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 | |
5d458dd8 |
8 | This page describes the syntax of regular expressions in Perl. |
91e0c79e |
9 | |
cc46d5f2 |
10 | If you haven't used regular expressions before, a quick-start |
91e0c79e |
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 |
5d458dd8 |
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. |
5d458dd8 |
248 | \1 Backreference to a specific group. |
c74340f9 |
249 | '1' may actually be any positive integer. |
2bf803e2 |
250 | \g1 Backreference to a specific or previous group, |
251 | \g{-1} number may be negative indicating a previous buffer and may |
252 | optionally be wrapped in curly brackets for safer parsing. |
1f1031fe |
253 | \g{name} Named backreference |
81714fb9 |
254 | \k<name> Named backreference |
5d458dd8 |
255 | \N{name} Named unicode character, or unicode escape |
81714fb9 |
256 | \x12 Hexadecimal escape sequence |
257 | \x{1234} Long hexadecimal escape sequence |
ee9b8eae |
258 | \K Keep the stuff left of the \K, don't include it in $& |
259 | \v Shortcut for (*PRUNE) |
260 | \V Shortcut for (*SKIP) |
a0d0e21e |
261 | |
08ce8fc6 |
262 | A C<\w> matches a single alphanumeric character (an alphabetic |
263 | character, or a decimal digit) or C<_>, not a whole word. Use C<\w+> |
264 | to match a string of Perl-identifier characters (which isn't the same |
265 | as matching an English word). If C<use locale> is in effect, the list |
266 | of alphabetic characters generated by C<\w> is taken from the current |
267 | locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>, |
1209ba90 |
268 | C<\d>, and C<\D> within character classes, but if you try to use them |
08ce8fc6 |
269 | as endpoints of a range, that's not a range, the "-" is understood |
270 | literally. If Unicode is in effect, C<\s> matches also "\x{85}", |
271 | "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about |
491fd90a |
272 | C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general. |
fa11829f |
273 | You can define your own C<\p> and C<\P> properties, see L<perlunicode>. |
d74e8afc |
274 | X<\w> X<\W> X<word> |
a0d0e21e |
275 | |
b8c5462f |
276 | The POSIX character class syntax |
d74e8afc |
277 | X<character class> |
b8c5462f |
278 | |
820475bd |
279 | [:class:] |
b8c5462f |
280 | |
5496314a |
281 | is also available. Note that the C<[> and C<]> braces are I<literal>; |
282 | they must always be used within a character class expression. |
283 | |
284 | # this is correct: |
285 | $string =~ /[[:alpha:]]/; |
286 | |
287 | # this is not, and will generate a warning: |
288 | $string =~ /[:alpha:]/; |
289 | |
290 | The available classes and their backslash equivalents (if available) are |
291 | as follows: |
d74e8afc |
292 | X<character class> |
293 | X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph> |
294 | X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit> |
b8c5462f |
295 | |
296 | alpha |
297 | alnum |
298 | ascii |
aaa51d5e |
299 | blank [1] |
b8c5462f |
300 | cntrl |
301 | digit \d |
302 | graph |
303 | lower |
304 | print |
305 | punct |
aaa51d5e |
306 | space \s [2] |
b8c5462f |
307 | upper |
aaa51d5e |
308 | word \w [3] |
b8c5462f |
309 | xdigit |
310 | |
07698885 |
311 | =over |
312 | |
313 | =item [1] |
314 | |
b432a672 |
315 | A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace". |
07698885 |
316 | |
317 | =item [2] |
318 | |
319 | Not exactly equivalent to C<\s> since the C<[[:space:]]> includes |
b432a672 |
320 | also the (very rare) "vertical tabulator", "\ck", chr(11). |
07698885 |
321 | |
322 | =item [3] |
323 | |
08ce8fc6 |
324 | A Perl extension, see above. |
07698885 |
325 | |
326 | =back |
aaa51d5e |
327 | |
26b44a0a |
328 | For example use C<[:upper:]> to match all the uppercase characters. |
aaa51d5e |
329 | Note that the C<[]> are part of the C<[::]> construct, not part of the |
330 | whole character class. For example: |
b8c5462f |
331 | |
820475bd |
332 | [01[:alpha:]%] |
b8c5462f |
333 | |
593df60c |
334 | matches zero, one, any alphabetic character, and the percentage sign. |
b8c5462f |
335 | |
72ff2908 |
336 | The following equivalences to Unicode \p{} constructs and equivalent |
337 | backslash character classes (if available), will hold: |
d74e8afc |
338 | X<character class> X<\p> X<\p{}> |
72ff2908 |
339 | |
5496314a |
340 | [[:...:]] \p{...} backslash |
b8c5462f |
341 | |
342 | alpha IsAlpha |
343 | alnum IsAlnum |
344 | ascii IsASCII |
b432a672 |
345 | blank IsSpace |
b8c5462f |
346 | cntrl IsCntrl |
3bec3564 |
347 | digit IsDigit \d |
b8c5462f |
348 | graph IsGraph |
349 | lower IsLower |
350 | print IsPrint |
351 | punct IsPunct |
352 | space IsSpace |
3bec3564 |
353 | IsSpacePerl \s |
b8c5462f |
354 | upper IsUpper |
355 | word IsWord |
356 | xdigit IsXDigit |
357 | |
5496314a |
358 | For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent. |
b8c5462f |
359 | |
360 | If the C<utf8> pragma is not used but the C<locale> pragma is, the |
aaa51d5e |
361 | classes correlate with the usual isalpha(3) interface (except for |
b432a672 |
362 | "word" and "blank"). |
b8c5462f |
363 | |
364 | The assumedly non-obviously named classes are: |
365 | |
366 | =over 4 |
367 | |
368 | =item cntrl |
d74e8afc |
369 | X<cntrl> |
b8c5462f |
370 | |
820475bd |
371 | Any control character. Usually characters that don't produce output as |
372 | such but instead control the terminal somehow: for example newline and |
373 | backspace are control characters. All characters with ord() less than |
593df60c |
374 | 32 are most often classified as control characters (assuming ASCII, |
7be5a6cf |
375 | the ISO Latin character sets, and Unicode), as is the character with |
376 | the ord() value of 127 (C<DEL>). |
b8c5462f |
377 | |
378 | =item graph |
d74e8afc |
379 | X<graph> |
b8c5462f |
380 | |
f1cbbd6e |
381 | Any alphanumeric or punctuation (special) character. |
b8c5462f |
382 | |
383 | =item print |
d74e8afc |
384 | X<print> |
b8c5462f |
385 | |
f79b3095 |
386 | Any alphanumeric or punctuation (special) character or the space character. |
b8c5462f |
387 | |
388 | =item punct |
d74e8afc |
389 | X<punct> |
b8c5462f |
390 | |
f1cbbd6e |
391 | Any punctuation (special) character. |
b8c5462f |
392 | |
393 | =item xdigit |
d74e8afc |
394 | X<xdigit> |
b8c5462f |
395 | |
593df60c |
396 | Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would |
820475bd |
397 | work just fine) it is included for completeness. |
b8c5462f |
398 | |
b8c5462f |
399 | =back |
400 | |
401 | You can negate the [::] character classes by prefixing the class name |
402 | with a '^'. This is a Perl extension. For example: |
d74e8afc |
403 | X<character class, negation> |
b8c5462f |
404 | |
5496314a |
405 | POSIX traditional Unicode |
93733859 |
406 | |
5496314a |
407 | [[:^digit:]] \D \P{IsDigit} |
408 | [[:^space:]] \S \P{IsSpace} |
409 | [[:^word:]] \W \P{IsWord} |
b8c5462f |
410 | |
54c18d04 |
411 | Perl respects the POSIX standard in that POSIX character classes are |
412 | only supported within a character class. The POSIX character classes |
413 | [.cc.] and [=cc=] are recognized but B<not> supported and trying to |
414 | use them will cause an error. |
b8c5462f |
415 | |
04838cea |
416 | =head3 Assertions |
417 | |
a0d0e21e |
418 | Perl defines the following zero-width assertions: |
d74e8afc |
419 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> |
420 | X<regexp, zero-width assertion> |
421 | X<regular expression, zero-width assertion> |
422 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> |
a0d0e21e |
423 | |
424 | \b Match a word boundary |
425 | \B Match a non-(word boundary) |
b85d18e9 |
426 | \A Match only at beginning of string |
427 | \Z Match only at end of string, or before newline at the end |
428 | \z Match only at end of string |
9da458fc |
429 | \G Match only at pos() (e.g. at the end-of-match position |
430 | of prior m//g) |
a0d0e21e |
431 | |
14218588 |
432 | A word boundary (C<\b>) is a spot between two characters |
19799a22 |
433 | that has a C<\w> on one side of it and a C<\W> on the other side |
434 | of it (in either order), counting the imaginary characters off the |
