Commit | Line | Data |
a0d0e21e |
1 | =head1 NAME |
d74e8afc |
2 | X<regular expression> X<regex> X<regexp> |
a0d0e21e |
3 | |
4 | perlre - Perl regular expressions |
5 | |
6 | =head1 DESCRIPTION |
7 | |
91e0c79e |
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 |
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 |
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 |
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 |
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 "$" from matching |
14218588 |
40 | the start or end of the string to matching the start or end of any |
7f761169 |
41 | line anywhere within the string. |
55497cff |
42 | |
54310121 |
43 | =item s |
d74e8afc |
44 | X</s> X<regex, single-line> X<regexp, single-line> |
45 | X<regular expression, single-line> |
55497cff |
46 | |
47 | Treat string as single line. That is, change "." to match any character |
19799a22 |
48 | whatsoever, even a newline, which normally it would not match. |
55497cff |
49 | |
f02c194e |
50 | Used together, as /ms, they let the "." match any character whatsoever, |
fb55449c |
51 | while still allowing "^" and "$" 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 |
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 |
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 |
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 |
82 | |
83 | =head2 Regular Expressions |
84 | |
04838cea |
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 |
89 | (distantly) from Henry Spencer's freely redistributable reimplementation |
90 | of the V8 routines.) See L<Version 8 Regular Expressions> for |
91 | details. |
a0d0e21e |
92 | |
93 | In particular the following metacharacters have their standard I<egrep>-ish |
94 | meanings: |
d74e8afc |
95 | X<metacharacter> |
96 | X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]> |
97 | |
a0d0e21e |
98 | |
54310121 |
99 | \ Quote the next metacharacter |
a0d0e21e |
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 |
103 | | Alternation |
104 | () Grouping |
105 | [] Character class |
106 | |
14218588 |
107 | By default, the "^" character is guaranteed to match only the |
108 | beginning of the string, the "$" character only the end (or before the |
109 | newline at the end), and Perl does certain optimizations with the |
a0d0e21e |
110 | assumption that the string contains only one line. Embedded newlines |
111 | will not be matched by "^" or "$". You may, however, wish to treat a |
4a6725af |
112 | string as a multi-line buffer, such that the "^" will match after any |
0d520e8e |
113 | newline within the string (except if the newline is the last character in |
114 | the string), and "$" will match before any newline. At the |
a0d0e21e |
115 | cost of a little more overhead, you can do this by using the /m modifier |
116 | on the pattern match operator. (Older programs did this by setting C<$*>, |
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 |
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 |
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 |
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 |
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 |
152 | X<metacharacter> X<greedy> X<greedyness> |
153 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?> |
a0d0e21e |
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 |
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 |
169 | ?+ Match 0 or 1 time and give nothing back |
170 | {n}+ Match exactly n times and give nothing back (redundant) |
04838cea |
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 |
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 |
193 | =head3 Escape sequences |
194 | |
5f05dabc |
195 | Because patterns are processed as double quoted strings, the following |
a0d0e21e |
196 | also work: |
d74e8afc |
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 |
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 |
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 |
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 |
227 | =head3 Character classes |
228 | |
a0d0e21e |
229 | In addition, Perl defines the following: |
d74e8afc |
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 |
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 |
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 |
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 |
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 |
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 |
288 | |
289 | alpha |
290 | alnum |
291 | ascii |
aaa51d5e |
292 | blank [1] |
b8c5462f |
293 | cntrl |
294 | digit \d |
295 | graph |
296 | lower |
297 | print |
298 | punct |
aaa51d5e |
299 | space \s [2] |
b8c5462f |
300 | upper |
aaa51d5e |
301 | word \w [3] |
b8c5462f |
302 | xdigit |
303 | |
07698885 |
304 | =over |
305 | |
306 | =item [1] |
307 | |
b432a672 |
308 | A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace". |
07698885 |
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 |
314 | |
315 | =item [3] |
316 | |
