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 | ||
91e0c79e MJD |
8 | This page describes the syntax of regular expressions in Perl. |
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 | |
19799a22 | 19 | Matching operations can have various modifiers. Modifiers |
5a964f20 | 20 | that relate to the interpretation of the regular expression inside |
19799a22 GS |
21 | are listed below. Modifiers that alter the way a regular expression |
22 | is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and | |
1e66bd83 | 23 | L<perlop/"Gory details of parsing quoted constructs">. |
a0d0e21e | 24 | |
55497cff | 25 | =over 4 |
26 | ||
27 | =item i | |
d74e8afc ITB |
28 | X</i> X<regex, case-insensitive> X<regexp, case-insensitive> |
29 | X<regular expression, case-insensitive> | |
55497cff | 30 | |
31 | Do case-insensitive pattern matching. | |
32 | ||
a034a98d DD |
33 | If C<use locale> is in effect, the case map is taken from the current |
34 | locale. See L<perllocale>. | |
35 | ||
54310121 | 36 | =item m |
d74e8afc | 37 | X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline> |
55497cff | 38 | |
39 | Treat string as multiple lines. That is, change "^" and "$" from matching | |
14218588 | 40 | the start or end of the string to matching the start or end of any |
7f761169 | 41 | line anywhere within the string. |
55497cff | 42 | |
54310121 | 43 | =item s |
d74e8afc ITB |
44 | X</s> X<regex, single-line> X<regexp, single-line> |
45 | X<regular expression, single-line> | |
55497cff | 46 | |
47 | Treat string as single line. That is, change "." to match any character | |
19799a22 | 48 | whatsoever, even a newline, which normally it would not match. |
55497cff | 49 | |
f02c194e | 50 | Used together, as /ms, they let the "." match any character whatsoever, |
fb55449c | 51 | while still allowing "^" and "$" to match, respectively, just after |
19799a22 | 52 | and just before newlines within the string. |
7b8d334a | 53 | |
54310121 | 54 | =item x |
d74e8afc | 55 | X</x> |
55497cff | 56 | |
57 | Extend your pattern's legibility by permitting whitespace and comments. | |
58 | ||
59 | =back | |
a0d0e21e LW |
60 | |
61 | These are usually written as "the C</x> modifier", even though the delimiter | |
14218588 | 62 | in question might not really be a slash. Any of these |
a0d0e21e | 63 | modifiers may also be embedded within the regular expression itself using |
14218588 | 64 | the C<(?...)> construct. See below. |
a0d0e21e | 65 | |
4633a7c4 | 66 | The C</x> modifier itself needs a little more explanation. It tells |
55497cff | 67 | the regular expression parser to ignore whitespace that is neither |
68 | backslashed nor within a character class. You can use this to break up | |
4633a7c4 | 69 | your regular expression into (slightly) more readable parts. The C<#> |
54310121 | 70 | character is also treated as a metacharacter introducing a comment, |
55497cff | 71 | just as in ordinary Perl code. This also means that if you want real |
14218588 | 72 | whitespace or C<#> characters in the pattern (outside a character |
f9a3ff1a RGS |
73 | class, where they are unaffected by C</x>), then you'll either have to |
74 | escape them (using backslashes or C<\Q...\E>) or encode them using octal | |
8933a740 RGS |
75 | or hex escapes. Taken together, these features go a long way towards |
76 | making Perl's regular expressions more readable. Note that you have to | |
77 | be careful not to include the pattern delimiter in the comment--perl has | |
78 | no way of knowing you did not intend to close the pattern early. See | |
79 | the C-comment deletion code in L<perlop>. Also note that anything inside | |
1031e5db | 80 | a C<\Q...\E> stays unaffected by C</x>. |
d74e8afc | 81 | X</x> |
a0d0e21e LW |
82 | |
83 | =head2 Regular Expressions | |
84 | ||
04838cea RGS |
85 | =head3 Metacharacters |
86 | ||
19799a22 | 87 | The patterns used in Perl pattern matching derive from supplied in |
14218588 | 88 | the Version 8 regex routines. (The routines are derived |
19799a22 GS |
89 | (distantly) from Henry Spencer's freely redistributable reimplementation |
90 | of the V8 routines.) See L<Version 8 Regular Expressions> for | |
91 | details. | |
a0d0e21e LW |
92 | |
93 | In particular the following metacharacters have their standard I<egrep>-ish | |
94 | meanings: | |
d74e8afc ITB |
95 | X<metacharacter> |
96 | X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]> | |
97 | ||
a0d0e21e | 98 | |
54310121 | 99 | \ Quote the next metacharacter |
a0d0e21e LW |
100 | ^ Match the beginning of the line |
101 | . Match any character (except newline) | |
c07a80fd | 102 | $ Match the end of the line (or before newline at the end) |
a0d0e21e LW |
103 | | Alternation |
104 | () Grouping | |
105 | [] Character class | |
106 | ||
14218588 GS |
107 | By default, the "^" character is guaranteed to match only the |
108 | beginning of the string, the "$" character only the end (or before the | |
109 | newline at the end), and Perl does certain optimizations with the | |
a0d0e21e LW |
110 | assumption that the string contains only one line. Embedded newlines |
111 | will not be matched by "^" or "$". You may, however, wish to treat a | |
4a6725af | 112 | string as a multi-line buffer, such that the "^" will match after any |
0d520e8e YO |
113 | newline within the string (except if the newline is the last character in |
114 | the string), and "$" will match before any newline. At the | |
a0d0e21e LW |
115 | cost of a little more overhead, you can do this by using the /m modifier |
116 | on the pattern match operator. (Older programs did this by setting C<$*>, | |
f02c194e | 117 | but this practice has been removed in perl 5.9.) |
d74e8afc | 118 | X<^> X<$> X</m> |
a0d0e21e | 119 | |
14218588 | 120 | To simplify multi-line substitutions, the "." character never matches a |
55497cff | 121 | newline unless you use the C</s> modifier, which in effect tells Perl to pretend |
f02c194e | 122 | the string is a single line--even if it isn't. |
d74e8afc | 123 | X<.> X</s> |
a0d0e21e | 124 | |
04838cea RGS |
125 | =head3 Quantifiers |
126 | ||
a0d0e21e | 127 | The following standard quantifiers are recognized: |
d74e8afc | 128 | X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}> |
a0d0e21e LW |
129 | |
130 | * Match 0 or more times | |
131 | + Match 1 or more times | |
132 | ? Match 1 or 0 times | |
133 | {n} Match exactly n times | |
134 | {n,} Match at least n times | |
135 | {n,m} Match at least n but not more than m times | |
136 | ||
137 | (If a curly bracket occurs in any other context, it is treated | |
b975c076 JL |
138 | as a regular character. In particular, the lower bound |
139 | is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+" | |
25f94b33 | 140 | modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited |
9c79236d GS |
141 | to integral values less than a preset limit defined when perl is built. |
142 | This is usually 32766 on the most common platforms. The actual limit can | |
143 | be seen in the error message generated by code such as this: | |
144 | ||
820475bd | 145 | $_ **= $_ , / {$_} / for 2 .. 42; |
a0d0e21e | 146 | |
54310121 | 147 | By default, a quantified subpattern is "greedy", that is, it will match as |
148 | many times as possible (given a particular starting location) while still | |
149 | allowing the rest of the pattern to match. If you want it to match the | |
150 | minimum number of times possible, follow the quantifier with a "?". Note | |
151 | that the meanings don't change, just the "greediness": | |
d74e8afc ITB |
152 | X<metacharacter> X<greedy> X<greedyness> |
153 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?> | |
a0d0e21e LW |
154 | |
155 | *? Match 0 or more times | |
156 | +? Match 1 or more times | |
157 | ?? Match 0 or 1 time | |
158 | {n}? Match exactly n times | |
159 | {n,}? Match at least n times | |
160 | {n,m}? Match at least n but not more than m times | |
161 | ||
b9b4dddf YO |
162 | By default, when a quantified subpattern does not allow the rest of the |
163 | overall pattern to match, Perl will backtrack. However, this behaviour is | |
164 | sometimes undesirable. Thus Perl provides the "possesive" quantifier form | |
165 | as well. | |
166 | ||
167 | *+ Match 0 or more times and give nothing back | |
04838cea | 168 | ++ Match 1 or more times and give nothing back |
b9b4dddf YO |
169 | ?+ Match 0 or 1 time and give nothing back |
170 | {n}+ Match exactly n times and give nothing back (redundant) | |
04838cea RGS |
171 | {n,}+ Match at least n times and give nothing back |
172 | {n,m}+ Match at least n but not more than m times and give nothing back | |
b9b4dddf YO |
173 | |
174 | For instance, | |
175 | ||
176 | 'aaaa' =~ /a++a/ | |
177 | ||
178 | will never match, as the C<a++> will gobble up all the C<a>'s in the | |
179 | string and won't leave any for the remaining part of the pattern. This | |
180 | feature can be extremely useful to give perl hints about where it | |
181 | shouldn't backtrack. For instance, the typical "match a double-quoted | |
182 | string" problem can be most efficiently performed when written as: | |
183 | ||
184 | /"(?:[^"\\]++|\\.)*+"/ | |
185 | ||
186 | as we know that if the final quote does not match, bactracking will not | |
187 | help. See the independent subexpression C<< (?>...) >> for more details; | |
188 | possessive quantifiers are just syntactic sugar for that construct. For | |
189 | instance the above example could also be written as follows: | |
190 | ||
191 | /"(?>(?:(?>[^"\\]+)|\\.)*)"/ | |
192 | ||
04838cea RGS |
193 | =head3 Escape sequences |
194 | ||
5f05dabc | 195 | Because patterns are processed as double quoted strings, the following |
a0d0e21e | 196 | also work: |
d74e8afc ITB |
197 | X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q> |
198 | X<\0> X<\c> X<\N> X<\x> | |
a0d0e21e | 199 | |
0f36ee90 | 200 | \t tab (HT, TAB) |
201 | \n newline (LF, NL) | |
202 | \r return (CR) | |
203 | \f form feed (FF) | |
204 | \a alarm (bell) (BEL) | |
205 | \e escape (think troff) (ESC) | |
cb1a09d0 AD |
206 | \033 octal char (think of a PDP-11) |
207 | \x1B hex char | |
a0ed51b3 | 208 | \x{263a} wide hex char (Unicode SMILEY) |
a0d0e21e | 209 | \c[ control char |
4a2d328f | 210 | \N{name} named char |
cb1a09d0 AD |
211 | \l lowercase next char (think vi) |
212 | \u uppercase next char (think vi) | |
213 | \L lowercase till \E (think vi) | |
214 | \U uppercase till \E (think vi) | |
215 | \E end case modification (think vi) | |
5a964f20 | 216 | \Q quote (disable) pattern metacharacters till \E |
a0d0e21e | 217 | |
a034a98d | 218 | If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u> |
423cee85 | 219 | and C<\U> is taken from the current locale. See L<perllocale>. For |
4a2d328f | 220 | documentation of C<\N{name}>, see L<charnames>. |
a034a98d | 221 | |
1d2dff63 GS |
222 | You cannot include a literal C<$> or C<@> within a C<\Q> sequence. |
223 | An unescaped C<$> or C<@> interpolates the corresponding variable, | |
224 | while escaping will cause the literal string C<\$> to be matched. | |
225 | You'll need to write something like C<m/\Quser\E\@\Qhost/>. | |
226 | ||
04838cea RGS |
227 | =head3 Character classes |
228 | ||
a0d0e21e | 229 | In addition, Perl defines the following: |
d74e8afc ITB |
230 | X<metacharacter> |
231 | X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C> | |
232 | X<word> X<whitespace> | |
a0d0e21e | 233 | |
81714fb9 YO |
