuation (special) character. |
b8c5462f JH |
402 | |
403 | =item xdigit | |
d74e8afc | 404 | X<xdigit> |
b8c5462f | 405 | |
593df60c | 406 | Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would |
820475bd | 407 | work just fine) it is included for completeness. |
b8c5462f | 408 | |
b8c5462f JH |
409 | =back |
410 | ||
411 | You can negate the [::] character classes by prefixing the class name | |
412 | with a '^'. This is a Perl extension. For example: | |
d74e8afc | 413 | X<character class, negation> |
b8c5462f | 414 | |
5496314a | 415 | POSIX traditional Unicode |
93733859 | 416 | |
5496314a SP |
417 | [[:^digit:]] \D \P{IsDigit} |
418 | [[:^space:]] \S \P{IsSpace} | |
419 | [[:^word:]] \W \P{IsWord} | |
b8c5462f | 420 | |
54c18d04 MK |
421 | Perl respects the POSIX standard in that POSIX character classes are |
422 | only supported within a character class. The POSIX character classes | |
423 | [.cc.] and [=cc=] are recognized but B<not> supported and trying to | |
424 | use them will cause an error. | |
b8c5462f | 425 | |
04838cea RGS |
426 | =head3 Assertions |
427 | ||
a0d0e21e | 428 | Perl defines the following zero-width assertions: |
d74e8afc ITB |
429 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> |
430 | X<regexp, zero-width assertion> | |
431 | X<regular expression, zero-width assertion> | |
432 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> | |
a0d0e21e LW |
433 | |
434 | \b Match a word boundary | |
0d017f4d | 435 | \B Match except at a word boundary |
b85d18e9 IZ |
436 | \A Match only at beginning of string |
437 | \Z Match only at end of string, or before newline at the end | |
438 | \z Match only at end of string | |
9da458fc IZ |
439 | \G Match only at pos() (e.g. at the end-of-match position |
440 | of prior m//g) | |
a0d0e21e | 441 | |
14218588 | 442 | A word boundary (C<\b>) is a spot between two characters |
19799a22 GS |
443 | that has a C<\w> on one side of it and a C<\W> on the other side |
444 | of it (in either order), counting the imaginary characters off the | |
445 | beginning and end of the string as matching a C<\W>. (Within | |
446 | character classes C<\b> represents backspace rather than a word | |
447 | boundary, just as it normally does in any double-quoted string.) | |
448 | The C<\A> and C<\Z> are just like "^" and "$", except that they | |
449 | won't match multiple times when the C</m> modifier is used, while | |
450 | "^" and "$" will match at every internal line boundary. To match | |
451 | the actual end of the string and not ignore an optional trailing | |
452 | newline, use C<\z>. | |
d74e8afc | 453 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 GS |
454 | |
455 | The C<\G> assertion can be used to chain global matches (using | |
456 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. | |
457 | It is also useful when writing C<lex>-like scanners, when you have | |
458 | several patterns that you want to match against consequent substrings | |
459 | of your string, see the previous reference. The actual location | |
460 | where C<\G> will match can also be influenced by using C<pos()> as | |
58e23c8d YO |
461 | an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length |
462 | matches is modified somewhat, in that contents to the left of C<\G> is | |
463 | not counted when determining the length of the match. Thus the following | |
464 | will not match forever: | |
d74e8afc | 465 | X<\G> |
c47ff5f1 | 466 | |
58e23c8d YO |
467 | $str = 'ABC'; |
468 | pos($str) = 1; | |
469 | while (/.\G/g) { | |
470 | print $&; | |
471 | } | |
472 | ||
473 | It will print 'A' and then terminate, as it considers the match to | |
474 | be zero-width, and thus will not match at the same position twice in a | |
475 | row. | |
476 | ||
477 | It is worth noting that C<\G> improperly used can result in an infinite | |
478 | loop. Take care when using patterns that include C<\G> in an alternation. | |
479 | ||
04838cea RGS |
480 | =head3 Capture buffers |
481 | ||
0d017f4d WL |
482 | The bracketing construct C<( ... )> creates capture buffers. To refer |
483 | to the current contents of a buffer later on, within the same pattern, | |
484 | use \1 for the first, \2 for the second, and so on. | |
485 | Outside the match use "$" instead of "\". (The | |
81714fb9 | 486 | \<digit> notation works in certain circumstances outside |
14218588 GS |
487 | the match. See the warning below about \1 vs $1 for details.) |
488 | Referring back to another part of the match is called a | |
489 | I<backreference>. | |
d74e8afc ITB |
490 | X<regex, capture buffer> X<regexp, capture buffer> |
491 | X<regular expression, capture buffer> X<backreference> | |
14218588 GS |
492 | |
493 | There is no limit to the number of captured substrings that you may | |
494 | use. However Perl also uses \10, \11, etc. as aliases for \010, | |
fb55449c JH |
495 | \011, etc. (Recall that 0 means octal, so \011 is the character at |
496 | number 9 in your coded character set; which would be the 10th character, | |
81714fb9 YO |
497 | a horizontal tab under ASCII.) Perl resolves this |
498 | ambiguity by interpreting \10 as a backreference only if at least 10 | |
499 | left parentheses have opened before it. Likewise \11 is a | |
500 | backreference only if at least 11 left parentheses have opened | |
501 | before it. And so on. \1 through \9 are always interpreted as | |
5624f11d | 502 | backreferences. |
c74340f9 | 503 | |
1f1031fe | 504 | X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference> |
2bf803e2 | 505 | In order to provide a safer and easier way to construct patterns using |
0d017f4d | 506 | backreferences, Perl 5.10 provides the C<\g{N}> notation. The curly |
2bf803e2 YO |
507 | brackets are optional, however omitting them is less safe as the meaning |
508 | of the pattern can be changed by text (such as digits) following it. | |
509 | When N is a positive integer the C<\g{N}> notation is exactly equivalent | |
510 | to using normal backreferences. When N is a negative integer then it is | |
511 | a relative backreference referring to the previous N'th capturing group. | |
1f1031fe YO |
512 | When the bracket form is used and N is not an integer, it is treated as a |
513 | reference to a named buffer. | |
2bf803e2 YO |
514 | |
515 | Thus C<\g{-1}> refers to the last buffer, C<\g{-2}> refers to the | |
516 | buffer before that. For example: | |
5624f11d YO |
517 | |
518 | / | |
519 | (Y) # buffer 1 | |
520 | ( # buffer 2 | |
521 | (X) # buffer 3 | |
2bf803e2 YO |
522 | \g{-1} # backref to buffer 3 |
523 | \g{-3} # backref to buffer 1 | |
5624f11d YO |
524 | ) |
525 | /x | |
526 | ||
2bf803e2 | 527 | and would match the same as C</(Y) ( (X) \3 \1 )/x>. |
14218588 | 528 | |
81714fb9 | 529 | Additionally, as of Perl 5.10 you may use named capture buffers and named |
1f1031fe | 530 | backreferences. The notation is C<< (?<name>...) >> to declare and C<< \k<name> >> |
0d017f4d WL |
531 | to reference. You may also use apostrophes instead of angle brackets to delimit the |
532 | name; and you may use the bracketed C<< \g{name} >> backreference syntax. | |
533 | It's possible to refer to a named capture buffer by absolute and relative number as well. | |
534 | Outside the pattern, a named capture buffer is available via the C<%+> hash. | |
535 | When different buffers within the same pattern have the same name, C<$+{name}> | |
536 | and C<< \k<name> >> refer to the leftmost defined group. (Thus it's possible | |
537 | to do things with named capture buffers that would otherwise require C<(??{})> | |
538 | code to accomplish.) | |
539 | X<named capture buffer> X<regular expression, named capture buffer> | |
540 | X<%+> X<$+{name}> X<\k{name}> | |
81714fb9 | 541 | |
14218588 | 542 | Examples: |
a0d0e21e LW |
543 | |
544 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
545 | ||
81714fb9 YO |
546 | /(.)\1/ # find first doubled char |
547 | and print "'$1' is the first doubled character\n"; | |
548 | ||
549 | /(?<char>.)\k<char>/ # ... a different way | |
550 | and print "'$+{char}' is the first doubled character\n"; | |
551 | ||
0d017f4d | 552 | /(?'char'.)\1/ # ... mix and match |
81714fb9 | 553 | and print "'$1' is the first doubled character\n"; |
c47ff5f1 | 554 | |
14218588 | 555 | if (/Time: (..):(..):(..)/) { # parse out values |
a0d0e21e LW |
556 | $hours = $1; |
557 | $minutes = $2; | |
558 | $seconds = $3; | |
559 | } | |
c47ff5f1 | 560 | |
14218588 GS |
561 | Several special variables also refer back to portions of the previous |
562 | match. C<$+> returns whatever the last bracket match matched. | |
563 | C<$&> returns the entire matched string. (At one point C<$0> did | |
564 | also, but now it returns the name of the program.) C<$`> returns | |
77ea4f6d JV |
565 | everything before the matched string. C<$'> returns everything |
566 | after the matched string. And C<$^N> contains whatever was matched by | |
567 | the most-recently closed group (submatch). C<$^N> can be used in | |
568 | extended patterns (see below), for example to assign a submatch to a | |
81714fb9 | 569 | variable. |
d74e8afc | 570 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 | 571 | |
665e98b9 | 572 | The numbered match variables ($1, $2, $3, etc.) and the related punctuation |
77ea4f6d | 573 | set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped |
14218588 GS |
574 | until the end of the enclosing block or until the next successful |
575 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
d74e8afc ITB |
576 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
577 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> | |
578 | ||
14218588 | 579 | |
0d017f4d | 580 | B<NOTE>: Failed matches in Perl do not reset the match variables, |
5146ce24 | 581 | which makes it easier to write code that tests for a series of more |
665e98b9 JH |
582 | specific cases and remembers the best match. |
583 | ||
14218588 GS |
584 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or |
585 | C<$'> anywhere in the program, it has to provide them for every | |
586 | pattern match. This may substantially slow your program. Perl | |
587 | uses the same mechanism to produce $1, $2, etc, so you also pay a | |
588 | price for each pattern that contains capturing parentheses. (To | |
589 | avoid this cost while retaining the grouping behaviour, use the | |
590 | extended regular expression C<(?: ... )> instead.) But if you never | |
591 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
592 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
593 | if you can, but if you can't (and some algorithms really appreciate | |
594 | them), once you've used them once, use them at will, because you've | |
595 | already paid the price. As of 5.005, C<$&> is not so costly as the | |
596 | other two. | |
d74e8afc | 597 | X<$&> X<$`> X<$'> |
68dc0745 | 598 | |