435 | beginning and end of the string as matching a C<\W>. (Within |
436 | character classes C<\b> represents backspace rather than a word |
437 | boundary, just as it normally does in any double-quoted string.) |
438 | The C<\A> and C<\Z> are just like "^" and "$", except that they |
439 | won't match multiple times when the C</m> modifier is used, while |
440 | "^" and "$" will match at every internal line boundary. To match |
441 | the actual end of the string and not ignore an optional trailing |
442 | newline, use C<\z>. |
d74e8afc |
443 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 |
444 | |
445 | The C<\G> assertion can be used to chain global matches (using |
446 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. |
447 | It is also useful when writing C<lex>-like scanners, when you have |
448 | several patterns that you want to match against consequent substrings |
449 | of your string, see the previous reference. The actual location |
450 | where C<\G> will match can also be influenced by using C<pos()> as |
58e23c8d |
451 | an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length |
452 | matches is modified somewhat, in that contents to the left of C<\G> is |
453 | not counted when determining the length of the match. Thus the following |
454 | will not match forever: |
d74e8afc |
455 | X<\G> |
c47ff5f1 |
456 | |
58e23c8d |
457 | $str = 'ABC'; |
458 | pos($str) = 1; |
459 | while (/.\G/g) { |
460 | print $&; |
461 | } |
462 | |
463 | It will print 'A' and then terminate, as it considers the match to |
464 | be zero-width, and thus will not match at the same position twice in a |
465 | row. |
466 | |
467 | It is worth noting that C<\G> improperly used can result in an infinite |
468 | loop. Take care when using patterns that include C<\G> in an alternation. |
469 | |
04838cea |
470 | =head3 Capture buffers |
471 | |
14218588 |
472 | The bracketing construct C<( ... )> creates capture buffers. To |
c47ff5f1 |
473 | refer to the digit'th buffer use \<digit> within the |
14218588 |
474 | match. Outside the match use "$" instead of "\". (The |
81714fb9 |
475 | \<digit> notation works in certain circumstances outside |
14218588 |
476 | the match. See the warning below about \1 vs $1 for details.) |
477 | Referring back to another part of the match is called a |
478 | I<backreference>. |
d74e8afc |
479 | X<regex, capture buffer> X<regexp, capture buffer> |
480 | X<regular expression, capture buffer> X<backreference> |
14218588 |
481 | |
482 | There is no limit to the number of captured substrings that you may |
483 | use. However Perl also uses \10, \11, etc. as aliases for \010, |
fb55449c |
484 | \011, etc. (Recall that 0 means octal, so \011 is the character at |
485 | number 9 in your coded character set; which would be the 10th character, |
81714fb9 |
486 | a horizontal tab under ASCII.) Perl resolves this |
487 | ambiguity by interpreting \10 as a backreference only if at least 10 |
488 | left parentheses have opened before it. Likewise \11 is a |
489 | backreference only if at least 11 left parentheses have opened |
490 | before it. And so on. \1 through \9 are always interpreted as |
5624f11d |
491 | backreferences. |
c74340f9 |
492 | |
1f1031fe |
493 | X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference> |
2bf803e2 |
494 | In order to provide a safer and easier way to construct patterns using |
495 | backrefs, in Perl 5.10 the C<\g{N}> notation is provided. The curly |
496 | brackets are optional, however omitting them is less safe as the meaning |
497 | of the pattern can be changed by text (such as digits) following it. |
498 | When N is a positive integer the C<\g{N}> notation is exactly equivalent |
499 | to using normal backreferences. When N is a negative integer then it is |
500 | a relative backreference referring to the previous N'th capturing group. |
1f1031fe |
501 | When the bracket form is used and N is not an integer, it is treated as a |
502 | reference to a named buffer. |
2bf803e2 |
503 | |
504 | Thus C<\g{-1}> refers to the last buffer, C<\g{-2}> refers to the |
505 | buffer before that. For example: |
5624f11d |
506 | |
507 | / |
508 | (Y) # buffer 1 |
509 | ( # buffer 2 |
510 | (X) # buffer 3 |
2bf803e2 |
511 | \g{-1} # backref to buffer 3 |
512 | \g{-3} # backref to buffer 1 |
5624f11d |
513 | ) |
514 | /x |
515 | |
2bf803e2 |
516 | and would match the same as C</(Y) ( (X) \3 \1 )/x>. |
14218588 |
517 | |
81714fb9 |
518 | Additionally, as of Perl 5.10 you may use named capture buffers and named |
1f1031fe |
519 | backreferences. The notation is C<< (?<name>...) >> to declare and C<< \k<name> >> |
520 | to reference. You may also use single quotes instead of angle brackets to quote the |
521 | name; and you may use the bracketed C<< \g{name} >> back reference syntax. |
522 | The only difference between named capture buffers and unnamed ones is |
81714fb9 |
523 | that multiple buffers may have the same name and that the contents of |
1f1031fe |
524 | named capture buffers are available via the C<%+> hash. When multiple |
81714fb9 |
525 | groups share the same name C<$+{name}> and C<< \k<name> >> refer to the |
526 | leftmost defined group, thus it's possible to do things with named capture |
527 | buffers that would otherwise require C<(??{})> code to accomplish. Named |
528 | capture buffers are numbered just as normal capture buffers are and may be |
529 | referenced via the magic numeric variables or via numeric backreferences |
530 | as well as by name. |
531 | |
14218588 |
532 | Examples: |
a0d0e21e |
533 | |
534 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words |
535 | |
81714fb9 |
536 | /(.)\1/ # find first doubled char |
537 | and print "'$1' is the first doubled character\n"; |
538 | |
539 | /(?<char>.)\k<char>/ # ... a different way |
540 | and print "'$+{char}' is the first doubled character\n"; |
541 | |
542 | /(?<char>.)\1/ # ... mix and match |
543 | and print "'$1' is the first doubled character\n"; |
c47ff5f1 |
544 | |
14218588 |
545 | if (/Time: (..):(..):(..)/) { # parse out values |
a0d0e21e |
546 | $hours = $1; |
547 | $minutes = $2; |
548 | $seconds = $3; |
549 | } |
c47ff5f1 |
550 | |
14218588 |
551 | Several special variables also refer back to portions of the previous |
552 | match. C<$+> returns whatever the last bracket match matched. |
553 | C<$&> returns the entire matched string. (At one point C<$0> did |
554 | also, but now it returns the name of the program.) C<$`> returns |
77ea4f6d |
555 | everything before the matched string. C<$'> returns everything |
556 | after the matched string. And C<$^N> contains whatever was matched by |
557 | the most-recently closed group (submatch). C<$^N> can be used in |
558 | extended patterns (see below), for example to assign a submatch to a |
81714fb9 |
559 | variable. |
d74e8afc |
560 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 |
561 | |
665e98b9 |
562 | The numbered match variables ($1, $2, $3, etc.) and the related punctuation |
77ea4f6d |
563 | set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped |
14218588 |
564 | until the end of the enclosing block or until the next successful |
565 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) |
d74e8afc |
566 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
567 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> |
568 | |
14218588 |
569 | |
665e98b9 |
570 | B<NOTE>: failed matches in Perl do not reset the match variables, |
5146ce24 |
571 | which makes it easier to write code that tests for a series of more |
665e98b9 |
572 | specific cases and remembers the best match. |
573 | |
14218588 |
574 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or |
575 | C<$'> anywhere in the program, it has to provide them for every |
576 | pattern match. This may substantially slow your program. Perl |
577 | uses the same mechanism to produce $1, $2, etc, so you also pay a |
578 | price for each pattern that contains capturing parentheses. (To |
579 | avoid this cost while retaining the grouping behaviour, use the |
580 | extended regular expression C<(?: ... )> instead.) But if you never |
