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 |
a0d0e21e LW |
113 | newline within the string, and "$" will match before any newline. At the |
114 | cost of a little more overhead, you can do this by using the /m modifier | |
115 | on the pattern match operator. (Older programs did this by setting C<$*>, | |
f02c194e | 116 | but this practice has been removed in perl 5.9.) |
d74e8afc | 117 | X<^> X<$> X</m> |
a0d0e21e | 118 | |
14218588 | 119 | To simplify multi-line substitutions, the "." character never matches a |
55497cff | 120 | newline unless you use the C</s> modifier, which in effect tells Perl to pretend |
f02c194e | 121 | the string is a single line--even if it isn't. |
d74e8afc | 122 | X<.> X</s> |
a0d0e21e | 123 | |
04838cea RGS |
124 | =head3 Quantifiers |
125 | ||
a0d0e21e | 126 | The following standard quantifiers are recognized: |
d74e8afc | 127 | X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}> |
a0d0e21e LW |
128 | |
129 | * Match 0 or more times | |
130 | + Match 1 or more times | |
131 | ? Match 1 or 0 times | |
132 | {n} Match exactly n times | |
133 | {n,} Match at least n times | |
134 | {n,m} Match at least n but not more than m times | |
135 | ||
136 | (If a curly bracket occurs in any other context, it is treated | |
b975c076 JL |
137 | as a regular character. In particular, the lower bound |
138 | is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+" | |
25f94b33 | 139 | modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited |
9c79236d GS |
140 | to integral values less than a preset limit defined when perl is built. |
141 | This is usually 32766 on the most common platforms. The actual limit can | |
142 | be seen in the error message generated by code such as this: | |
143 | ||
820475bd | 144 | $_ **= $_ , / {$_} / for 2 .. 42; |
a0d0e21e | 145 | |
54310121 | 146 | By default, a quantified subpattern is "greedy", that is, it will match as |
147 | many times as possible (given a particular starting location) while still | |
148 | allowing the rest of the pattern to match. If you want it to match the | |
149 | minimum number of times possible, follow the quantifier with a "?". Note | |
150 | that the meanings don't change, just the "greediness": | |
d74e8afc ITB |
151 | X<metacharacter> X<greedy> X<greedyness> |
152 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?> | |
a0d0e21e LW |
153 | |
154 | *? Match 0 or more times | |
155 | +? Match 1 or more times | |
156 | ?? Match 0 or 1 time | |
157 | {n}? Match exactly n times | |
158 | {n,}? Match at least n times | |
159 | {n,m}? Match at least n but not more than m times | |
160 | ||
b9b4dddf YO |
161 | By default, when a quantified subpattern does not allow the rest of the |
162 | overall pattern to match, Perl will backtrack. However, this behaviour is | |
163 | sometimes undesirable. Thus Perl provides the "possesive" quantifier form | |
164 | as well. | |
165 | ||
166 | *+ Match 0 or more times and give nothing back | |
04838cea | 167 | ++ Match 1 or more times and give nothing back |
b9b4dddf YO |
168 | ?+ Match 0 or 1 time and give nothing back |
169 | {n}+ Match exactly n times and give nothing back (redundant) | |
04838cea RGS |
170 | {n,}+ Match at least n times and give nothing back |
171 | {n,m}+ Match at least n but not more than m times and give nothing back | |
b9b4dddf YO |
172 | |
173 | For instance, | |
174 | ||
175 | 'aaaa' =~ /a++a/ | |
176 | ||
177 | will never match, as the C<a++> will gobble up all the C<a>'s in the | |
178 | string and won't leave any for the remaining part of the pattern. This | |
179 | feature can be extremely useful to give perl hints about where it | |
180 | shouldn't backtrack. For instance, the typical "match a double-quoted | |
181 | string" problem can be most efficiently performed when written as: | |
182 | ||
183 | /"(?:[^"\\]++|\\.)*+"/ | |
184 | ||
185 | as we know that if the final quote does not match, bactracking will not | |
186 | help. See the independent subexpression C<< (?>...) >> for more details; | |
187 | possessive quantifiers are just syntactic sugar for that construct. For | |
188 | instance the above example could also be written as follows: | |
189 | ||
190 | /"(?>(?:(?>[^"\\]+)|\\.)*)"/ | |
191 | ||
04838cea RGS |
192 | =head3 Escape sequences |
193 | ||
5f05dabc | 194 | Because patterns are processed as double quoted strings, the following |
a0d0e21e | 195 | also work: |
d74e8afc ITB |
196 | X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q> |
197 | X<\0> X<\c> X<\N> X<\x> | |
a0d0e21e | 198 | |
0f36ee90 | 199 | \t tab (HT, TAB) |
200 | \n newline (LF, NL) | |
201 | \r return (CR) | |
202 | \f form feed (FF) | |
203 | \a alarm (bell) (BEL) | |
204 | \e escape (think troff) (ESC) | |
cb1a09d0 AD |
205 | \033 octal char (think of a PDP-11) |
206 | \x1B hex char | |
a0ed51b3 | 207 | \x{263a} wide hex char (Unicode SMILEY) |
a0d0e21e | 208 | \c[ control char |
4a2d328f | 209 | \N{name} named char |
cb1a09d0 AD |
210 | \l lowercase next char (think vi) |
211 | \u uppercase next char (think vi) | |
212 | \L lowercase till \E (think vi) | |
213 | \U uppercase till \E (think vi) | |
214 | \E end case modification (think vi) | |
5a964f20 | 215 | \Q quote (disable) pattern metacharacters till \E |
a0d0e21e | 216 | |
a034a98d | 217 | If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u> |
423cee85 | 218 | and C<\U> is taken from the current locale. See L<perllocale>. For |
4a2d328f | 219 | documentation of C<\N{name}>, see L<charnames>. |
a034a98d | 220 | |
1d2dff63 GS |
221 | You cannot include a literal C<$> or C<@> within a C<\Q> sequence. |
222 | An unescaped C<$> or C<@> interpolates the corresponding variable, | |
223 | while escaping will cause the literal string C<\$> to be matched. | |
224 | You'll need to write something like C<m/\Quser\E\@\Qhost/>. | |
225 | ||
04838cea RGS |
226 | =head3 Character classes |
227 | ||
a0d0e21e | 228 | In addition, Perl defines the following: |
d74e8afc ITB |
229 | X<metacharacter> |
230 | X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C> | |
231 | X<word> X<whitespace> | |
a0d0e21e | 232 | |
81714fb9 YO |
233 | \w Match a "word" character (alphanumeric plus "_") |
234 | \W Match a non-"word" character | |
235 | \s Match a whitespace character | |
236 | \S Match a non-whitespace character | |
237 | \d Match a digit character | |
238 | \D Match a non-digit character | |
239 | \pP Match P, named property. Use \p{Prop} for longer names. | |
240 | \PP Match non-P | |
241 | \X Match eXtended Unicode "combining character sequence", | |
242 | equivalent to (?:\PM\pM*) | |
243 | \C Match a single C char (octet) even under Unicode. | |
244 | NOTE: breaks up characters into their UTF-8 bytes, | |
245 | so you may end up with malformed pieces of UTF-8. | |
246 | Unsupported in lookbehind. | |
247 | \1 Backreference to a a specific group. | |
248 | '1' may actually be any positive integer | |
249 | \k<name> Named backreference | |
250 | \N{name} Named unicode character, or unicode escape. | |
251 | \x12 Hexadecimal escape sequence | |
252 | \x{1234} Long hexadecimal escape sequence | |
a0d0e21e | 253 | |
08ce8fc6 JH |
254 | A C<\w> matches a single alphanumeric character (an alphabetic |
255 | character, or a decimal digit) or C<_>, not a whole word. Use C<\w+> | |
256 | to match a string of Perl-identifier characters (which isn't the same | |
257 | as matching an English word). If C<use locale> is in effect, the list | |
258 | of alphabetic characters generated by C<\w> is taken from the current | |
259 | locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>, | |
1209ba90 | 260 | C<\d>, and C<\D> within character classes, but if you try to use them |
08ce8fc6 JH |
261 | as endpoints of a range, that's not a range, the "-" is understood |
262 | literally. If Unicode is in effect, C<\s> matches also "\x{85}", | |
263 | "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about | |
491fd90a | 264 | C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general. |
fa11829f | 265 | You can define your own C<\p> and C<\P> properties, see L<perlunicode>. |
d74e8afc | 266 | X<\w> X<\W> X<word> |
a0d0e21e | 267 | |
b8c5462f | 268 | The POSIX character class syntax |
d74e8afc | 269 | X<character class> |
b8c5462f | 270 | |
820475bd | 271 | [:class:] |
b8c5462f | 272 | |
5496314a SP |
273 | is also available. Note that the C<[> and C<]> braces are I<literal>; |
274 | they must always be used within a character class expression. | |
275 | ||
276 | # this is correct: | |
277 | $string =~ /[[:alpha:]]/; | |
278 | ||
279 | # this is not, and will generate a warning: | |
280 | $string =~ /[:alpha:]/; | |
281 | ||
282 | The available classes and their backslash equivalents (if available) are | |
283 | as follows: | |
d74e8afc ITB |
284 | X<character class> |
285 | X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph> | |
286 | X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit> | |
b8c5462f JH |
287 | |
288 | alpha | |
289 | alnum | |
290 | ascii | |
aaa51d5e | 291 | blank [1] |
b8c5462f JH |
292 | cntrl |
293 | digit \d | |
294 | graph | |
295 | lower | |
296 | ||
297 | punct | |
aaa51d5e | 298 | space \s [2] |
b8c5462f | 299 | upper |
aaa51d5e | 300 | word \w [3] |
b8c5462f JH |
301 | xdigit |
302 | ||
07698885 RGS |
303 | =over |
304 | ||
305 | =item [1] | |
306 | ||
b432a672 | 307 | A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace". |
07698885 RGS |
308 | |
309 | =item [2] | |
310 | ||
311 | Not exactly equivalent to C<\s> since the C<[[:space:]]> includes | |
b432a672 | 312 | also the (very rare) "vertical tabulator", "\ck", chr(11). |
07698885 RGS |
313 | |
314 | =item [3] | |
315 | ||
08ce8fc6 | 316 | A Perl extension, see above. |
07698885 RGS |
317 | |
318 | =back | |
aaa51d5e | 319 | |
26b44a0a | 320 | For example use C<[:upper:]> to match all the uppercase characters. |
aaa51d5e JF |
321 | Note that the C<[]> are part of the C<[::]> construct, not part of the |
322 | whole character class. For example: | |
b8c5462f | 323 | |
820475bd | 324 | [01[:alpha:]%] |
b8c5462f | 325 | |
593df60c | 326 | matches zero, one, any alphabetic character, and the percentage sign. |