cde0cee5 YO |
599 | As a workaround for this problem, Perl 5.10 introduces C<${^PREMATCH}>, |
600 | C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&> | |
601 | and C<$'>, B<except> that they are only guaranteed to be defined after a | |
87e95b7f | 602 | successful match that was executed with the C</p> (preserve) modifier. |
cde0cee5 YO |
603 | The use of these variables incurs no global performance penalty, unlike |
604 | their punctuation char equivalents, however at the trade-off that you | |
605 | have to tell perl when you want to use them. | |
87e95b7f | 606 | X</p> X<p modifier> |
cde0cee5 | 607 | |
19799a22 GS |
608 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
609 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
610 | are no backslashed symbols that aren't alphanumeric. So anything | |
c47ff5f1 | 611 | that looks like \\, \(, \), \<, \>, \{, or \} is always |
19799a22 GS |
612 | interpreted as a literal character, not a metacharacter. This was |
613 | once used in a common idiom to disable or quote the special meanings | |
614 | of regular expression metacharacters in a string that you want to | |
36bbe248 | 615 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e LW |
616 | |
617 | $pattern =~ s/(\W)/\\$1/g; | |
618 | ||
f1cbbd6e | 619 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 GS |
620 | Today it is more common to use the quotemeta() function or the C<\Q> |
621 | metaquoting escape sequence to disable all metacharacters' special | |
622 | meanings like this: | |
a0d0e21e LW |
623 | |
624 | /$unquoted\Q$quoted\E$unquoted/ | |
625 | ||
9da458fc IZ |
626 | Beware that if you put literal backslashes (those not inside |
627 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
628 | backslash interpolation may lead to confusing results. If you | |
629 | I<need> to use literal backslashes within C<\Q...\E>, | |
630 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
631 | ||
19799a22 GS |
632 | =head2 Extended Patterns |
633 | ||
14218588 GS |
634 | Perl also defines a consistent extension syntax for features not |
635 | found in standard tools like B<awk> and B<lex>. The syntax is a | |
636 | pair of parentheses with a question mark as the first thing within | |
637 | the parentheses. The character after the question mark indicates | |
638 | the extension. | |
19799a22 | 639 | |
14218588 GS |
640 | The stability of these extensions varies widely. Some have been |
641 | part of the core language for many years. Others are experimental | |
642 | and may change without warning or be completely removed. Check | |
643 | the documentation on an individual feature to verify its current | |
644 | status. | |
19799a22 | 645 | |
14218588 GS |
646 | A question mark was chosen for this and for the minimal-matching |
647 | construct because 1) question marks are rare in older regular | |
648 | expressions, and 2) whenever you see one, you should stop and | |
649 | "question" exactly what is going on. That's psychology... | |
a0d0e21e LW |
650 | |
651 | =over 10 | |
652 | ||
cc6b7395 | 653 | =item C<(?#text)> |
d74e8afc | 654 | X<(?#)> |
a0d0e21e | 655 | |
14218588 | 656 | A comment. The text is ignored. If the C</x> modifier enables |
19799a22 | 657 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes |
259138e3 GS |
658 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
659 | C<)> in the comment. | |
a0d0e21e | 660 | |
cde0cee5 | 661 | =item C<(?kimsx-imsx)> |
d74e8afc | 662 | X<(?)> |
19799a22 | 663 | |
0b6d1084 JH |
664 | One or more embedded pattern-match modifiers, to be turned on (or |
665 | turned off, if preceded by C<->) for the remainder of the pattern or | |
666 | the remainder of the enclosing pattern group (if any). This is | |
667 | particularly useful for dynamic patterns, such as those read in from a | |
0d017f4d WL |
668 | configuration file, taken from an argument, or specified in a table |
669 | somewhere. Consider the case where some patterns want to be case | |
670 | sensitive and some do not: The case insensitive ones merely need to | |
671 | include C<(?i)> at the front of the pattern. For example: | |
19799a22 GS |
672 | |
673 | $pattern = "foobar"; | |
5d458dd8 | 674 | if ( /$pattern/i ) { } |
19799a22 GS |
675 | |
676 | # more flexible: | |
677 | ||
678 | $pattern = "(?i)foobar"; | |
5d458dd8 | 679 | if ( /$pattern/ ) { } |
19799a22 | 680 | |
0b6d1084 | 681 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 GS |
682 | |
683 | ( (?i) blah ) \s+ \1 | |
684 | ||
0d017f4d WL |
685 | will match C<blah> in any case, some spaces, and an exact (I<including the case>!) |
686 | repetition of the previous word, assuming the C</x> modifier, and no C</i> | |
687 | modifier outside this group. | |
19799a22 | 688 | |
cde0cee5 YO |
689 | Note that the C<k> modifier is special in that it can only be enabled, |
690 | not disabled, and that its presence anywhere in a pattern has a global | |
691 | effect. Thus C<(?-k)> and C<(?-k:...)> are meaningless and will warn | |
692 | when executed under C<use warnings>. | |
693 | ||
5a964f20 | 694 | =item C<(?:pattern)> |
d74e8afc | 695 | X<(?:)> |
a0d0e21e | 696 | |
ca9dfc88 IZ |
697 | =item C<(?imsx-imsx:pattern)> |
698 | ||
5a964f20 TC |
699 | This is for clustering, not capturing; it groups subexpressions like |
700 | "()", but doesn't make backreferences as "()" does. So | |
a0d0e21e | 701 | |
5a964f20 | 702 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e LW |
703 | |
704 | is like | |
705 | ||
5a964f20 | 706 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e | 707 | |
19799a22 GS |
708 | but doesn't spit out extra fields. It's also cheaper not to capture |
709 | characters if you don't need to. | |
a0d0e21e | 710 | |
19799a22 | 711 | Any letters between C<?> and C<:> act as flags modifiers as with |
5d458dd8 | 712 | C<(?imsx-imsx)>. For example, |
ca9dfc88 IZ |
713 | |
714 | /(?s-i:more.*than).*million/i | |
715 | ||
14218588 | 716 | is equivalent to the more verbose |
ca9dfc88 IZ |
717 | |
718 | /(?:(?s-i)more.*than).*million/i | |
719 | ||
594d7033 YO |
720 | =item C<(?|pattern)> |
721 | X<(?|)> X<Branch reset> | |
722 | ||
723 | This is the "branch reset" pattern, which has the special property | |
724 | that the capture buffers are numbered from the same starting point | |
693596a8 | 725 | in each alternation branch. It is available starting from perl 5.10. |
4deaaa80 | 726 | |
693596a8 RGS |
727 | Capture buffers are numbered from left to right, but inside this |
728 | construct the numbering is restarted for each branch. | |
4deaaa80 PJ |
729 | |
730 | The numbering within each branch will be as normal, and any buffers | |
731 | following this construct will be numbered as though the construct | |
732 | contained only one branch, that being the one with the most capture | |
733 | buffers in it. | |
734 | ||
735 | This construct will be useful when you want to capture one of a | |
736 | number of alternative matches. | |
737 | ||
738 | Consider the following pattern. The numbers underneath show in | |
739 | which buffer the captured content will be stored. | |
594d7033 YO |
740 | |
741 | ||
742 | # before ---------------branch-reset----------- after | |
743 | / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x | |
744 | # 1 2 2 3 2 3 4 | |
745 | ||
ee9b8eae YO |
746 | =item Look-Around Assertions |
747 | X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround> | |
748 | ||
749 | Look-around assertions are zero width patterns which match a specific | |
750 | pattern without including it in C<$&>. Positive assertions match when | |
751 | their subpattern matches, negative assertions match when their subpattern | |
752 | fails. Look-behind matches text up to the current match position, | |
753 | look-ahead matches text following the current match position. | |
754 | ||
755 | =over 4 | |
756 | ||
5a964f20 | 757 | =item C<(?=pattern)> |
d74e8afc | 758 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> |
a0d0e21e | 759 | |
19799a22 | 760 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e LW |
761 | matches a word followed by a tab, without including the tab in C<$&>. |
762 | ||
5a964f20 | 763 | =item C<(?!pattern)> |
d74e8afc | 764 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> |
a0d0e21e | 765 | |
19799a22 | 766 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e | 767 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 GS |
768 | however that look-ahead and look-behind are NOT the same thing. You cannot |
769 | use this for look-behind. | |
7b8d334a | 770 | |
5a964f20 | 771 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a GS |
772 | will not do what you want. That's because the C<(?!foo)> is just saying that |
773 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
774 | match. You would have to do something like C</(?!foo)...bar/> for that. We | |
775 | say "like" because there's the case of your "bar" not having three characters | |
776 | before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>. | |
777 | Sometimes it's still easier just to say: | |
a0d0e21e | 778 | |
a3cb178b | 779 | if (/bar/ && $` !~ /foo$/) |
a0d0e21e | 780 | |
19799a22 | 781 | For look-behind see below. |
c277df42 | 782 | |
ee9b8eae YO |
783 | =item C<(?<=pattern)> C<\K> |
784 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K> | |
c277df42 | 785 | |
c47ff5f1 | 786 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 GS |
787 | matches a word that follows a tab, without including the tab in C<$&>. |
788 | Works only for fixed-width look-behind. | |
c277df42 | 789 | |
ee9b8eae YO |
790 | There is a special form of this construct, called C<\K>, which causes the |
791 | regex engine to "keep" everything it had matched prior to the C<\K> and | |
792 | not include it in C<$&>. This effectively provides variable length | |
793 | look-behind. The use of C<\K> inside of another look-around assertion | |
794 | is allowed, but the behaviour is currently not well defined. | |
795 | ||
796 | For various reasons C<\K> may be signifigantly more efficient than the | |
797 | equivalent C<< (?<=...) >> construct, and it is especially useful in | |
798 | situations where you want to efficiently remove something following | |
799 | something else in a string. For instance | |
800 | ||
801 | s/(foo)bar/$1/g; | |
802 | ||
803 | can be rewritten as the much more efficient | |
804 | ||
805 | s/foo\Kbar//g; | |
806 | ||
5a964f20 | 807 | =item C<(?<!pattern)> |
d74e8afc | 808 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> |
c277df42 | 809 | |
19799a22 GS |
810 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
811 | matches any occurrence of "foo" that does not follow "bar". Works | |
812 | only for fixed-width look-behind. | |
c277df42 | 813 | |
ee9b8eae YO |
814 | =back |
815 | ||
81714fb9 YO |
816 | =item C<(?'NAME'pattern)> |