234 | \w Match a "word" character (alphanumeric plus "_") |
235 | \W Match a non-"word" character | |
236 | \s Match a whitespace character | |
237 | \S Match a non-whitespace character | |
238 | \d Match a digit character | |
239 | \D Match a non-digit character | |
240 | \pP Match P, named property. Use \p{Prop} for longer names. | |
241 | \PP Match non-P | |
242 | \X Match eXtended Unicode "combining character sequence", | |
243 | equivalent to (?:\PM\pM*) | |
244 | \C Match a single C char (octet) even under Unicode. | |
245 | NOTE: breaks up characters into their UTF-8 bytes, | |
246 | so you may end up with malformed pieces of UTF-8. | |
247 | Unsupported in lookbehind. | |
248 | \1 Backreference to a a specific group. | |
249 | '1' may actually be any positive integer | |
250 | \k<name> Named backreference | |
251 | \N{name} Named unicode character, or unicode escape. | |
252 | \x12 Hexadecimal escape sequence | |
253 | \x{1234} Long hexadecimal escape sequence | |
a0d0e21e | 254 | |
08ce8fc6 JH |
255 | A C<\w> matches a single alphanumeric character (an alphabetic |
256 | character, or a decimal digit) or C<_>, not a whole word. Use C<\w+> | |
257 | to match a string of Perl-identifier characters (which isn't the same | |
258 | as matching an English word). If C<use locale> is in effect, the list | |
259 | of alphabetic characters generated by C<\w> is taken from the current | |
260 | locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>, | |
1209ba90 | 261 | C<\d>, and C<\D> within character classes, but if you try to use them |
08ce8fc6 JH |
262 | as endpoints of a range, that's not a range, the "-" is understood |
263 | literally. If Unicode is in effect, C<\s> matches also "\x{85}", | |
264 | "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about | |
491fd90a | 265 | C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general. |
fa11829f | 266 | You can define your own C<\p> and C<\P> properties, see L<perlunicode>. |
d74e8afc | 267 | X<\w> X<\W> X<word> |
a0d0e21e | 268 | |
b8c5462f | 269 | The POSIX character class syntax |
d74e8afc | 270 | X<character class> |
b8c5462f | 271 | |
820475bd | 272 | [:class:] |
b8c5462f | 273 | |
5496314a SP |
274 | is also available. Note that the C<[> and C<]> braces are I<literal>; |
275 | they must always be used within a character class expression. | |
276 | ||
277 | # this is correct: | |
278 | $string =~ /[[:alpha:]]/; | |
279 | ||
280 | # this is not, and will generate a warning: | |
281 | $string =~ /[:alpha:]/; | |
282 | ||
283 | The available classes and their backslash equivalents (if available) are | |
284 | as follows: | |
d74e8afc ITB |
285 | X<character class> |
286 | X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph> | |
287 | X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit> | |
b8c5462f JH |
288 | |
289 | alpha | |
290 | alnum | |
291 | ascii | |
aaa51d5e | 292 | blank [1] |
b8c5462f JH |
293 | cntrl |
294 | digit \d | |
295 | graph | |
296 | lower | |
297 | ||
298 | punct | |
aaa51d5e | 299 | space \s [2] |
b8c5462f | 300 | upper |
aaa51d5e | 301 | word \w [3] |
b8c5462f JH |
302 | xdigit |
303 | ||
07698885 RGS |
304 | =over |
305 | ||
306 | =item [1] | |
307 | ||
b432a672 | 308 | A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace". |
07698885 RGS |
309 | |
310 | =item [2] | |
311 | ||
312 | Not exactly equivalent to C<\s> since the C<[[:space:]]> includes | |
b432a672 | 313 | also the (very rare) "vertical tabulator", "\ck", chr(11). |
07698885 RGS |
314 | |
315 | =item [3] | |
316 | ||
08ce8fc6 | 317 | A Perl extension, see above. |
07698885 RGS |
318 | |
319 | =back | |
aaa51d5e | 320 | |
26b44a0a | 321 | For example use C<[:upper:]> to match all the uppercase characters. |
aaa51d5e JF |
322 | Note that the C<[]> are part of the C<[::]> construct, not part of the |
323 | whole character class. For example: | |
b8c5462f | 324 | |
820475bd | 325 | [01[:alpha:]%] |
b8c5462f | 326 | |
593df60c | 327 | matches zero, one, any alphabetic character, and the percentage sign. |
b8c5462f | 328 | |
72ff2908 JH |
329 | The following equivalences to Unicode \p{} constructs and equivalent |
330 | backslash character classes (if available), will hold: | |
d74e8afc | 331 | X<character class> X<\p> X<\p{}> |
72ff2908 | 332 | |
5496314a | 333 | [[:...:]] \p{...} backslash |
b8c5462f JH |
334 | |
335 | alpha IsAlpha | |
336 | alnum IsAlnum | |
337 | ascii IsASCII | |
b432a672 | 338 | blank IsSpace |
b8c5462f | 339 | cntrl IsCntrl |
3bec3564 | 340 | digit IsDigit \d |
b8c5462f JH |
341 | graph IsGraph |
342 | lower IsLower | |
343 | print IsPrint | |
344 | punct IsPunct | |
345 | space IsSpace | |
3bec3564 | 346 | IsSpacePerl \s |
b8c5462f JH |
347 | upper IsUpper |
348 | word IsWord | |
349 | xdigit IsXDigit | |
350 | ||
5496314a | 351 | For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent. |
b8c5462f JH |
352 | |
353 | If the C<utf8> pragma is not used but the C<locale> pragma is, the | |
aaa51d5e | 354 | classes correlate with the usual isalpha(3) interface (except for |
b432a672 | 355 | "word" and "blank"). |
b8c5462f JH |
356 | |
357 | The assumedly non-obviously named classes are: | |
358 | ||
359 | =over 4 | |
360 | ||
361 | =item cntrl | |
d74e8afc | 362 | X<cntrl> |
b8c5462f | 363 | |
820475bd GS |
364 | Any control character. Usually characters that don't produce output as |
365 | such but instead control the terminal somehow: for example newline and | |
366 | backspace are control characters. All characters with ord() less than | |
593df60c | 367 | 32 are most often classified as control characters (assuming ASCII, |
7be5a6cf JF |
368 | the ISO Latin character sets, and Unicode), as is the character with |
369 | the ord() value of 127 (C<DEL>). | |
b8c5462f JH |
370 | |
371 | =item graph | |
d74e8afc | 372 | X<graph> |
b8c5462f | 373 | |
f1cbbd6e | 374 | Any alphanumeric or punctuation (special) character. |
b8c5462f JH |
375 | |
376 | =item print | |
d74e8afc | 377 | X<print> |
b8c5462f | 378 | |
f79b3095 | 379 | Any alphanumeric or punctuation (special) character or the space character. |
b8c5462f JH |
380 | |
381 | =item punct | |
d74e8afc | 382 | X<punct> |
b8c5462f | 383 | |
f1cbbd6e | 384 | Any punctuation (special) character. |
b8c5462f JH |
385 | |
386 | =item xdigit | |
d74e8afc | 387 | X<xdigit> |
b8c5462f | 388 | |
593df60c | 389 | Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would |
820475bd | 390 | work just fine) it is included for completeness. |
b8c5462f | 391 | |
b8c5462f JH |
392 | =back |
393 | ||
394 | You can negate the [::] character classes by prefixing the class name | |
395 | with a '^'. This is a Perl extension. For example: | |
d74e8afc | 396 | X<character class, negation> |
b8c5462f | 397 | |
5496314a | 398 | POSIX traditional Unicode |
93733859 | 399 | |
5496314a SP |
400 | [[:^digit:]] \D \P{IsDigit} |
401 | [[:^space:]] \S \P{IsSpace} | |
402 | [[:^word:]] \W \P{IsWord} | |
b8c5462f | 403 | |
54c18d04 MK |
404 | Perl respects the POSIX standard in that POSIX character classes are |
405 | only supported within a character class. The POSIX character classes | |
406 | [.cc.] and [=cc=] are recognized but B<not> supported and trying to | |
407 | use them will cause an error. | |
b8c5462f | 408 | |
04838cea RGS |
409 | =head3 Assertions |
410 | ||
a0d0e21e | 411 | Perl defines the following zero-width assertions: |
d74e8afc ITB |
412 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> |
413 | X<regexp, zero-width assertion> | |
414 | X<regular expression, zero-width assertion> | |
415 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> | |
a0d0e21e LW |
416 | |
417 | \b Match a word boundary | |
418 | \B Match a non-(word boundary) | |
b85d18e9 IZ |
419 | \A Match only at beginning of string |
420 | \Z Match only at end of string, or before newline at the end | |
421 | \z Match only at end of string | |
9da458fc IZ |
422 | \G Match only at pos() (e.g. at the end-of-match position |
423 | of prior m//g) | |
a0d0e21e | 424 | |
14218588 | 425 | A word boundary (C<\b>) is a spot between two characters |
19799a22 GS |
426 | that has a C<\w> on one side of it and a C<\W> on the other side |
427 | of it (in either order), counting the imaginary characters off the | |
428 | beginning and end of the string as matching a C<\W>. (Within | |
429 | character classes C<\b> represents backspace rather than a word | |
430 | boundary, just as it normally does in any double-quoted string.) | |
431 | The C<\A> and C<\Z> are just like "^" and "$", except that they | |
432 | won't match multiple times when the C</m> modifier is used, while | |
433 | "^" and "$" will match at every internal line boundary. To match | |
434 | the actual end of the string and not ignore an optional trailing | |
435 | newline, use C<\z>. | |
d74e8afc | 436 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 GS |
437 | |
438 | The C<\G> assertion can be used to chain global matches (using | |
439 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. | |
440 | It is also useful when writing C<lex>-like scanners, when you have | |
441 | several patterns that you want to match against consequent substrings | |
442 | of your string, see the previous reference. The actual location | |
443 | where C<\G> will match can also be influenced by using C<pos()> as | |
25cf8c22 HS |
444 | an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully |
445 | supported when anchored to the start of the pattern; while it | |
446 | is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some | |
447 | such uses (C</.\G/g>, for example) currently cause problems, and | |
448 | it is recommended that you avoid such usage for now. | |
d74e8afc | 449 | X<\G> |
c47ff5f1 | 450 | |
04838cea RGS |
451 | =head3 Capture buffers |
452 | ||
14218588 | 453 | The bracketing construct C<( ... )> creates capture buffers. To |
c47ff5f1 | 454 | refer to the digit'th buffer use \<digit> within the |
14218588 | 455 | match. Outside the match use "$" instead of "\". (The |
81714fb9 | 456 | \<digit> notation works in certain circumstances outside |
14218588 GS |
457 | the match. See the warning below about \1 vs $1 for details.) |
458 | Referring back to another part of the match is called a | |
459 | I<backreference>. | |
d74e8afc ITB |
460 | X<regex, capture buffer> X<regexp, capture buffer> |
461 | X<regular expression, capture buffer> X<backreference> | |
14218588 GS |
462 | |
463 | There is no limit to the number of captured substrings that you may | |
464 | use. However Perl also uses \10, \11, etc. as aliases for \010, | |
fb55449c JH |
465 | \011, etc. (Recall that 0 means octal, so \011 is the character at |
466 | number 9 in your coded character set; which would be the 10th character, | |
81714fb9 YO |
467 | a horizontal tab under ASCII.) Perl resolves this |
468 | ambiguity by interpreting \10 as a backreference only if at least 10 | |
469 | left parentheses have opened before it. Likewise \11 is a | |
470 | backreference only if at least 11 left parentheses have opened | |
471 | before it. And so on. \1 through \9 are always interpreted as | |
fb55449c | 472 | backreferences. |
14218588 | 473 | |
81714fb9 YO |
474 | Additionally, as of Perl 5.10 you may use named capture buffers and named |
475 | backreferences. The notation is C<< (?<name>...) >> and C<< \k<name> >> | |
476 | (you may also use single quotes instead of angle brackets to quote the | |