b8c5462f | 327 | |
72ff2908 JH |
328 | The following equivalences to Unicode \p{} constructs and equivalent |
329 | backslash character classes (if available), will hold: | |
d74e8afc | 330 | X<character class> X<\p> X<\p{}> |
72ff2908 | 331 | |
5496314a | 332 | [[:...:]] \p{...} backslash |
b8c5462f JH |
333 | |
334 | alpha IsAlpha | |
335 | alnum IsAlnum | |
336 | ascii IsASCII | |
b432a672 | 337 | blank IsSpace |
b8c5462f | 338 | cntrl IsCntrl |
3bec3564 | 339 | digit IsDigit \d |
b8c5462f JH |
340 | graph IsGraph |
341 | lower IsLower | |
342 | print IsPrint | |
343 | punct IsPunct | |
344 | space IsSpace | |
3bec3564 | 345 | IsSpacePerl \s |
b8c5462f JH |
346 | upper IsUpper |
347 | word IsWord | |
348 | xdigit IsXDigit | |
349 | ||
5496314a | 350 | For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent. |
b8c5462f JH |
351 | |
352 | If the C<utf8> pragma is not used but the C<locale> pragma is, the | |
aaa51d5e | 353 | classes correlate with the usual isalpha(3) interface (except for |
b432a672 | 354 | "word" and "blank"). |
b8c5462f JH |
355 | |
356 | The assumedly non-obviously named classes are: | |
357 | ||
358 | =over 4 | |
359 | ||
360 | =item cntrl | |
d74e8afc | 361 | X<cntrl> |
b8c5462f | 362 | |
820475bd GS |
363 | Any control character. Usually characters that don't produce output as |
364 | such but instead control the terminal somehow: for example newline and | |
365 | backspace are control characters. All characters with ord() less than | |
593df60c | 366 | 32 are most often classified as control characters (assuming ASCII, |
7be5a6cf JF |
367 | the ISO Latin character sets, and Unicode), as is the character with |
368 | the ord() value of 127 (C<DEL>). | |
b8c5462f JH |
369 | |
370 | =item graph | |
d74e8afc | 371 | X<graph> |
b8c5462f | 372 | |
f1cbbd6e | 373 | Any alphanumeric or punctuation (special) character. |
b8c5462f JH |
374 | |
375 | =item print | |
d74e8afc | 376 | X<print> |
b8c5462f | 377 | |
f79b3095 | 378 | Any alphanumeric or punctuation (special) character or the space character. |
b8c5462f JH |
379 | |
380 | =item punct | |
d74e8afc | 381 | X<punct> |
b8c5462f | 382 | |
f1cbbd6e | 383 | Any punctuation (special) character. |
b8c5462f JH |
384 | |
385 | =item xdigit | |
d74e8afc | 386 | X<xdigit> |
b8c5462f | 387 | |
593df60c | 388 | Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would |
820475bd | 389 | work just fine) it is included for completeness. |
b8c5462f | 390 | |
b8c5462f JH |
391 | =back |
392 | ||
393 | You can negate the [::] character classes by prefixing the class name | |
394 | with a '^'. This is a Perl extension. For example: | |
d74e8afc | 395 | X<character class, negation> |
b8c5462f | 396 | |
5496314a | 397 | POSIX traditional Unicode |
93733859 | 398 | |
5496314a SP |
399 | [[:^digit:]] \D \P{IsDigit} |
400 | [[:^space:]] \S \P{IsSpace} | |
401 | [[:^word:]] \W \P{IsWord} | |
b8c5462f | 402 | |
54c18d04 MK |
403 | Perl respects the POSIX standard in that POSIX character classes are |
404 | only supported within a character class. The POSIX character classes | |
405 | [.cc.] and [=cc=] are recognized but B<not> supported and trying to | |
406 | use them will cause an error. | |
b8c5462f | 407 | |
04838cea RGS |
408 | =head3 Assertions |
409 | ||
a0d0e21e | 410 | Perl defines the following zero-width assertions: |
d74e8afc ITB |
411 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> |
412 | X<regexp, zero-width assertion> | |
413 | X<regular expression, zero-width assertion> | |
414 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> | |
a0d0e21e LW |
415 | |
416 | \b Match a word boundary | |
417 | \B Match a non-(word boundary) | |
b85d18e9 IZ |
418 | \A Match only at beginning of string |
419 | \Z Match only at end of string, or before newline at the end | |
420 | \z Match only at end of string | |
9da458fc IZ |
421 | \G Match only at pos() (e.g. at the end-of-match position |
422 | of prior m//g) | |
a0d0e21e | 423 | |
14218588 | 424 | A word boundary (C<\b>) is a spot between two characters |
19799a22 GS |
425 | that has a C<\w> on one side of it and a C<\W> on the other side |
426 | of it (in either order), counting the imaginary characters off the | |
427 | beginning and end of the string as matching a C<\W>. (Within | |
428 | character classes C<\b> represents backspace rather than a word | |
429 | boundary, just as it normally does in any double-quoted string.) | |
430 | The C<\A> and C<\Z> are just like "^" and "$", except that they | |
431 | won't match multiple times when the C</m> modifier is used, while | |
432 | "^" and "$" will match at every internal line boundary. To match | |
433 | the actual end of the string and not ignore an optional trailing | |
434 | newline, use C<\z>. | |
d74e8afc | 435 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 GS |
436 | |
437 | The C<\G> assertion can be used to chain global matches (using | |
438 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. | |
439 | It is also useful when writing C<lex>-like scanners, when you have | |
440 | several patterns that you want to match against consequent substrings | |
441 | of your string, see the previous reference. The actual location | |
442 | where C<\G> will match can also be influenced by using C<pos()> as | |
25cf8c22 HS |
443 | an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully |
444 | supported when anchored to the start of the pattern; while it | |
445 | is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some | |
446 | such uses (C</.\G/g>, for example) currently cause problems, and | |
447 | it is recommended that you avoid such usage for now. | |
d74e8afc | 448 | X<\G> |
c47ff5f1 | 449 | |
04838cea RGS |
450 | =head3 Capture buffers |
451 | ||
14218588 | 452 | The bracketing construct C<( ... )> creates capture buffers. To |
c47ff5f1 | 453 | refer to the digit'th buffer use \<digit> within the |
14218588 | 454 | match. Outside the match use "$" instead of "\". (The |
81714fb9 | 455 | \<digit> notation works in certain circumstances outside |
14218588 GS |
456 | the match. See the warning below about \1 vs $1 for details.) |
457 | Referring back to another part of the match is called a | |
458 | I<backreference>. | |
d74e8afc ITB |
459 | X<regex, capture buffer> X<regexp, capture buffer> |
460 | X<regular expression, capture buffer> X<backreference> | |
14218588 GS |
461 | |
462 | There is no limit to the number of captured substrings that you may | |
463 | use. However Perl also uses \10, \11, etc. as aliases for \010, | |
fb55449c JH |
464 | \011, etc. (Recall that 0 means octal, so \011 is the character at |
465 | number 9 in your coded character set; which would be the 10th character, | |
81714fb9 YO |
466 | a horizontal tab under ASCII.) Perl resolves this |
467 | ambiguity by interpreting \10 as a backreference only if at least 10 | |
468 | left parentheses have opened before it. Likewise \11 is a | |
469 | backreference only if at least 11 left parentheses have opened | |
470 | before it. And so on. \1 through \9 are always interpreted as | |
fb55449c | 471 | backreferences. |
14218588 | 472 | |
81714fb9 YO |
473 | Additionally, as of Perl 5.10 you may use named capture buffers and named |
474 | backreferences. The notation is C<< (?<name>...) >> and C<< \k<name> >> | |
475 | (you may also use single quotes instead of angle brackets to quote the | |
476 | name). The only difference with named capture buffers and unnamed ones is | |
477 | that multiple buffers may have the same name and that the contents of | |
478 | named capture buffers is available via the C<%+> hash. When multiple | |
479 | groups share the same name C<$+{name}> and C<< \k<name> >> refer to the | |
480 | leftmost defined group, thus it's possible to do things with named capture | |
481 | buffers that would otherwise require C<(??{})> code to accomplish. Named | |
482 | capture buffers are numbered just as normal capture buffers are and may be | |
483 | referenced via the magic numeric variables or via numeric backreferences | |
484 | as well as by name. | |
485 | ||
14218588 | 486 | Examples: |
a0d0e21e LW |
487 | |
488 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
489 | ||
81714fb9 YO |
490 | /(.)\1/ # find first doubled char |
491 | and print "'$1' is the first doubled character\n"; | |
492 | ||
493 | /(?<char>.)\k<char>/ # ... a different way | |
494 | and print "'$+{char}' is the first doubled character\n"; | |
495 | ||
496 | /(?<char>.)\1/ # ... mix and match | |
497 | and print "'$1' is the first doubled character\n"; | |
c47ff5f1 | 498 | |
14218588 | 499 | if (/Time: (..):(..):(..)/) { # parse out values |
a0d0e21e LW |
500 | $hours = $1; |
501 | $minutes = $2; | |
502 | $seconds = $3; | |
503 | } | |
c47ff5f1 | 504 | |
14218588 GS |
505 | Several special variables also refer back to portions of the previous |
506 | match. C<$+> returns whatever the last bracket match matched. | |
507 | C<$&> returns the entire matched string. (At one point C<$0> did | |
508 | also, but now it returns the name of the program.) C<$`> returns | |
77ea4f6d JV |
509 | everything before the matched string. C<$'> returns everything |
510 | after the matched string. And C<$^N> contains whatever was matched by | |
511 | the most-recently closed group (submatch). C<$^N> can be used in | |
512 | extended patterns (see below), for example to assign a submatch to a | |
81714fb9 | 513 | variable. |
d74e8afc | 514 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 | 515 | |
665e98b9 | 516 | The numbered match variables ($1, $2, $3, etc.) and the related punctuation |
77ea4f6d | 517 | set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped |
14218588 GS |
518 | until the end of the enclosing block or until the next successful |
519 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
d74e8afc ITB |
520 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
521 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> | |
522 | ||
14218588 | 523 | |
665e98b9 | 524 | B<NOTE>: failed matches in Perl do not reset the match variables, |
5146ce24 | 525 | which makes it easier to write code that tests for a series of more |