817 | ||
818 | =item C<< (?<NAME>pattern) >> | |
819 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> | |
820 | ||
821 | A named capture buffer. Identical in every respect to normal capturing | |
0d017f4d | 822 | parentheses C<()> but for the additional fact that C<%+> may be used after |
81714fb9 YO |
823 | a succesful match to refer to a named buffer. See C<perlvar> for more |
824 | details on the C<%+> hash. | |
825 | ||
826 | If multiple distinct capture buffers have the same name then the | |
827 | $+{NAME} will refer to the leftmost defined buffer in the match. | |
828 | ||
0d017f4d | 829 | The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent. |
81714fb9 YO |
830 | |
831 | B<NOTE:> While the notation of this construct is the same as the similar | |
0d017f4d | 832 | function in .NET regexes, the behavior is not. In Perl the buffers are |
81714fb9 YO |
833 | numbered sequentially regardless of being named or not. Thus in the |
834 | pattern | |
835 | ||
836 | /(x)(?<foo>y)(z)/ | |
837 | ||
838 | $+{foo} will be the same as $2, and $3 will contain 'z' instead of | |
839 | the opposite which is what a .NET regex hacker might expect. | |
840 | ||
1f1031fe YO |
841 | Currently NAME is restricted to simple identifiers only. |
842 | In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or | |
843 | its Unicode extension (see L<utf8>), | |
844 | though it isn't extended by the locale (see L<perllocale>). | |
81714fb9 | 845 | |
1f1031fe | 846 | B<NOTE:> In order to make things easier for programmers with experience |
0d017f4d WL |
847 | with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >> |
848 | may be used instead of C<< (?<NAME>pattern) >>; however this form does not | |
1f1031fe YO |
849 | support the use of single quotes as a delimiter for the name. This is |
850 | only available in Perl 5.10 or later. | |
81714fb9 | 851 | |
1f1031fe YO |
852 | =item C<< \k<NAME> >> |
853 | ||
854 | =item C<< \k'NAME' >> | |
81714fb9 YO |
855 | |
856 | Named backreference. Similar to numeric backreferences, except that | |
857 | the group is designated by name and not number. If multiple groups | |
858 | have the same name then it refers to the leftmost defined group in | |
859 | the current match. | |
860 | ||
0d017f4d | 861 | It is an error to refer to a name not defined by a C<< (?<NAME>) >> |
81714fb9 YO |
862 | earlier in the pattern. |
863 | ||
864 | Both forms are equivalent. | |
865 | ||
1f1031fe | 866 | B<NOTE:> In order to make things easier for programmers with experience |
0d017f4d WL |
867 | with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >> |
868 | may be used instead of C<< \k<NAME> >> in Perl 5.10 or later. | |
1f1031fe | 869 | |
cc6b7395 | 870 | =item C<(?{ code })> |
d74e8afc | 871 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 | 872 | |
19799a22 | 873 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
874 | experimental, and may be changed without notice. Code executed that |
875 | has side effects may not perform identically from version to version | |
876 | due to the effect of future optimisations in the regex engine. | |
c277df42 | 877 | |
cc46d5f2 | 878 | This zero-width assertion evaluates any embedded Perl code. It |
19799a22 GS |
879 | always succeeds, and its C<code> is not interpolated. Currently, |
880 | the rules to determine where the C<code> ends are somewhat convoluted. | |
881 | ||
77ea4f6d JV |
882 | This feature can be used together with the special variable C<$^N> to |
883 | capture the results of submatches in variables without having to keep | |
884 | track of the number of nested parentheses. For example: | |
885 | ||
886 | $_ = "The brown fox jumps over the lazy dog"; | |
887 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
888 | print "color = $color, animal = $animal\n"; | |
889 | ||
754091cb RGS |
890 | Inside the C<(?{...})> block, C<$_> refers to the string the regular |
891 | expression is matching against. You can also use C<pos()> to know what is | |
fa11829f | 892 | the current position of matching within this string. |
754091cb | 893 | |
19799a22 GS |
894 | The C<code> is properly scoped in the following sense: If the assertion |
895 | is backtracked (compare L<"Backtracking">), all changes introduced after | |
896 | C<local>ization are undone, so that | |
b9ac3b5b GS |
897 | |
898 | $_ = 'a' x 8; | |
5d458dd8 | 899 | m< |
b9ac3b5b GS |
900 | (?{ $cnt = 0 }) # Initialize $cnt. |
901 | ( | |
5d458dd8 | 902 | a |
b9ac3b5b GS |
903 | (?{ |
904 | local $cnt = $cnt + 1; # Update $cnt, backtracking-safe. | |
905 | }) | |
5d458dd8 | 906 | )* |
b9ac3b5b GS |
907 | aaaa |
908 | (?{ $res = $cnt }) # On success copy to non-localized | |
909 | # location. | |
910 | >x; | |
911 | ||
0d017f4d | 912 | will set C<$res = 4>. Note that after the match, C<$cnt> returns to the globally |
14218588 | 913 | introduced value, because the scopes that restrict C<local> operators |
b9ac3b5b GS |
914 | are unwound. |
915 | ||
19799a22 GS |
916 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> |
917 | switch. If I<not> used in this way, the result of evaluation of | |
918 | C<code> is put into the special variable C<$^R>. This happens | |
919 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions | |
920 | inside the same regular expression. | |
b9ac3b5b | 921 | |
19799a22 GS |
922 | The assignment to C<$^R> above is properly localized, so the old |
923 | value of C<$^R> is restored if the assertion is backtracked; compare | |
924 | L<"Backtracking">. | |
b9ac3b5b | 925 | |
61528107 SP |
926 | Due to an unfortunate implementation issue, the Perl code contained in these |
927 | blocks is treated as a compile time closure that can have seemingly bizarre | |
6bda09f9 | 928 | consequences when used with lexically scoped variables inside of subroutines |
61528107 SP |
929 | or loops. There are various workarounds for this, including simply using |
930 | global variables instead. If you are using this construct and strange results | |
6bda09f9 YO |
931 | occur then check for the use of lexically scoped variables. |
932 | ||
19799a22 GS |
933 | For reasons of security, this construct is forbidden if the regular |
934 | expression involves run-time interpolation of variables, unless the | |
935 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
936 | variables contain results of C<qr//> operator (see | |
5d458dd8 | 937 | L<perlop/"qr/STRING/imosx">). |
871b0233 | 938 | |
0d017f4d | 939 | This restriction is due to the wide-spread and remarkably convenient |
19799a22 | 940 | custom of using run-time determined strings as patterns. For example: |
871b0233 IZ |
941 | |
942 | $re = <>; | |
943 | chomp $re; | |
944 | $string =~ /$re/; | |
945 | ||
14218588 GS |
946 | Before Perl knew how to execute interpolated code within a pattern, |
947 | this operation was completely safe from a security point of view, | |
948 | although it could raise an exception from an illegal pattern. If | |
949 | you turn on the C<use re 'eval'>, though, it is no longer secure, | |
950 | so you should only do so if you are also using taint checking. | |
951 | Better yet, use the carefully constrained evaluation within a Safe | |
cc46d5f2 | 952 | compartment. See L<perlsec> for details about both these mechanisms. |
871b0233 | 953 | |
0d017f4d | 954 | Because Perl's regex engine is currently not re-entrant, interpolated |
8988a1bb DD |
955 | code may not invoke the regex engine either directly with C<m//> or C<s///>), |
956 | or indirectly with functions such as C<split>. | |
957 | ||
14455d6c | 958 | =item C<(??{ code })> |
d74e8afc ITB |
959 | X<(??{})> |
960 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> | |
0f5d15d6 | 961 | |
19799a22 | 962 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
963 | experimental, and may be changed without notice. Code executed that |
964 | has side effects may not perform identically from version to version | |
965 | due to the effect of future optimisations in the regex engine. | |
0f5d15d6 | 966 | |
19799a22 GS |
967 | This is a "postponed" regular subexpression. The C<code> is evaluated |
968 | at run time, at the moment this subexpression may match. The result | |
969 | of evaluation is considered as a regular expression and matched as | |
61528107 | 970 | if it were inserted instead of this construct. Note that this means |
6bda09f9 YO |
971 | that the contents of capture buffers defined inside an eval'ed pattern |
972 | are not available outside of the pattern, and vice versa, there is no | |
973 | way for the inner pattern to refer to a capture buffer defined outside. | |
974 | Thus, | |
975 | ||
976 | ('a' x 100)=~/(??{'(.)' x 100})/ | |
977 | ||
81714fb9 | 978 | B<will> match, it will B<not> set $1. |
0f5d15d6 | 979 | |
428594d9 | 980 | The C<code> is not interpolated. As before, the rules to determine |
19799a22 GS |
981 | where the C<code> ends are currently somewhat convoluted. |
982 | ||
983 | The following pattern matches a parenthesized group: | |
0f5d15d6 IZ |
984 | |
985 | $re = qr{ | |
986 | \( | |
987 | (?: | |
988 | (?> [^()]+ ) # Non-parens without backtracking | |
989 | | | |
14455d6c | 990 | (??{ $re }) # Group with matching parens |
0f5d15d6 IZ |
991 | )* |
992 | \) | |
993 | }x; | |
994 | ||
6bda09f9 YO |
995 | See also C<(?PARNO)> for a different, more efficient way to accomplish |
996 | the same task. | |
997 | ||
5d458dd8 | 998 | Because perl's regex engine is not currently re-entrant, delayed |
8988a1bb DD |
999 | code may not invoke the regex engine either directly with C<m//> or C<s///>), |
1000 | or indirectly with functions such as C<split>. | |
1001 | ||
5d458dd8 YO |
1002 | Recursing deeper than 50 times without consuming any input string will |
1003 | result in a fatal error. The maximum depth is compiled into perl, so | |
6bda09f9 YO |
1004 | changing it requires a custom build. |
1005 | ||
542fa716 YO |
1006 | =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)> |
1007 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)> | |
6bda09f9 | 1008 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
542fa716 | 1009 | X<regex, relative recursion> |
6bda09f9 | 1010 | |
81714fb9 YO |
1011 | Similar to C<(??{ code })> except it does not involve compiling any code, |
1012 | instead it treats the contents of a capture buffer as an independent | |
61528107 | 1013 | pattern that must match at the current position. Capture buffers |
81714fb9 | 1014 | contained by the pattern will have the value as determined by the |
6bda09f9 YO |
1015 | outermost recursion. |
1016 | ||
894be9b7 YO |
1017 | PARNO is a sequence of digits (not starting with 0) whose value reflects |
1018 | the paren-number of the capture buffer to recurse to. C<(?R)> recurses to | |
1019 | the beginning of the whole pattern. C<(?0)> is an alternate syntax for | |
542fa716 YO |