477 | name). The only difference with named capture buffers and unnamed ones is | |
478 | that multiple buffers may have the same name and that the contents of | |
479 | named capture buffers is available via the C<%+> hash. When multiple | |
480 | groups share the same name C<$+{name}> and C<< \k<name> >> refer to the | |
481 | leftmost defined group, thus it's possible to do things with named capture | |
482 | buffers that would otherwise require C<(??{})> code to accomplish. Named | |
483 | capture buffers are numbered just as normal capture buffers are and may be | |
484 | referenced via the magic numeric variables or via numeric backreferences | |
485 | as well as by name. | |
486 | ||
14218588 | 487 | Examples: |
a0d0e21e LW |
488 | |
489 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
490 | ||
81714fb9 YO |
491 | /(.)\1/ # find first doubled char |
492 | and print "'$1' is the first doubled character\n"; | |
493 | ||
494 | /(?<char>.)\k<char>/ # ... a different way | |
495 | and print "'$+{char}' is the first doubled character\n"; | |
496 | ||
497 | /(?<char>.)\1/ # ... mix and match | |
498 | and print "'$1' is the first doubled character\n"; | |
c47ff5f1 | 499 | |
14218588 | 500 | if (/Time: (..):(..):(..)/) { # parse out values |
a0d0e21e LW |
501 | $hours = $1; |
502 | $minutes = $2; | |
503 | $seconds = $3; | |
504 | } | |
c47ff5f1 | 505 | |
14218588 GS |
506 | Several special variables also refer back to portions of the previous |
507 | match. C<$+> returns whatever the last bracket match matched. | |
508 | C<$&> returns the entire matched string. (At one point C<$0> did | |
509 | also, but now it returns the name of the program.) C<$`> returns | |
77ea4f6d JV |
510 | everything before the matched string. C<$'> returns everything |
511 | after the matched string. And C<$^N> contains whatever was matched by | |
512 | the most-recently closed group (submatch). C<$^N> can be used in | |
513 | extended patterns (see below), for example to assign a submatch to a | |
81714fb9 | 514 | variable. |
d74e8afc | 515 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 | 516 | |
665e98b9 | 517 | The numbered match variables ($1, $2, $3, etc.) and the related punctuation |
77ea4f6d | 518 | set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped |
14218588 GS |
519 | until the end of the enclosing block or until the next successful |
520 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
d74e8afc ITB |
521 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
522 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> | |
523 | ||
14218588 | 524 | |
665e98b9 | 525 | B<NOTE>: failed matches in Perl do not reset the match variables, |
5146ce24 | 526 | which makes it easier to write code that tests for a series of more |
665e98b9 JH |
527 | specific cases and remembers the best match. |
528 | ||
14218588 GS |
529 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or |
530 | C<$'> anywhere in the program, it has to provide them for every | |
531 | pattern match. This may substantially slow your program. Perl | |
532 | uses the same mechanism to produce $1, $2, etc, so you also pay a | |
533 | price for each pattern that contains capturing parentheses. (To | |
534 | avoid this cost while retaining the grouping behaviour, use the | |
535 | extended regular expression C<(?: ... )> instead.) But if you never | |
536 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
537 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
538 | if you can, but if you can't (and some algorithms really appreciate | |
539 | them), once you've used them once, use them at will, because you've | |
540 | already paid the price. As of 5.005, C<$&> is not so costly as the | |
541 | other two. | |
d74e8afc | 542 | X<$&> X<$`> X<$'> |
68dc0745 | 543 | |
19799a22 GS |
544 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
545 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
546 | are no backslashed symbols that aren't alphanumeric. So anything | |
c47ff5f1 | 547 | that looks like \\, \(, \), \<, \>, \{, or \} is always |
19799a22 GS |
548 | interpreted as a literal character, not a metacharacter. This was |
549 | once used in a common idiom to disable or quote the special meanings | |
550 | of regular expression metacharacters in a string that you want to | |
36bbe248 | 551 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e LW |
552 | |
553 | $pattern =~ s/(\W)/\\$1/g; | |
554 | ||
f1cbbd6e | 555 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 GS |
556 | Today it is more common to use the quotemeta() function or the C<\Q> |
557 | metaquoting escape sequence to disable all metacharacters' special | |
558 | meanings like this: | |
a0d0e21e LW |
559 | |
560 | /$unquoted\Q$quoted\E$unquoted/ | |
561 | ||
9da458fc IZ |
562 | Beware that if you put literal backslashes (those not inside |
563 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
564 | backslash interpolation may lead to confusing results. If you | |
565 | I<need> to use literal backslashes within C<\Q...\E>, | |
566 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
567 | ||
19799a22 GS |
568 | =head2 Extended Patterns |
569 | ||
14218588 GS |
570 | Perl also defines a consistent extension syntax for features not |
571 | found in standard tools like B<awk> and B<lex>. The syntax is a | |
572 | pair of parentheses with a question mark as the first thing within | |
573 | the parentheses. The character after the question mark indicates | |
574 | the extension. | |
19799a22 | 575 | |
14218588 GS |
576 | The stability of these extensions varies widely. Some have been |
577 | part of the core language for many years. Others are experimental | |
578 | and may change without warning or be completely removed. Check | |
579 | the documentation on an individual feature to verify its current | |
580 | status. | |
19799a22 | 581 | |
14218588 GS |
582 | A question mark was chosen for this and for the minimal-matching |
583 | construct because 1) question marks are rare in older regular | |
584 | expressions, and 2) whenever you see one, you should stop and | |
585 | "question" exactly what is going on. That's psychology... | |
a0d0e21e LW |
586 | |
587 | =over 10 | |
588 | ||
cc6b7395 | 589 | =item C<(?#text)> |
d74e8afc | 590 | X<(?#)> |
a0d0e21e | 591 | |
14218588 | 592 | A comment. The text is ignored. If the C</x> modifier enables |
19799a22 | 593 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes |
259138e3 GS |
594 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
595 | C<)> in the comment. | |
a0d0e21e | 596 | |
19799a22 | 597 | =item C<(?imsx-imsx)> |
d74e8afc | 598 | X<(?)> |
19799a22 | 599 | |
0b6d1084 JH |
600 | One or more embedded pattern-match modifiers, to be turned on (or |
601 | turned off, if preceded by C<->) for the remainder of the pattern or | |
602 | the remainder of the enclosing pattern group (if any). This is | |
603 | particularly useful for dynamic patterns, such as those read in from a | |
604 | configuration file, read in as an argument, are specified in a table | |
605 | somewhere, etc. Consider the case that some of which want to be case | |
606 | sensitive and some do not. The case insensitive ones need to include | |
607 | merely C<(?i)> at the front of the pattern. For example: | |
19799a22 GS |
608 | |
609 | $pattern = "foobar"; | |
610 | if ( /$pattern/i ) { } | |
611 | ||
612 | # more flexible: | |
613 | ||
614 | $pattern = "(?i)foobar"; | |
615 | if ( /$pattern/ ) { } | |
616 | ||
0b6d1084 | 617 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 GS |
618 | |
619 | ( (?i) blah ) \s+ \1 | |
620 | ||
621 | will match a repeated (I<including the case>!) word C<blah> in any | |
14218588 | 622 | case, assuming C<x> modifier, and no C<i> modifier outside this |
19799a22 GS |
623 | group. |
624 | ||
5a964f20 | 625 | =item C<(?:pattern)> |
d74e8afc | 626 | X<(?:)> |
a0d0e21e | 627 | |
ca9dfc88 IZ |
628 | =item C<(?imsx-imsx:pattern)> |
629 | ||
5a964f20 TC |
630 | This is for clustering, not capturing; it groups subexpressions like |
631 | "()", but doesn't make backreferences as "()" does. So | |
a0d0e21e | 632 | |
5a964f20 | 633 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e LW |
634 | |
635 | is like | |
636 | ||
5a964f20 | 637 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e | 638 | |
19799a22 GS |
639 | but doesn't spit out extra fields. It's also cheaper not to capture |
640 | characters if you don't need to. | |
a0d0e21e | 641 | |
19799a22 GS |
642 | Any letters between C<?> and C<:> act as flags modifiers as with |
643 | C<(?imsx-imsx)>. For example, | |
ca9dfc88 IZ |
644 | |
645 | /(?s-i:more.*than).*million/i | |
646 | ||
14218588 | 647 | is equivalent to the more verbose |
ca9dfc88 IZ |
648 | |
649 | /(?:(?s-i)more.*than).*million/i | |
650 | ||
5a964f20 | 651 | =item C<(?=pattern)> |
d74e8afc | 652 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> |
a0d0e21e | 653 | |
19799a22 | 654 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e LW |
655 | matches a word followed by a tab, without including the tab in C<$&>. |
656 | ||
5a964f20 | 657 | =item C<(?!pattern)> |
d74e8afc | 658 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> |
a0d0e21e | 659 | |
19799a22 | 660 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e | 661 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 GS |
662 | however that look-ahead and look-behind are NOT the same thing. You cannot |
663 | use this for look-behind. | |
7b8d334a | 664 | |
5a964f20 | 665 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a GS |
666 | will not do what you want. That's because the C<(?!foo)> is just saying that |
667 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
668 | match. You would have to do something like C</(?!foo)...bar/> for that. We | |
669 | say "like" because there's the case of your "bar" not having three characters | |
670 | before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>. | |
671 | Sometimes it's still easier just to say: | |
a0d0e21e | 672 | |
a3cb178b | 673 | if (/bar/ && $` !~ /foo$/) |
a0d0e21e | 674 | |
19799a22 | 675 | For look-behind see below. |
c277df42 | 676 | |
c47ff5f1 | 677 | =item C<(?<=pattern)> |
d74e8afc | 678 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> |
c277df42 | 679 | |
c47ff5f1 | 680 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 GS |
681 | matches a word that follows a tab, without including the tab in C<$&>. |
682 | Works only for fixed-width look-behind. | |
c277df42 | 683 | |
5a964f20 | 684 | =item C<(?<!pattern)> |
d74e8afc | 685 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> |
c277df42 | 686 | |
19799a22 GS |
687 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
688 | matches any occurrence of "foo" that does not follow "bar". Works | |
689 | only for fixed-width look-behind. | |
c277df42 | 690 | |
81714fb9 YO |
691 | =item C<(?'NAME'pattern)> |
692 | ||
693 | =item C<< (?<NAME>pattern) >> | |
694 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> | |
695 | ||
696 | A named capture buffer. Identical in every respect to normal capturing | |
697 | parens C<()> but for the additional fact that C<%+> may be used after | |
698 | a succesful match to refer to a named buffer. See C<perlvar> for more | |
699 | details on the C<%+> hash. | |
700 | ||
701 | If multiple distinct capture buffers have the same name then the | |
702 | $+{NAME} will refer to the leftmost defined buffer in the match. | |