581 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing |
582 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> |
583 | if you can, but if you can't (and some algorithms really appreciate |
584 | them), once you've used them once, use them at will, because you've |
585 | already paid the price. As of 5.005, C<$&> is not so costly as the |
586 | other two. |
d74e8afc |
587 | X<$&> X<$`> X<$'> |
68dc0745 |
588 | |
cde0cee5 |
589 | As a workaround for this problem, Perl 5.10 introduces C<${^PREMATCH}>, |
590 | C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&> |
591 | and C<$'>, B<except> that they are only guaranteed to be defined after a |
592 | successful match that was executed with the C</k> (keep-copy) modifier. |
593 | The use of these variables incurs no global performance penalty, unlike |
594 | their punctuation char equivalents, however at the trade-off that you |
595 | have to tell perl when you want to use them. |
596 | X</k> X<k modifier> |
597 | |
19799a22 |
598 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
599 | C<\w>, C<\n>. Unlike some other regular expression languages, there |
600 | are no backslashed symbols that aren't alphanumeric. So anything |
c47ff5f1 |
601 | that looks like \\, \(, \), \<, \>, \{, or \} is always |
19799a22 |
602 | interpreted as a literal character, not a metacharacter. This was |
603 | once used in a common idiom to disable or quote the special meanings |
604 | of regular expression metacharacters in a string that you want to |
36bbe248 |
605 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e |
606 | |
607 | $pattern =~ s/(\W)/\\$1/g; |
608 | |
f1cbbd6e |
609 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 |
610 | Today it is more common to use the quotemeta() function or the C<\Q> |
611 | metaquoting escape sequence to disable all metacharacters' special |
612 | meanings like this: |
a0d0e21e |
613 | |
614 | /$unquoted\Q$quoted\E$unquoted/ |
615 | |
9da458fc |
616 | Beware that if you put literal backslashes (those not inside |
617 | interpolated variables) between C<\Q> and C<\E>, double-quotish |
618 | backslash interpolation may lead to confusing results. If you |
619 | I<need> to use literal backslashes within C<\Q...\E>, |
620 | consult L<perlop/"Gory details of parsing quoted constructs">. |
621 | |
19799a22 |
622 | =head2 Extended Patterns |
623 | |
14218588 |
624 | Perl also defines a consistent extension syntax for features not |
625 | found in standard tools like B<awk> and B<lex>. The syntax is a |
626 | pair of parentheses with a question mark as the first thing within |
627 | the parentheses. The character after the question mark indicates |
628 | the extension. |
19799a22 |
629 | |
14218588 |
630 | The stability of these extensions varies widely. Some have been |
631 | part of the core language for many years. Others are experimental |
632 | and may change without warning or be completely removed. Check |
633 | the documentation on an individual feature to verify its current |
634 | status. |
19799a22 |
635 | |
14218588 |
636 | A question mark was chosen for this and for the minimal-matching |
637 | construct because 1) question marks are rare in older regular |
638 | expressions, and 2) whenever you see one, you should stop and |
639 | "question" exactly what is going on. That's psychology... |
a0d0e21e |
640 | |
641 | =over 10 |
642 | |
cc6b7395 |
643 | =item C<(?#text)> |
d74e8afc |
644 | X<(?#)> |
a0d0e21e |
645 | |
14218588 |
646 | A comment. The text is ignored. If the C</x> modifier enables |
19799a22 |
647 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes |
259138e3 |
648 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
649 | C<)> in the comment. |
a0d0e21e |
650 | |
cde0cee5 |
651 | =item C<(?kimsx-imsx)> |
d74e8afc |
652 | X<(?)> |
19799a22 |
653 | |
0b6d1084 |
654 | One or more embedded pattern-match modifiers, to be turned on (or |
655 | turned off, if preceded by C<->) for the remainder of the pattern or |
656 | the remainder of the enclosing pattern group (if any). This is |
657 | particularly useful for dynamic patterns, such as those read in from a |
658 | configuration file, read in as an argument, are specified in a table |
659 | somewhere, etc. Consider the case that some of which want to be case |
660 | sensitive and some do not. The case insensitive ones need to include |
661 | merely C<(?i)> at the front of the pattern. For example: |
19799a22 |
662 | |
663 | $pattern = "foobar"; |
5d458dd8 |
664 | if ( /$pattern/i ) { } |
19799a22 |
665 | |
666 | # more flexible: |
667 | |
668 | $pattern = "(?i)foobar"; |
5d458dd8 |
669 | if ( /$pattern/ ) { } |
19799a22 |
670 | |
0b6d1084 |
671 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 |
672 | |
673 | ( (?i) blah ) \s+ \1 |
674 | |
675 | will match a repeated (I<including the case>!) word C<blah> in any |
14218588 |
676 | case, assuming C<x> modifier, and no C<i> modifier outside this |
19799a22 |
677 | group. |
678 | |
cde0cee5 |
679 | Note that the C<k> modifier is special in that it can only be enabled, |
680 | not disabled, and that its presence anywhere in a pattern has a global |
681 | effect. Thus C<(?-k)> and C<(?-k:...)> are meaningless and will warn |
682 | when executed under C<use warnings>. |
683 | |
5a964f20 |
684 | =item C<(?:pattern)> |
d74e8afc |
685 | X<(?:)> |
a0d0e21e |
686 | |
ca9dfc88 |
687 | =item C<(?imsx-imsx:pattern)> |
688 | |
5a964f20 |
689 | This is for clustering, not capturing; it groups subexpressions like |
690 | "()", but doesn't make backreferences as "()" does. So |
a0d0e21e |
691 | |
5a964f20 |
692 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e |
693 | |
694 | is like |
695 | |
5a964f20 |
696 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e |
697 | |
19799a22 |
698 | but doesn't spit out extra fields. It's also cheaper not to capture |
699 | characters if you don't need to. |
a0d0e21e |
700 | |
19799a22 |
701 | Any letters between C<?> and C<:> act as flags modifiers as with |
5d458dd8 |
702 | C<(?imsx-imsx)>. For example, |
ca9dfc88 |
703 | |
704 | /(?s-i:more.*than).*million/i |
705 | |
14218588 |
706 | is equivalent to the more verbose |
ca9dfc88 |
707 | |
708 | /(?:(?s-i)more.*than).*million/i |
709 | |
ee9b8eae |
710 | =item Look-Around Assertions |
711 | X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround> |
712 | |
713 | Look-around assertions are zero width patterns which match a specific |
714 | pattern without including it in C<$&>. Positive assertions match when |
715 | their subpattern matches, negative assertions match when their subpattern |
716 | fails. Look-behind matches text up to the current match position, |
717 | look-ahead matches text following the current match position. |
718 | |
719 | =over 4 |
720 | |
5a964f20 |
721 | =item C<(?=pattern)> |
d74e8afc |
722 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> |
a0d0e21e |
723 | |
19799a22 |
724 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e |
725 | matches a word followed by a tab, without including the tab in C<$&>. |
726 | |
5a964f20 |
727 | =item C<(?!pattern)> |
d74e8afc |
728 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> |
a0d0e21e |
729 | |
19799a22 |
730 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e |
731 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 |
732 | however that look-ahead and look-behind are NOT the same thing. You cannot |
733 | use this for look-behind. |
7b8d334a |
734 | |
5a964f20 |
735 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a |
736 | will not do what you want. That's because the C<(?!foo)> is just saying that |
737 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will |
738 | match. You would have to do something like C</(?!foo)...bar/> for that. We |
739 | say "like" because there's the case of your "bar" not having three characters |
740 | before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>. |
741 | Sometimes it's still easier just to say: |
a0d0e21e |
742 | |
a3cb178b |
743 | if (/bar/ && $` !~ /foo$/) |
a0d0e21e |
744 | |
19799a22 |
745 | For look-behind see below. |
c277df42 |
746 | |
ee9b8eae |