1020 | C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed |
1021 | to be relative, with negative numbers indicating preceding capture buffers | |
1022 | and positive ones following. Thus C<(?-1)> refers to the most recently | |
1023 | declared buffer, and C<(?+1)> indicates the next buffer to be declared. | |
c74340f9 YO |
1024 | Note that the counting for relative recursion differs from that of |
1025 | relative backreferences, in that with recursion unclosed buffers B<are> | |
1026 | included. | |
6bda09f9 | 1027 | |
81714fb9 | 1028 | The following pattern matches a function foo() which may contain |
f145b7e9 | 1029 | balanced parentheses as the argument. |
6bda09f9 YO |
1030 | |
1031 | $re = qr{ ( # paren group 1 (full function) | |
81714fb9 | 1032 | foo |
6bda09f9 YO |
1033 | ( # paren group 2 (parens) |
1034 | \( | |
1035 | ( # paren group 3 (contents of parens) | |
1036 | (?: | |
1037 | (?> [^()]+ ) # Non-parens without backtracking | |
1038 | | | |
1039 | (?2) # Recurse to start of paren group 2 | |
1040 | )* | |
1041 | ) | |
1042 | \) | |
1043 | ) | |
1044 | ) | |
1045 | }x; | |
1046 | ||
1047 | If the pattern was used as follows | |
1048 | ||
1049 | 'foo(bar(baz)+baz(bop))'=~/$re/ | |
1050 | and print "\$1 = $1\n", | |
1051 | "\$2 = $2\n", | |
1052 | "\$3 = $3\n"; | |
1053 | ||
1054 | the output produced should be the following: | |
1055 | ||
1056 | $1 = foo(bar(baz)+baz(bop)) | |
1057 | $2 = (bar(baz)+baz(bop)) | |
81714fb9 | 1058 | $3 = bar(baz)+baz(bop) |
6bda09f9 | 1059 | |
81714fb9 | 1060 | If there is no corresponding capture buffer defined, then it is a |
61528107 | 1061 | fatal error. Recursing deeper than 50 times without consuming any input |
81714fb9 | 1062 | string will also result in a fatal error. The maximum depth is compiled |
6bda09f9 YO |
1063 | into perl, so changing it requires a custom build. |
1064 | ||
542fa716 YO |
1065 | The following shows how using negative indexing can make it |
1066 | easier to embed recursive patterns inside of a C<qr//> construct | |
1067 | for later use: | |
1068 | ||
1069 | my $parens = qr/(\((?:[^()]++|(?-1))*+\))/; | |
1070 | if (/foo $parens \s+ + \s+ bar $parens/x) { | |
1071 | # do something here... | |
1072 | } | |
1073 | ||
81714fb9 | 1074 | B<Note> that this pattern does not behave the same way as the equivalent |
0d017f4d | 1075 | PCRE or Python construct of the same form. In Perl you can backtrack into |
6bda09f9 | 1076 | a recursed group, in PCRE and Python the recursed into group is treated |
542fa716 YO |
1077 | as atomic. Also, modifiers are resolved at compile time, so constructs |
1078 | like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will | |
1079 | be processed. | |
6bda09f9 | 1080 | |
894be9b7 YO |
1081 | =item C<(?&NAME)> |
1082 | X<(?&NAME)> | |
1083 | ||
0d017f4d WL |
1084 | Recurse to a named subpattern. Identical to C<(?PARNO)> except that the |
1085 | parenthesis to recurse to is determined by name. If multiple parentheses have | |
894be9b7 YO |
1086 | the same name, then it recurses to the leftmost. |
1087 | ||
1088 | It is an error to refer to a name that is not declared somewhere in the | |
1089 | pattern. | |
1090 | ||
1f1031fe YO |
1091 | B<NOTE:> In order to make things easier for programmers with experience |
1092 | with the Python or PCRE regex engines the pattern C<< (?P>NAME) >> | |
0d017f4d | 1093 | may be used instead of C<< (?&NAME) >> in Perl 5.10 or later. |
1f1031fe | 1094 | |
e2e6a0f1 YO |
1095 | =item C<(?(condition)yes-pattern|no-pattern)> |
1096 | X<(?()> | |
286f584a | 1097 | |
e2e6a0f1 | 1098 | =item C<(?(condition)yes-pattern)> |
286f584a | 1099 | |
e2e6a0f1 YO |
1100 | Conditional expression. C<(condition)> should be either an integer in |
1101 | parentheses (which is valid if the corresponding pair of parentheses | |
1102 | matched), a look-ahead/look-behind/evaluate zero-width assertion, a | |
1103 | name in angle brackets or single quotes (which is valid if a buffer | |
1104 | with the given name matched), or the special symbol (R) (true when | |
1105 | evaluated inside of recursion or eval). Additionally the R may be | |
1106 | followed by a number, (which will be true when evaluated when recursing | |
1107 | inside of the appropriate group), or by C<&NAME>, in which case it will | |
1108 | be true only when evaluated during recursion in the named group. | |
1109 | ||
1110 | Here's a summary of the possible predicates: | |
1111 | ||
1112 | =over 4 | |
1113 | ||
1114 | =item (1) (2) ... | |
1115 | ||
1116 | Checks if the numbered capturing buffer has matched something. | |
1117 | ||
1118 | =item (<NAME>) ('NAME') | |
1119 | ||
1120 | Checks if a buffer with the given name has matched something. | |
1121 | ||
1122 | =item (?{ CODE }) | |
1123 | ||
1124 | Treats the code block as the condition. | |
1125 | ||
1126 | =item (R) | |
1127 | ||
1128 | Checks if the expression has been evaluated inside of recursion. | |
1129 | ||
1130 | =item (R1) (R2) ... | |
1131 | ||
1132 | Checks if the expression has been evaluated while executing directly | |
1133 | inside of the n-th capture group. This check is the regex equivalent of | |
1134 | ||
1135 | if ((caller(0))[3] eq 'subname') { ... } | |
1136 | ||
1137 | In other words, it does not check the full recursion stack. | |
1138 | ||
1139 | =item (R&NAME) | |
1140 | ||
1141 | Similar to C<(R1)>, this predicate checks to see if we're executing | |
1142 | directly inside of the leftmost group with a given name (this is the same | |
1143 | logic used by C<(?&NAME)> to disambiguate). It does not check the full | |
1144 | stack, but only the name of the innermost active recursion. | |
1145 | ||
1146 | =item (DEFINE) | |
1147 | ||
1148 | In this case, the yes-pattern is never directly executed, and no | |
1149 | no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient. | |
1150 | See below for details. | |
1151 | ||
1152 | =back | |
1153 | ||
1154 | For example: | |
1155 | ||
1156 | m{ ( \( )? | |
1157 | [^()]+ | |
1158 | (?(1) \) ) | |
1159 | }x | |
1160 | ||
1161 | matches a chunk of non-parentheses, possibly included in parentheses | |
1162 | themselves. | |
1163 | ||
1164 | A special form is the C<(DEFINE)> predicate, which never executes directly | |
1165 | its yes-pattern, and does not allow a no-pattern. This allows to define | |
1166 | subpatterns which will be executed only by using the recursion mechanism. | |
1167 | This way, you can define a set of regular expression rules that can be | |
1168 | bundled into any pattern you choose. | |
1169 | ||
1170 | It is recommended that for this usage you put the DEFINE block at the | |
1171 | end of the pattern, and that you name any subpatterns defined within it. | |
1172 | ||
1173 | Also, it's worth noting that patterns defined this way probably will | |
1174 | not be as efficient, as the optimiser is not very clever about | |
1175 | handling them. | |
1176 | ||
1177 | An example of how this might be used is as follows: | |
1178 | ||
2bf803e2 | 1179 | /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT)) |
e2e6a0f1 | 1180 | (?(DEFINE) |
2bf803e2 YO |
1181 | (?<NAME_PAT>....) |
1182 | (?<ADRESS_PAT>....) | |
e2e6a0f1 YO |
1183 | )/x |
1184 | ||
1185 | Note that capture buffers matched inside of recursion are not accessible | |
0d017f4d | 1186 | after the recursion returns, so the extra layer of capturing buffers is |
e2e6a0f1 YO |
1187 | necessary. Thus C<$+{NAME_PAT}> would not be defined even though |
1188 | C<$+{NAME}> would be. | |
286f584a | 1189 | |
c47ff5f1 | 1190 | =item C<< (?>pattern) >> |
6bda09f9 | 1191 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 | 1192 | |
19799a22 GS |
1193 | An "independent" subexpression, one which matches the substring |
1194 | that a I<standalone> C<pattern> would match if anchored at the given | |
9da458fc | 1195 | position, and it matches I<nothing other than this substring>. This |
19799a22 GS |
1196 | construct is useful for optimizations of what would otherwise be |
1197 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). | |
9da458fc IZ |
1198 | It may also be useful in places where the "grab all you can, and do not |
1199 | give anything back" semantic is desirable. | |
19799a22 | 1200 | |
c47ff5f1 | 1201 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 GS |
1202 | (anchored at the beginning of string, as above) will match I<all> |
1203 | characters C<a> at the beginning of string, leaving no C<a> for | |
1204 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, | |
1205 | since the match of the subgroup C<a*> is influenced by the following | |
1206 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside | |
1207 | C<a*ab> will match fewer characters than a standalone C<a*>, since | |
1208 | this makes the tail match. | |
1209 | ||
c47ff5f1 | 1210 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
19799a22 GS |
1211 | C<(?=(pattern))\1>. This matches the same substring as a standalone |
1212 | C<a+>, and the following C<\1> eats the matched string; it therefore | |
c47ff5f1 | 1213 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 GS |
1214 | (The difference between these two constructs is that the second one |
1215 | uses a capturing group, thus shifting ordinals of backreferences | |
1216 | in the rest of a regular expression.) | |
1217 | ||
1218 | Consider this pattern: | |
c277df42 | 1219 | |
871b0233 | 1220 | m{ \( |
e2e6a0f1 YO |
1221 | ( |
1222 | [^()]+ # x+ | |
1223 | | | |
871b0233 IZ |
1224 | \( [^()]* \) |
1225 | )+ | |
e2e6a0f1 | 1226 | \) |
871b0233 | 1227 | }x |
5a964f20 | 1228 | |
19799a22 GS |
1229 | That will efficiently match a nonempty group with matching parentheses |
1230 | two levels deep or less. However, if there is no such group, it | |
1231 | will take virtually forever on a long string. That's because there | |
1232 | are so many different ways to split a long string into several | |
1233 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
1234 | to a subpattern of the above pattern. Consider how the pattern | |
1235 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
1236 | seconds, but that each extra letter doubles this time. This | |
1237 | exponential performance will make it appear that your program has | |
14218588 | 1238 | hung. However, a tiny change to this pattern |
5a964f20 | 1239 | |
e2e6a0f1 YO |
1240 | m{ \( |
1241 | ( | |
1242 | (?> [^()]+ ) # change x+ above to (?> x+ ) | |
1243 | | | |
871b0233 IZ |
1244 | \( [^()]* \) |
1245 | )+ | |
e2e6a0f1 | 1246 | \) |
871b0233 | 1247 | }x |
c277df42 | 1248 | |
c47ff5f1 | 1249 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 TC |
1250 | this yourself would be a productive exercise), but finishes in a fourth |
1251 | the time when used on a similar string with 1000000 C<a>s. Be aware, | |
1252 | however, that this pattern currently triggers a warning message under | |