703 | ||
704 | The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent. | |
705 | ||
706 | B<NOTE:> While the notation of this construct is the same as the similar | |
707 | function in .NET regexes, the behavior is not, in Perl the buffers are | |
708 | numbered sequentially regardless of being named or not. Thus in the | |
709 | pattern | |
710 | ||
711 | /(x)(?<foo>y)(z)/ | |
712 | ||
713 | $+{foo} will be the same as $2, and $3 will contain 'z' instead of | |
714 | the opposite which is what a .NET regex hacker might expect. | |
715 | ||
716 | Currently NAME is restricted to word chars only. In other words, it | |
717 | must match C</^\w+$/>. | |
718 | ||
719 | =item C<< \k<name> >> | |
720 | ||
721 | =item C<< \k'name' >> | |
722 | ||
723 | Named backreference. Similar to numeric backreferences, except that | |
724 | the group is designated by name and not number. If multiple groups | |
725 | have the same name then it refers to the leftmost defined group in | |
726 | the current match. | |
727 | ||
728 | It is an error to refer to a name not defined by a C<(?<NAME>)> | |
729 | earlier in the pattern. | |
730 | ||
731 | Both forms are equivalent. | |
732 | ||
cc6b7395 | 733 | =item C<(?{ code })> |
d74e8afc | 734 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 | 735 | |
19799a22 | 736 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
737 | experimental, and may be changed without notice. Code executed that |
738 | has side effects may not perform identically from version to version | |
739 | due to the effect of future optimisations in the regex engine. | |
c277df42 | 740 | |
cc46d5f2 | 741 | This zero-width assertion evaluates any embedded Perl code. It |
19799a22 GS |
742 | always succeeds, and its C<code> is not interpolated. Currently, |
743 | the rules to determine where the C<code> ends are somewhat convoluted. | |
744 | ||
77ea4f6d JV |
745 | This feature can be used together with the special variable C<$^N> to |
746 | capture the results of submatches in variables without having to keep | |
747 | track of the number of nested parentheses. For example: | |
748 | ||
749 | $_ = "The brown fox jumps over the lazy dog"; | |
750 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
751 | print "color = $color, animal = $animal\n"; | |
752 | ||
754091cb RGS |
753 | Inside the C<(?{...})> block, C<$_> refers to the string the regular |
754 | expression is matching against. You can also use C<pos()> to know what is | |
fa11829f | 755 | the current position of matching within this string. |
754091cb | 756 | |
19799a22 GS |
757 | The C<code> is properly scoped in the following sense: If the assertion |
758 | is backtracked (compare L<"Backtracking">), all changes introduced after | |
759 | C<local>ization are undone, so that | |
b9ac3b5b GS |
760 | |
761 | $_ = 'a' x 8; | |
762 | m< | |
763 | (?{ $cnt = 0 }) # Initialize $cnt. | |
764 | ( | |
765 | a | |
766 | (?{ | |
767 | local $cnt = $cnt + 1; # Update $cnt, backtracking-safe. | |
768 | }) | |
769 | )* | |
770 | aaaa | |
771 | (?{ $res = $cnt }) # On success copy to non-localized | |
772 | # location. | |
773 | >x; | |
774 | ||
19799a22 | 775 | will set C<$res = 4>. Note that after the match, $cnt returns to the globally |
14218588 | 776 | introduced value, because the scopes that restrict C<local> operators |
b9ac3b5b GS |
777 | are unwound. |
778 | ||
19799a22 GS |
779 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> |
780 | switch. If I<not> used in this way, the result of evaluation of | |
781 | C<code> is put into the special variable C<$^R>. This happens | |
782 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions | |
783 | inside the same regular expression. | |
b9ac3b5b | 784 | |
19799a22 GS |
785 | The assignment to C<$^R> above is properly localized, so the old |
786 | value of C<$^R> is restored if the assertion is backtracked; compare | |
787 | L<"Backtracking">. | |
b9ac3b5b | 788 | |
61528107 SP |
789 | Due to an unfortunate implementation issue, the Perl code contained in these |
790 | blocks is treated as a compile time closure that can have seemingly bizarre | |
6bda09f9 | 791 | consequences when used with lexically scoped variables inside of subroutines |
61528107 SP |
792 | or loops. There are various workarounds for this, including simply using |
793 | global variables instead. If you are using this construct and strange results | |
6bda09f9 YO |
794 | occur then check for the use of lexically scoped variables. |
795 | ||
19799a22 GS |
796 | For reasons of security, this construct is forbidden if the regular |
797 | expression involves run-time interpolation of variables, unless the | |
798 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
799 | variables contain results of C<qr//> operator (see | |
800 | L<perlop/"qr/STRING/imosx">). | |
871b0233 | 801 | |
14218588 | 802 | This restriction is because of the wide-spread and remarkably convenient |
19799a22 | 803 | custom of using run-time determined strings as patterns. For example: |
871b0233 IZ |
804 | |
805 | $re = <>; | |
806 | chomp $re; | |
807 | $string =~ /$re/; | |
808 | ||
14218588 GS |
809 | Before Perl knew how to execute interpolated code within a pattern, |
810 | this operation was completely safe from a security point of view, | |
811 | although it could raise an exception from an illegal pattern. If | |
812 | you turn on the C<use re 'eval'>, though, it is no longer secure, | |
813 | so you should only do so if you are also using taint checking. | |
814 | Better yet, use the carefully constrained evaluation within a Safe | |
cc46d5f2 | 815 | compartment. See L<perlsec> for details about both these mechanisms. |
871b0233 | 816 | |
8988a1bb DD |
817 | Because perl's regex engine is not currently re-entrant, interpolated |
818 | code may not invoke the regex engine either directly with C<m//> or C<s///>), | |
819 | or indirectly with functions such as C<split>. | |
820 | ||
14455d6c | 821 | =item C<(??{ code })> |
d74e8afc ITB |
822 | X<(??{})> |
823 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> | |
0f5d15d6 | 824 | |
19799a22 | 825 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
826 | experimental, and may be changed without notice. Code executed that |
827 | has side effects may not perform identically from version to version | |
828 | due to the effect of future optimisations in the regex engine. | |
0f5d15d6 | 829 | |
19799a22 GS |
830 | This is a "postponed" regular subexpression. The C<code> is evaluated |
831 | at run time, at the moment this subexpression may match. The result | |
832 | of evaluation is considered as a regular expression and matched as | |
61528107 | 833 | if it were inserted instead of this construct. Note that this means |
6bda09f9 YO |
834 | that the contents of capture buffers defined inside an eval'ed pattern |
835 | are not available outside of the pattern, and vice versa, there is no | |
836 | way for the inner pattern to refer to a capture buffer defined outside. | |
837 | Thus, | |
838 | ||
839 | ('a' x 100)=~/(??{'(.)' x 100})/ | |
840 | ||
81714fb9 | 841 | B<will> match, it will B<not> set $1. |
0f5d15d6 | 842 | |
428594d9 | 843 | The C<code> is not interpolated. As before, the rules to determine |
19799a22 GS |
844 | where the C<code> ends are currently somewhat convoluted. |
845 | ||
846 | The following pattern matches a parenthesized group: | |
0f5d15d6 IZ |
847 | |
848 | $re = qr{ | |
849 | \( | |
850 | (?: | |
851 | (?> [^()]+ ) # Non-parens without backtracking | |
852 | | | |
14455d6c | 853 | (??{ $re }) # Group with matching parens |
0f5d15d6 IZ |
854 | )* |
855 | \) | |
856 | }x; | |
857 | ||
6bda09f9 YO |
858 | See also C<(?PARNO)> for a different, more efficient way to accomplish |
859 | the same task. | |
860 | ||
8988a1bb DD |
861 | Because perl's regex engine is not currently re-entrant, delayed |
862 | code may not invoke the regex engine either directly with C<m//> or C<s///>), | |
863 | or indirectly with functions such as C<split>. | |
864 | ||
6bda09f9 | 865 | Recursing deeper than 50 times without consuming any input string will |
61528107 | 866 | result in a fatal error. The maximum depth is compiled into perl, so |
6bda09f9 YO |
867 | changing it requires a custom build. |
868 | ||
894be9b7 YO |
869 | =item C<(?PARNO)> C<(?R)> C<(?0)> |
870 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> | |
6bda09f9 YO |
871 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
872 | ||
81714fb9 YO |
873 | Similar to C<(??{ code })> except it does not involve compiling any code, |
874 | instead it treats the contents of a capture buffer as an independent | |
61528107 | 875 | pattern that must match at the current position. Capture buffers |
81714fb9 | 876 | contained by the pattern will have the value as determined by the |
6bda09f9 YO |
877 | outermost recursion. |
878 | ||
894be9b7 YO |
879 | PARNO is a sequence of digits (not starting with 0) whose value reflects |
880 | the paren-number of the capture buffer to recurse to. C<(?R)> recurses to | |
881 | the beginning of the whole pattern. C<(?0)> is an alternate syntax for | |
882 | C<(?R)>. | |
6bda09f9 | 883 | |
81714fb9 | 884 | The following pattern matches a function foo() which may contain |
f145b7e9 | 885 | balanced parentheses as the argument. |
6bda09f9 YO |
886 | |
887 | $re = qr{ ( # paren group 1 (full function) | |
81714fb9 | 888 | foo |
6bda09f9 YO |
889 | ( # paren group 2 (parens) |
890 | \( | |
891 | ( # paren group 3 (contents of parens) | |
892 | (?: | |
893 | (?> [^()]+ ) # Non-parens without backtracking | |
894 | | | |
895 | (?2) # Recurse to start of paren group 2 | |
896 | )* | |
897 | ) | |
898 | \) | |
899 | ) | |
900 | ) | |
901 | }x; | |
902 | ||
903 | If the pattern was used as follows | |
904 | ||
905 | 'foo(bar(baz)+baz(bop))'=~/$re/ | |
906 | and print "\$1 = $1\n", | |
907 | "\$2 = $2\n", | |
908 | "\$3 = $3\n"; | |
909 | ||
910 | the output produced should be the following: | |
911 | ||
912 | $1 = foo(bar(baz)+baz(bop)) | |
913 | $2 = (bar(baz)+baz(bop)) | |
81714fb9 | 914 | $3 = bar(baz)+baz(bop) |
6bda09f9 | 915 | |
81714fb9 | 916 | If there is no corresponding capture buffer defined, then it is a |
61528107 | 917 | fatal error. Recursing deeper than 50 times without consuming any input |
81714fb9 | 918 | string will also result in a fatal error. The maximum depth is compiled |
6bda09f9 YO |
919 | into perl, so changing it requires a custom build. |
920 | ||
81714fb9 | 921 | B<Note> that this pattern does not behave the same way as the equivalent |
6bda09f9 YO |
922 | PCRE or Python construct of the same form. In perl you can backtrack into |
923 | a recursed group, in PCRE and Python the recursed into group is treated | |
81714fb9 YO |
924 | as atomic. Also, constructs like (?i:(?1)) or (?:(?i)(?1)) do not affect |
925 | the pattern being recursed into. | |
6bda09f9 | 926 | |
894be9b7 YO |
927 | =item C<(?&NAME)> |
928 | X<(?&NAME)> | |
929 | ||
930 | Recurse to a named subpattern. Identical to (?PARNO) except that the | |
931 | parenthesis to recurse to is determined by name. If multiple parens have | |
932 | the same name, then it recurses to the leftmost. | |
933 | ||
934 | It is an error to refer to a name that is not declared somewhere in the | |
935 | pattern. | |
936 | ||
e2e6a0f1 YO |
937 | =item C<(?(condition)yes-pattern|no-pattern)> |
938 | X<(?()> | |
286f584a | 939 | |