08ce8fc6 |
317 | A Perl extension, see above. |
07698885 |
318 | |
319 | =back |
aaa51d5e |
320 | |
26b44a0a |
321 | For example use C<[:upper:]> to match all the uppercase characters. |
aaa51d5e |
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 |
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 |
334 | |
335 | alpha IsAlpha |
336 | alnum IsAlnum |
337 | ascii IsASCII |
b432a672 |
338 | blank IsSpace |
b8c5462f |
339 | cntrl IsCntrl |
3bec3564 |
340 | digit IsDigit \d |
b8c5462f |
341 | graph IsGraph |
342 | lower IsLower |
343 | print IsPrint |
344 | punct IsPunct |
345 | space IsSpace |
3bec3564 |
346 | IsSpacePerl \s |
b8c5462f |
347 | upper IsUpper |
348 | word IsWord |
349 | xdigit IsXDigit |
350 | |
5496314a |
351 | For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent. |
b8c5462f |
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 |
356 | |
357 | The assumedly non-obviously named classes are: |
358 | |
359 | =over 4 |
360 | |
361 | =item cntrl |
d74e8afc |
362 | X<cntrl> |
b8c5462f |
363 | |
820475bd |
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 |
368 | the ISO Latin character sets, and Unicode), as is the character with |
369 | the ord() value of 127 (C<DEL>). |
b8c5462f |
370 | |
371 | =item graph |
d74e8afc |
372 | X<graph> |
b8c5462f |
373 | |
f1cbbd6e |
374 | Any alphanumeric or punctuation (special) character. |
b8c5462f |
375 | |
376 | =item print |
d74e8afc |
377 | X<print> |
b8c5462f |
378 | |
f79b3095 |
379 | Any alphanumeric or punctuation (special) character or the space character. |
b8c5462f |
380 | |
381 | =item punct |
d74e8afc |
382 | X<punct> |
b8c5462f |
383 | |
f1cbbd6e |
384 | Any punctuation (special) character. |
b8c5462f |
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 |
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 |
400 | [[:^digit:]] \D \P{IsDigit} |
401 | [[:^space:]] \S \P{IsSpace} |
402 | [[:^word:]] \W \P{IsWord} |
b8c5462f |
403 | |
54c18d04 |
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 |
409 | =head3 Assertions |
410 | |
a0d0e21e |
411 | Perl defines the following zero-width assertions: |
d74e8afc |
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 |
416 | |
417 | \b Match a word boundary |
418 | \B Match a non-(word boundary) |
b85d18e9 |
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 |
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 |
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 "$" will match at every internal line boundary. To match |
434 | the actual end of the string and not ignore an optional trailing |
435 | newline, use C<\z>. |
d74e8afc |
436 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 |
437 | |
438 | The C<\G> assertion can be used to chain global matches (using |
439 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. |
440 | It is also useful when writing C<lex>-like scanners, when you have |
441 | several patterns that you want to match against consequent substrings |
442 | of your string, see the previous reference. The actual location |
443 | where C<\G> will match can also be influenced by using C<pos()> as |
25cf8c22 |
444 | an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully |
445 | supported when anchored to the start of the pattern; while it |
446 | is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some |
447 | such uses (C</.\G/g>, for example) currently cause problems, and |
448 | it is recommended that you avoid such usage for now. |
d74e8afc |
449 | X<\G> |
c47ff5f1 |
450 | |
04838cea |
451 | =head3 Capture buffers |
452 | |
14218588 |
453 | The bracketing construct C<( ... )> creates capture buffers. To |
c47ff5f1 |
454 | refer to the digit'th buffer use \<digit> within the |
14218588 |
455 | match. Outside the match use "$" instead of "\". (The |
81714fb9 |
456 | \<digit> notation works in certain circumstances outside |
14218588 |
457 | the match. See the warning below about \1 vs $1 for details.) |
458 | Referring back to another part of the match is called a |
459 | I<backreference>. |
d74e8afc |
460 | X<regex, capture buffer> X<regexp, capture buffer> |
461 | X<regular expression, capture buffer> X<backreference> |
14218588 |
462 | |
463 | There is no limit to the number of captured substrings that you may |
464 | use. However Perl also uses \10, \11, etc. as aliases for \010, |
fb55449c |
465 | \011, etc. (Recall that 0 means octal, so \011 is the character at |
466 | number 9 in your coded character set; which would be the 10th character, |
81714fb9 |
467 | a horizontal tab under ASCII.) Perl resolves this |
468 | ambiguity by interpreting \10 as a backreference only if at least 10 |
469 | left parentheses have opened before it. Likewise \11 is a |
470 | backreference only if at least 11 left parentheses have opened |
471 | before it. And so on. \1 through \9 are always interpreted as |
fb55449c |
472 | backreferences. |
14218588 |
473 | |
81714fb9 |
474 | Additionally, as of Perl 5.10 you may use named capture buffers and named |