9f1b1f2d | 1253 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b | 1254 | C<"matches null string many times in regex">. |
c277df42 | 1255 | |
c47ff5f1 | 1256 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 | 1257 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 IZ |
1258 | This was only 4 times slower on a string with 1000000 C<a>s. |
1259 | ||
9da458fc IZ |
1260 | The "grab all you can, and do not give anything back" semantic is desirable |
1261 | in many situations where on the first sight a simple C<()*> looks like | |
1262 | the correct solution. Suppose we parse text with comments being delimited | |
1263 | by C<#> followed by some optional (horizontal) whitespace. Contrary to | |
4375e838 | 1264 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc IZ |
1265 | the comment delimiter, because it may "give up" some whitespace if |
1266 | the remainder of the pattern can be made to match that way. The correct | |
1267 | answer is either one of these: | |
1268 | ||
1269 | (?>#[ \t]*) | |
1270 | #[ \t]*(?![ \t]) | |
1271 | ||
1272 | For example, to grab non-empty comments into $1, one should use either | |
1273 | one of these: | |
1274 | ||
1275 | / (?> \# [ \t]* ) ( .+ ) /x; | |
1276 | / \# [ \t]* ( [^ \t] .* ) /x; | |
1277 | ||
1278 | Which one you pick depends on which of these expressions better reflects | |
1279 | the above specification of comments. | |
1280 | ||
6bda09f9 YO |
1281 | In some literature this construct is called "atomic matching" or |
1282 | "possessive matching". | |
1283 | ||
b9b4dddf YO |
1284 | Possessive quantifiers are equivalent to putting the item they are applied |
1285 | to inside of one of these constructs. The following equivalences apply: | |
1286 | ||
1287 | Quantifier Form Bracketing Form | |
1288 | --------------- --------------- | |
1289 | PAT*+ (?>PAT*) | |
1290 | PAT++ (?>PAT+) | |
1291 | PAT?+ (?>PAT?) | |
1292 | PAT{min,max}+ (?>PAT{min,max}) | |
1293 | ||
e2e6a0f1 YO |
1294 | =back |
1295 | ||
1296 | =head2 Special Backtracking Control Verbs | |
1297 | ||
1298 | B<WARNING:> These patterns are experimental and subject to change or | |
0d017f4d | 1299 | removal in a future version of Perl. Their usage in production code should |
e2e6a0f1 YO |
1300 | be noted to avoid problems during upgrades. |
1301 | ||
1302 | These special patterns are generally of the form C<(*VERB:ARG)>. Unless | |
1303 | otherwise stated the ARG argument is optional; in some cases, it is | |
1304 | forbidden. | |
1305 | ||
1306 | Any pattern containing a special backtracking verb that allows an argument | |
1307 | has the special behaviour that when executed it sets the current packages' | |
5d458dd8 YO |
1308 | C<$REGERROR> and C<$REGMARK> variables. When doing so the following |
1309 | rules apply: | |
e2e6a0f1 | 1310 | |
5d458dd8 YO |
1311 | On failure, the C<$REGERROR> variable will be set to the ARG value of the |
1312 | verb pattern, if the verb was involved in the failure of the match. If the | |
1313 | ARG part of the pattern was omitted, then C<$REGERROR> will be set to the | |
1314 | name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was | |
1315 | none. Also, the C<$REGMARK> variable will be set to FALSE. | |
e2e6a0f1 | 1316 | |
5d458dd8 YO |
1317 | On a successful match, the C<$REGERROR> variable will be set to FALSE, and |
1318 | the C<$REGMARK> variable will be set to the name of the last | |
1319 | C<(*MARK:NAME)> pattern executed. See the explanation for the | |
1320 | C<(*MARK:NAME)> verb below for more details. | |
e2e6a0f1 | 1321 | |
5d458dd8 YO |
1322 | B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1> |
1323 | and most other regex related variables. They are not local to a scope, nor | |
1324 | readonly, but instead are volatile package variables similar to C<$AUTOLOAD>. | |
1325 | Use C<local> to localize changes to them to a specific scope if necessary. | |
e2e6a0f1 YO |
1326 | |
1327 | If a pattern does not contain a special backtracking verb that allows an | |
5d458dd8 | 1328 | argument, then C<$REGERROR> and C<$REGMARK> are not touched at all. |
e2e6a0f1 YO |
1329 | |
1330 | =over 4 | |
1331 | ||
1332 | =item Verbs that take an argument | |
1333 | ||
1334 | =over 4 | |
1335 | ||
5d458dd8 | 1336 | =item C<(*PRUNE)> C<(*PRUNE:NAME)> |
ee9b8eae | 1337 | X<(*PRUNE)> X<(*PRUNE:NAME)> X<\v> |
54612592 | 1338 | |
5d458dd8 YO |
1339 | This zero-width pattern prunes the backtracking tree at the current point |
1340 | when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>, | |
1341 | where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached, | |
1342 | A may backtrack as necessary to match. Once it is reached, matching | |
1343 | continues in B, which may also backtrack as necessary; however, should B | |
1344 | not match, then no further backtracking will take place, and the pattern | |
1345 | will fail outright at the current starting position. | |
54612592 | 1346 | |
0d017f4d | 1347 | As a shortcut, C<\v> is exactly equivalent to C<(*PRUNE)>. |
ee9b8eae | 1348 | |
54612592 YO |
1349 | The following example counts all the possible matching strings in a |
1350 | pattern (without actually matching any of them). | |
1351 | ||
e2e6a0f1 | 1352 | 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1353 | print "Count=$count\n"; |
1354 | ||
1355 | which produces: | |
1356 | ||
1357 | aaab | |
1358 | aaa | |
1359 | aa | |
1360 | a | |
1361 | aab | |
1362 | aa | |
1363 | a | |
1364 | ab | |
1365 | a | |
1366 | Count=9 | |
1367 | ||
5d458dd8 | 1368 | If we add a C<(*PRUNE)> before the count like the following |
54612592 | 1369 | |
5d458dd8 | 1370 | 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1371 | print "Count=$count\n"; |
1372 | ||
1373 | we prevent backtracking and find the count of the longest matching | |
1374 | at each matching startpoint like so: | |
1375 | ||
1376 | aaab | |
1377 | aab | |
1378 | ab | |
1379 | Count=3 | |
1380 | ||
5d458dd8 | 1381 | Any number of C<(*PRUNE)> assertions may be used in a pattern. |
54612592 | 1382 | |
5d458dd8 YO |
1383 | See also C<< (?>pattern) >> and possessive quantifiers for other ways to |
1384 | control backtracking. In some cases, the use of C<(*PRUNE)> can be | |
1385 | replaced with a C<< (?>pattern) >> with no functional difference; however, | |
1386 | C<(*PRUNE)> can be used to handle cases that cannot be expressed using a | |
1387 | C<< (?>pattern) >> alone. | |
54612592 | 1388 | |
e2e6a0f1 | 1389 | |
5d458dd8 YO |
1390 | =item C<(*SKIP)> C<(*SKIP:NAME)> |
1391 | X<(*SKIP)> | |
e2e6a0f1 | 1392 | |
5d458dd8 | 1393 | This zero-width pattern is similar to C<(*PRUNE)>, except that on |
e2e6a0f1 | 1394 | failure it also signifies that whatever text that was matched leading up |
5d458dd8 YO |
1395 | to the C<(*SKIP)> pattern being executed cannot be part of I<any> match |
1396 | of this pattern. This effectively means that the regex engine "skips" forward | |
1397 | to this position on failure and tries to match again, (assuming that | |
1398 | there is sufficient room to match). | |
1399 | ||
0d017f4d | 1400 | As a shortcut C<\V> is exactly equivalent to C<(*SKIP)>. |
ee9b8eae | 1401 | |
5d458dd8 YO |
1402 | The name of the C<(*SKIP:NAME)> pattern has special significance. If a |
1403 | C<(*MARK:NAME)> was encountered while matching, then it is that position | |
1404 | which is used as the "skip point". If no C<(*MARK)> of that name was | |
1405 | encountered, then the C<(*SKIP)> operator has no effect. When used | |
1406 | without a name the "skip point" is where the match point was when | |
1407 | executing the (*SKIP) pattern. | |
1408 | ||
1409 | Compare the following to the examples in C<(*PRUNE)>, note the string | |
24b23f37 YO |
1410 | is twice as long: |
1411 | ||
5d458dd8 | 1412 | 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 YO |
1413 | print "Count=$count\n"; |
1414 | ||
1415 | outputs | |
1416 | ||
1417 | aaab | |
1418 | aaab | |
1419 | Count=2 | |
1420 | ||
5d458dd8 | 1421 | Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)> |
e2e6a0f1 | 1422 | executed, the next startpoint will be where the cursor was when the |
5d458dd8 YO |
1423 | C<(*SKIP)> was executed. |
1424 | ||
5d458dd8 YO |
1425 | =item C<(*MARK:NAME)> C<(*:NAME)> |
1426 | X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)> | |
1427 | ||
1428 | This zero-width pattern can be used to mark the point reached in a string | |
1429 | when a certain part of the pattern has been successfully matched. This | |
1430 | mark may be given a name. A later C<(*SKIP)> pattern will then skip | |
1431 | forward to that point if backtracked into on failure. Any number of | |
1432 | C<(*MARK)> patterns are allowed, and the NAME portion is optional and may | |
1433 | be duplicated. | |
1434 | ||
1435 | In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)> | |
1436 | can be used to "label" a pattern branch, so that after matching, the | |
1437 | program can determine which branches of the pattern were involved in the | |
1438 | match. | |
1439 | ||
1440 | When a match is successful, the C<$REGMARK> variable will be set to the | |
1441 | name of the most recently executed C<(*MARK:NAME)> that was involved | |
1442 | in the match. | |
1443 | ||
1444 | This can be used to determine which branch of a pattern was matched | |
1445 | without using a seperate capture buffer for each branch, which in turn | |
1446 | can result in a performance improvement, as perl cannot optimize | |
1447 | C</(?:(x)|(y)|(z))/> as efficiently as something like | |
1448 | C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>. | |
1449 | ||
1450 | When a match has failed, and unless another verb has been involved in | |
1451 | failing the match and has provided its own name to use, the C<$REGERROR> | |
1452 | variable will be set to the name of the most recently executed | |
1453 | C<(*MARK:NAME)>. | |
1454 | ||
1455 | See C<(*SKIP)> for more details. | |
1456 | ||
b62d2d15 YO |
1457 | As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>. |
1458 | ||
5d458dd8 YO |
1459 | =item C<(*THEN)> C<(*THEN:NAME)> |
1460 | ||
1461 | This is similar to the "cut group" operator C<::> from Perl6. Like | |
1462 | C<(*PRUNE)>, this verb always matches, and when backtracked into on | |
1463 | failure, it causes the regex engine to try the next alternation in the | |
1464 | innermost enclosing group (capturing or otherwise). | |
1465 | ||
1466 | Its name comes from the observation that this operation combined with the | |
1467 | alternation operator (C<|>) can be used to create what is essentially a | |
1468 | pattern-based if/then/else block: | |
1469 | ||
1470 | ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) | |
1471 | ||
1472 | Note that if this operator is used and NOT inside of an alternation then | |
1473 | it acts exactly like the C<(*PRUNE)> operator. | |
1474 | ||