747 | =item C<(?<=pattern)> C<\K> |
748 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K> |
c277df42 |
749 | |
c47ff5f1 |
750 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 |
751 | matches a word that follows a tab, without including the tab in C<$&>. |
752 | Works only for fixed-width look-behind. |
c277df42 |
753 | |
ee9b8eae |
754 | There is a special form of this construct, called C<\K>, which causes the |
755 | regex engine to "keep" everything it had matched prior to the C<\K> and |
756 | not include it in C<$&>. This effectively provides variable length |
757 | look-behind. The use of C<\K> inside of another look-around assertion |
758 | is allowed, but the behaviour is currently not well defined. |
759 | |
760 | For various reasons C<\K> may be signifigantly more efficient than the |
761 | equivalent C<< (?<=...) >> construct, and it is especially useful in |
762 | situations where you want to efficiently remove something following |
763 | something else in a string. For instance |
764 | |
765 | s/(foo)bar/$1/g; |
766 | |
767 | can be rewritten as the much more efficient |
768 | |
769 | s/foo\Kbar//g; |
770 | |
5a964f20 |
771 | =item C<(?<!pattern)> |
d74e8afc |
772 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> |
c277df42 |
773 | |
19799a22 |
774 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
775 | matches any occurrence of "foo" that does not follow "bar". Works |
776 | only for fixed-width look-behind. |
c277df42 |
777 | |
ee9b8eae |
778 | =back |
779 | |
81714fb9 |
780 | =item C<(?'NAME'pattern)> |
781 | |
782 | =item C<< (?<NAME>pattern) >> |
783 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> |
784 | |
785 | A named capture buffer. Identical in every respect to normal capturing |
786 | parens C<()> but for the additional fact that C<%+> may be used after |
787 | a succesful match to refer to a named buffer. See C<perlvar> for more |
788 | details on the C<%+> hash. |
789 | |
790 | If multiple distinct capture buffers have the same name then the |
791 | $+{NAME} will refer to the leftmost defined buffer in the match. |
792 | |
793 | The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent. |
794 | |
795 | B<NOTE:> While the notation of this construct is the same as the similar |
796 | function in .NET regexes, the behavior is not, in Perl the buffers are |
797 | numbered sequentially regardless of being named or not. Thus in the |
798 | pattern |
799 | |
800 | /(x)(?<foo>y)(z)/ |
801 | |
802 | $+{foo} will be the same as $2, and $3 will contain 'z' instead of |
803 | the opposite which is what a .NET regex hacker might expect. |
804 | |
1f1031fe |
805 | Currently NAME is restricted to simple identifiers only. |
806 | In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or |
807 | its Unicode extension (see L<utf8>), |
808 | though it isn't extended by the locale (see L<perllocale>). |
81714fb9 |
809 | |
1f1031fe |
810 | B<NOTE:> In order to make things easier for programmers with experience |
ee9b8eae |
811 | with the Python or PCRE regex engines the pattern C<< (?PE<lt>NAMEE<gt>pattern) >> |
1f1031fe |
812 | maybe be used instead of C<< (?<NAME>pattern) >>; however this form does not |
813 | support the use of single quotes as a delimiter for the name. This is |
814 | only available in Perl 5.10 or later. |
81714fb9 |
815 | |
1f1031fe |
816 | =item C<< \k<NAME> >> |
817 | |
818 | =item C<< \k'NAME' >> |
81714fb9 |
819 | |
820 | Named backreference. Similar to numeric backreferences, except that |
821 | the group is designated by name and not number. If multiple groups |
822 | have the same name then it refers to the leftmost defined group in |
823 | the current match. |
824 | |
825 | It is an error to refer to a name not defined by a C<(?<NAME>)> |
826 | earlier in the pattern. |
827 | |
828 | Both forms are equivalent. |
829 | |
1f1031fe |
830 | B<NOTE:> In order to make things easier for programmers with experience |
831 | with the Python or PCRE regex engines the pattern C<< (?P=NAME) >> |
832 | maybe be used instead of C<< \k<NAME> >> in Perl 5.10 or later. |
833 | |
cc6b7395 |
834 | =item C<(?{ code })> |
d74e8afc |
835 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 |
836 | |
19799a22 |
837 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf |
838 | experimental, and may be changed without notice. Code executed that |
839 | has side effects may not perform identically from version to version |
840 | due to the effect of future optimisations in the regex engine. |
c277df42 |
841 | |
cc46d5f2 |
842 | This zero-width assertion evaluates any embedded Perl code. It |
19799a22 |
843 | always succeeds, and its C<code> is not interpolated. Currently, |
844 | the rules to determine where the C<code> ends are somewhat convoluted. |
845 | |
77ea4f6d |
846 | This feature can be used together with the special variable C<$^N> to |
847 | capture the results of submatches in variables without having to keep |
848 | track of the number of nested parentheses. For example: |
849 | |
850 | $_ = "The brown fox jumps over the lazy dog"; |
851 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; |
852 | print "color = $color, animal = $animal\n"; |
853 | |
754091cb |
854 | Inside the C<(?{...})> block, C<$_> refers to the string the regular |
855 | expression is matching against. You can also use C<pos()> to know what is |
fa11829f |
856 | the current position of matching within this string. |
754091cb |
857 | |
19799a22 |
858 | The C<code> is properly scoped in the following sense: If the assertion |
859 | is backtracked (compare L<"Backtracking">), all changes introduced after |
860 | C<local>ization are undone, so that |
b9ac3b5b |
861 | |
862 | $_ = 'a' x 8; |
5d458dd8 |
863 | m< |
b9ac3b5b |
864 | (?{ $cnt = 0 }) # Initialize $cnt. |
865 | ( |
5d458dd8 |
866 | a |
b9ac3b5b |
867 | (?{ |
868 | local $cnt = $cnt + 1; # Update $cnt, backtracking-safe. |
869 | }) |
5d458dd8 |
870 | )* |
b9ac3b5b |
871 | aaaa |
872 | (?{ $res = $cnt }) # On success copy to non-localized |
873 | # location. |
874 | >x; |
875 | |
19799a22 |
876 | will set C<$res = 4>. Note that after the match, $cnt returns to the globally |
14218588 |
877 | introduced value, because the scopes that restrict C<local> operators |
b9ac3b5b |
878 | are unwound. |
879 | |
19799a22 |
880 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> |
881 | switch. If I<not> used in this way, the result of evaluation of |
882 | C<code> is put into the special variable C<$^R>. This happens |
883 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions |
884 | inside the same regular expression. |
b9ac3b5b |
885 | |
19799a22 |
886 | The assignment to C<$^R> above is properly localized, so the old |
887 | value of C<$^R> is restored if the assertion is backtracked; compare |
888 | L<"Backtracking">. |
b9ac3b5b |
889 | |
61528107 |
890 | Due to an unfortunate implementation issue, the Perl code contained in these |
891 | blocks is treated as a compile time closure that can have seemingly bizarre |
6bda09f9 |
892 | consequences when used with lexically scoped variables inside of subroutines |
61528107 |
893 | or loops. There are various workarounds for this, including simply using |
894 | global variables instead. If you are using this construct and strange results |
6bda09f9 |
895 | occur then check for the use of lexically scoped variables. |
896 | |
19799a22 |
897 | For reasons of security, this construct is forbidden if the regular |
898 | expression involves run-time interpolation of variables, unless the |
899 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the |
900 | variables contain results of C<qr//> operator (see |
5d458dd8 |
901 | L<perlop/"qr/STRING/imosx">). |
871b0233 |
902 | |
14218588 |
903 | This restriction is because of the wide-spread and remarkably convenient |
19799a22 |
904 | custom of using run-time determined strings as patterns. For example: |
871b0233 |
905 | |
906 | $re = <>; |
907 | chomp $re; |
908 | $string =~ /$re/; |
909 | |
14218588 |
910 | Before Perl knew how to execute interpolated code within a pattern, |
911 | this operation was completely safe from a security point of view, |