e2e6a0f1 | 940 | =item C<(?(condition)yes-pattern)> |
286f584a | 941 | |
e2e6a0f1 YO |
942 | Conditional expression. C<(condition)> should be either an integer in |
943 | parentheses (which is valid if the corresponding pair of parentheses | |
944 | matched), a look-ahead/look-behind/evaluate zero-width assertion, a | |
945 | name in angle brackets or single quotes (which is valid if a buffer | |
946 | with the given name matched), or the special symbol (R) (true when | |
947 | evaluated inside of recursion or eval). Additionally the R may be | |
948 | followed by a number, (which will be true when evaluated when recursing | |
949 | inside of the appropriate group), or by C<&NAME>, in which case it will | |
950 | be true only when evaluated during recursion in the named group. | |
951 | ||
952 | Here's a summary of the possible predicates: | |
953 | ||
954 | =over 4 | |
955 | ||
956 | =item (1) (2) ... | |
957 | ||
958 | Checks if the numbered capturing buffer has matched something. | |
959 | ||
960 | =item (<NAME>) ('NAME') | |
961 | ||
962 | Checks if a buffer with the given name has matched something. | |
963 | ||
964 | =item (?{ CODE }) | |
965 | ||
966 | Treats the code block as the condition. | |
967 | ||
968 | =item (R) | |
969 | ||
970 | Checks if the expression has been evaluated inside of recursion. | |
971 | ||
972 | =item (R1) (R2) ... | |
973 | ||
974 | Checks if the expression has been evaluated while executing directly | |
975 | inside of the n-th capture group. This check is the regex equivalent of | |
976 | ||
977 | if ((caller(0))[3] eq 'subname') { ... } | |
978 | ||
979 | In other words, it does not check the full recursion stack. | |
980 | ||
981 | =item (R&NAME) | |
982 | ||
983 | Similar to C<(R1)>, this predicate checks to see if we're executing | |
984 | directly inside of the leftmost group with a given name (this is the same | |
985 | logic used by C<(?&NAME)> to disambiguate). It does not check the full | |
986 | stack, but only the name of the innermost active recursion. | |
987 | ||
988 | =item (DEFINE) | |
989 | ||
990 | In this case, the yes-pattern is never directly executed, and no | |
991 | no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient. | |
992 | See below for details. | |
993 | ||
994 | =back | |
995 | ||
996 | For example: | |
997 | ||
998 | m{ ( \( )? | |
999 | [^()]+ | |
1000 | (?(1) \) ) | |
1001 | }x | |
1002 | ||
1003 | matches a chunk of non-parentheses, possibly included in parentheses | |
1004 | themselves. | |
1005 | ||
1006 | A special form is the C<(DEFINE)> predicate, which never executes directly | |
1007 | its yes-pattern, and does not allow a no-pattern. This allows to define | |
1008 | subpatterns which will be executed only by using the recursion mechanism. | |
1009 | This way, you can define a set of regular expression rules that can be | |
1010 | bundled into any pattern you choose. | |
1011 | ||
1012 | It is recommended that for this usage you put the DEFINE block at the | |
1013 | end of the pattern, and that you name any subpatterns defined within it. | |
1014 | ||
1015 | Also, it's worth noting that patterns defined this way probably will | |
1016 | not be as efficient, as the optimiser is not very clever about | |
1017 | handling them. | |
1018 | ||
1019 | An example of how this might be used is as follows: | |
1020 | ||
1021 | /(?<NAME>(&NAME_PAT))(?<ADDR>(&ADDRESS_PAT)) | |
1022 | (?(DEFINE) | |
1023 | (<NAME_PAT>....) | |
1024 | (<ADRESS_PAT>....) | |
1025 | )/x | |
1026 | ||
1027 | Note that capture buffers matched inside of recursion are not accessible | |
1028 | after the recursion returns, so the extra layer of capturing buffers are | |
1029 | necessary. Thus C<$+{NAME_PAT}> would not be defined even though | |
1030 | C<$+{NAME}> would be. | |
286f584a | 1031 | |
c47ff5f1 | 1032 | =item C<< (?>pattern) >> |
6bda09f9 | 1033 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 | 1034 | |
19799a22 GS |
1035 | An "independent" subexpression, one which matches the substring |
1036 | that a I<standalone> C<pattern> would match if anchored at the given | |
9da458fc | 1037 | position, and it matches I<nothing other than this substring>. This |
19799a22 GS |
1038 | construct is useful for optimizations of what would otherwise be |
1039 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). | |
9da458fc IZ |
1040 | It may also be useful in places where the "grab all you can, and do not |
1041 | give anything back" semantic is desirable. | |
19799a22 | 1042 | |
c47ff5f1 | 1043 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 GS |
1044 | (anchored at the beginning of string, as above) will match I<all> |
1045 | characters C<a> at the beginning of string, leaving no C<a> for | |
1046 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, | |
1047 | since the match of the subgroup C<a*> is influenced by the following | |
1048 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside | |
1049 | C<a*ab> will match fewer characters than a standalone C<a*>, since | |
1050 | this makes the tail match. | |
1051 | ||
c47ff5f1 | 1052 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
19799a22 GS |
1053 | C<(?=(pattern))\1>. This matches the same substring as a standalone |
1054 | C<a+>, and the following C<\1> eats the matched string; it therefore | |
c47ff5f1 | 1055 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 GS |
1056 | (The difference between these two constructs is that the second one |
1057 | uses a capturing group, thus shifting ordinals of backreferences | |
1058 | in the rest of a regular expression.) | |
1059 | ||
1060 | Consider this pattern: | |
c277df42 | 1061 | |
871b0233 | 1062 | m{ \( |
e2e6a0f1 YO |
1063 | ( |
1064 | [^()]+ # x+ | |
1065 | | | |
871b0233 IZ |
1066 | \( [^()]* \) |
1067 | )+ | |
e2e6a0f1 | 1068 | \) |
871b0233 | 1069 | }x |
5a964f20 | 1070 | |
19799a22 GS |
1071 | That will efficiently match a nonempty group with matching parentheses |
1072 | two levels deep or less. However, if there is no such group, it | |
1073 | will take virtually forever on a long string. That's because there | |
1074 | are so many different ways to split a long string into several | |
1075 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
1076 | to a subpattern of the above pattern. Consider how the pattern | |
1077 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
1078 | seconds, but that each extra letter doubles this time. This | |
1079 | exponential performance will make it appear that your program has | |
14218588 | 1080 | hung. However, a tiny change to this pattern |
5a964f20 | 1081 | |
e2e6a0f1 YO |
1082 | m{ \( |
1083 | ( | |
1084 | (?> [^()]+ ) # change x+ above to (?> x+ ) | |
1085 | | | |
871b0233 IZ |
1086 | \( [^()]* \) |
1087 | )+ | |
e2e6a0f1 | 1088 | \) |
871b0233 | 1089 | }x |
c277df42 | 1090 | |
c47ff5f1 | 1091 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 TC |
1092 | this yourself would be a productive exercise), but finishes in a fourth |
1093 | the time when used on a similar string with 1000000 C<a>s. Be aware, | |
1094 | however, that this pattern currently triggers a warning message under | |
9f1b1f2d | 1095 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b | 1096 | C<"matches null string many times in regex">. |
c277df42 | 1097 | |
c47ff5f1 | 1098 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 | 1099 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 IZ |
1100 | This was only 4 times slower on a string with 1000000 C<a>s. |
1101 | ||
9da458fc IZ |
1102 | The "grab all you can, and do not give anything back" semantic is desirable |
1103 | in many situations where on the first sight a simple C<()*> looks like | |
1104 | the correct solution. Suppose we parse text with comments being delimited | |
1105 | by C<#> followed by some optional (horizontal) whitespace. Contrary to | |
4375e838 | 1106 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc IZ |
1107 | the comment delimiter, because it may "give up" some whitespace if |
1108 | the remainder of the pattern can be made to match that way. The correct | |
1109 | answer is either one of these: | |
1110 | ||
1111 | (?>#[ \t]*) | |
1112 | #[ \t]*(?![ \t]) | |
1113 | ||
1114 | For example, to grab non-empty comments into $1, one should use either | |
1115 | one of these: | |
1116 | ||
1117 | / (?> \# [ \t]* ) ( .+ ) /x; | |
1118 | / \# [ \t]* ( [^ \t] .* ) /x; | |
1119 | ||
1120 | Which one you pick depends on which of these expressions better reflects | |
1121 | the above specification of comments. | |
1122 | ||
6bda09f9 YO |
1123 | In some literature this construct is called "atomic matching" or |
1124 | "possessive matching". | |
1125 | ||
b9b4dddf YO |
1126 | Possessive quantifiers are equivalent to putting the item they are applied |
1127 | to inside of one of these constructs. The following equivalences apply: | |
1128 | ||
1129 | Quantifier Form Bracketing Form | |
1130 | --------------- --------------- | |
1131 | PAT*+ (?>PAT*) | |
1132 | PAT++ (?>PAT+) | |
1133 | PAT?+ (?>PAT?) | |
1134 | PAT{min,max}+ (?>PAT{min,max}) | |
1135 | ||
e2e6a0f1 YO |
1136 | =back |
1137 | ||
1138 | =head2 Special Backtracking Control Verbs | |
1139 | ||
1140 | B<WARNING:> These patterns are experimental and subject to change or | |
1141 | removal in a future version of perl. Their usage in production code should | |
1142 | be noted to avoid problems during upgrades. | |
1143 | ||
1144 | These special patterns are generally of the form C<(*VERB:ARG)>. Unless | |
1145 | otherwise stated the ARG argument is optional; in some cases, it is | |
1146 | forbidden. | |
1147 | ||
1148 | Any pattern containing a special backtracking verb that allows an argument | |
1149 | has the special behaviour that when executed it sets the current packages' | |
1150 | C<$REGERROR> variable. In this case, the following rules apply: | |
1151 | ||
1152 | On failure, this variable will be set to the ARG value of the verb | |
1153 | pattern, if the verb was involved in the failure of the match. If the ARG | |
1154 | part of the pattern was omitted, then C<$REGERROR> will be set to TRUE. | |
1155 | ||
1156 | On a successful match this variable will be set to FALSE. | |
1157 | ||
1158 | B<NOTE:> C<$REGERROR> is not a magic variable in the same sense than | |
1159 | C<$1> and most other regex related variables. It is not local to a | |
1160 | scope, nor readonly but instead a volatile package variable similar to | |
1161 | C<$AUTOLOAD>. Use C<local> to localize changes to it to a specific scope | |
1162 | if necessary. | |
1163 | ||
1164 | If a pattern does not contain a special backtracking verb that allows an | |
1165 | argument, then C<$REGERROR> is not touched at all. | |
1166 | ||
1167 | =over 4 | |
1168 | ||
1169 | =item Verbs that take an argument | |
1170 | ||
1171 | =over 4 | |
1172 | ||
1173 | =item C<(*NOMATCH)> C<(*NOMATCH:NAME)> | |
1174 | X<(*NOMATCH)> X<(*NOMATCH:NAME)> | |
54612592 YO |
1175 | |
1176 | This zero-width pattern commits the match at the current point, preventing | |
e2e6a0f1 YO |
1177 | the engine from backtracking on failure to the left of the this point. |
1178 | Consider the pattern C<A (*NOMATCH) B>, where A and B are complex patterns. | |
1179 | Until the C<(*NOMATCH)> is reached, A may backtrack as necessary to match. | |
54612592 YO |
1180 | Once it is reached, matching continues in B, which may also backtrack as |
1181 | necessary; however, should B not match, then no further backtracking will | |