665e98b9 JH |
526 | specific cases and remembers the best match. |
527 | ||
14218588 GS |
528 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or |
529 | C<$'> anywhere in the program, it has to provide them for every | |
530 | pattern match. This may substantially slow your program. Perl | |
531 | uses the same mechanism to produce $1, $2, etc, so you also pay a | |
532 | price for each pattern that contains capturing parentheses. (To | |
533 | avoid this cost while retaining the grouping behaviour, use the | |
534 | extended regular expression C<(?: ... )> instead.) But if you never | |
535 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
536 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
537 | if you can, but if you can't (and some algorithms really appreciate | |
538 | them), once you've used them once, use them at will, because you've | |
539 | already paid the price. As of 5.005, C<$&> is not so costly as the | |
540 | other two. | |
d74e8afc | 541 | X<$&> X<$`> X<$'> |
68dc0745 | 542 | |
19799a22 GS |
543 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
544 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
545 | are no backslashed symbols that aren't alphanumeric. So anything | |
c47ff5f1 | 546 | that looks like \\, \(, \), \<, \>, \{, or \} is always |
19799a22 GS |
547 | interpreted as a literal character, not a metacharacter. This was |
548 | once used in a common idiom to disable or quote the special meanings | |
549 | of regular expression metacharacters in a string that you want to | |
36bbe248 | 550 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e LW |
551 | |
552 | $pattern =~ s/(\W)/\\$1/g; | |
553 | ||
f1cbbd6e | 554 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 GS |
555 | Today it is more common to use the quotemeta() function or the C<\Q> |
556 | metaquoting escape sequence to disable all metacharacters' special | |
557 | meanings like this: | |
a0d0e21e LW |
558 | |
559 | /$unquoted\Q$quoted\E$unquoted/ | |
560 | ||
9da458fc IZ |
561 | Beware that if you put literal backslashes (those not inside |
562 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
563 | backslash interpolation may lead to confusing results. If you | |
564 | I<need> to use literal backslashes within C<\Q...\E>, | |
565 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
566 | ||
19799a22 GS |
567 | =head2 Extended Patterns |
568 | ||
14218588 GS |
569 | Perl also defines a consistent extension syntax for features not |
570 | found in standard tools like B<awk> and B<lex>. The syntax is a | |
571 | pair of parentheses with a question mark as the first thing within | |
572 | the parentheses. The character after the question mark indicates | |
573 | the extension. | |
19799a22 | 574 | |
14218588 GS |
575 | The stability of these extensions varies widely. Some have been |
576 | part of the core language for many years. Others are experimental | |
577 | and may change without warning or be completely removed. Check | |
578 | the documentation on an individual feature to verify its current | |
579 | status. | |
19799a22 | 580 | |
14218588 GS |
581 | A question mark was chosen for this and for the minimal-matching |
582 | construct because 1) question marks are rare in older regular | |
583 | expressions, and 2) whenever you see one, you should stop and | |
584 | "question" exactly what is going on. That's psychology... | |
a0d0e21e LW |
585 | |
586 | =over 10 | |
587 | ||
cc6b7395 | 588 | =item C<(?#text)> |
d74e8afc | 589 | X<(?#)> |
a0d0e21e | 590 | |
14218588 | 591 | A comment. The text is ignored. If the C</x> modifier enables |
19799a22 | 592 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes |
259138e3 GS |
593 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
594 | C<)> in the comment. | |
a0d0e21e | 595 | |
19799a22 | 596 | =item C<(?imsx-imsx)> |
d74e8afc | 597 | X<(?)> |
19799a22 | 598 | |
0b6d1084 JH |
599 | One or more embedded pattern-match modifiers, to be turned on (or |
600 | turned off, if preceded by C<->) for the remainder of the pattern or | |
601 | the remainder of the enclosing pattern group (if any). This is | |
602 | particularly useful for dynamic patterns, such as those read in from a | |
603 | configuration file, read in as an argument, are specified in a table | |
604 | somewhere, etc. Consider the case that some of which want to be case | |
605 | sensitive and some do not. The case insensitive ones need to include | |
606 | merely C<(?i)> at the front of the pattern. For example: | |
19799a22 GS |
607 | |
608 | $pattern = "foobar"; | |
609 | if ( /$pattern/i ) { } | |
610 | ||
611 | # more flexible: | |
612 | ||
613 | $pattern = "(?i)foobar"; | |
614 | if ( /$pattern/ ) { } | |
615 | ||
0b6d1084 | 616 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 GS |
617 | |
618 | ( (?i) blah ) \s+ \1 | |
619 | ||
620 | will match a repeated (I<including the case>!) word C<blah> in any | |
14218588 | 621 | case, assuming C<x> modifier, and no C<i> modifier outside this |
19799a22 GS |
622 | group. |
623 | ||
5a964f20 | 624 | =item C<(?:pattern)> |
d74e8afc | 625 | X<(?:)> |
a0d0e21e | 626 | |
ca9dfc88 IZ |
627 | =item C<(?imsx-imsx:pattern)> |
628 | ||
5a964f20 TC |
629 | This is for clustering, not capturing; it groups subexpressions like |
630 | "()", but doesn't make backreferences as "()" does. So | |
a0d0e21e | 631 | |
5a964f20 | 632 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e LW |
633 | |
634 | is like | |
635 | ||
5a964f20 | 636 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e | 637 | |
19799a22 GS |
638 | but doesn't spit out extra fields. It's also cheaper not to capture |
639 | characters if you don't need to. | |
a0d0e21e | 640 | |
19799a22 GS |
641 | Any letters between C<?> and C<:> act as flags modifiers as with |
642 | C<(?imsx-imsx)>. For example, | |
ca9dfc88 IZ |
643 | |
644 | /(?s-i:more.*than).*million/i | |
645 | ||
14218588 | 646 | is equivalent to the more verbose |
ca9dfc88 IZ |
647 | |
648 | /(?:(?s-i)more.*than).*million/i | |
649 | ||
5a964f20 | 650 | =item C<(?=pattern)> |
d74e8afc | 651 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> |
a0d0e21e | 652 | |
19799a22 | 653 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e LW |
654 | matches a word followed by a tab, without including the tab in C<$&>. |
655 | ||
5a964f20 | 656 | =item C<(?!pattern)> |
d74e8afc | 657 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> |
a0d0e21e | 658 | |
19799a22 | 659 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e | 660 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 GS |
661 | however that look-ahead and look-behind are NOT the same thing. You cannot |
662 | use this for look-behind. | |
7b8d334a | 663 | |
5a964f20 | 664 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a GS |
665 | will not do what you want. That's because the C<(?!foo)> is just saying that |
666 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
667 | match. You would have to do something like C</(?!foo)...bar/> for that. We | |
668 | say "like" because there's the case of your "bar" not having three characters | |
669 | before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>. | |
670 | Sometimes it's still easier just to say: | |
a0d0e21e | 671 | |
a3cb178b | 672 | if (/bar/ && $` !~ /foo$/) |
a0d0e21e | 673 | |
19799a22 | 674 | For look-behind see below. |
c277df42 | 675 | |
c47ff5f1 | 676 | =item C<(?<=pattern)> |
d74e8afc | 677 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> |
c277df42 | 678 | |
c47ff5f1 | 679 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 GS |
680 | matches a word that follows a tab, without including the tab in C<$&>. |
681 | Works only for fixed-width look-behind. | |
c277df42 | 682 | |
5a964f20 | 683 | =item C<(?<!pattern)> |
d74e8afc | 684 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> |
c277df42 | 685 | |
19799a22 GS |
686 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
687 | matches any occurrence of "foo" that does not follow "bar". Works | |
688 | only for fixed-width look-behind. | |
c277df42 | 689 | |
81714fb9 YO |
690 | =item C<(?'NAME'pattern)> |
691 | ||
692 | =item C<< (?<NAME>pattern) >> | |
693 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> | |
694 | ||
695 | A named capture buffer. Identical in every respect to normal capturing | |
696 | parens C<()> but for the additional fact that C<%+> may be used after | |
697 | a succesful match to refer to a named buffer. See C<perlvar> for more | |
698 | details on the C<%+> hash. | |
699 | ||
700 | If multiple distinct capture buffers have the same name then the | |
701 | $+{NAME} will refer to the leftmost defined buffer in the match. | |
702 | ||
703 | The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent. | |
704 | ||
705 | B<NOTE:> While the notation of this construct is the same as the similar | |
706 | function in .NET regexes, the behavior is not, in Perl the buffers are | |
707 | numbered sequentially regardless of being named or not. Thus in the | |
708 | pattern | |
709 | ||
710 | /(x)(?<foo>y)(z)/ | |
711 | ||
712 | $+{foo} will be the same as $2, and $3 will contain 'z' instead of | |
713 | the opposite which is what a .NET regex hacker might expect. | |
714 | ||
715 | Currently NAME is restricted to word chars only. In other words, it | |
716 | must match C</^\w+$/>. | |
717 | ||
718 | =item C<< \k<name> >> | |
719 | ||
720 | =item C<< \k'name' >> | |
721 | ||
722 | Named backreference. Similar to numeric backreferences, except that | |
723 | the group is designated by name and not number. If multiple groups | |
724 | have the same name then it refers to the leftmost defined group in | |
725 | the current match. | |
726 | ||
727 | It is an error to refer to a name not defined by a C<(?<NAME>)> | |
728 | earlier in the pattern. | |
729 | ||
730 | Both forms are equivalent. | |
731 | ||
cc6b7395 | 732 | =item C<(?{ code })> |
d74e8afc | 733 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 | 734 | |
19799a22 | 735 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
736 | experimental, and may be changed without notice. Code executed that |