912 | although it could raise an exception from an illegal pattern. If |
913 | you turn on the C<use re 'eval'>, though, it is no longer secure, |
914 | so you should only do so if you are also using taint checking. |
915 | Better yet, use the carefully constrained evaluation within a Safe |
cc46d5f2 |
916 | compartment. See L<perlsec> for details about both these mechanisms. |
871b0233 |
917 | |
5d458dd8 |
918 | Because perl's regex engine is not currently re-entrant, interpolated |
8988a1bb |
919 | code may not invoke the regex engine either directly with C<m//> or C<s///>), |
920 | or indirectly with functions such as C<split>. |
921 | |
14455d6c |
922 | =item C<(??{ code })> |
d74e8afc |
923 | X<(??{})> |
924 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> |
0f5d15d6 |
925 | |
19799a22 |
926 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf |
927 | experimental, and may be changed without notice. Code executed that |
928 | has side effects may not perform identically from version to version |
929 | due to the effect of future optimisations in the regex engine. |
0f5d15d6 |
930 | |
19799a22 |
931 | This is a "postponed" regular subexpression. The C<code> is evaluated |
932 | at run time, at the moment this subexpression may match. The result |
933 | of evaluation is considered as a regular expression and matched as |
61528107 |
934 | if it were inserted instead of this construct. Note that this means |
6bda09f9 |
935 | that the contents of capture buffers defined inside an eval'ed pattern |
936 | are not available outside of the pattern, and vice versa, there is no |
937 | way for the inner pattern to refer to a capture buffer defined outside. |
938 | Thus, |
939 | |
940 | ('a' x 100)=~/(??{'(.)' x 100})/ |
941 | |
81714fb9 |
942 | B<will> match, it will B<not> set $1. |
0f5d15d6 |
943 | |
428594d9 |
944 | The C<code> is not interpolated. As before, the rules to determine |
19799a22 |
945 | where the C<code> ends are currently somewhat convoluted. |
946 | |
947 | The following pattern matches a parenthesized group: |
0f5d15d6 |
948 | |
949 | $re = qr{ |
950 | \( |
951 | (?: |
952 | (?> [^()]+ ) # Non-parens without backtracking |
953 | | |
14455d6c |
954 | (??{ $re }) # Group with matching parens |
0f5d15d6 |
955 | )* |
956 | \) |
957 | }x; |
958 | |
6bda09f9 |
959 | See also C<(?PARNO)> for a different, more efficient way to accomplish |
960 | the same task. |
961 | |
5d458dd8 |
962 | Because perl's regex engine is not currently re-entrant, delayed |
8988a1bb |
963 | code may not invoke the regex engine either directly with C<m//> or C<s///>), |
964 | or indirectly with functions such as C<split>. |
965 | |
5d458dd8 |
966 | Recursing deeper than 50 times without consuming any input string will |
967 | result in a fatal error. The maximum depth is compiled into perl, so |
6bda09f9 |
968 | changing it requires a custom build. |
969 | |
542fa716 |
970 | =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)> |
971 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)> |
6bda09f9 |
972 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
542fa716 |
973 | X<regex, relative recursion> |
6bda09f9 |
974 | |
81714fb9 |
975 | Similar to C<(??{ code })> except it does not involve compiling any code, |
976 | instead it treats the contents of a capture buffer as an independent |
61528107 |
977 | pattern that must match at the current position. Capture buffers |
81714fb9 |
978 | contained by the pattern will have the value as determined by the |
6bda09f9 |
979 | outermost recursion. |
980 | |
894be9b7 |
981 | PARNO is a sequence of digits (not starting with 0) whose value reflects |
982 | the paren-number of the capture buffer to recurse to. C<(?R)> recurses to |
983 | the beginning of the whole pattern. C<(?0)> is an alternate syntax for |
542fa716 |
984 | C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed |
985 | to be relative, with negative numbers indicating preceding capture buffers |
986 | and positive ones following. Thus C<(?-1)> refers to the most recently |
987 | declared buffer, and C<(?+1)> indicates the next buffer to be declared. |
c74340f9 |
988 | Note that the counting for relative recursion differs from that of |
989 | relative backreferences, in that with recursion unclosed buffers B<are> |
990 | included. |
6bda09f9 |
991 | |
81714fb9 |
992 | The following pattern matches a function foo() which may contain |
f145b7e9 |
993 | balanced parentheses as the argument. |
6bda09f9 |
994 | |
995 | $re = qr{ ( # paren group 1 (full function) |
81714fb9 |
996 | foo |
6bda09f9 |
997 | ( # paren group 2 (parens) |
998 | \( |
999 | ( # paren group 3 (contents of parens) |
1000 | (?: |
1001 | (?> [^()]+ ) # Non-parens without backtracking |
1002 | | |
1003 | (?2) # Recurse to start of paren group 2 |
1004 | )* |
1005 | ) |
1006 | \) |
1007 | ) |
1008 | ) |
1009 | }x; |
1010 | |
1011 | If the pattern was used as follows |
1012 | |
1013 | 'foo(bar(baz)+baz(bop))'=~/$re/ |
1014 | and print "\$1 = $1\n", |
1015 | "\$2 = $2\n", |
1016 | "\$3 = $3\n"; |
1017 | |
1018 | the output produced should be the following: |
1019 | |
1020 | $1 = foo(bar(baz)+baz(bop)) |
1021 | $2 = (bar(baz)+baz(bop)) |
81714fb9 |
1022 | $3 = bar(baz)+baz(bop) |
6bda09f9 |
1023 | |
81714fb9 |
1024 | If there is no corresponding capture buffer defined, then it is a |
61528107 |
1025 | fatal error. Recursing deeper than 50 times without consuming any input |
81714fb9 |
1026 | string will also result in a fatal error. The maximum depth is compiled |
6bda09f9 |
1027 | into perl, so changing it requires a custom build. |
1028 | |
542fa716 |
1029 | The following shows how using negative indexing can make it |
1030 | easier to embed recursive patterns inside of a C<qr//> construct |
1031 | for later use: |
1032 | |
1033 | my $parens = qr/(\((?:[^()]++|(?-1))*+\))/; |
1034 | if (/foo $parens \s+ + \s+ bar $parens/x) { |
1035 | # do something here... |
1036 | } |
1037 | |
81714fb9 |
1038 | B<Note> that this pattern does not behave the same way as the equivalent |
6bda09f9 |
1039 | PCRE or Python construct of the same form. In perl you can backtrack into |
1040 | a recursed group, in PCRE and Python the recursed into group is treated |
542fa716 |
1041 | as atomic. Also, modifiers are resolved at compile time, so constructs |
1042 | like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will |
1043 | be processed. |
6bda09f9 |
1044 | |
894be9b7 |
1045 | =item C<(?&NAME)> |
1046 | X<(?&NAME)> |
1047 | |
1048 | Recurse to a named subpattern. Identical to (?PARNO) except that the |
1049 | parenthesis to recurse to is determined by name. If multiple parens have |
1050 | the same name, then it recurses to the leftmost. |
1051 | |
1052 | It is an error to refer to a name that is not declared somewhere in the |
1053 | pattern. |
1054 | |
1f1031fe |
1055 | B<NOTE:> In order to make things easier for programmers with experience |
1056 | with the Python or PCRE regex engines the pattern C<< (?P>NAME) >> |
1057 | maybe be used instead of C<< (?&NAME) >> as of Perl 5.10. |
1058 | |
e2e6a0f1 |
1059 | =item C<(?(condition)yes-pattern|no-pattern)> |
1060 | X<(?()> |
286f584a |
1061 | |
e2e6a0f1 |
1062 | =item C<(?(condition)yes-pattern)> |
286f584a |
1063 | |
e2e6a0f1 |
1064 | Conditional expression. C<(condition)> should be either an integer in |
1065 | parentheses (which is valid if the corresponding pair of parentheses |
1066 | matched), a look-ahead/look-behind/evaluate zero-width assertion, a |
1067 | name in angle brackets or single quotes (which is valid if a buffer |
1068 | with the given name matched), or the special symbol (R) (true when |
1069 | evaluated inside of recursion or eval). Additionally the R may be |
1070 | followed by a number, (which will be true when evaluated when recursing |
1071 | inside of the appropriate group), or by C<&NAME>, in which case it will |
1072 | be true only when evaluated during recursion in the named group. |
1073 | |
1074 | Here's a summary of the possible predicates: |
1075 | |
1076 | =over 4 |
1077 | |
1078 | =item (1) (2) ... |
1079 | |