1475 | / A (*PRUNE) B / | |
1476 | ||
1477 | is the same as | |
1478 | ||
1479 | / A (*THEN) B / | |
1480 | ||
1481 | but | |
1482 | ||
1483 | / ( A (*THEN) B | C (*THEN) D ) / | |
1484 | ||
1485 | is not the same as | |
1486 | ||
1487 | / ( A (*PRUNE) B | C (*PRUNE) D ) / | |
1488 | ||
1489 | as after matching the A but failing on the B the C<(*THEN)> verb will | |
1490 | backtrack and try C; but the C<(*PRUNE)> verb will simply fail. | |
24b23f37 | 1491 | |
e2e6a0f1 YO |
1492 | =item C<(*COMMIT)> |
1493 | X<(*COMMIT)> | |
24b23f37 | 1494 | |
5d458dd8 YO |
1495 | This is the Perl6 "commit pattern" C<< <commit> >> or C<:::>. It's a |
1496 | zero-width pattern similar to C<(*SKIP)>, except that when backtracked | |
1497 | into on failure it causes the match to fail outright. No further attempts | |
1498 | to find a valid match by advancing the start pointer will occur again. | |
1499 | For example, | |
24b23f37 | 1500 | |
e2e6a0f1 | 1501 | 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 YO |
1502 | print "Count=$count\n"; |
1503 | ||
1504 | outputs | |
1505 | ||
1506 | aaab | |
1507 | Count=1 | |
1508 | ||
e2e6a0f1 YO |
1509 | In other words, once the C<(*COMMIT)> has been entered, and if the pattern |
1510 | does not match, the regex engine will not try any further matching on the | |
1511 | rest of the string. | |
c277df42 | 1512 | |
e2e6a0f1 | 1513 | =back |
9af228c6 | 1514 | |
e2e6a0f1 | 1515 | =item Verbs without an argument |
9af228c6 YO |
1516 | |
1517 | =over 4 | |
1518 | ||
e2e6a0f1 YO |
1519 | =item C<(*FAIL)> C<(*F)> |
1520 | X<(*FAIL)> X<(*F)> | |
9af228c6 | 1521 | |
e2e6a0f1 YO |
1522 | This pattern matches nothing and always fails. It can be used to force the |
1523 | engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In | |
1524 | fact, C<(?!)> gets optimised into C<(*FAIL)> internally. | |
9af228c6 | 1525 | |
e2e6a0f1 | 1526 | It is probably useful only when combined with C<(?{})> or C<(??{})>. |
9af228c6 | 1527 | |
e2e6a0f1 YO |
1528 | =item C<(*ACCEPT)> |
1529 | X<(*ACCEPT)> | |
9af228c6 | 1530 | |
e2e6a0f1 YO |
1531 | B<WARNING:> This feature is highly experimental. It is not recommended |
1532 | for production code. | |
9af228c6 | 1533 | |
e2e6a0f1 YO |
1534 | This pattern matches nothing and causes the end of successful matching at |
1535 | the point at which the C<(*ACCEPT)> pattern was encountered, regardless of | |
1536 | whether there is actually more to match in the string. When inside of a | |
0d017f4d | 1537 | nested pattern, such as recursion, or in a subpattern dynamically generated |
e2e6a0f1 | 1538 | via C<(??{})>, only the innermost pattern is ended immediately. |
9af228c6 | 1539 | |
e2e6a0f1 YO |
1540 | If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are |
1541 | marked as ended at the point at which the C<(*ACCEPT)> was encountered. | |
1542 | For instance: | |
9af228c6 | 1543 | |
e2e6a0f1 | 1544 | 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x; |
9af228c6 | 1545 | |
e2e6a0f1 | 1546 | will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not |
0d017f4d | 1547 | be set. If another branch in the inner parentheses were matched, such as in the |
e2e6a0f1 | 1548 | string 'ACDE', then the C<D> and C<E> would have to be matched as well. |
9af228c6 YO |
1549 | |
1550 | =back | |
c277df42 | 1551 | |
a0d0e21e LW |
1552 | =back |
1553 | ||
c07a80fd | 1554 | =head2 Backtracking |
d74e8afc | 1555 | X<backtrack> X<backtracking> |
c07a80fd | 1556 | |
35a734be IZ |
1557 | NOTE: This section presents an abstract approximation of regular |
1558 | expression behavior. For a more rigorous (and complicated) view of | |
1559 | the rules involved in selecting a match among possible alternatives, | |
0d017f4d | 1560 | see L<Combining RE Pieces>. |
35a734be | 1561 | |
c277df42 | 1562 | A fundamental feature of regular expression matching involves the |
5a964f20 | 1563 | notion called I<backtracking>, which is currently used (when needed) |
0d017f4d | 1564 | by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc IZ |
1565 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
1566 | internally, but the general principle outlined here is valid. | |
c07a80fd | 1567 | |
1568 | For a regular expression to match, the I<entire> regular expression must | |
1569 | match, not just part of it. So if the beginning of a pattern containing a | |
1570 | quantifier succeeds in a way that causes later parts in the pattern to | |
1571 | fail, the matching engine backs up and recalculates the beginning | |
1572 | part--that's why it's called backtracking. | |
1573 | ||
1574 | Here is an example of backtracking: Let's say you want to find the | |
1575 | word following "foo" in the string "Food is on the foo table.": | |
1576 | ||
1577 | $_ = "Food is on the foo table."; | |
1578 | if ( /\b(foo)\s+(\w+)/i ) { | |
1579 | print "$2 follows $1.\n"; | |
1580 | } | |
1581 | ||
1582 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
1583 | finds a possible match right at the beginning of the string, and loads up | |
1584 | $1 with "Foo". However, as soon as the matching engine sees that there's | |
1585 | no whitespace following the "Foo" that it had saved in $1, it realizes its | |
68dc0745 | 1586 | mistake and starts over again one character after where it had the |
c07a80fd | 1587 | tentative match. This time it goes all the way until the next occurrence |
1588 | of "foo". The complete regular expression matches this time, and you get | |
1589 | the expected output of "table follows foo." | |
1590 | ||
1591 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
1592 | everything between "foo" and "bar". Initially, you write something | |
1593 | like this: | |
1594 | ||
1595 | $_ = "The food is under the bar in the barn."; | |
1596 | if ( /foo(.*)bar/ ) { | |
1597 | print "got <$1>\n"; | |
1598 | } | |
1599 | ||
1600 | Which perhaps unexpectedly yields: | |
1601 | ||
1602 | got <d is under the bar in the > | |
1603 | ||
1604 | That's because C<.*> was greedy, so you get everything between the | |
14218588 | 1605 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd | 1606 | to use minimal matching to make sure you get the text between a "foo" |
1607 | and the first "bar" thereafter. | |
1608 | ||
1609 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
1610 | got <d is under the > | |
1611 | ||
0d017f4d | 1612 | Here's another example. Let's say you'd like to match a number at the end |
b6e13d97 | 1613 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd | 1614 | So you write this: |
1615 | ||
1616 | $_ = "I have 2 numbers: 53147"; | |
1617 | if ( /(.*)(\d*)/ ) { # Wrong! | |
1618 | print "Beginning is <$1>, number is <$2>.\n"; | |
1619 | } | |
1620 | ||
1621 | That won't work at all, because C<.*> was greedy and gobbled up the | |
1622 | whole string. As C<\d*> can match on an empty string the complete | |
1623 | regular expression matched successfully. | |
1624 | ||
8e1088bc | 1625 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd | 1626 | |
1627 | Here are some variants, most of which don't work: | |
1628 | ||
1629 | $_ = "I have 2 numbers: 53147"; | |
1630 | @pats = qw{ | |
1631 | (.*)(\d*) | |
1632 | (.*)(\d+) | |
1633 | (.*?)(\d*) | |
1634 | (.*?)(\d+) | |
1635 | (.*)(\d+)$ | |
1636 | (.*?)(\d+)$ | |
1637 | (.*)\b(\d+)$ | |
1638 | (.*\D)(\d+)$ | |
1639 | }; | |
1640 | ||
1641 | for $pat (@pats) { | |
1642 | printf "%-12s ", $pat; | |
1643 | if ( /$pat/ ) { | |
1644 | print "<$1> <$2>\n"; | |
1645 | } else { | |
1646 | print "FAIL\n"; | |
1647 | } | |
1648 | } | |
1649 | ||
1650 | That will print out: | |
1651 | ||
1652 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
1653 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
1654 | (.*?)(\d*) <> <> | |
1655 | (.*?)(\d+) <I have > <2> | |
1656 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
1657 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
1658 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
1659 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
1660 | ||
1661 | As you see, this can be a bit tricky. It's important to realize that a | |
1662 | regular expression is merely a set of assertions that gives a definition | |
1663 | of success. There may be 0, 1, or several different ways that the | |
1664 | definition might succeed against a particular string. And if there are | |
5a964f20 TC |
1665 | multiple ways it might succeed, you need to understand backtracking to |
1666 | know which variety of success you will achieve. | |
c07a80fd | 1667 | |
19799a22 | 1668 | When using look-ahead assertions and negations, this can all get even |
8b19b778 | 1669 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd | 1670 | followed by "123". You might try to write that as |
1671 | ||
871b0233 IZ |
1672 | $_ = "ABC123"; |
1673 | if ( /^\D*(?!123)/ ) { # Wrong! | |
1674 | print "Yup, no 123 in $_\n"; | |
1675 | } | |
c07a80fd | 1676 | |
1677 | But that isn't going to match; at least, not the way you're hoping. It | |
1678 | claims that there is no 123 in the string. Here's a clearer picture of | |
9b9391b2 | 1679 | why that pattern matches, contrary to popular expectations: |
c07a80fd | 1680 | |
4358a253 SS |
1681 | $x = 'ABC123'; |
1682 | $y = 'ABC445'; | |
c07a80fd | 1683 | |
4358a253 SS |
1684 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
1685 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; | |
c07a80fd | 1686 | |
4358a253 SS |
1687 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; |
1688 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; | |
c07a80fd | 1689 | |
1690 | This prints | |
1691 | ||
1692 | 2: got ABC | |
1693 | 3: got AB | |
1694 | 4: got ABC | |
1695 | ||
5f05dabc | 1696 | You might have expected test 3 to fail because it seems to a more |
c07a80fd | 1697 | general purpose version of test 1. The important difference between |
1698 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
1699 | backtracking, whereas test 1 will not. What's happening is | |
1700 | that you've asked "Is it true that at the start of $x, following 0 or more | |
5f05dabc | 1701 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd | 1702 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 | 1703 | fail. |
14218588 | 1704 | |
c07a80fd | 1705 | The search engine will initially match C<\D*> with "ABC". Then it will |
14218588 | 1706 | try to match C<(?!123> with "123", which fails. But because |
c07a80fd | 1707 | a quantifier (C<\D*>) has been used in the regular expression, the |
1708 | search engine can backtrack and retry the match differently | |
54310121 | 1709 | in the hope of matching the complete regular expression. |
c07a80fd | 1710 | |
5a964f20 TC |
1711 | The pattern really, I<really> wants to succeed, so it uses the |
1712 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
c07a80fd | 1713 | time. Now there's indeed something following "AB" that is not |