737 | has side effects may not perform identically from version to version | |
738 | due to the effect of future optimisations in the regex engine. | |
c277df42 | 739 | |
cc46d5f2 | 740 | This zero-width assertion evaluates any embedded Perl code. It |
19799a22 GS |
741 | always succeeds, and its C<code> is not interpolated. Currently, |
742 | the rules to determine where the C<code> ends are somewhat convoluted. | |
743 | ||
77ea4f6d JV |
744 | This feature can be used together with the special variable C<$^N> to |
745 | capture the results of submatches in variables without having to keep | |
746 | track of the number of nested parentheses. For example: | |
747 | ||
748 | $_ = "The brown fox jumps over the lazy dog"; | |
749 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
750 | print "color = $color, animal = $animal\n"; | |
751 | ||
754091cb RGS |
752 | Inside the C<(?{...})> block, C<$_> refers to the string the regular |
753 | expression is matching against. You can also use C<pos()> to know what is | |
fa11829f | 754 | the current position of matching within this string. |
754091cb | 755 | |
19799a22 GS |
756 | The C<code> is properly scoped in the following sense: If the assertion |
757 | is backtracked (compare L<"Backtracking">), all changes introduced after | |
758 | C<local>ization are undone, so that | |
b9ac3b5b GS |
759 | |
760 | $_ = 'a' x 8; | |
761 | m< | |
762 | (?{ $cnt = 0 }) # Initialize $cnt. | |
763 | ( | |
764 | a | |
765 | (?{ | |
766 | local $cnt = $cnt + 1; # Update $cnt, backtracking-safe. | |
767 | }) | |
768 | )* | |
769 | aaaa | |
770 | (?{ $res = $cnt }) # On success copy to non-localized | |
771 | # location. | |
772 | >x; | |
773 | ||
19799a22 | 774 | will set C<$res = 4>. Note that after the match, $cnt returns to the globally |
14218588 | 775 | introduced value, because the scopes that restrict C<local> operators |
b9ac3b5b GS |
776 | are unwound. |
777 | ||
19799a22 GS |
778 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> |
779 | switch. If I<not> used in this way, the result of evaluation of | |
780 | C<code> is put into the special variable C<$^R>. This happens | |
781 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions | |
782 | inside the same regular expression. | |
b9ac3b5b | 783 | |
19799a22 GS |
784 | The assignment to C<$^R> above is properly localized, so the old |
785 | value of C<$^R> is restored if the assertion is backtracked; compare | |
786 | L<"Backtracking">. | |
b9ac3b5b | 787 | |
61528107 SP |
788 | Due to an unfortunate implementation issue, the Perl code contained in these |
789 | blocks is treated as a compile time closure that can have seemingly bizarre | |
6bda09f9 | 790 | consequences when used with lexically scoped variables inside of subroutines |
61528107 SP |
791 | or loops. There are various workarounds for this, including simply using |
792 | global variables instead. If you are using this construct and strange results | |
6bda09f9 YO |
793 | occur then check for the use of lexically scoped variables. |
794 | ||
19799a22 GS |
795 | For reasons of security, this construct is forbidden if the regular |
796 | expression involves run-time interpolation of variables, unless the | |
797 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
798 | variables contain results of C<qr//> operator (see | |
799 | L<perlop/"qr/STRING/imosx">). | |
871b0233 | 800 | |
14218588 | 801 | This restriction is because of the wide-spread and remarkably convenient |
19799a22 | 802 | custom of using run-time determined strings as patterns. For example: |
871b0233 IZ |
803 | |
804 | $re = <>; | |
805 | chomp $re; | |
806 | $string =~ /$re/; | |
807 | ||
14218588 GS |
808 | Before Perl knew how to execute interpolated code within a pattern, |
809 | this operation was completely safe from a security point of view, | |
810 | although it could raise an exception from an illegal pattern. If | |
811 | you turn on the C<use re 'eval'>, though, it is no longer secure, | |
812 | so you should only do so if you are also using taint checking. | |
813 | Better yet, use the carefully constrained evaluation within a Safe | |
cc46d5f2 | 814 | compartment. See L<perlsec> for details about both these mechanisms. |
871b0233 | 815 | |
8988a1bb DD |
816 | Because perl's regex engine is not currently re-entrant, interpolated |
817 | code may not invoke the regex engine either directly with C<m//> or C<s///>), | |
818 | or indirectly with functions such as C<split>. | |
819 | ||
14455d6c | 820 | =item C<(??{ code })> |
d74e8afc ITB |
821 | X<(??{})> |
822 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> | |
0f5d15d6 | 823 | |
19799a22 | 824 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
825 | experimental, and may be changed without notice. Code executed that |
826 | has side effects may not perform identically from version to version | |
827 | due to the effect of future optimisations in the regex engine. | |
0f5d15d6 | 828 | |
19799a22 GS |
829 | This is a "postponed" regular subexpression. The C<code> is evaluated |
830 | at run time, at the moment this subexpression may match. The result | |
831 | of evaluation is considered as a regular expression and matched as | |
61528107 | 832 | if it were inserted instead of this construct. Note that this means |
6bda09f9 YO |
833 | that the contents of capture buffers defined inside an eval'ed pattern |
834 | are not available outside of the pattern, and vice versa, there is no | |
835 | way for the inner pattern to refer to a capture buffer defined outside. | |
836 | Thus, | |
837 | ||
838 | ('a' x 100)=~/(??{'(.)' x 100})/ | |
839 | ||
81714fb9 | 840 | B<will> match, it will B<not> set $1. |
0f5d15d6 | 841 | |
428594d9 | 842 | The C<code> is not interpolated. As before, the rules to determine |
19799a22 GS |
843 | where the C<code> ends are currently somewhat convoluted. |
844 | ||
845 | The following pattern matches a parenthesized group: | |
0f5d15d6 IZ |
846 | |
847 | $re = qr{ | |
848 | \( | |
849 | (?: | |
850 | (?> [^()]+ ) # Non-parens without backtracking | |
851 | | | |
14455d6c | 852 | (??{ $re }) # Group with matching parens |
0f5d15d6 IZ |
853 | )* |
854 | \) | |
855 | }x; | |
856 | ||
6bda09f9 YO |
857 | See also C<(?PARNO)> for a different, more efficient way to accomplish |
858 | the same task. | |
859 | ||
8988a1bb DD |
860 | Because perl's regex engine is not currently re-entrant, delayed |
861 | code may not invoke the regex engine either directly with C<m//> or C<s///>), | |
862 | or indirectly with functions such as C<split>. | |
863 | ||
6bda09f9 | 864 | Recursing deeper than 50 times without consuming any input string will |
61528107 | 865 | result in a fatal error. The maximum depth is compiled into perl, so |
6bda09f9 YO |
866 | changing it requires a custom build. |
867 | ||
894be9b7 YO |
868 | =item C<(?PARNO)> C<(?R)> C<(?0)> |
869 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> | |
6bda09f9 YO |
870 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
871 | ||
81714fb9 YO |
872 | Similar to C<(??{ code })> except it does not involve compiling any code, |
873 | instead it treats the contents of a capture buffer as an independent | |
61528107 | 874 | pattern that must match at the current position. Capture buffers |
81714fb9 | 875 | contained by the pattern will have the value as determined by the |
6bda09f9 YO |
876 | outermost recursion. |
877 | ||
894be9b7 YO |
878 | PARNO is a sequence of digits (not starting with 0) whose value reflects |
879 | the paren-number of the capture buffer to recurse to. C<(?R)> recurses to | |
880 | the beginning of the whole pattern. C<(?0)> is an alternate syntax for | |
881 | C<(?R)>. | |
6bda09f9 | 882 | |
81714fb9 | 883 | The following pattern matches a function foo() which may contain |
f145b7e9 | 884 | balanced parentheses as the argument. |
6bda09f9 YO |
885 | |
886 | $re = qr{ ( # paren group 1 (full function) | |
81714fb9 | 887 | foo |
6bda09f9 YO |
888 | ( # paren group 2 (parens) |
889 | \( | |
890 | ( # paren group 3 (contents of parens) | |
891 | (?: | |
892 | (?> [^()]+ ) # Non-parens without backtracking | |
893 | | | |
894 | (?2) # Recurse to start of paren group 2 | |
895 | )* | |
896 | ) | |
897 | \) | |
898 | ) | |
899 | ) | |
900 | }x; | |
901 | ||
902 | If the pattern was used as follows | |
903 | ||
904 | 'foo(bar(baz)+baz(bop))'=~/$re/ | |
905 | and print "\$1 = $1\n", | |
906 | "\$2 = $2\n", | |
907 | "\$3 = $3\n"; | |
908 | ||
909 | the output produced should be the following: | |
910 | ||
911 | $1 = foo(bar(baz)+baz(bop)) | |
912 | $2 = (bar(baz)+baz(bop)) | |
81714fb9 | 913 | $3 = bar(baz)+baz(bop) |
6bda09f9 | 914 | |
81714fb9 | 915 | If there is no corresponding capture buffer defined, then it is a |
61528107 | 916 | fatal error. Recursing deeper than 50 times without consuming any input |
81714fb9 | 917 | string will also result in a fatal error. The maximum depth is compiled |
6bda09f9 YO |
918 | into perl, so changing it requires a custom build. |
919 | ||
81714fb9 | 920 | B<Note> that this pattern does not behave the same way as the equivalent |
6bda09f9 YO |
921 | PCRE or Python construct of the same form. In perl you can backtrack into |
922 | a recursed group, in PCRE and Python the recursed into group is treated | |
81714fb9 YO |
923 | as atomic. Also, constructs like (?i:(?1)) or (?:(?i)(?1)) do not affect |
924 | the pattern being recursed into. | |
6bda09f9 | 925 | |
894be9b7 YO |
926 | =item C<(?&NAME)> |
927 | X<(?&NAME)> | |
928 | ||
929 | Recurse to a named subpattern. Identical to (?PARNO) except that the | |
930 | parenthesis to recurse to is determined by name. If multiple parens have | |
931 | the same name, then it recurses to the leftmost. | |
932 | ||
933 | It is an error to refer to a name that is not declared somewhere in the | |
934 | pattern. | |
935 | ||
286f584a RGS |
936 | =item C<(?FAIL)> C<(?F)> |
937 | X<(?FAIL)> X<(?F)> | |
938 | ||
939 | This pattern matches nothing and always fails. It can be used to force the | |
940 | engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In | |
941 | fact, C<(?!)> gets optimised into C<(?FAIL)> internally. | |