1080 | Checks if the numbered capturing buffer has matched something. |
1081 | |
1082 | =item (<NAME>) ('NAME') |
1083 | |
1084 | Checks if a buffer with the given name has matched something. |
1085 | |
1086 | =item (?{ CODE }) |
1087 | |
1088 | Treats the code block as the condition. |
1089 | |
1090 | =item (R) |
1091 | |
1092 | Checks if the expression has been evaluated inside of recursion. |
1093 | |
1094 | =item (R1) (R2) ... |
1095 | |
1096 | Checks if the expression has been evaluated while executing directly |
1097 | inside of the n-th capture group. This check is the regex equivalent of |
1098 | |
1099 | if ((caller(0))[3] eq 'subname') { ... } |
1100 | |
1101 | In other words, it does not check the full recursion stack. |
1102 | |
1103 | =item (R&NAME) |
1104 | |
1105 | Similar to C<(R1)>, this predicate checks to see if we're executing |
1106 | directly inside of the leftmost group with a given name (this is the same |
1107 | logic used by C<(?&NAME)> to disambiguate). It does not check the full |
1108 | stack, but only the name of the innermost active recursion. |
1109 | |
1110 | =item (DEFINE) |
1111 | |
1112 | In this case, the yes-pattern is never directly executed, and no |
1113 | no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient. |
1114 | See below for details. |
1115 | |
1116 | =back |
1117 | |
1118 | For example: |
1119 | |
1120 | m{ ( \( )? |
1121 | [^()]+ |
1122 | (?(1) \) ) |
1123 | }x |
1124 | |
1125 | matches a chunk of non-parentheses, possibly included in parentheses |
1126 | themselves. |
1127 | |
1128 | A special form is the C<(DEFINE)> predicate, which never executes directly |
1129 | its yes-pattern, and does not allow a no-pattern. This allows to define |
1130 | subpatterns which will be executed only by using the recursion mechanism. |
1131 | This way, you can define a set of regular expression rules that can be |
1132 | bundled into any pattern you choose. |
1133 | |
1134 | It is recommended that for this usage you put the DEFINE block at the |
1135 | end of the pattern, and that you name any subpatterns defined within it. |
1136 | |
1137 | Also, it's worth noting that patterns defined this way probably will |
1138 | not be as efficient, as the optimiser is not very clever about |
1139 | handling them. |
1140 | |
1141 | An example of how this might be used is as follows: |
1142 | |
2bf803e2 |
1143 | /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT)) |
e2e6a0f1 |
1144 | (?(DEFINE) |
2bf803e2 |
1145 | (?<NAME_PAT>....) |
1146 | (?<ADRESS_PAT>....) |
e2e6a0f1 |
1147 | )/x |
1148 | |
1149 | Note that capture buffers matched inside of recursion are not accessible |
1150 | after the recursion returns, so the extra layer of capturing buffers are |
1151 | necessary. Thus C<$+{NAME_PAT}> would not be defined even though |
1152 | C<$+{NAME}> would be. |
286f584a |
1153 | |
c47ff5f1 |
1154 | =item C<< (?>pattern) >> |
6bda09f9 |
1155 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 |
1156 | |
19799a22 |
1157 | An "independent" subexpression, one which matches the substring |
1158 | that a I<standalone> C<pattern> would match if anchored at the given |
9da458fc |
1159 | position, and it matches I<nothing other than this substring>. This |
19799a22 |
1160 | construct is useful for optimizations of what would otherwise be |
1161 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). |
9da458fc |
1162 | It may also be useful in places where the "grab all you can, and do not |
1163 | give anything back" semantic is desirable. |
19799a22 |
1164 | |
c47ff5f1 |
1165 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 |
1166 | (anchored at the beginning of string, as above) will match I<all> |
1167 | characters C<a> at the beginning of string, leaving no C<a> for |
1168 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, |
1169 | since the match of the subgroup C<a*> is influenced by the following |
1170 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside |
1171 | C<a*ab> will match fewer characters than a standalone C<a*>, since |
1172 | this makes the tail match. |
1173 | |
c47ff5f1 |
1174 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
19799a22 |
1175 | C<(?=(pattern))\1>. This matches the same substring as a standalone |
1176 | C<a+>, and the following C<\1> eats the matched string; it therefore |
c47ff5f1 |
1177 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 |
1178 | (The difference between these two constructs is that the second one |
1179 | uses a capturing group, thus shifting ordinals of backreferences |
1180 | in the rest of a regular expression.) |
1181 | |
1182 | Consider this pattern: |
c277df42 |
1183 | |
871b0233 |
1184 | m{ \( |
e2e6a0f1 |
1185 | ( |
1186 | [^()]+ # x+ |
1187 | | |
871b0233 |
1188 | \( [^()]* \) |
1189 | )+ |
e2e6a0f1 |
1190 | \) |
871b0233 |
1191 | }x |
5a964f20 |
1192 | |
19799a22 |
1193 | That will efficiently match a nonempty group with matching parentheses |
1194 | two levels deep or less. However, if there is no such group, it |
1195 | will take virtually forever on a long string. That's because there |
1196 | are so many different ways to split a long string into several |
1197 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar |
1198 | to a subpattern of the above pattern. Consider how the pattern |
1199 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several |
1200 | seconds, but that each extra letter doubles this time. This |
1201 | exponential performance will make it appear that your program has |
14218588 |
1202 | hung. However, a tiny change to this pattern |
5a964f20 |
1203 | |
e2e6a0f1 |
1204 | m{ \( |
1205 | ( |
1206 | (?> [^()]+ ) # change x+ above to (?> x+ ) |
1207 | | |
871b0233 |
1208 | \( [^()]* \) |
1209 | )+ |
e2e6a0f1 |
1210 | \) |
871b0233 |
1211 | }x |
c277df42 |
1212 | |
c47ff5f1 |
1213 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 |
1214 | this yourself would be a productive exercise), but finishes in a fourth |
1215 | the time when used on a similar string with 1000000 C<a>s. Be aware, |
1216 | however, that this pattern currently triggers a warning message under |
9f1b1f2d |
1217 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b |
1218 | C<"matches null string many times in regex">. |
c277df42 |
1219 | |
c47ff5f1 |
1220 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 |
1221 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 |
1222 | This was only 4 times slower on a string with 1000000 C<a>s. |
1223 | |
9da458fc |
1224 | The "grab all you can, and do not give anything back" semantic is desirable |
1225 | in many situations where on the first sight a simple C<()*> looks like |
1226 | the correct solution. Suppose we parse text with comments being delimited |
1227 | by C<#> followed by some optional (horizontal) whitespace. Contrary to |
4375e838 |
1228 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc |
1229 | the comment delimiter, because it may "give up" some whitespace if |
1230 | the remainder of the pattern can be made to match that way. The correct |
1231 | answer is either one of these: |
1232 | |
1233 | (?>#[ \t]*) |
1234 | #[ \t]*(?![ \t]) |
1235 | |
1236 | For example, to grab non-empty comments into $1, one should use either |
1237 | one of these: |
1238 | |
1239 | / (?> \# [ \t]* ) ( .+ ) /x; |
1240 | / \# [ \t]* ( [^ \t] .* ) /x; |
1241 | |
1242 | Which one you pick depends on which of these expressions better reflects |
1243 | the above specification of comments. |
1244 | |
6bda09f9 |
1245 | In some literature this construct is called "atomic matching" or |
1246 | "possessive matching". |
1247 | |
b9b4dddf |
1248 | Possessive quantifiers are equivalent to putting the item they are applied |
1249 | to inside of one of these constructs. The following equivalences apply: |
1250 | |
1251 | Quantifier Form Bracketing Form |
1252 | --------------- --------------- |
1253 | PAT*+ (?>PAT*) |
1254 | PAT++ (?>PAT+) |
1255 | PAT?+ (?>PAT?) |
1256 | PAT{min,max}+ (?>PAT{min,max}) |
1257 | |
e2e6a0f1 |
1258 | =back |
1259 | |
1260 | =head2 Special Backtracking Control Verbs |
1261 | |
1262 | B<WARNING:> These patterns are experimental and subject to change or |