1182 | take place, and the pattern will fail outright at that starting position. | |
1183 | ||
1184 | The following example counts all the possible matching strings in a | |
1185 | pattern (without actually matching any of them). | |
1186 | ||
e2e6a0f1 | 1187 | 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1188 | print "Count=$count\n"; |
1189 | ||
1190 | which produces: | |
1191 | ||
1192 | aaab | |
1193 | aaa | |
1194 | aa | |
1195 | a | |
1196 | aab | |
1197 | aa | |
1198 | a | |
1199 | ab | |
1200 | a | |
1201 | Count=9 | |
1202 | ||
e2e6a0f1 | 1203 | If we add a C<(*NOMATCH)> before the count like the following |
54612592 | 1204 | |
e2e6a0f1 | 1205 | 'aaab' =~ /a+b?(*NOMATCH)(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1206 | print "Count=$count\n"; |
1207 | ||
1208 | we prevent backtracking and find the count of the longest matching | |
1209 | at each matching startpoint like so: | |
1210 | ||
1211 | aaab | |
1212 | aab | |
1213 | ab | |
1214 | Count=3 | |
1215 | ||
e2e6a0f1 | 1216 | Any number of C<(*NOMATCH)> assertions may be used in a pattern. |
54612592 YO |
1217 | |
1218 | See also C<< (?>pattern) >> and possessive quantifiers for other | |
1219 | ways to control backtracking. | |
1220 | ||
e2e6a0f1 YO |
1221 | =item C<(*MARK)> C<(*MARK:NAME)> |
1222 | X<(*MARK)> | |
1223 | ||
1224 | This zero-width pattern can be used to mark the point in a string | |
1225 | reached when a certain part of the pattern has been successfully | |
1226 | matched. This mark may be given a name. A later C<(*CUT)> pattern | |
1227 | will then cut at that point if backtracked into on failure. Any | |
1228 | number of (*MARK) patterns are allowed, and the NAME portion is | |
1229 | optional and may be duplicated. | |
1230 | ||
1231 | See C<*CUT> for more detail. | |
1232 | ||
1233 | =item C<(*CUT)> C<(*CUT:NAME)> | |
1234 | X<(*CUT)> | |
1235 | ||
1236 | This zero-width pattern is similar to C<(*NOMATCH)>, except that on | |
1237 | failure it also signifies that whatever text that was matched leading up | |
1238 | to the C<(*CUT)> pattern being executed cannot be part of a match, I<even | |
1239 | if started from a later point>. This effectively means that the regex | |
1240 | engine moves forward to this position on failure and tries to match | |
1241 | again, (assuming that there is sufficient room to match). | |
1242 | ||
1243 | The name of the C<(*CUT:NAME)> pattern has special significance. If a | |
1244 | C<(*MARK:NAME)> was encountered while matching, then it is the position | |
1245 | where that pattern was executed that is used for the "cut point" in the | |
1246 | string. If no mark of that name was encountered, then the cut is done at | |
1247 | the point where the C<(*CUT)> was. Similarly if no NAME is specified in | |
1248 | the C<(*CUT)>, and if a C<(*MARK)> with any name (or none) is encountered, | |
1249 | then that C<(*MARK)>'s cursor point will be used. If the C<(*CUT)> is not | |
1250 | preceded by a C<(*MARK)>, then the cut point is where the string was when | |
1251 | the C<(*CUT)> was encountered. | |
1252 | ||
1253 | Compare the following to the examples in C<(*NOMATCH)>, note the string | |
24b23f37 YO |
1254 | is twice as long: |
1255 | ||
e2e6a0f1 | 1256 | 'aaabaaab' =~ /a+b?(*CUT)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 YO |
1257 | print "Count=$count\n"; |
1258 | ||
1259 | outputs | |
1260 | ||
1261 | aaab | |
1262 | aaab | |
1263 | Count=2 | |
1264 | ||
e2e6a0f1 YO |
1265 | Once the 'aaab' at the start of the string has matched, and the C<(*CUT)> |
1266 | executed, the next startpoint will be where the cursor was when the | |
1267 | C<(*CUT)> was executed. | |
24b23f37 | 1268 | |
e2e6a0f1 YO |
1269 | =item C<(*COMMIT)> |
1270 | X<(*COMMIT)> | |
24b23f37 | 1271 | |
e2e6a0f1 | 1272 | This zero-width pattern is similar to C<(*CUT)> except that it causes |
24b23f37 YO |
1273 | the match to fail outright. No attempts to match will occur again. |
1274 | ||
e2e6a0f1 | 1275 | 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 YO |
1276 | print "Count=$count\n"; |
1277 | ||
1278 | outputs | |
1279 | ||
1280 | aaab | |
1281 | Count=1 | |
1282 | ||
e2e6a0f1 YO |
1283 | In other words, once the C<(*COMMIT)> has been entered, and if the pattern |
1284 | does not match, the regex engine will not try any further matching on the | |
1285 | rest of the string. | |
c277df42 | 1286 | |
e2e6a0f1 | 1287 | =back |
9af228c6 | 1288 | |
e2e6a0f1 | 1289 | =item Verbs without an argument |
9af228c6 YO |
1290 | |
1291 | =over 4 | |
1292 | ||
e2e6a0f1 YO |
1293 | =item C<(*FAIL)> C<(*F)> |
1294 | X<(*FAIL)> X<(*F)> | |
9af228c6 | 1295 | |
e2e6a0f1 YO |
1296 | This pattern matches nothing and always fails. It can be used to force the |
1297 | engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In | |
1298 | fact, C<(?!)> gets optimised into C<(*FAIL)> internally. | |
9af228c6 | 1299 | |
e2e6a0f1 | 1300 | It is probably useful only when combined with C<(?{})> or C<(??{})>. |
9af228c6 | 1301 | |
e2e6a0f1 YO |
1302 | =item C<(*ACCEPT)> |
1303 | X<(*ACCEPT)> | |
9af228c6 | 1304 | |
e2e6a0f1 YO |
1305 | B<WARNING:> This feature is highly experimental. It is not recommended |
1306 | for production code. | |
9af228c6 | 1307 | |
e2e6a0f1 YO |
1308 | This pattern matches nothing and causes the end of successful matching at |
1309 | the point at which the C<(*ACCEPT)> pattern was encountered, regardless of | |
1310 | whether there is actually more to match in the string. When inside of a | |
1311 | nested pattern, such as recursion or a dynamically generated subbpattern | |
1312 | via C<(??{})>, only the innermost pattern is ended immediately. | |
9af228c6 | 1313 | |
e2e6a0f1 YO |
1314 | If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are |
1315 | marked as ended at the point at which the C<(*ACCEPT)> was encountered. | |
1316 | For instance: | |
9af228c6 | 1317 | |
e2e6a0f1 | 1318 | 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x; |
9af228c6 | 1319 | |
e2e6a0f1 YO |
1320 | will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not |
1321 | be set. If another branch in the inner parens were matched, such as in the | |
1322 | string 'ACDE', then the C<D> and C<E> would have to be matched as well. | |
9af228c6 YO |
1323 | |
1324 | =back | |
c277df42 | 1325 | |
a0d0e21e LW |
1326 | =back |
1327 | ||
c07a80fd | 1328 | =head2 Backtracking |
d74e8afc | 1329 | X<backtrack> X<backtracking> |
c07a80fd | 1330 | |
35a734be IZ |
1331 | NOTE: This section presents an abstract approximation of regular |
1332 | expression behavior. For a more rigorous (and complicated) view of | |
1333 | the rules involved in selecting a match among possible alternatives, | |
1334 | see L<Combining pieces together>. | |
1335 | ||
c277df42 | 1336 | A fundamental feature of regular expression matching involves the |
5a964f20 | 1337 | notion called I<backtracking>, which is currently used (when needed) |
c277df42 | 1338 | by all regular expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc IZ |
1339 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
1340 | internally, but the general principle outlined here is valid. | |
c07a80fd | 1341 | |
1342 | For a regular expression to match, the I<entire> regular expression must | |
1343 | match, not just part of it. So if the beginning of a pattern containing a | |
1344 | quantifier succeeds in a way that causes later parts in the pattern to | |
1345 | fail, the matching engine backs up and recalculates the beginning | |
1346 | part--that's why it's called backtracking. | |
1347 | ||
1348 | Here is an example of backtracking: Let's say you want to find the | |
1349 | word following "foo" in the string "Food is on the foo table.": | |
1350 | ||
1351 | $_ = "Food is on the foo table."; | |
1352 | if ( /\b(foo)\s+(\w+)/i ) { | |
1353 | print "$2 follows $1.\n"; | |
1354 | } | |
1355 | ||
1356 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
1357 | finds a possible match right at the beginning of the string, and loads up | |
1358 | $1 with "Foo". However, as soon as the matching engine sees that there's | |
1359 | no whitespace following the "Foo" that it had saved in $1, it realizes its | |
68dc0745 | 1360 | mistake and starts over again one character after where it had the |
c07a80fd | 1361 | tentative match. This time it goes all the way until the next occurrence |
1362 | of "foo". The complete regular expression matches this time, and you get | |
1363 | the expected output of "table follows foo." | |
1364 | ||
1365 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
1366 | everything between "foo" and "bar". Initially, you write something | |
1367 | like this: | |
1368 | ||
1369 | $_ = "The food is under the bar in the barn."; | |
1370 | if ( /foo(.*)bar/ ) { | |
1371 | print "got <$1>\n"; | |
1372 | } | |
1373 | ||
1374 | Which perhaps unexpectedly yields: | |
1375 | ||
1376 | got <d is under the bar in the > | |
1377 | ||
1378 | That's because C<.*> was greedy, so you get everything between the | |
14218588 | 1379 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd | 1380 | to use minimal matching to make sure you get the text between a "foo" |
1381 | and the first "bar" thereafter. | |
1382 | ||
1383 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
1384 | got <d is under the > | |
1385 | ||
1386 | Here's another example: let's say you'd like to match a number at the end | |
b6e13d97 | 1387 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd | 1388 | So you write this: |
1389 | ||
1390 | $_ = "I have 2 numbers: 53147"; | |
1391 | if ( /(.*)(\d*)/ ) { # Wrong! | |
1392 | print "Beginning is <$1>, number is <$2>.\n"; | |
1393 | } | |
1394 | ||
1395 | That won't work at all, because C<.*> was greedy and gobbled up the | |
1396 | whole string. As C<\d*> can match on an empty string the complete | |
1397 | regular expression matched successfully. | |
1398 | ||
8e1088bc | 1399 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd | 1400 | |
1401 | Here are some variants, most of which don't work: | |
1402 | ||
1403 | $_ = "I have 2 numbers: 53147"; | |
1404 | @pats = qw{ | |
1405 | (.*)(\d*) | |
1406 | (.*)(\d+) | |
1407 | (.*?)(\d*) | |
1408 | (.*?)(\d+) | |
1409 | (.*)(\d+)$ | |
1410 | (.*?)(\d+)$ | |
1411 | (.*)\b(\d+)$ | |
1412 | (.*\D)(\d+)$ | |
1413 | }; | |
1414 | ||
1415 | for $pat (@pats) { | |
1416 | printf "%-12s ", $pat; | |
1417 | if ( /$pat/ ) { | |
1418 | print "<$1> <$2>\n"; | |
1419 | } else { | |
1420 | print "FAIL\n"; | |
1421 | } | |
1422 | } | |
1423 | ||
1424 | That will print out: | |
1425 | ||
1426 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
1427 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
1428 | (.*?)(\d*) <> <> | |
1429 | (.*?)(\d+) <I have > <2> | |
1430 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
1431 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
1432 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
1433 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
1434 | ||
1435 | As you see, this can be a bit tricky. It's important to realize that a | |
1436 | regular expression is merely a set of assertions that gives a definition | |
1437 | of success. There may be 0, 1, or several different ways that the | |
1438 | definition might succeed against a particular string. And if there are | |
5a964f20 TC |
1439 | multiple ways it might succeed, you need to understand backtracking to |