942 | ||
943 | It is probably useful only when combined with C<(?{})> or C<(??{})>. | |
944 | ||
c47ff5f1 | 945 | =item C<< (?>pattern) >> |
6bda09f9 | 946 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 | 947 | |
19799a22 GS |
948 | An "independent" subexpression, one which matches the substring |
949 | that a I<standalone> C<pattern> would match if anchored at the given | |
9da458fc | 950 | position, and it matches I<nothing other than this substring>. This |
19799a22 GS |
951 | construct is useful for optimizations of what would otherwise be |
952 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). | |
9da458fc IZ |
953 | It may also be useful in places where the "grab all you can, and do not |
954 | give anything back" semantic is desirable. | |
19799a22 | 955 | |
c47ff5f1 | 956 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 GS |
957 | (anchored at the beginning of string, as above) will match I<all> |
958 | characters C<a> at the beginning of string, leaving no C<a> for | |
959 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, | |
960 | since the match of the subgroup C<a*> is influenced by the following | |
961 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside | |
962 | C<a*ab> will match fewer characters than a standalone C<a*>, since | |
963 | this makes the tail match. | |
964 | ||
c47ff5f1 | 965 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
19799a22 GS |
966 | C<(?=(pattern))\1>. This matches the same substring as a standalone |
967 | C<a+>, and the following C<\1> eats the matched string; it therefore | |
c47ff5f1 | 968 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 GS |
969 | (The difference between these two constructs is that the second one |
970 | uses a capturing group, thus shifting ordinals of backreferences | |
971 | in the rest of a regular expression.) | |
972 | ||
973 | Consider this pattern: | |
c277df42 | 974 | |
871b0233 IZ |
975 | m{ \( |
976 | ( | |
9da458fc | 977 | [^()]+ # x+ |
871b0233 IZ |
978 | | |
979 | \( [^()]* \) | |
980 | )+ | |
981 | \) | |
982 | }x | |
5a964f20 | 983 | |
19799a22 GS |
984 | That will efficiently match a nonempty group with matching parentheses |
985 | two levels deep or less. However, if there is no such group, it | |
986 | will take virtually forever on a long string. That's because there | |
987 | are so many different ways to split a long string into several | |
988 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
989 | to a subpattern of the above pattern. Consider how the pattern | |
990 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
991 | seconds, but that each extra letter doubles this time. This | |
992 | exponential performance will make it appear that your program has | |
14218588 | 993 | hung. However, a tiny change to this pattern |
5a964f20 | 994 | |
871b0233 IZ |
995 | m{ \( |
996 | ( | |
9da458fc | 997 | (?> [^()]+ ) # change x+ above to (?> x+ ) |
871b0233 IZ |
998 | | |
999 | \( [^()]* \) | |
1000 | )+ | |
1001 | \) | |
1002 | }x | |
c277df42 | 1003 | |
c47ff5f1 | 1004 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 TC |
1005 | this yourself would be a productive exercise), but finishes in a fourth |
1006 | the time when used on a similar string with 1000000 C<a>s. Be aware, | |
1007 | however, that this pattern currently triggers a warning message under | |
9f1b1f2d | 1008 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b | 1009 | C<"matches null string many times in regex">. |
c277df42 | 1010 | |
c47ff5f1 | 1011 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 | 1012 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 IZ |
1013 | This was only 4 times slower on a string with 1000000 C<a>s. |
1014 | ||
9da458fc IZ |
1015 | The "grab all you can, and do not give anything back" semantic is desirable |
1016 | in many situations where on the first sight a simple C<()*> looks like | |
1017 | the correct solution. Suppose we parse text with comments being delimited | |
1018 | by C<#> followed by some optional (horizontal) whitespace. Contrary to | |
4375e838 | 1019 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc IZ |
1020 | the comment delimiter, because it may "give up" some whitespace if |
1021 | the remainder of the pattern can be made to match that way. The correct | |
1022 | answer is either one of these: | |
1023 | ||
1024 | (?>#[ \t]*) | |
1025 | #[ \t]*(?![ \t]) | |
1026 | ||
1027 | For example, to grab non-empty comments into $1, one should use either | |
1028 | one of these: | |
1029 | ||
1030 | / (?> \# [ \t]* ) ( .+ ) /x; | |
1031 | / \# [ \t]* ( [^ \t] .* ) /x; | |
1032 | ||
1033 | Which one you pick depends on which of these expressions better reflects | |
1034 | the above specification of comments. | |
1035 | ||
6bda09f9 YO |
1036 | In some literature this construct is called "atomic matching" or |
1037 | "possessive matching". | |
1038 | ||
b9b4dddf YO |
1039 | Possessive quantifiers are equivalent to putting the item they are applied |
1040 | to inside of one of these constructs. The following equivalences apply: | |
1041 | ||
1042 | Quantifier Form Bracketing Form | |
1043 | --------------- --------------- | |
1044 | PAT*+ (?>PAT*) | |
1045 | PAT++ (?>PAT+) | |
1046 | PAT?+ (?>PAT?) | |
1047 | PAT{min,max}+ (?>PAT{min,max}) | |
1048 | ||
54612592 YO |
1049 | =item C<(?COMMIT)> |
1050 | X<(?COMMIT)> | |
1051 | ||
1052 | This zero-width pattern commits the match at the current point, preventing | |
1053 | the engine from back-tracking on failure to the left of the commit point. | |
1054 | Consider the pattern C<A (?COMMIT) B>, where A and B are complex patterns. | |
1055 | Until the C<(?COMMIT)> is reached, A may backtrack as necessary to match. | |
1056 | Once it is reached, matching continues in B, which may also backtrack as | |
1057 | necessary; however, should B not match, then no further backtracking will | |
1058 | take place, and the pattern will fail outright at that starting position. | |
1059 | ||
1060 | The following example counts all the possible matching strings in a | |
1061 | pattern (without actually matching any of them). | |
1062 | ||
1063 | 'aaab'=~/a+b?(?{print "$&\n"; $count++})(?FAIL)/; | |
1064 | print "Count=$count\n"; | |
1065 | ||
1066 | which produces: | |
1067 | ||
1068 | aaab | |
1069 | aaa | |
1070 | aa | |
1071 | a | |
1072 | aab | |
1073 | aa | |
1074 | a | |
1075 | ab | |
1076 | a | |
1077 | Count=9 | |
1078 | ||
1079 | If we add a C<(?COMMIT)> before the count like the following | |
1080 | ||
1081 | 'aaab'=~/a+b?(?COMMIT)(?{print "$&\n"; $count++})(?FAIL)/; | |
1082 | print "Count=$count\n"; | |
1083 | ||
1084 | we prevent backtracking and find the count of the longest matching | |
1085 | at each matching startpoint like so: | |
1086 | ||
1087 | aaab | |
1088 | aab | |
1089 | ab | |
1090 | Count=3 | |
1091 | ||
1092 | Any number of C<(?COMMIT)> assertions may be used in a pattern. | |
1093 | ||
1094 | See also C<< (?>pattern) >> and possessive quantifiers for other | |
1095 | ways to control backtracking. | |
1096 | ||
24b23f37 YO |
1097 | =item C<(?CUT)> |
1098 | X<(?CUT)> | |
1099 | ||
1100 | This zero-width pattern is similar to C<(?COMMIT)>, except that on | |
1101 | failure it also signifies that whatever text that was matched leading | |
1102 | up to the C<(?CUT)> pattern cannot match, I<even from another | |
1103 | starting point>. | |
1104 | ||
1105 | Compare the following to the examples in C<(?COMMIT)>, note the string | |
1106 | is twice as long: | |
1107 | ||
1108 | 'aaabaaab'=~/a+b?(?CUT)(?{print "$&\n"; $count++})(?FAIL)/; | |
1109 | print "Count=$count\n"; | |
1110 | ||
1111 | outputs | |
1112 | ||
1113 | aaab | |
1114 | aaab | |
1115 | Count=2 | |
1116 | ||
1117 | Once the 'aaab' at the start of the string has matched and the C<(?CUT)> | |
1118 | executed the next startpoint will be where the cursor was when the | |
1119 | C<(?CUT)> was executed. | |
1120 | ||
1121 | =item C<(?ERROR)> | |
1122 | X<(?ERROR)> | |
1123 | ||
1124 | This zero-width pattern is similar to C<(?CUT)> except that it causes | |
1125 | the match to fail outright. No attempts to match will occur again. | |
1126 | ||
1127 | 'aaabaaab'=~/a+b?(?ERROR)(?{print "$&\n"; $count++})(?FAIL)/; | |
1128 | print "Count=$count\n"; | |
1129 | ||
1130 | outputs | |
1131 | ||
1132 | aaab | |
1133 | Count=1 | |
1134 | ||
1135 | In other words, once the C<(?ERROR)> has been entered and then pattern | |
1136 | does not match then the regex engine will not try any further matching at | |
1137 | all on the rest of the string. | |
1138 | ||
5a964f20 | 1139 | =item C<(?(condition)yes-pattern|no-pattern)> |
d74e8afc | 1140 | X<(?()> |
c277df42 | 1141 | |
5a964f20 | 1142 | =item C<(?(condition)yes-pattern)> |
c277df42 IZ |
1143 | |
1144 | Conditional expression. C<(condition)> should be either an integer in | |
1145 | parentheses (which is valid if the corresponding pair of parentheses | |
0a4db386 YO |
1146 | matched), a look-ahead/look-behind/evaluate zero-width assertion, a |
1147 | name in angle brackets or single quotes (which is valid if a buffer | |
1148 | with the given name matched), the special symbol (R) (true when | |
1149 | evaluated inside of recursion or eval). Additionally the R may be | |
1150 | followed by a number, (which will be true when evaluated when recursing | |
1151 | inside of the appropriate group), or by C<&NAME> in which case it will | |
9af228c6 YO |
1152 | be true only when evaluated during recursion in the named group. |
1153 | ||
1154 | Here's a summary of the possible predicates: | |
1155 | ||
1156 | =over 4 | |
1157 | ||
1158 | =item (1) (2) ... | |
1159 | ||
1160 | Checks if the numbered capturing buffer has matched something. | |
1161 | ||
1162 | =item (<NAME>) ('NAME') | |
1163 | ||
1164 | Checks if a buffer with the given name has matched something. | |
1165 | ||
1166 | =item (?{ CODE }) | |
1167 | ||
1168 | Treats the code block as the condition | |
1169 | ||
1170 | =item (R) | |
1171 | ||
1172 | Checks if the expression has been evaluated inside of recursion. | |
1173 | ||
1174 | =item (R1) (R2) ... | |
1175 | ||
1176 | Checks if the expression has been evaluated while executing directly | |