14218588 | 1714 | "123". It's "C123", which suffices. |
c07a80fd | 1715 | |
14218588 GS |
1716 | We can deal with this by using both an assertion and a negation. |
1717 | We'll say that the first part in $1 must be followed both by a digit | |
1718 | and by something that's not "123". Remember that the look-aheads | |
1719 | are zero-width expressions--they only look, but don't consume any | |
1720 | of the string in their match. So rewriting this way produces what | |
c07a80fd | 1721 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
1722 | ||
4358a253 SS |
1723 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; |
1724 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; | |
c07a80fd | 1725 | |
1726 | 6: got ABC | |
1727 | ||
5a964f20 | 1728 | In other words, the two zero-width assertions next to each other work as though |
19799a22 | 1729 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd | 1730 | matches only if you're at the beginning of the line AND the end of the |
1731 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
1732 | regular expressions always means AND, except when you write an explicit OR | |
1733 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
1734 | although the attempted matches are made at different positions because "a" | |
1735 | is not a zero-width assertion, but a one-width assertion. | |
1736 | ||
0d017f4d | 1737 | B<WARNING>: Particularly complicated regular expressions can take |
14218588 | 1738 | exponential time to solve because of the immense number of possible |
0d017f4d | 1739 | ways they can use backtracking to try for a match. For example, without |
9da458fc IZ |
1740 | internal optimizations done by the regular expression engine, this will |
1741 | take a painfully long time to run: | |
c07a80fd | 1742 | |
e1901655 IZ |
1743 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
1744 | ||
1745 | And if you used C<*>'s in the internal groups instead of limiting them | |
1746 | to 0 through 5 matches, then it would take forever--or until you ran | |
1747 | out of stack space. Moreover, these internal optimizations are not | |
1748 | always applicable. For example, if you put C<{0,5}> instead of C<*> | |
1749 | on the external group, no current optimization is applicable, and the | |
1750 | match takes a long time to finish. | |
c07a80fd | 1751 | |
9da458fc IZ |
1752 | A powerful tool for optimizing such beasts is what is known as an |
1753 | "independent group", | |
c47ff5f1 | 1754 | which does not backtrack (see L<C<< (?>pattern) >>>). Note also that |
9da458fc | 1755 | zero-length look-ahead/look-behind assertions will not backtrack to make |
5d458dd8 | 1756 | the tail match, since they are in "logical" context: only |
14218588 | 1757 | whether they match is considered relevant. For an example |
9da458fc | 1758 | where side-effects of look-ahead I<might> have influenced the |
c47ff5f1 | 1759 | following match, see L<C<< (?>pattern) >>>. |
c277df42 | 1760 | |
a0d0e21e | 1761 | =head2 Version 8 Regular Expressions |
d74e8afc | 1762 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> |
a0d0e21e | 1763 | |
5a964f20 | 1764 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e LW |
1765 | routines, here are the pattern-matching rules not described above. |
1766 | ||
54310121 | 1767 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e | 1768 | with a special meaning described here or above. You can cause |
5a964f20 | 1769 | characters that normally function as metacharacters to be interpreted |
5f05dabc | 1770 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
0d017f4d WL |
1771 | character; "\\" matches a "\"). This escape mechanism is also required |
1772 | for the character used as the pattern delimiter. | |
1773 | ||
1774 | A series of characters matches that series of characters in the target | |
1775 | string, so the pattern C<blurfl> would match "blurfl" in the target | |
1776 | string. | |
a0d0e21e LW |
1777 | |
1778 | You can specify a character class, by enclosing a list of characters | |
5d458dd8 | 1779 | in C<[]>, which will match any character from the list. If the |
a0d0e21e | 1780 | first character after the "[" is "^", the class matches any character not |
14218588 | 1781 | in the list. Within a list, the "-" character specifies a |
5a964f20 | 1782 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 GS |
1783 | inclusive. If you want either "-" or "]" itself to be a member of a |
1784 | class, put it at the start of the list (possibly after a "^"), or | |
1785 | escape it with a backslash. "-" is also taken literally when it is | |
1786 | at the end of the list, just before the closing "]". (The | |
84850974 DD |
1787 | following all specify the same class of three characters: C<[-az]>, |
1788 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which | |
5d458dd8 YO |
1789 | specifies a class containing twenty-six characters, even on EBCDIC-based |
1790 | character sets.) Also, if you try to use the character | |
1791 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of | |
1792 | a range, the "-" is understood literally. | |
a0d0e21e | 1793 | |
8ada0baa JH |
1794 | Note also that the whole range idea is rather unportable between |
1795 | character sets--and even within character sets they may cause results | |
1796 | you probably didn't expect. A sound principle is to use only ranges | |
0d017f4d | 1797 | that begin from and end at either alphabetics of equal case ([a-e], |
8ada0baa JH |
1798 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, |
1799 | spell out the character sets in full. | |
1800 | ||
54310121 | 1801 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e LW |
1802 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
1803 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string | |
5d458dd8 YO |
1804 | of octal digits, matches the character whose coded character set value |
1805 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, | |
1806 | matches the character whose numeric value is I<nn>. The expression \cI<x> | |
1807 | matches the character control-I<x>. Finally, the "." metacharacter | |
fb55449c | 1808 | matches any character except "\n" (unless you use C</s>). |
a0d0e21e LW |
1809 | |
1810 | You can specify a series of alternatives for a pattern using "|" to | |
1811 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", | |
5a964f20 | 1812 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e LW |
1813 | first alternative includes everything from the last pattern delimiter |
1814 | ("(", "[", or the beginning of the pattern) up to the first "|", and | |
1815 | the last alternative contains everything from the last "|" to the next | |
14218588 GS |
1816 | pattern delimiter. That's why it's common practice to include |
1817 | alternatives in parentheses: to minimize confusion about where they | |
a3cb178b GS |
1818 | start and end. |
1819 | ||
5a964f20 | 1820 | Alternatives are tried from left to right, so the first |
a3cb178b GS |
1821 | alternative found for which the entire expression matches, is the one that |
1822 | is chosen. This means that alternatives are not necessarily greedy. For | |
628afcb5 | 1823 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b GS |
1824 | part will match, as that is the first alternative tried, and it successfully |
1825 | matches the target string. (This might not seem important, but it is | |
1826 | important when you are capturing matched text using parentheses.) | |
1827 | ||
5a964f20 | 1828 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b | 1829 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e | 1830 | |
14218588 GS |
1831 | Within a pattern, you may designate subpatterns for later reference |
1832 | by enclosing them in parentheses, and you may refer back to the | |
1833 | I<n>th subpattern later in the pattern using the metacharacter | |
1834 | \I<n>. Subpatterns are numbered based on the left to right order | |
1835 | of their opening parenthesis. A backreference matches whatever | |
1836 | actually matched the subpattern in the string being examined, not | |
1837 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will | |
1838 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern | |
1839 | 1 matched "0x", even though the rule C<0|0x> could potentially match | |
1840 | the leading 0 in the second number. | |
cb1a09d0 | 1841 | |
0d017f4d | 1842 | =head2 Warning on \1 Instead of $1 |
cb1a09d0 | 1843 | |
5a964f20 | 1844 | Some people get too used to writing things like: |
cb1a09d0 AD |
1845 | |
1846 | $pattern =~ s/(\W)/\\\1/g; | |
1847 | ||
1848 | This is grandfathered for the RHS of a substitute to avoid shocking the | |
1849 | B<sed> addicts, but it's a dirty habit to get into. That's because in | |
d1be9408 | 1850 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 AD |
1851 | the usual double-quoted string means a control-A. The customary Unix |
1852 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
1853 | of doing that, you get yourself into trouble if you then add an C</e> | |
1854 | modifier. | |
1855 | ||
5a964f20 | 1856 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 AD |
1857 | |
1858 | Or if you try to do | |
1859 | ||
1860 | s/(\d+)/\1000/; | |
1861 | ||
1862 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
14218588 | 1863 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 AD |
1864 | with the operation of matching a backreference. Certainly they mean two |
1865 | different things on the I<left> side of the C<s///>. | |
9fa51da4 | 1866 | |
0d017f4d | 1867 | =head2 Repeated Patterns Matching a Zero-length Substring |
c84d73f1 | 1868 | |
19799a22 | 1869 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 IZ |
1870 | |
1871 | Regular expressions provide a terse and powerful programming language. As | |
1872 | with most other power tools, power comes together with the ability | |
1873 | to wreak havoc. | |
1874 | ||
1875 | A common abuse of this power stems from the ability to make infinite | |
628afcb5 | 1876 | loops using regular expressions, with something as innocuous as: |
c84d73f1 IZ |
1877 | |
1878 | 'foo' =~ m{ ( o? )* }x; | |
1879 | ||
0d017f4d | 1880 | The C<o?> matches at the beginning of C<'foo'>, and since the position |
c84d73f1 | 1881 | in the string is not moved by the match, C<o?> would match again and again |
14218588 | 1882 | because of the C<*> modifier. Another common way to create a similar cycle |
c84d73f1 IZ |
1883 | is with the looping modifier C<//g>: |
1884 | ||
1885 | @matches = ( 'foo' =~ m{ o? }xg ); | |
1886 | ||
1887 | or | |
1888 | ||
1889 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
1890 | ||
1891 | or the loop implied by split(). | |
1892 | ||
1893 | However, long experience has shown that many programming tasks may | |
14218588 GS |