475 | backreferences. The notation is C<< (?<name>...) >> and C<< \k<name> >> |
476 | (you may also use single quotes instead of angle brackets to quote the |
477 | name). The only difference with named capture buffers and unnamed ones is |
478 | that multiple buffers may have the same name and that the contents of |
479 | named capture buffers is available via the C<%+> hash. When multiple |
480 | groups share the same name C<$+{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 |
488 | |
489 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words |
490 | |
81714fb9 |
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 |
501 | $hours = $1; |
502 | $minutes = $2; |
503 | $seconds = $3; |
504 | } |
c47ff5f1 |
505 | |
14218588 |
506 | Several special variables also refer back to portions of the previous |
507 | match. C<$+> 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 |
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 |
519 | until the end of the enclosing block or until the next successful |
520 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) |
d74e8afc |
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<$9> |
523 | |
14218588 |
524 | |
665e98b9 |
525 | B<NOTE>: failed matches in Perl do not reset the match variables, |
5146ce24 |
526 | which makes it easier to write code that tests for a series of more |
665e98b9 |
527 | specific cases and remembers the best match. |
528 | |
14218588 |
529 | B<WARNING>: Once Perl sees that you need one of C<$&>, 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 |
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 |
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 |
552 | |
553 | $pattern =~ s/(\W)/\\$1/g; |
554 | |
f1cbbd6e |
555 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 |
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 |
559 | |
560 | /$unquoted\Q$quoted\E$unquoted/ |
561 | |
9da458fc |
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 |
568 | =head2 Extended Patterns |
569 | |
14218588 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
623 | group. |
624 | |
5a964f20 |
625 | =item C<(?:pattern)> |
d74e8afc |
626 | X<(?:)> |
a0d0e21e |
627 | |
ca9dfc88 |
628 | =item C<(?imsx-imsx:pattern)> |
629 | |
5a964f20 |
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 |
634 | |
635 | is like |
636 | |
5a964f20 |
637 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e |
638 | |
19799a22 |
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 |
642 | Any letters between C<?> and C<:> act as flags modifiers as with |
643 | C<(?imsx-imsx)>. For example, |
ca9dfc88 |
644 | |
645 | /(?s-i:more.*than).*million/i |
646 | |
14218588 |
647 | is equivalent to the more verbose |
ca9dfc88 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 = $animal\n"; |
752 | |
754091cb |
753 | Inside the C<(?{...})> block, C<$_> 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 |
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 |
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, $cnt returns to the globally |
14218588 |
776 | introduced value, because the scopes that restrict C<local> operators |
b9ac3b5b |
777 | are unwound. |
778 | |
19799a22 |
779 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> |
780 | switch. If I<not> used in this way, the result of evaluation of |
781 | C<code> is put into the special variable C<$^R>. This happens |
782 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions |
783 | inside the same regular expression. |
b9ac3b5b |
784 | |
19799a22 |
785 | The assignment to C<$^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 |
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 |
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 |
794 | occur then check for the use of lexically scoped variables. |
795 | |
19799a22 |
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 |
804 | |
805 | $re = <>; |
806 | chomp $re; |
807 | $string =~ /$re/; |
808 | |
14218588 |
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 |
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 |
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 |
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 |
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 |
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 |
844 | where the C<code> ends are currently somewhat convoluted. |
845 | |
846 | The following pattern matches a parenthesized group: |
0f5d15d6 |
847 | |
848 | $re = qr{ |
849 | \( |
850 | (?: |
851 | (?> [^()]+ ) # Non-parens without backtracking |
852 | | |
14455d6c |
853 | (??{ $re }) # Group with matching parens |
0f5d15d6 |
854 | )* |
855 | \) |
856 | }x; |
857 | |
6bda09f9 |
858 | See also C<(?PARNO)> for a different, more efficient way to accomplish |
859 | the same task. |
860 | |
8988a1bb |
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 |
867 | changing it requires a custom build. |
868 | |
894be9b7 |
869 | =item C<(?PARNO)> C<(?R)> C<(?0)> |
870 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> |
6bda09f9 |