1177 | inside of the n-th capture group. This check is the regex equivalent of | |
1178 | ||
1179 | if ((caller(0))[3] eq 'subname') { .. } | |
1180 | ||
1181 | In other words, it does not check the full recursion stack. | |
1182 | ||
1183 | =item (R&NAME) | |
1184 | ||
1185 | Similar to C<(R1)>, this predicate checks to see if we're executing | |
1186 | directly inside of the leftmost group with a given name (this is the same | |
1187 | logic used by C<(?&NAME)> to disambiguate). It does not check the full | |
1188 | stack, but only the name of the innermost active recursion. | |
1189 | ||
1190 | =item (DEFINE) | |
1191 | ||
1192 | In this case, the yes-pattern is never directly executed, and no | |
1193 | no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient. | |
1194 | See below for details. | |
1195 | ||
1196 | =back | |
c277df42 | 1197 | |
19799a22 | 1198 | For example: |
c277df42 | 1199 | |
0a4db386 YO |
1200 | m{ ( \( )? |
1201 | [^()]+ | |
1202 | (?(1) \) ) | |
871b0233 | 1203 | }x |
c277df42 IZ |
1204 | |
1205 | matches a chunk of non-parentheses, possibly included in parentheses | |
1206 | themselves. | |
a0d0e21e | 1207 | |
9af228c6 YO |
1208 | A special form is the C<(DEFINE)> predicate, which never executes directly |
1209 | its yes-pattern, and does not allow a no-pattern. This allows to define | |
1210 | subpatterns which will be executed only by using the recursion mechanism. | |
1211 | This way, you can define a set of regular expression rules that can be | |
1212 | bundled into any pattern you choose. | |
1213 | ||
1214 | It is recommended that for this usage you put the DEFINE block at the | |
1215 | end of the pattern, and that you name any subpatterns defined within it. | |
1216 | ||
1217 | Also, it's worth noting that patterns defined this way probably will | |
1218 | not be as efficient, as the optimiser is not very clever about | |
1219 | handling them. YMMV. | |
1220 | ||
1221 | An example of how this might be used is as follows: | |
1222 | ||
1223 | /(?<NAME>(&NAME_PAT))(?<ADDR>(&ADDRESS_PAT)) | |
1224 | (?(DEFINE) | |
1225 | (<NAME_PAT>....) | |
1226 | (<ADRESS_PAT>....) | |
1227 | )/x | |
1228 | ||
1229 | Note that capture buffers matched inside of recursion are not accessible | |
1230 | after the recursion returns, so the extra layer of capturing buffers are | |
1231 | necessary. Thus C<$+{NAME_PAT}> would not be defined even though | |
1232 | C<$+{NAME}> would be. | |
0a4db386 | 1233 | |
a0d0e21e LW |
1234 | =back |
1235 | ||
c07a80fd | 1236 | =head2 Backtracking |
d74e8afc | 1237 | X<backtrack> X<backtracking> |
c07a80fd | 1238 | |
35a734be IZ |
1239 | NOTE: This section presents an abstract approximation of regular |
1240 | expression behavior. For a more rigorous (and complicated) view of | |
1241 | the rules involved in selecting a match among possible alternatives, | |
1242 | see L<Combining pieces together>. | |
1243 | ||
c277df42 | 1244 | A fundamental feature of regular expression matching involves the |
5a964f20 | 1245 | notion called I<backtracking>, which is currently used (when needed) |
c277df42 | 1246 | by all regular expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc IZ |
1247 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
1248 | internally, but the general principle outlined here is valid. | |
c07a80fd | 1249 | |
1250 | For a regular expression to match, the I<entire> regular expression must | |
1251 | match, not just part of it. So if the beginning of a pattern containing a | |
1252 | quantifier succeeds in a way that causes later parts in the pattern to | |
1253 | fail, the matching engine backs up and recalculates the beginning | |
1254 | part--that's why it's called backtracking. | |
1255 | ||
1256 | Here is an example of backtracking: Let's say you want to find the | |
1257 | word following "foo" in the string "Food is on the foo table.": | |
1258 | ||
1259 | $_ = "Food is on the foo table."; | |
1260 | if ( /\b(foo)\s+(\w+)/i ) { | |
1261 | print "$2 follows $1.\n"; | |
1262 | } | |
1263 | ||
1264 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
1265 | finds a possible match right at the beginning of the string, and loads up | |
1266 | $1 with "Foo". However, as soon as the matching engine sees that there's | |
1267 | no whitespace following the "Foo" that it had saved in $1, it realizes its | |
68dc0745 | 1268 | mistake and starts over again one character after where it had the |
c07a80fd | 1269 | tentative match. This time it goes all the way until the next occurrence |
1270 | of "foo". The complete regular expression matches this time, and you get | |
1271 | the expected output of "table follows foo." | |
1272 | ||
1273 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
1274 | everything between "foo" and "bar". Initially, you write something | |
1275 | like this: | |
1276 | ||
1277 | $_ = "The food is under the bar in the barn."; | |
1278 | if ( /foo(.*)bar/ ) { | |
1279 | print "got <$1>\n"; | |
1280 | } | |
1281 | ||
1282 | Which perhaps unexpectedly yields: | |
1283 | ||
1284 | got <d is under the bar in the > | |
1285 | ||
1286 | That's because C<.*> was greedy, so you get everything between the | |
14218588 | 1287 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd | 1288 | to use minimal matching to make sure you get the text between a "foo" |
1289 | and the first "bar" thereafter. | |
1290 | ||
1291 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
1292 | got <d is under the > | |
1293 | ||
1294 | Here's another example: let's say you'd like to match a number at the end | |
b6e13d97 | 1295 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd | 1296 | So you write this: |
1297 | ||
1298 | $_ = "I have 2 numbers: 53147"; | |
1299 | if ( /(.*)(\d*)/ ) { # Wrong! | |
1300 | print "Beginning is <$1>, number is <$2>.\n"; | |
1301 | } | |
1302 | ||
1303 | That won't work at all, because C<.*> was greedy and gobbled up the | |
1304 | whole string. As C<\d*> can match on an empty string the complete | |
1305 | regular expression matched successfully. | |
1306 | ||
8e1088bc | 1307 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd | 1308 | |
1309 | Here are some variants, most of which don't work: | |
1310 | ||
1311 | $_ = "I have 2 numbers: 53147"; | |
1312 | @pats = qw{ | |
1313 | (.*)(\d*) | |
1314 | (.*)(\d+) | |
1315 | (.*?)(\d*) | |
1316 | (.*?)(\d+) | |
1317 | (.*)(\d+)$ | |
1318 | (.*?)(\d+)$ | |
1319 | (.*)\b(\d+)$ | |
1320 | (.*\D)(\d+)$ | |
1321 | }; | |
1322 | ||
1323 | for $pat (@pats) { | |
1324 | printf "%-12s ", $pat; | |
1325 | if ( /$pat/ ) { | |
1326 | print "<$1> <$2>\n"; | |
1327 | } else { | |
1328 | print "FAIL\n"; | |
1329 | } | |
1330 | } | |
1331 | ||
1332 | That will print out: | |
1333 | ||
1334 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
1335 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
1336 | (.*?)(\d*) <> <> | |
1337 | (.*?)(\d+) <I have > <2> | |
1338 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
1339 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
1340 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
1341 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
1342 | ||
1343 | As you see, this can be a bit tricky. It's important to realize that a | |
1344 | regular expression is merely a set of assertions that gives a definition | |
1345 | of success. There may be 0, 1, or several different ways that the | |
1346 | definition might succeed against a particular string. And if there are | |
5a964f20 TC |
1347 | multiple ways it might succeed, you need to understand backtracking to |
1348 | know which variety of success you will achieve. | |
c07a80fd | 1349 | |
19799a22 | 1350 | When using look-ahead assertions and negations, this can all get even |
8b19b778 | 1351 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd | 1352 | followed by "123". You might try to write that as |
1353 | ||
871b0233 IZ |
1354 | $_ = "ABC123"; |
1355 | if ( /^\D*(?!123)/ ) { # Wrong! | |
1356 | print "Yup, no 123 in $_\n"; | |
1357 | } | |
c07a80fd | 1358 | |
1359 | But that isn't going to match; at least, not the way you're hoping. It | |
1360 | claims that there is no 123 in the string. Here's a clearer picture of | |
9b9391b2 | 1361 | why that pattern matches, contrary to popular expectations: |
c07a80fd | 1362 | |
4358a253 SS |
1363 | $x = 'ABC123'; |
1364 | $y = 'ABC445'; | |
c07a80fd | 1365 | |
4358a253 SS |
1366 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
1367 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; | |
c07a80fd | 1368 | |
4358a253 SS |
1369 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; |
1370 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; | |
c07a80fd | 1371 | |
1372 | This prints | |
1373 | ||
1374 | 2: got ABC | |
1375 | 3: got AB | |
1376 | 4: got ABC | |
1377 | ||
5f05dabc | 1378 | You might have expected test 3 to fail because it seems to a more |
c07a80fd | 1379 | general purpose version of test 1. The important difference between |
1380 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
1381 | backtracking, whereas test 1 will not. What's happening is | |
1382 | that you've asked "Is it true that at the start of $x, following 0 or more | |
5f05dabc | 1383 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd | 1384 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 | 1385 | fail. |
14218588 | 1386 | |
c07a80fd | 1387 | The search engine will initially match C<\D*> with "ABC". Then it will |
14218588 | 1388 | try to match C<(?!123> with "123", which fails. But because |
c07a80fd | 1389 | a quantifier (C<\D*>) has been used in the regular expression, the |
1390 | search engine can backtrack and retry the match differently | |
54310121 | 1391 | in the hope of matching the complete regular expression. |
c07a80fd | 1392 | |
5a964f20 TC |
1393 | The pattern really, I<really> wants to succeed, so it uses the |
1394 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
c07a80fd | 1395 | time. Now there's indeed something following "AB" that is not |
14218588 | 1396 | "123". It's "C123", which suffices. |
c07a80fd | 1397 | |
14218588 GS |