1263 | removal in a future version of perl. Their usage in production code should |
1264 | be noted to avoid problems during upgrades. |
1265 | |
1266 | These special patterns are generally of the form C<(*VERB:ARG)>. Unless |
1267 | otherwise stated the ARG argument is optional; in some cases, it is |
1268 | forbidden. |
1269 | |
1270 | Any pattern containing a special backtracking verb that allows an argument |
1271 | has the special behaviour that when executed it sets the current packages' |
5d458dd8 |
1272 | C<$REGERROR> and C<$REGMARK> variables. When doing so the following |
1273 | rules apply: |
e2e6a0f1 |
1274 | |
5d458dd8 |
1275 | On failure, the C<$REGERROR> variable will be set to the ARG value of the |
1276 | verb pattern, if the verb was involved in the failure of the match. If the |
1277 | ARG part of the pattern was omitted, then C<$REGERROR> will be set to the |
1278 | name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was |
1279 | none. Also, the C<$REGMARK> variable will be set to FALSE. |
e2e6a0f1 |
1280 | |
5d458dd8 |
1281 | On a successful match, the C<$REGERROR> variable will be set to FALSE, and |
1282 | the C<$REGMARK> variable will be set to the name of the last |
1283 | C<(*MARK:NAME)> pattern executed. See the explanation for the |
1284 | C<(*MARK:NAME)> verb below for more details. |
e2e6a0f1 |
1285 | |
5d458dd8 |
1286 | B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1> |
1287 | and most other regex related variables. They are not local to a scope, nor |
1288 | readonly, but instead are volatile package variables similar to C<$AUTOLOAD>. |
1289 | Use C<local> to localize changes to them to a specific scope if necessary. |
e2e6a0f1 |
1290 | |
1291 | If a pattern does not contain a special backtracking verb that allows an |
5d458dd8 |
1292 | argument, then C<$REGERROR> and C<$REGMARK> are not touched at all. |
e2e6a0f1 |
1293 | |
1294 | =over 4 |
1295 | |
1296 | =item Verbs that take an argument |
1297 | |
1298 | =over 4 |
1299 | |
5d458dd8 |
1300 | =item C<(*PRUNE)> C<(*PRUNE:NAME)> |
ee9b8eae |
1301 | X<(*PRUNE)> X<(*PRUNE:NAME)> X<\v> |
54612592 |
1302 | |
5d458dd8 |
1303 | This zero-width pattern prunes the backtracking tree at the current point |
1304 | when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>, |
1305 | where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached, |
1306 | A may backtrack as necessary to match. Once it is reached, matching |
1307 | continues in B, which may also backtrack as necessary; however, should B |
1308 | not match, then no further backtracking will take place, and the pattern |
1309 | will fail outright at the current starting position. |
54612592 |
1310 | |
ee9b8eae |
1311 | As a shortcut, X<\v> is exactly equivalent to C<(*PRUNE)>. |
1312 | |
54612592 |
1313 | The following example counts all the possible matching strings in a |
1314 | pattern (without actually matching any of them). |
1315 | |
e2e6a0f1 |
1316 | 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 |
1317 | print "Count=$count\n"; |
1318 | |
1319 | which produces: |
1320 | |
1321 | aaab |
1322 | aaa |
1323 | aa |
1324 | a |
1325 | aab |
1326 | aa |
1327 | a |
1328 | ab |
1329 | a |
1330 | Count=9 |
1331 | |
5d458dd8 |
1332 | If we add a C<(*PRUNE)> before the count like the following |
54612592 |
1333 | |
5d458dd8 |
1334 | 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 |
1335 | print "Count=$count\n"; |
1336 | |
1337 | we prevent backtracking and find the count of the longest matching |
1338 | at each matching startpoint like so: |
1339 | |
1340 | aaab |
1341 | aab |
1342 | ab |
1343 | Count=3 |
1344 | |
5d458dd8 |
1345 | Any number of C<(*PRUNE)> assertions may be used in a pattern. |
54612592 |
1346 | |
5d458dd8 |
1347 | See also C<< (?>pattern) >> and possessive quantifiers for other ways to |
1348 | control backtracking. In some cases, the use of C<(*PRUNE)> can be |
1349 | replaced with a C<< (?>pattern) >> with no functional difference; however, |
1350 | C<(*PRUNE)> can be used to handle cases that cannot be expressed using a |
1351 | C<< (?>pattern) >> alone. |
54612592 |
1352 | |
e2e6a0f1 |
1353 | |
5d458dd8 |
1354 | =item C<(*SKIP)> C<(*SKIP:NAME)> |
1355 | X<(*SKIP)> |
e2e6a0f1 |
1356 | |
5d458dd8 |
1357 | This zero-width pattern is similar to C<(*PRUNE)>, except that on |
e2e6a0f1 |
1358 | failure it also signifies that whatever text that was matched leading up |
5d458dd8 |
1359 | to the C<(*SKIP)> pattern being executed cannot be part of I<any> match |
1360 | of this pattern. This effectively means that the regex engine "skips" forward |
1361 | to this position on failure and tries to match again, (assuming that |
1362 | there is sufficient room to match). |
1363 | |
ee9b8eae |
1364 | As a shortcut X<\V> is exactly equivalent to C<(*SKIP)>. |
1365 | |
5d458dd8 |
1366 | The name of the C<(*SKIP:NAME)> pattern has special significance. If a |
1367 | C<(*MARK:NAME)> was encountered while matching, then it is that position |
1368 | which is used as the "skip point". If no C<(*MARK)> of that name was |
1369 | encountered, then the C<(*SKIP)> operator has no effect. When used |
1370 | without a name the "skip point" is where the match point was when |
1371 | executing the (*SKIP) pattern. |
1372 | |
1373 | Compare the following to the examples in C<(*PRUNE)>, note the string |
24b23f37 |
1374 | is twice as long: |
1375 | |
5d458dd8 |
1376 | 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 |
1377 | print "Count=$count\n"; |
1378 | |
1379 | outputs |
1380 | |
1381 | aaab |
1382 | aaab |
1383 | Count=2 |
1384 | |
5d458dd8 |
1385 | Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)> |
e2e6a0f1 |
1386 | executed, the next startpoint will be where the cursor was when the |
5d458dd8 |
1387 | C<(*SKIP)> was executed. |
1388 | |
5d458dd8 |
1389 | =item C<(*MARK:NAME)> C<(*:NAME)> |
1390 | X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)> |
1391 | |
1392 | This zero-width pattern can be used to mark the point reached in a string |
1393 | when a certain part of the pattern has been successfully matched. This |
1394 | mark may be given a name. A later C<(*SKIP)> pattern will then skip |
1395 | forward to that point if backtracked into on failure. Any number of |
1396 | C<(*MARK)> patterns are allowed, and the NAME portion is optional and may |
1397 | be duplicated. |
1398 | |
1399 | In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)> |
1400 | can be used to "label" a pattern branch, so that after matching, the |
1401 | program can determine which branches of the pattern were involved in the |
1402 | match. |
1403 | |
1404 | When a match is successful, the C<$REGMARK> variable will be set to the |
1405 | name of the most recently executed C<(*MARK:NAME)> that was involved |
1406 | in the match. |
1407 | |
1408 | This can be used to determine which branch of a pattern was matched |
1409 | without using a seperate capture buffer for each branch, which in turn |
1410 | can result in a performance improvement, as perl cannot optimize |
1411 | C</(?:(x)|(y)|(z))/> as efficiently as something like |
1412 | C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>. |
1413 | |
1414 | When a match has failed, and unless another verb has been involved in |
1415 | failing the match and has provided its own name to use, the C<$REGERROR> |
1416 | variable will be set to the name of the most recently executed |
1417 | C<(*MARK:NAME)>. |
1418 | |
1419 | See C<(*SKIP)> for more details. |
1420 | |
b62d2d15 |
1421 | As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>. |
1422 | |
5d458dd8 |
1423 | =item C<(*THEN)> C<(*THEN:NAME)> |
1424 | |
1425 | This is similar to the "cut group" operator C<::> from Perl6. Like |
1426 | C<(*PRUNE)>, this verb always matches, and when backtracked into on |
1427 | failure, it causes the regex engine to try the next alternation in the |
1428 | innermost enclosing group (capturing or otherwise). |
1429 | |
1430 | Its name comes from the observation that this operation combined with the |
1431 | alternation operator (C<|>) can be used to create what is essentially a |