871 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
872 | |
81714fb9 |
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 |
877 | outermost recursion. |
878 | |
894be9b7 |
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 |
886 | |
887 | $re = qr{ ( # paren group 1 (full function) |
81714fb9 |
888 | foo |
6bda09f9 |
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 |
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 |
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 |
924 | as atomic. Also, constructs like (?i:(?1)) or (?:(?i)(?1)) do not affect |
925 | the pattern being recursed into. |
6bda09f9 |
926 | |
894be9b7 |
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 |
937 | =item C<(?(condition)yes-pattern|no-pattern)> |
938 | X<(?()> |
286f584a |
939 | |
e2e6a0f1 |
940 | =item C<(?(condition)yes-pattern)> |
286f584a |
941 | |
e2e6a0f1 |
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 |
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 |
1038 | construct is useful for optimizations of what would otherwise be |
1039 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). |
9da458fc |
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 |
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 |
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 |
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 |
1063 | ( |
1064 | [^()]+ # x+ |
1065 | | |
871b0233 |
1066 | \( [^()]* \) |
1067 | )+ |
e2e6a0f1 |
1068 | \) |
871b0233 |
1069 | }x |
5a964f20 |
1070 | |
19799a22 |
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 |
1082 | m{ \( |
1083 | ( |
1084 | (?> [^()]+ ) # change x+ above to (?> x+ ) |
1085 | | |
871b0233 |
1086 | \( [^()]* \) |
1087 | )+ |
e2e6a0f1 |
1088 | \) |
871b0233 |
1089 | }x |
c277df42 |
1090 | |
c47ff5f1 |
1091 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 |
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 |
1100 | This was only 4 times slower on a string with 1000000 C<a>s. |
1101 | |
9da458fc |
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 |
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 $1, one should use either |
1115 | one of these: |
1116 | |
1117 | / (?> \# [ \t]* ) ( .+ ) /x; |
1118 | / \# [ \t]* ( [^ \t] .* ) /x; |
1119 | |
1120 | Which one you pick depends on which of these expressions better reflects |
1121 | the above specification of comments. |
1122 | |
6bda09f9 |
1123 | In some literature this construct is called "atomic matching" or |
1124 | "possessive matching". |
1125 | |
b9b4dddf |
1126 | Possessive quantifiers are equivalent to putting the item they are applied |
1127 | to inside of one of these constructs. The following equivalences apply: |
1128 | |
1129 | Quantifier Form Bracketing Form |
1130 | --------------- --------------- |
1131 | PAT*+ (?>PAT*) |
1132 | PAT++ (?>PAT+) |
1133 | PAT?+ (?>PAT?) |
1134 | PAT{min,max}+ (?>PAT{min,max}) |
1135 | |
e2e6a0f1 |
1136 | =back |
1137 | |
1138 | =head2 Special Backtracking Control Verbs |
1139 | |
1140 | B<WARNING:> These patterns are experimental and subject to change or |
1141 | removal in a future version of perl. Their usage in production code should |
1142 | be noted to avoid problems during upgrades. |
1143 | |
1144 | These special patterns are generally of the form C<(*VERB:ARG)>. Unless |
1145 | otherwise stated the ARG argument is optional; in some cases, it is |
1146 | forbidden. |
1147 | |
1148 | Any pattern containing a special backtracking verb that allows an argument |
1149 | has the special behaviour that when executed it sets the current packages' |
1150 | C<$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<$REGERROR> is not touched at all. |
1166 | |
1167 | =over 4 |
1168 | |
1169 | =item Verbs that take an argument |
1170 | |
1171 | =over 4 |
1172 | |
1173 | =item C<(*NOMATCH)> C<(*NOMATCH:NAME)> |
1174 | X<(*NOMATCH)> X<(*NOMATCH:NAME)> |
54612592 |
1175 | |
1176 | This zero-width pattern commits the match at the current point, preventing |
e2e6a0f1 |
1177 | the engine from backtracking on failure to the left of the this point. |
1178 | Consider the pattern C<A (*NOMATCH) B>, where A and B are complex patterns. |
1179 | Until the C<(*NOMATCH)> is reached, A may backtrack as necessary to match. |
54612592 |
1180 | Once it is reached, matching continues in B, which may also backtrack as |
1181 | necessary; however, should B not match, then no further backtracking will |
1182 | take place, and the pattern will fail outright at that starting position. |
1183 | |
1184 | The following example counts all the possible matching strings in a |
1185 | pattern (without actually matching any of them). |
1186 | |
e2e6a0f1 |
1187 | 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 |
1188 | print "Count=$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"; $count++})(*FAIL)/; |
54612592 |
1206 | print "Count=$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 |
1217 | |
1218 | See also C<< (?>pattern) >> and possessive quantifiers for other |
1219 | ways to control backtracking. |
1220 | |
e2e6a0f1 |