1398 | We can deal with this by using both an assertion and a negation. |
1399 | We'll say that the first part in $1 must be followed both by a digit | |
1400 | and by something that's not "123". Remember that the look-aheads | |
1401 | are zero-width expressions--they only look, but don't consume any | |
1402 | of the string in their match. So rewriting this way produces what | |
c07a80fd | 1403 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
1404 | ||
4358a253 SS |
1405 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; |
1406 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; | |
c07a80fd | 1407 | |
1408 | 6: got ABC | |
1409 | ||
5a964f20 | 1410 | In other words, the two zero-width assertions next to each other work as though |
19799a22 | 1411 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd | 1412 | matches only if you're at the beginning of the line AND the end of the |
1413 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
1414 | regular expressions always means AND, except when you write an explicit OR | |
1415 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
1416 | although the attempted matches are made at different positions because "a" | |
1417 | is not a zero-width assertion, but a one-width assertion. | |
1418 | ||
19799a22 | 1419 | B<WARNING>: particularly complicated regular expressions can take |
14218588 | 1420 | exponential time to solve because of the immense number of possible |
9da458fc IZ |
1421 | ways they can use backtracking to try match. For example, without |
1422 | internal optimizations done by the regular expression engine, this will | |
1423 | take a painfully long time to run: | |
c07a80fd | 1424 | |
e1901655 IZ |
1425 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
1426 | ||
1427 | And if you used C<*>'s in the internal groups instead of limiting them | |
1428 | to 0 through 5 matches, then it would take forever--or until you ran | |
1429 | out of stack space. Moreover, these internal optimizations are not | |
1430 | always applicable. For example, if you put C<{0,5}> instead of C<*> | |
1431 | on the external group, no current optimization is applicable, and the | |
1432 | match takes a long time to finish. | |
c07a80fd | 1433 | |
9da458fc IZ |
1434 | A powerful tool for optimizing such beasts is what is known as an |
1435 | "independent group", | |
c47ff5f1 | 1436 | which does not backtrack (see L<C<< (?>pattern) >>>). Note also that |
9da458fc | 1437 | zero-length look-ahead/look-behind assertions will not backtrack to make |
14218588 GS |
1438 | the tail match, since they are in "logical" context: only |
1439 | whether they match is considered relevant. For an example | |
9da458fc | 1440 | where side-effects of look-ahead I<might> have influenced the |
c47ff5f1 | 1441 | following match, see L<C<< (?>pattern) >>>. |
c277df42 | 1442 | |
a0d0e21e | 1443 | =head2 Version 8 Regular Expressions |
d74e8afc | 1444 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> |
a0d0e21e | 1445 | |
5a964f20 | 1446 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e LW |
1447 | routines, here are the pattern-matching rules not described above. |
1448 | ||
54310121 | 1449 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e | 1450 | with a special meaning described here or above. You can cause |
5a964f20 | 1451 | characters that normally function as metacharacters to be interpreted |
5f05dabc | 1452 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
a0d0e21e LW |
1453 | character; "\\" matches a "\"). A series of characters matches that |
1454 | series of characters in the target string, so the pattern C<blurfl> | |
1455 | would match "blurfl" in the target string. | |
1456 | ||
1457 | You can specify a character class, by enclosing a list of characters | |
5a964f20 | 1458 | in C<[]>, which will match any one character from the list. If the |
a0d0e21e | 1459 | first character after the "[" is "^", the class matches any character not |
14218588 | 1460 | in the list. Within a list, the "-" character specifies a |
5a964f20 | 1461 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 GS |
1462 | inclusive. If you want either "-" or "]" itself to be a member of a |
1463 | class, put it at the start of the list (possibly after a "^"), or | |
1464 | escape it with a backslash. "-" is also taken literally when it is | |
1465 | at the end of the list, just before the closing "]". (The | |
84850974 DD |
1466 | following all specify the same class of three characters: C<[-az]>, |
1467 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which | |
fb55449c JH |
1468 | specifies a class containing twenty-six characters, even on EBCDIC |
1469 | based coded character sets.) Also, if you try to use the character | |
1470 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of | |
1471 | a range, that's not a range, the "-" is understood literally. | |
a0d0e21e | 1472 | |
8ada0baa JH |
1473 | Note also that the whole range idea is rather unportable between |
1474 | character sets--and even within character sets they may cause results | |
1475 | you probably didn't expect. A sound principle is to use only ranges | |
1476 | that begin from and end at either alphabets of equal case ([a-e], | |
1477 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, | |
1478 | spell out the character sets in full. | |
1479 | ||
54310121 | 1480 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e LW |
1481 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
1482 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string | |
fb55449c JH |
1483 | of octal digits, matches the character whose coded character set value |
1484 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, | |
1485 | matches the character whose numeric value is I<nn>. The expression \cI<x> | |
1486 | matches the character control-I<x>. Finally, the "." metacharacter | |
1487 | matches any character except "\n" (unless you use C</s>). | |
a0d0e21e LW |
1488 | |
1489 | You can specify a series of alternatives for a pattern using "|" to | |
1490 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", | |
5a964f20 | 1491 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e LW |
1492 | first alternative includes everything from the last pattern delimiter |
1493 | ("(", "[", or the beginning of the pattern) up to the first "|", and | |
1494 | the last alternative contains everything from the last "|" to the next | |
14218588 GS |
1495 | pattern delimiter. That's why it's common practice to include |
1496 | alternatives in parentheses: to minimize confusion about where they | |
a3cb178b GS |
1497 | start and end. |
1498 | ||
5a964f20 | 1499 | Alternatives are tried from left to right, so the first |
a3cb178b GS |
1500 | alternative found for which the entire expression matches, is the one that |
1501 | is chosen. This means that alternatives are not necessarily greedy. For | |
628afcb5 | 1502 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b GS |
1503 | part will match, as that is the first alternative tried, and it successfully |
1504 | matches the target string. (This might not seem important, but it is | |
1505 | important when you are capturing matched text using parentheses.) | |
1506 | ||
5a964f20 | 1507 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b | 1508 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e | 1509 | |
14218588 GS |
1510 | Within a pattern, you may designate subpatterns for later reference |
1511 | by enclosing them in parentheses, and you may refer back to the | |
1512 | I<n>th subpattern later in the pattern using the metacharacter | |
1513 | \I<n>. Subpatterns are numbered based on the left to right order | |
1514 | of their opening parenthesis. A backreference matches whatever | |
1515 | actually matched the subpattern in the string being examined, not | |
1516 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will | |
1517 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern | |
1518 | 1 matched "0x", even though the rule C<0|0x> could potentially match | |
1519 | the leading 0 in the second number. | |
cb1a09d0 | 1520 | |
19799a22 | 1521 | =head2 Warning on \1 vs $1 |
cb1a09d0 | 1522 | |
5a964f20 | 1523 | Some people get too used to writing things like: |
cb1a09d0 AD |
1524 | |
1525 | $pattern =~ s/(\W)/\\\1/g; | |
1526 | ||
1527 | This is grandfathered for the RHS of a substitute to avoid shocking the | |
1528 | B<sed> addicts, but it's a dirty habit to get into. That's because in | |
d1be9408 | 1529 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 AD |
1530 | the usual double-quoted string means a control-A. The customary Unix |
1531 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
1532 | of doing that, you get yourself into trouble if you then add an C</e> | |
1533 | modifier. | |
1534 | ||
5a964f20 | 1535 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 AD |
1536 | |
1537 | Or if you try to do | |
1538 | ||
1539 | s/(\d+)/\1000/; | |
1540 | ||
1541 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
14218588 | 1542 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 AD |
1543 | with the operation of matching a backreference. Certainly they mean two |
1544 | different things on the I<left> side of the C<s///>. | |
9fa51da4 | 1545 | |
c84d73f1 IZ |
1546 | =head2 Repeated patterns matching zero-length substring |
1547 | ||
19799a22 | 1548 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 IZ |
1549 | |
1550 | Regular expressions provide a terse and powerful programming language. As | |
1551 | with most other power tools, power comes together with the ability | |
1552 | to wreak havoc. | |
1553 | ||
1554 | A common abuse of this power stems from the ability to make infinite | |
628afcb5 | 1555 | loops using regular expressions, with something as innocuous as: |
c84d73f1 IZ |
1556 | |
1557 | 'foo' =~ m{ ( o? )* }x; | |
1558 | ||
1559 | The C<o?> can match at the beginning of C<'foo'>, and since the position | |
1560 | in the string is not moved by the match, C<o?> would match again and again | |
14218588 | 1561 | because of the C<*> modifier. Another common way to create a similar cycle |
c84d73f1 IZ |
1562 | is with the looping modifier C<//g>: |