1440 | know which variety of success you will achieve. | |
c07a80fd | 1441 | |
19799a22 | 1442 | When using look-ahead assertions and negations, this can all get even |
8b19b778 | 1443 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd | 1444 | followed by "123". You might try to write that as |
1445 | ||
871b0233 IZ |
1446 | $_ = "ABC123"; |
1447 | if ( /^\D*(?!123)/ ) { # Wrong! | |
1448 | print "Yup, no 123 in $_\n"; | |
1449 | } | |
c07a80fd | 1450 | |
1451 | But that isn't going to match; at least, not the way you're hoping. It | |
1452 | claims that there is no 123 in the string. Here's a clearer picture of | |
9b9391b2 | 1453 | why that pattern matches, contrary to popular expectations: |
c07a80fd | 1454 | |
4358a253 SS |
1455 | $x = 'ABC123'; |
1456 | $y = 'ABC445'; | |
c07a80fd | 1457 | |
4358a253 SS |
1458 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
1459 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; | |
c07a80fd | 1460 | |
4358a253 SS |
1461 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; |
1462 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; | |
c07a80fd | 1463 | |
1464 | This prints | |
1465 | ||
1466 | 2: got ABC | |
1467 | 3: got AB | |
1468 | 4: got ABC | |
1469 | ||
5f05dabc | 1470 | You might have expected test 3 to fail because it seems to a more |
c07a80fd | 1471 | general purpose version of test 1. The important difference between |
1472 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
1473 | backtracking, whereas test 1 will not. What's happening is | |
1474 | that you've asked "Is it true that at the start of $x, following 0 or more | |
5f05dabc | 1475 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd | 1476 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 | 1477 | fail. |
14218588 | 1478 | |
c07a80fd | 1479 | The search engine will initially match C<\D*> with "ABC". Then it will |
14218588 | 1480 | try to match C<(?!123> with "123", which fails. But because |
c07a80fd | 1481 | a quantifier (C<\D*>) has been used in the regular expression, the |
1482 | search engine can backtrack and retry the match differently | |
54310121 | 1483 | in the hope of matching the complete regular expression. |
c07a80fd | 1484 | |
5a964f20 TC |
1485 | The pattern really, I<really> wants to succeed, so it uses the |
1486 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
c07a80fd | 1487 | time. Now there's indeed something following "AB" that is not |
14218588 | 1488 | "123". It's "C123", which suffices. |
c07a80fd | 1489 | |
14218588 GS |
1490 | We can deal with this by using both an assertion and a negation. |
1491 | We'll say that the first part in $1 must be followed both by a digit | |
1492 | and by something that's not "123". Remember that the look-aheads | |
1493 | are zero-width expressions--they only look, but don't consume any | |
1494 | of the string in their match. So rewriting this way produces what | |
c07a80fd | 1495 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
1496 | ||
4358a253 SS |
1497 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; |
1498 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; | |
c07a80fd | 1499 | |
1500 | 6: got ABC | |
1501 | ||
5a964f20 | 1502 | In other words, the two zero-width assertions next to each other work as though |
19799a22 | 1503 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd | 1504 | matches only if you're at the beginning of the line AND the end of the |
1505 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
1506 | regular expressions always means AND, except when you write an explicit OR | |
1507 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
1508 | although the attempted matches are made at different positions because "a" | |
1509 | is not a zero-width assertion, but a one-width assertion. | |
1510 | ||
19799a22 | 1511 | B<WARNING>: particularly complicated regular expressions can take |
14218588 | 1512 | exponential time to solve because of the immense number of possible |
9da458fc IZ |
1513 | ways they can use backtracking to try match. For example, without |
1514 | internal optimizations done by the regular expression engine, this will | |
1515 | take a painfully long time to run: | |
c07a80fd | 1516 | |
e1901655 IZ |
1517 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
1518 | ||
1519 | And if you used C<*>'s in the internal groups instead of limiting them | |
1520 | to 0 through 5 matches, then it would take forever--or until you ran | |
1521 | out of stack space. Moreover, these internal optimizations are not | |
1522 | always applicable. For example, if you put C<{0,5}> instead of C<*> | |
1523 | on the external group, no current optimization is applicable, and the | |
1524 | match takes a long time to finish. | |
c07a80fd | 1525 | |
9da458fc IZ |
1526 | A powerful tool for optimizing such beasts is what is known as an |
1527 | "independent group", | |
c47ff5f1 | 1528 | which does not backtrack (see L<C<< (?>pattern) >>>). Note also that |
9da458fc | 1529 | zero-length look-ahead/look-behind assertions will not backtrack to make |
14218588 GS |
1530 | the tail match, since they are in "logical" context: only |
1531 | whether they match is considered relevant. For an example | |
9da458fc | 1532 | where side-effects of look-ahead I<might> have influenced the |
c47ff5f1 | 1533 | following match, see L<C<< (?>pattern) >>>. |
c277df42 | 1534 | |
a0d0e21e | 1535 | =head2 Version 8 Regular Expressions |
d74e8afc | 1536 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> |
a0d0e21e | 1537 | |
5a964f20 | 1538 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e LW |
1539 | routines, here are the pattern-matching rules not described above. |
1540 | ||
54310121 | 1541 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e | 1542 | with a special meaning described here or above. You can cause |
5a964f20 | 1543 | characters that normally function as metacharacters to be interpreted |
5f05dabc | 1544 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
a0d0e21e LW |
1545 | character; "\\" matches a "\"). A series of characters matches that |
1546 | series of characters in the target string, so the pattern C<blurfl> | |
1547 | would match "blurfl" in the target string. | |
1548 | ||
1549 | You can specify a character class, by enclosing a list of characters | |
5a964f20 | 1550 | in C<[]>, which will match any one character from the list. If the |
a0d0e21e | 1551 | first character after the "[" is "^", the class matches any character not |
14218588 | 1552 | in the list. Within a list, the "-" character specifies a |
5a964f20 | 1553 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 GS |
1554 | inclusive. If you want either "-" or "]" itself to be a member of a |
1555 | class, put it at the start of the list (possibly after a "^"), or | |
1556 | escape it with a backslash. "-" is also taken literally when it is | |
1557 | at the end of the list, just before the closing "]". (The | |
84850974 DD |
1558 | following all specify the same class of three characters: C<[-az]>, |
1559 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which | |
fb55449c JH |
1560 | specifies a class containing twenty-six characters, even on EBCDIC |
1561 | based coded character sets.) Also, if you try to use the character | |
1562 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of | |
1563 | a range, that's not a range, the "-" is understood literally. | |
a0d0e21e | 1564 | |
8ada0baa JH |
1565 | Note also that the whole range idea is rather unportable between |
1566 | character sets--and even within character sets they may cause results | |
1567 | you probably didn't expect. A sound principle is to use only ranges | |
1568 | that begin from and end at either alphabets of equal case ([a-e], | |
1569 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, | |
1570 | spell out the character sets in full. | |
1571 | ||
54310121 | 1572 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e LW |
1573 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
1574 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string | |
fb55449c JH |
1575 | of octal digits, matches the character whose coded character set value |
1576 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, | |
1577 | matches the character whose numeric value is I<nn>. The expression \cI<x> | |
1578 | matches the character control-I<x>. Finally, the "." metacharacter | |
1579 | matches any character except "\n" (unless you use C</s>). | |
a0d0e21e LW |
1580 | |
1581 | You can specify a series of alternatives for a pattern using "|" to | |
1582 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", | |
5a964f20 | 1583 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e LW |
1584 | first alternative includes everything from the last pattern delimiter |
1585 | ("(", "[", or the beginning of the pattern) up to the first "|", and | |
1586 | the last alternative contains everything from the last "|" to the next | |
14218588 GS |
1587 | pattern delimiter. That's why it's common practice to include |
1588 | alternatives in parentheses: to minimize confusion about where they | |
a3cb178b GS |
1589 | start and end. |
1590 | ||
5a964f20 | 1591 | Alternatives are tried from left to right, so the first |
a3cb178b GS |
1592 | alternative found for which the entire expression matches, is the one that |
1593 | is chosen. This means that alternatives are not necessarily greedy. For | |
628afcb5 | 1594 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b GS |
1595 | part will match, as that is the first alternative tried, and it successfully |
1596 | matches the target string. (This might not seem important, but it is | |
1597 | important when you are capturing matched text using parentheses.) | |
1598 | ||
5a964f20 | 1599 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b | 1600 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e | 1601 | |
14218588 GS |
1602 | Within a pattern, you may designate subpatterns for later reference |
1603 | by enclosing them in parentheses, and you may refer back to the | |
1604 | I<n>th subpattern later in the pattern using the metacharacter | |
1605 | \I<n>. Subpatterns are numbered based on the left to right order | |
1606 | of their opening parenthesis. A backreference matches whatever | |
1607 | actually matched the subpattern in the string being examined, not | |
1608 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will | |
1609 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern | |
1610 | 1 matched "0x", even though the rule C<0|0x> could potentially match | |
1611 | the leading 0 in the second number. | |
cb1a09d0 | 1612 | |
19799a22 | 1613 | =head2 Warning on \1 vs $1 |
cb1a09d0 | 1614 | |
5a964f20 | 1615 | Some people get too used to writing things like: |
cb1a09d0 AD |
1616 | |
1617 | $pattern =~ s/(\W)/\\\1/g; | |
1618 | ||
1619 | This is grandfathered for the RHS of a substitute to avoid shocking the | |
1620 | B<sed> addicts, but it's a dirty habit to get into. That's because in | |
d1be9408 | 1621 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 AD |
1622 | the usual double-quoted string means a control-A. The customary Unix |
1623 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
1624 | of doing that, you get yourself into trouble if you then add an C</e> | |
1625 | modifier. | |
1626 | ||
5a964f20 | 1627 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 AD |
1628 | |
1629 | Or if you try to do | |