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 |
1254 | is twice as long: |
1255 | |
e2e6a0f1 |
1256 | 'aaabaaab' =~ /a+b?(*CUT)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 |
1257 | print "Count=$count\n"; |
1258 | |
1259 | outputs |
1260 | |
1261 | aaab |
1262 | aaab |
1263 | Count=2 |
1264 | |
e2e6a0f1 |
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 |
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 |
1273 | the match to fail outright. No attempts to match will occur again. |
1274 | |
e2e6a0f1 |
1275 | 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 |
1276 | print "Count=$count\n"; |
1277 | |
1278 | outputs |
1279 | |
1280 | aaab |
1281 | Count=1 |
1282 | |
e2e6a0f1 |
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 |
1290 | |
1291 | =over 4 |
1292 | |
e2e6a0f1 |
1293 | =item C<(*FAIL)> C<(*F)> |
1294 | X<(*FAIL)> X<(*F)> |
9af228c6 |
1295 | |
e2e6a0f1 |
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 |
1302 | =item C<(*ACCEPT)> |
1303 | X<(*ACCEPT)> |
9af228c6 |
1304 | |
e2e6a0f1 |
1305 | B<WARNING:> This feature is highly experimental. It is not recommended |
1306 | for production code. |
9af228c6 |
1307 | |
e2e6a0f1 |
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 |
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 |
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 |
1323 | |
1324 | =back |
c277df42 |
1325 | |
a0d0e21e |
1326 | =back |
1327 | |
c07a80fd |
1328 | =head2 Backtracking |
d74e8afc |
1329 | X<backtrack> X<backtracking> |
c07a80fd |
1330 | |
35a734be |
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 |
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 |
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 |
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 |
1455 | $x = 'ABC123'; |
1456 | $y = 'ABC445'; |
c07a80fd |
1457 | |
4358a253 |
1458 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
1459 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; |
c07a80fd |
1460 | |
4358a253 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
1587 | pattern delimiter. That's why it's common practice to include |
1588 | alternatives in parentheses: to minimize confusion about where they |
a3cb178b |
1589 | start and end. |
1590 | |
5a964f20 |
1591 | Alternatives are tried from left to right, so the first |
a3cb178b |
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 |
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 |
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 |
1616 | |
1617 | $pattern =~ s/(\W)/\\\1/g; |
1618 | |
1619 | This is grandfathered for the RHS of a substitute to avoid shocking the |
1620 | B<sed> addicts, but it's a dirty habit to get into. That's because in |
d1be9408 |
1621 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 |
1622 | the usual double-quoted string means a control-A. The customary Unix |
1623 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit |
1624 | of doing that, you get yourself into trouble if you then add an C</e> |
1625 | modifier. |
1626 | |
5a964f20 |
1627 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 |
1628 | |
1629 | Or if you try to do |
1630 | |
1631 | s/(\d+)/\1000/; |
1632 | |
1633 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with |
14218588 |
1634 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 |
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 |
1638 | =head2 Repeated patterns matching zero-length substring |
1639 | |
19799a22 |
1640 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 |
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 |
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 |
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 |
1665 | be significantly simplified by using repeated subexpressions that |
1666 | may match zero-length substrings. Here's a simple example being: |
c84d73f1 |
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 |
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 |
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 |
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 |
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 |
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 |
1711 | Zero-length matches at the end of the previous match are ignored |
1712 | during C<split>. |
c84d73f1 |
1713 | |
35a734be |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
1864 | =head1 BUGS |
1865 | |
9da458fc |
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 |
1872 | |
1873 | =head1 SEE ALSO |
9fa51da4 |
1874 | |
91e0c79e |
1875 | L<perlrequick>. |
1876 | |
1877 | L<perlretut>. |
1878 | |
9b599b2a |
1879 | L<perlop/"Regexp Quote-Like Operators">. |
1880 | |
1e66bd83 |
1881 | L<perlop/"Gory details of parsing quoted constructs">. |
1882 | |
14218588 |
1883 | L<perlfaq6>. |
1884 | |
9b599b2a |
1885 | L<perlfunc/pos>. |
1886 | |
1887 | L<perllocale>. |
1888 | |
fb55449c |
1889 | L<perlebcdic>. |
1890 | |
14218588 |
1891 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
1892 | by O'Reilly and Associates. |