1563 | ||
1564 | @matches = ( 'foo' =~ m{ o? }xg ); | |
1565 | ||
1566 | or | |
1567 | ||
1568 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
1569 | ||
1570 | or the loop implied by split(). | |
1571 | ||
1572 | However, long experience has shown that many programming tasks may | |
14218588 GS |
1573 | be significantly simplified by using repeated subexpressions that |
1574 | may match zero-length substrings. Here's a simple example being: | |
c84d73f1 IZ |
1575 | |
1576 | @chars = split //, $string; # // is not magic in split | |
1577 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / | |
1578 | ||
9da458fc | 1579 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 IZ |
1580 | the infinite loop>. The rules for this are different for lower-level |
1581 | loops given by the greedy modifiers C<*+{}>, and for higher-level | |
1582 | ones like the C</g> modifier or split() operator. | |
1583 | ||
19799a22 GS |
1584 | The lower-level loops are I<interrupted> (that is, the loop is |
1585 | broken) when Perl detects that a repeated expression matched a | |
1586 | zero-length substring. Thus | |
c84d73f1 IZ |
1587 | |
1588 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
1589 | ||
1590 | is made equivalent to | |
1591 | ||
1592 | m{ (?: NON_ZERO_LENGTH )* | |
1593 | | | |
1594 | (?: ZERO_LENGTH )? | |
1595 | }x; | |
1596 | ||
1597 | The higher level-loops preserve an additional state between iterations: | |
1598 | whether the last match was zero-length. To break the loop, the following | |
1599 | match after a zero-length match is prohibited to have a length of zero. | |
1600 | This prohibition interacts with backtracking (see L<"Backtracking">), | |
1601 | and so the I<second best> match is chosen if the I<best> match is of | |
1602 | zero length. | |
1603 | ||
19799a22 | 1604 | For example: |
c84d73f1 IZ |
1605 | |
1606 | $_ = 'bar'; | |
1607 | s/\w??/<$&>/g; | |
1608 | ||
20fb949f | 1609 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
c84d73f1 IZ |
1610 | match given by non-greedy C<??> is the zero-length match, and the I<second |
1611 | best> match is what is matched by C<\w>. Thus zero-length matches | |
1612 | alternate with one-character-long matches. | |
1613 | ||
1614 | Similarly, for repeated C<m/()/g> the second-best match is the match at the | |
1615 | position one notch further in the string. | |
1616 | ||
19799a22 | 1617 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 | 1618 | the matched string, and is reset by each assignment to pos(). |
9da458fc IZ |
1619 | Zero-length matches at the end of the previous match are ignored |
1620 | during C<split>. | |
c84d73f1 | 1621 | |
35a734be IZ |
1622 | =head2 Combining pieces together |
1623 | ||
1624 | Each of the elementary pieces of regular expressions which were described | |
1625 | before (such as C<ab> or C<\Z>) could match at most one substring | |
1626 | at the given position of the input string. However, in a typical regular | |
1627 | expression these elementary pieces are combined into more complicated | |
1628 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc | |
1629 | (in these examples C<S> and C<T> are regular subexpressions). | |
1630 | ||
1631 | Such combinations can include alternatives, leading to a problem of choice: | |
1632 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
1633 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
1634 | actually matched is the concept of backtracking (see L<"Backtracking">). | |
1635 | However, this description is too low-level and makes you think | |
1636 | in terms of a particular implementation. | |
1637 | ||
1638 | Another description starts with notions of "better"/"worse". All the | |
1639 | substrings which may be matched by the given regular expression can be | |
1640 | sorted from the "best" match to the "worst" match, and it is the "best" | |
1641 | match which is chosen. This substitutes the question of "what is chosen?" | |
1642 | by the question of "which matches are better, and which are worse?". | |
1643 | ||
1644 | Again, for elementary pieces there is no such question, since at most | |
1645 | one match at a given position is possible. This section describes the | |
1646 | notion of better/worse for combining operators. In the description | |
1647 | below C<S> and C<T> are regular subexpressions. | |
1648 | ||
13a2d996 | 1649 | =over 4 |
35a734be IZ |
1650 | |
1651 | =item C<ST> | |
1652 | ||
1653 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are | |
1654 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings | |
1655 | which can be matched by C<T>. | |
1656 | ||
1657 | If C<A> is better match for C<S> than C<A'>, C<AB> is a better | |
1658 | match than C<A'B'>. | |
1659 | ||
1660 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if | |
1661 | C<B> is better match for C<T> than C<B'>. | |
1662 | ||
1663 | =item C<S|T> | |
1664 | ||
1665 | When C<S> can match, it is a better match than when only C<T> can match. | |
1666 | ||
1667 | Ordering of two matches for C<S> is the same as for C<S>. Similar for | |
1668 | two matches for C<T>. | |
1669 | ||
1670 | =item C<S{REPEAT_COUNT}> | |
1671 | ||
1672 | Matches as C<SSS...S> (repeated as many times as necessary). | |
1673 | ||
1674 | =item C<S{min,max}> | |
1675 | ||
1676 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
1677 | ||
1678 | =item C<S{min,max}?> | |
1679 | ||
1680 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
1681 | ||
1682 | =item C<S?>, C<S*>, C<S+> | |
1683 | ||
1684 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
1685 | ||
1686 | =item C<S??>, C<S*?>, C<S+?> | |
1687 | ||
1688 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
1689 | ||
c47ff5f1 | 1690 | =item C<< (?>S) >> |
35a734be IZ |
1691 | |
1692 | Matches the best match for C<S> and only that. | |
1693 | ||
1694 | =item C<(?=S)>, C<(?<=S)> | |
1695 | ||
1696 | Only the best match for C<S> is considered. (This is important only if | |
1697 | C<S> has capturing parentheses, and backreferences are used somewhere | |
1698 | else in the whole regular expression.) | |
1699 | ||
1700 | =item C<(?!S)>, C<(?<!S)> | |
1701 | ||
1702 | For this grouping operator there is no need to describe the ordering, since | |
1703 | only whether or not C<S> can match is important. | |
1704 | ||
6bda09f9 | 1705 | =item C<(??{ EXPR })>, C<(?PARNO)> |
35a734be IZ |
1706 | |
1707 | The ordering is the same as for the regular expression which is | |
6bda09f9 | 1708 | the result of EXPR, or the pattern contained by capture buffer PARNO. |
35a734be IZ |
1709 | |
1710 | =item C<(?(condition)yes-pattern|no-pattern)> | |
1711 | ||
1712 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is | |
1713 | already determined. The ordering of the matches is the same as for the | |
1714 | chosen subexpression. | |
1715 | ||
1716 | =back | |
1717 | ||
1718 | The above recipes describe the ordering of matches I<at a given position>. | |
1719 | One more rule is needed to understand how a match is determined for the | |
1720 | whole regular expression: a match at an earlier position is always better | |
1721 | than a match at a later position. | |
1722 | ||
c84d73f1 IZ |
1723 | =head2 Creating custom RE engines |
1724 | ||
1725 | Overloaded constants (see L<overload>) provide a simple way to extend | |
1726 | the functionality of the RE engine. | |
1727 | ||
1728 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
6b0ac556 | 1729 | matches at boundary between whitespace characters and non-whitespace |
c84d73f1 IZ |
1730 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
1731 | at these positions, so we want to have each C<\Y|> in the place of the | |
1732 | more complicated version. We can create a module C<customre> to do | |
1733 | this: | |
1734 | ||
1735 | package customre; | |
1736 | use overload; | |
1737 | ||
1738 | sub import { | |
1739 | shift; | |
1740 | die "No argument to customre::import allowed" if @_; | |
1741 | overload::constant 'qr' => \&convert; | |
1742 | } | |
1743 | ||
1744 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
1745 | ||
580a9fe1 RGS |
1746 | # We must also take care of not escaping the legitimate \\Y| |
1747 | # sequence, hence the presence of '\\' in the conversion rules. | |
141db969 | 1748 | my %rules = ( '\\' => '\\\\', |
c84d73f1 IZ |
1749 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
1750 | sub convert { | |
1751 | my $re = shift; | |
1752 | $re =~ s{ | |
1753 | \\ ( \\ | Y . ) | |
1754 | } | |
1755 | { $rules{$1} or invalid($re,$1) }sgex; | |
1756 | return $re; | |
1757 | } | |
1758 | ||
1759 | Now C<use customre> enables the new escape in constant regular | |
1760 | expressions, i.e., those without any runtime variable interpolations. | |
1761 | As documented in L<overload>, this conversion will work only over | |
1762 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
1763 | part of this regular expression needs to be converted explicitly | |
1764 | (but only if the special meaning of C<\Y|> should be enabled inside $re): | |
1765 | ||
1766 | use customre; | |
1767 | $re = <>; | |
1768 | chomp $re; | |
1769 | $re = customre::convert $re; | |
1770 | /\Y|$re\Y|/; | |
1771 | ||
19799a22 GS |
1772 | =head1 BUGS |
1773 | ||
9da458fc IZ |
1774 | This document varies from difficult to understand to completely |
1775 | and utterly opaque. The wandering prose riddled with jargon is | |
1776 | hard to fathom in several places. | |
1777 | ||
1778 | This document needs a rewrite that separates the tutorial content | |
1779 | from the reference content. | |
19799a22 GS |
1780 | |
1781 | =head1 SEE ALSO | |
9fa51da4 | 1782 | |
91e0c79e MJD |
1783 | L<perlrequick>. |
1784 | ||
1785 | L<perlretut>. | |
1786 | ||
9b599b2a GS |
1787 | L<perlop/"Regexp Quote-Like Operators">. |
1788 | ||
1e66bd83 PP |
1789 | L<perlop/"Gory details of parsing quoted constructs">. |
1790 | ||
14218588 GS |
1791 | L<perlfaq6>. |
1792 | ||
9b599b2a GS |
1793 | L<perlfunc/pos>. |
1794 | ||
1795 | L<perllocale>. | |
1796 | ||
fb55449c JH |
1797 | L<perlebcdic>. |
1798 | ||
14218588 GS |
1799 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
1800 | by O'Reilly and Associates. |