1630 | ||
1631 | s/(\d+)/\1000/; | |
1632 | ||
1633 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
14218588 | 1634 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 AD |
1635 | with the operation of matching a backreference. Certainly they mean two |
1636 | different things on the I<left> side of the C<s///>. | |
9fa51da4 | 1637 | |
c84d73f1 IZ |
1638 | =head2 Repeated patterns matching zero-length substring |
1639 | ||
19799a22 | 1640 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 IZ |
1641 | |
1642 | Regular expressions provide a terse and powerful programming language. As | |
1643 | with most other power tools, power comes together with the ability | |
1644 | to wreak havoc. | |
1645 | ||
1646 | A common abuse of this power stems from the ability to make infinite | |
628afcb5 | 1647 | loops using regular expressions, with something as innocuous as: |
c84d73f1 IZ |
1648 | |
1649 | 'foo' =~ m{ ( o? )* }x; | |
1650 | ||
1651 | The C<o?> can match at the beginning of C<'foo'>, and since the position | |
1652 | in the string is not moved by the match, C<o?> would match again and again | |
14218588 | 1653 | because of the C<*> modifier. Another common way to create a similar cycle |
c84d73f1 IZ |
1654 | is with the looping modifier C<//g>: |
1655 | ||
1656 | @matches = ( 'foo' =~ m{ o? }xg ); | |
1657 | ||
1658 | or | |
1659 | ||
1660 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
1661 | ||
1662 | or the loop implied by split(). | |
1663 | ||
1664 | However, long experience has shown that many programming tasks may | |
14218588 GS |
1665 | be significantly simplified by using repeated subexpressions that |
1666 | may match zero-length substrings. Here's a simple example being: | |
c84d73f1 IZ |
1667 | |
1668 | @chars = split //, $string; # // is not magic in split | |
1669 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / | |
1670 | ||
9da458fc | 1671 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 IZ |
1672 | the infinite loop>. The rules for this are different for lower-level |
1673 | loops given by the greedy modifiers C<*+{}>, and for higher-level | |
1674 | ones like the C</g> modifier or split() operator. | |
1675 | ||
19799a22 GS |
1676 | The lower-level loops are I<interrupted> (that is, the loop is |
1677 | broken) when Perl detects that a repeated expression matched a | |
1678 | zero-length substring. Thus | |
c84d73f1 IZ |
1679 | |
1680 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
1681 | ||
1682 | is made equivalent to | |
1683 | ||
1684 | m{ (?: NON_ZERO_LENGTH )* | |
1685 | | | |
1686 | (?: ZERO_LENGTH )? | |
1687 | }x; | |
1688 | ||
1689 | The higher level-loops preserve an additional state between iterations: | |
1690 | whether the last match was zero-length. To break the loop, the following | |
1691 | match after a zero-length match is prohibited to have a length of zero. | |
1692 | This prohibition interacts with backtracking (see L<"Backtracking">), | |
1693 | and so the I<second best> match is chosen if the I<best> match is of | |
1694 | zero length. | |
1695 | ||
19799a22 | 1696 | For example: |
c84d73f1 IZ |
1697 | |
1698 | $_ = 'bar'; | |
1699 | s/\w??/<$&>/g; | |
1700 | ||
20fb949f | 1701 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
c84d73f1 IZ |
1702 | match given by non-greedy C<??> is the zero-length match, and the I<second |
1703 | best> match is what is matched by C<\w>. Thus zero-length matches | |
1704 | alternate with one-character-long matches. | |
1705 | ||
1706 | Similarly, for repeated C<m/()/g> the second-best match is the match at the | |
1707 | position one notch further in the string. | |
1708 | ||
19799a22 | 1709 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 | 1710 | the matched string, and is reset by each assignment to pos(). |
9da458fc IZ |
1711 | Zero-length matches at the end of the previous match are ignored |
1712 | during C<split>. | |
c84d73f1 | 1713 | |
35a734be IZ |
1714 | =head2 Combining pieces together |
1715 | ||
1716 | Each of the elementary pieces of regular expressions which were described | |
1717 | before (such as C<ab> or C<\Z>) could match at most one substring | |
1718 | at the given position of the input string. However, in a typical regular | |
1719 | expression these elementary pieces are combined into more complicated | |
1720 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc | |
1721 | (in these examples C<S> and C<T> are regular subexpressions). | |
1722 | ||
1723 | Such combinations can include alternatives, leading to a problem of choice: | |
1724 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
1725 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
1726 | actually matched is the concept of backtracking (see L<"Backtracking">). | |
1727 | However, this description is too low-level and makes you think | |
1728 | in terms of a particular implementation. | |
1729 | ||
1730 | Another description starts with notions of "better"/"worse". All the | |
1731 | substrings which may be matched by the given regular expression can be | |
1732 | sorted from the "best" match to the "worst" match, and it is the "best" | |
1733 | match which is chosen. This substitutes the question of "what is chosen?" | |
1734 | by the question of "which matches are better, and which are worse?". | |
1735 | ||
1736 | Again, for elementary pieces there is no such question, since at most | |
1737 | one match at a given position is possible. This section describes the | |
1738 | notion of better/worse for combining operators. In the description | |
1739 | below C<S> and C<T> are regular subexpressions. | |
1740 | ||
13a2d996 | 1741 | =over 4 |
35a734be IZ |
1742 | |
1743 | =item C<ST> | |
1744 | ||
1745 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are | |
1746 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings | |
1747 | which can be matched by C<T>. | |
1748 | ||
1749 | If C<A> is better match for C<S> than C<A'>, C<AB> is a better | |
1750 | match than C<A'B'>. | |
1751 | ||
1752 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if | |
1753 | C<B> is better match for C<T> than C<B'>. | |
1754 | ||
1755 | =item C<S|T> | |
1756 | ||
1757 | When C<S> can match, it is a better match than when only C<T> can match. | |
1758 | ||
1759 | Ordering of two matches for C<S> is the same as for C<S>. Similar for | |
1760 | two matches for C<T>. | |
1761 | ||
1762 | =item C<S{REPEAT_COUNT}> | |
1763 | ||
1764 | Matches as C<SSS...S> (repeated as many times as necessary). | |
1765 | ||
1766 | =item C<S{min,max}> | |
1767 | ||
1768 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
1769 | ||
1770 | =item C<S{min,max}?> | |
1771 | ||
1772 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
1773 | ||
1774 | =item C<S?>, C<S*>, C<S+> | |
1775 | ||
1776 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
1777 | ||
1778 | =item C<S??>, C<S*?>, C<S+?> | |
1779 | ||
1780 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
1781 | ||
c47ff5f1 | 1782 | =item C<< (?>S) >> |
35a734be IZ |
1783 | |
1784 | Matches the best match for C<S> and only that. | |
1785 | ||
1786 | =item C<(?=S)>, C<(?<=S)> | |
1787 | ||
1788 | Only the best match for C<S> is considered. (This is important only if | |
1789 | C<S> has capturing parentheses, and backreferences are used somewhere | |
1790 | else in the whole regular expression.) | |
1791 | ||
1792 | =item C<(?!S)>, C<(?<!S)> | |
1793 | ||
1794 | For this grouping operator there is no need to describe the ordering, since | |
1795 | only whether or not C<S> can match is important. | |
1796 | ||
6bda09f9 | 1797 | =item C<(??{ EXPR })>, C<(?PARNO)> |
35a734be IZ |
1798 | |
1799 | The ordering is the same as for the regular expression which is | |
6bda09f9 | 1800 | the result of EXPR, or the pattern contained by capture buffer PARNO. |
35a734be IZ |
1801 | |
1802 | =item C<(?(condition)yes-pattern|no-pattern)> | |
1803 | ||
1804 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is | |
1805 | already determined. The ordering of the matches is the same as for the | |
1806 | chosen subexpression. | |
1807 | ||
1808 | =back | |
1809 | ||
1810 | The above recipes describe the ordering of matches I<at a given position>. | |
1811 | One more rule is needed to understand how a match is determined for the | |
1812 | whole regular expression: a match at an earlier position is always better | |
1813 | than a match at a later position. | |
1814 | ||
c84d73f1 IZ |
1815 | =head2 Creating custom RE engines |
1816 | ||
1817 | Overloaded constants (see L<overload>) provide a simple way to extend | |
1818 | the functionality of the RE engine. | |
1819 | ||
1820 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
6b0ac556 | 1821 | matches at boundary between whitespace characters and non-whitespace |
c84d73f1 IZ |
1822 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
1823 | at these positions, so we want to have each C<\Y|> in the place of the | |
1824 | more complicated version. We can create a module C<customre> to do | |
1825 | this: | |
1826 | ||
1827 | package customre; | |
1828 | use overload; | |
1829 | ||
1830 | sub import { | |
1831 | shift; | |
1832 | die "No argument to customre::import allowed" if @_; | |
1833 | overload::constant 'qr' => \&convert; | |
1834 | } | |
1835 | ||
1836 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
1837 | ||
580a9fe1 RGS |
1838 | # We must also take care of not escaping the legitimate \\Y| |
1839 | # sequence, hence the presence of '\\' in the conversion rules. | |
141db969 | 1840 | my %rules = ( '\\' => '\\\\', |
c84d73f1 IZ |
1841 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
1842 | sub convert { | |
1843 | my $re = shift; | |
1844 | $re =~ s{ | |
1845 | \\ ( \\ | Y . ) | |
1846 | } | |
1847 | { $rules{$1} or invalid($re,$1) }sgex; | |
1848 | return $re; | |
1849 | } | |
1850 | ||
1851 | Now C<use customre> enables the new escape in constant regular | |
1852 | expressions, i.e., those without any runtime variable interpolations. | |
1853 | As documented in L<overload>, this conversion will work only over | |
1854 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
1855 | part of this regular expression needs to be converted explicitly | |
1856 | (but only if the special meaning of C<\Y|> should be enabled inside $re): | |
1857 | ||
1858 | use customre; | |
1859 | $re = <>; | |
1860 | chomp $re; | |
1861 | $re = customre::convert $re; | |
1862 | /\Y|$re\Y|/; | |
1863 | ||
19799a22 GS |
1864 | =head1 BUGS |
1865 | ||
9da458fc IZ |
1866 | This document varies from difficult to understand to completely |
1867 | and utterly opaque. The wandering prose riddled with jargon is | |
1868 | hard to fathom in several places. | |
1869 | ||
1870 | This document needs a rewrite that separates the tutorial content | |
1871 | from the reference content. | |
19799a22 GS |
1872 | |
1873 | =head1 SEE ALSO | |
9fa51da4 | 1874 | |
91e0c79e MJD |
1875 | L<perlrequick>. |
1876 | ||
1877 | L<perlretut>. | |
1878 | ||
9b599b2a GS |
1879 | L<perlop/"Regexp Quote-Like Operators">. |
1880 | ||
1e66bd83 PP |
1881 | L<perlop/"Gory details of parsing quoted constructs">. |
1882 | ||
14218588 GS |
1883 | L<perlfaq6>. |
1884 | ||
9b599b2a GS |
1885 | L<perlfunc/pos>. |
1886 | ||
1887 | L<perllocale>. | |
1888 | ||
fb55449c JH |
1889 | L<perlebcdic>. |
1890 | ||
14218588 GS |
1891 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
1892 | by O'Reilly and Associates. |