1894 | be significantly simplified by using repeated subexpressions that |
1895 | may match zero-length substrings. Here's a simple example being: | |
c84d73f1 IZ |
1896 | |
1897 | @chars = split //, $string; # // is not magic in split | |
1898 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / | |
1899 | ||
9da458fc | 1900 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 IZ |
1901 | the infinite loop>. The rules for this are different for lower-level |
1902 | loops given by the greedy modifiers C<*+{}>, and for higher-level | |
1903 | ones like the C</g> modifier or split() operator. | |
1904 | ||
19799a22 GS |
1905 | The lower-level loops are I<interrupted> (that is, the loop is |
1906 | broken) when Perl detects that a repeated expression matched a | |
1907 | zero-length substring. Thus | |
c84d73f1 IZ |
1908 | |
1909 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
1910 | ||
5d458dd8 | 1911 | is made equivalent to |
c84d73f1 | 1912 | |
5d458dd8 YO |
1913 | m{ (?: NON_ZERO_LENGTH )* |
1914 | | | |
1915 | (?: ZERO_LENGTH )? | |
c84d73f1 IZ |
1916 | }x; |
1917 | ||
1918 | The higher level-loops preserve an additional state between iterations: | |
5d458dd8 | 1919 | whether the last match was zero-length. To break the loop, the following |
c84d73f1 | 1920 | match after a zero-length match is prohibited to have a length of zero. |
5d458dd8 | 1921 | This prohibition interacts with backtracking (see L<"Backtracking">), |
c84d73f1 IZ |
1922 | and so the I<second best> match is chosen if the I<best> match is of |
1923 | zero length. | |
1924 | ||
19799a22 | 1925 | For example: |
c84d73f1 IZ |
1926 | |
1927 | $_ = 'bar'; | |
1928 | s/\w??/<$&>/g; | |
1929 | ||
20fb949f | 1930 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
5d458dd8 | 1931 | match given by non-greedy C<??> is the zero-length match, and the I<second |
c84d73f1 IZ |
1932 | best> match is what is matched by C<\w>. Thus zero-length matches |
1933 | alternate with one-character-long matches. | |
1934 | ||
5d458dd8 | 1935 | Similarly, for repeated C<m/()/g> the second-best match is the match at the |
c84d73f1 IZ |
1936 | position one notch further in the string. |
1937 | ||
19799a22 | 1938 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 | 1939 | the matched string, and is reset by each assignment to pos(). |
9da458fc IZ |
1940 | Zero-length matches at the end of the previous match are ignored |
1941 | during C<split>. | |
c84d73f1 | 1942 | |
0d017f4d | 1943 | =head2 Combining RE Pieces |
35a734be IZ |
1944 | |
1945 | Each of the elementary pieces of regular expressions which were described | |
1946 | before (such as C<ab> or C<\Z>) could match at most one substring | |
1947 | at the given position of the input string. However, in a typical regular | |
1948 | expression these elementary pieces are combined into more complicated | |
1949 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc | |
1950 | (in these examples C<S> and C<T> are regular subexpressions). | |
1951 | ||
1952 | Such combinations can include alternatives, leading to a problem of choice: | |
1953 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
1954 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
1955 | actually matched is the concept of backtracking (see L<"Backtracking">). | |
1956 | However, this description is too low-level and makes you think | |
1957 | in terms of a particular implementation. | |
1958 | ||
1959 | Another description starts with notions of "better"/"worse". All the | |
1960 | substrings which may be matched by the given regular expression can be | |
1961 | sorted from the "best" match to the "worst" match, and it is the "best" | |
1962 | match which is chosen. This substitutes the question of "what is chosen?" | |
1963 | by the question of "which matches are better, and which are worse?". | |
1964 | ||
1965 | Again, for elementary pieces there is no such question, since at most | |
1966 | one match at a given position is possible. This section describes the | |
1967 | notion of better/worse for combining operators. In the description | |
1968 | below C<S> and C<T> are regular subexpressions. | |
1969 | ||
13a2d996 | 1970 | =over 4 |
35a734be IZ |
1971 | |
1972 | =item C<ST> | |
1973 | ||
1974 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are | |
1975 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings | |
5d458dd8 | 1976 | which can be matched by C<T>. |
35a734be IZ |
1977 | |
1978 | If C<A> is better match for C<S> than C<A'>, C<AB> is a better | |
1979 | match than C<A'B'>. | |
1980 | ||
1981 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if | |
1982 | C<B> is better match for C<T> than C<B'>. | |
1983 | ||
1984 | =item C<S|T> | |
1985 | ||
1986 | When C<S> can match, it is a better match than when only C<T> can match. | |
1987 | ||
1988 | Ordering of two matches for C<S> is the same as for C<S>. Similar for | |
1989 | two matches for C<T>. | |
1990 | ||
1991 | =item C<S{REPEAT_COUNT}> | |
1992 | ||
1993 | Matches as C<SSS...S> (repeated as many times as necessary). | |
1994 | ||
1995 | =item C<S{min,max}> | |
1996 | ||
1997 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
1998 | ||
1999 | =item C<S{min,max}?> | |
2000 | ||
2001 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
2002 | ||
2003 | =item C<S?>, C<S*>, C<S+> | |
2004 | ||
2005 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
2006 | ||
2007 | =item C<S??>, C<S*?>, C<S+?> | |
2008 | ||
2009 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
2010 | ||
c47ff5f1 | 2011 | =item C<< (?>S) >> |
35a734be IZ |
2012 | |
2013 | Matches the best match for C<S> and only that. | |
2014 | ||
2015 | =item C<(?=S)>, C<(?<=S)> | |
2016 | ||
2017 | Only the best match for C<S> is considered. (This is important only if | |
2018 | C<S> has capturing parentheses, and backreferences are used somewhere | |
2019 | else in the whole regular expression.) | |
2020 | ||
2021 | =item C<(?!S)>, C<(?<!S)> | |
2022 | ||
2023 | For this grouping operator there is no need to describe the ordering, since | |
2024 | only whether or not C<S> can match is important. | |
2025 | ||
6bda09f9 | 2026 | =item C<(??{ EXPR })>, C<(?PARNO)> |
35a734be IZ |
2027 | |
2028 | The ordering is the same as for the regular expression which is | |
6bda09f9 | 2029 | the result of EXPR, or the pattern contained by capture buffer PARNO. |
35a734be IZ |
2030 | |
2031 | =item C<(?(condition)yes-pattern|no-pattern)> | |
2032 | ||
2033 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is | |
2034 | already determined. The ordering of the matches is the same as for the | |
2035 | chosen subexpression. | |
2036 | ||
2037 | =back | |
2038 | ||
2039 | The above recipes describe the ordering of matches I<at a given position>. | |
2040 | One more rule is needed to understand how a match is determined for the | |
2041 | whole regular expression: a match at an earlier position is always better | |
2042 | than a match at a later position. | |
2043 | ||
0d017f4d | 2044 | =head2 Creating Custom RE Engines |
c84d73f1 IZ |
2045 | |
2046 | Overloaded constants (see L<overload>) provide a simple way to extend | |
2047 | the functionality of the RE engine. | |
2048 | ||
2049 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
0d017f4d | 2050 | matches at a boundary between whitespace characters and non-whitespace |
c84d73f1 IZ |
2051 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
2052 | at these positions, so we want to have each C<\Y|> in the place of the | |
2053 | more complicated version. We can create a module C<customre> to do | |
2054 | this: | |
2055 | ||
2056 | package customre; | |
2057 | use overload; | |
2058 | ||
2059 | sub import { | |
2060 | shift; | |
2061 | die "No argument to customre::import allowed" if @_; | |
2062 | overload::constant 'qr' => \&convert; | |
2063 | } | |
2064 | ||
2065 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
2066 | ||
580a9fe1 RGS |
2067 | # We must also take care of not escaping the legitimate \\Y| |
2068 | # sequence, hence the presence of '\\' in the conversion rules. | |
5d458dd8 | 2069 | my %rules = ( '\\' => '\\\\', |
c84d73f1 IZ |
2070 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
2071 | sub convert { | |
2072 | my $re = shift; | |
5d458dd8 | 2073 | $re =~ s{ |
c84d73f1 IZ |
2074 | \\ ( \\ | Y . ) |
2075 | } | |
5d458dd8 | 2076 | { $rules{$1} or invalid($re,$1) }sgex; |
c84d73f1 IZ |
2077 | return $re; |
2078 | } | |
2079 | ||
2080 | Now C<use customre> enables the new escape in constant regular | |
2081 | expressions, i.e., those without any runtime variable interpolations. | |
2082 | As documented in L<overload>, this conversion will work only over | |
2083 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
2084 | part of this regular expression needs to be converted explicitly | |
2085 | (but only if the special meaning of C<\Y|> should be enabled inside $re): | |
2086 | ||
2087 | use customre; | |
2088 | $re = <>; | |
2089 | chomp $re; | |
2090 | $re = customre::convert $re; | |
2091 | /\Y|$re\Y|/; | |
2092 | ||
1f1031fe YO |
2093 | =head1 PCRE/Python Support |
2094 | ||
2095 | As of Perl 5.10 Perl supports several Python/PCRE specific extensions | |
2096 | to the regex syntax. While Perl programmers are encouraged to use the | |
2097 | Perl specific syntax, the following are legal in Perl 5.10: | |
2098 | ||
2099 | =over 4 | |
2100 | ||
ee9b8eae | 2101 | =item C<< (?PE<lt>NAMEE<gt>pattern) >> |
1f1031fe YO |
2102 | |
2103 | Define a named capture buffer. Equivalent to C<< (?<NAME>pattern) >>. | |
2104 | ||
2105 | =item C<< (?P=NAME) >> | |
2106 | ||
2107 | Backreference to a named capture buffer. Equivalent to C<< \g{NAME} >>. | |
2108 | ||
2109 | =item C<< (?P>NAME) >> | |
2110 | ||
2111 | Subroutine call to a named capture buffer. Equivalent to C<< (?&NAME) >>. | |
2112 | ||
ee9b8eae | 2113 | =back |
1f1031fe | 2114 | |
19799a22 GS |
2115 | =head1 BUGS |
2116 | ||
9da458fc IZ |
2117 | This document varies from difficult to understand to completely |
2118 | and utterly opaque. The wandering prose riddled with jargon is | |
2119 | hard to fathom in several places. | |
2120 | ||
2121 | This document needs a rewrite that separates the tutorial content | |
2122 | from the reference content. | |
19799a22 GS |
2123 | |
2124 | =head1 SEE ALSO | |
9fa51da4 | 2125 | |
91e0c79e MJD |
2126 | L<perlrequick>. |
2127 | ||
2128 | L<perlretut>. | |
2129 | ||
9b599b2a GS |
2130 | L<perlop/"Regexp Quote-Like Operators">. |
2131 | ||
1e66bd83 PP |
2132 | L<perlop/"Gory details of parsing quoted constructs">. |
2133 | ||
14218588 GS |
2134 | L<perlfaq6>. |
2135 | ||
9b599b2a GS |
2136 | L<perlfunc/pos>. |
2137 | ||
2138 | L<perllocale>. | |
2139 | ||
fb55449c JH |
2140 | L<perlebcdic>. |
2141 | ||
14218588 GS |
2142 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
2143 | by O'Reilly and Associates. |