1432 | pattern-based if/then/else block: |
1433 | |
1434 | ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) |
1435 | |
1436 | Note that if this operator is used and NOT inside of an alternation then |
1437 | it acts exactly like the C<(*PRUNE)> operator. |
1438 | |
1439 | / A (*PRUNE) B / |
1440 | |
1441 | is the same as |
1442 | |
1443 | / A (*THEN) B / |
1444 | |
1445 | but |
1446 | |
1447 | / ( A (*THEN) B | C (*THEN) D ) / |
1448 | |
1449 | is not the same as |
1450 | |
1451 | / ( A (*PRUNE) B | C (*PRUNE) D ) / |
1452 | |
1453 | as after matching the A but failing on the B the C<(*THEN)> verb will |
1454 | backtrack and try C; but the C<(*PRUNE)> verb will simply fail. |
24b23f37 |
1455 | |
e2e6a0f1 |
1456 | =item C<(*COMMIT)> |
1457 | X<(*COMMIT)> |
24b23f37 |
1458 | |
5d458dd8 |
1459 | This is the Perl6 "commit pattern" C<< <commit> >> or C<:::>. It's a |
1460 | zero-width pattern similar to C<(*SKIP)>, except that when backtracked |
1461 | into on failure it causes the match to fail outright. No further attempts |
1462 | to find a valid match by advancing the start pointer will occur again. |
1463 | For example, |
24b23f37 |
1464 | |
e2e6a0f1 |
1465 | 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 |
1466 | print "Count=$count\n"; |
1467 | |
1468 | outputs |
1469 | |
1470 | aaab |
1471 | Count=1 |
1472 | |
e2e6a0f1 |
1473 | In other words, once the C<(*COMMIT)> has been entered, and if the pattern |
1474 | does not match, the regex engine will not try any further matching on the |
1475 | rest of the string. |
c277df42 |
1476 | |
e2e6a0f1 |
1477 | =back |
9af228c6 |
1478 | |
e2e6a0f1 |
1479 | =item Verbs without an argument |
9af228c6 |
1480 | |
1481 | =over 4 |
1482 | |
e2e6a0f1 |
1483 | =item C<(*FAIL)> C<(*F)> |
1484 | X<(*FAIL)> X<(*F)> |
9af228c6 |
1485 | |
e2e6a0f1 |
1486 | This pattern matches nothing and always fails. It can be used to force the |
1487 | engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In |
1488 | fact, C<(?!)> gets optimised into C<(*FAIL)> internally. |
9af228c6 |
1489 | |
e2e6a0f1 |
1490 | It is probably useful only when combined with C<(?{})> or C<(??{})>. |
9af228c6 |
1491 | |
e2e6a0f1 |
1492 | =item C<(*ACCEPT)> |
1493 | X<(*ACCEPT)> |
9af228c6 |
1494 | |
e2e6a0f1 |
1495 | B<WARNING:> This feature is highly experimental. It is not recommended |
1496 | for production code. |
9af228c6 |
1497 | |
e2e6a0f1 |
1498 | This pattern matches nothing and causes the end of successful matching at |
1499 | the point at which the C<(*ACCEPT)> pattern was encountered, regardless of |
1500 | whether there is actually more to match in the string. When inside of a |
1501 | nested pattern, such as recursion or a dynamically generated subbpattern |
1502 | via C<(??{})>, only the innermost pattern is ended immediately. |
9af228c6 |
1503 | |
e2e6a0f1 |
1504 | If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are |
1505 | marked as ended at the point at which the C<(*ACCEPT)> was encountered. |
1506 | For instance: |
9af228c6 |
1507 | |
e2e6a0f1 |
1508 | 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x; |
9af228c6 |
1509 | |
e2e6a0f1 |
1510 | will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not |
1511 | be set. If another branch in the inner parens were matched, such as in the |
1512 | string 'ACDE', then the C<D> and C<E> would have to be matched as well. |
9af228c6 |
1513 | |
1514 | =back |
c277df42 |
1515 | |
a0d0e21e |
1516 | =back |
1517 | |
c07a80fd |
1518 | =head2 Backtracking |
d74e8afc |
1519 | X<backtrack> X<backtracking> |
c07a80fd |
1520 | |
35a734be |
1521 | NOTE: This section presents an abstract approximation of regular |
1522 | expression behavior. For a more rigorous (and complicated) view of |
1523 | the rules involved in selecting a match among possible alternatives, |
1524 | see L<Combining pieces together>. |
1525 | |
c277df42 |
1526 | A fundamental feature of regular expression matching involves the |
5a964f20 |
1527 | notion called I<backtracking>, which is currently used (when needed) |
c277df42 |
1528 | by all regular expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc |
1529 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
1530 | internally, but the general principle outlined here is valid. |
c07a80fd |
1531 | |
1532 | For a regular expression to match, the I<entire> regular expression must |
1533 | match, not just part of it. So if the beginning of a pattern containing a |
1534 | quantifier succeeds in a way that causes later parts in the pattern to |
1535 | fail, the matching engine backs up and recalculates the beginning |
1536 | part--that's why it's called backtracking. |
1537 | |
1538 | Here is an example of backtracking: Let's say you want to find the |
1539 | word following "foo" in the string "Food is on the foo table.": |
1540 | |
1541 | $_ = "Food is on the foo table."; |
1542 | if ( /\b(foo)\s+(\w+)/i ) { |
1543 | print "$2 follows $1.\n"; |
1544 | } |
1545 | |
1546 | When the match runs, the first part of the regular expression (C<\b(foo)>) |
1547 | finds a possible match right at the beginning of the string, and loads up |
1548 | $1 with "Foo". However, as soon as the matching engine sees that there's |
1549 | no whitespace following the "Foo" that it had saved in $1, it realizes its |
68dc0745 |
1550 | mistake and starts over again one character after where it had the |
c07a80fd |
1551 | tentative match. This time it goes all the way until the next occurrence |
1552 | of "foo". The complete regular expression matches this time, and you get |
1553 | the expected output of "table follows foo." |
1554 | |
1555 | Sometimes minimal matching can help a lot. Imagine you'd like to match |
1556 | everything between "foo" and "bar". Initially, you write something |
1557 | like this: |
1558 | |
1559 | $_ = "The food is under the bar in the barn."; |
1560 | if ( /foo(.*)bar/ ) { |
1561 | print "got <$1>\n"; |
1562 | } |
1563 | |
1564 | Which perhaps unexpectedly yields: |
1565 | |
1566 | got <d is under the bar in the > |
1567 | |
1568 | That's because C<.*> was greedy, so you get everything between the |
14218588 |
1569 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd |
1570 | to use minimal matching to make sure you get the text between a "foo" |
1571 | and the first "bar" thereafter. |
1572 | |
1573 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } |
1574 | got <d is under the > |
1575 | |
1576 | Here's another example: let's say you'd like to match a number at the end |
b6e13d97 |
1577 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd |
1578 | So you write this: |
1579 | |
1580 | $_ = "I have 2 numbers: 53147"; |
1581 | if ( /(.*)(\d*)/ ) { # Wrong! |
1582 | print "Beginning is <$1>, number is <$2>.\n"; |
1583 | } |
1584 | |
1585 | That won't work at all, because C<.*> was greedy and gobbled up the |
1586 | whole string. As C<\d*> can match on an empty string the complete |
1587 | regular expression matched successfully. |
1588 | |
8e1088bc |
1589 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd |
1590 | |
1591 | Here are some variants, most of which don't work: |
1592 | |
1593 | $_ = "I have 2 numbers: 53147"; |
1594 | @pats = qw{ |
1595 | (.*)(\d*) |
1596 | (.*)(\d+) |
1597 | (.*?)(\d*) |
1598 | (.*?)(\d+) |
1599 | (.*)(\d+)$ |
1600 | (.*?)(\d+)$ |
1601 | (.*)\b(\d+)$ |
1602 | (.*\D)(\d+)$ |
1603 | }; |
1604 | |
1605 | for $pat (@pats) { |
1606 | printf "%-12s ", $pat; |
1607 | if ( /$pat/ ) { |
1608 | print "<$1> <$2>\n"; |
1609 | } else { |
1610 | print "FAIL\n"; |
1611 | } |
1612 | } |
1613 | |
1614 | That will print out: |
1615 | |
1616 | (.*)(\d*) <I have 2 numbers: 53147> <> |
1617 | (.*)(\d+) <I have 2 numbers: 5314> <7> |
1618 | (.*?)(\d*) <> <> |
1619 | (.*?)(\d+) <I have > <2> |
1620 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> |
1621 | (.*?)(\d+)$ <I have 2 numbers: > <53147> |
1622 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> |
1623 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> |
1624 | |
1625 | As you see, this can be a bit tricky. It's important to realize that a |
1626 | regular expression is merely a set of assertions that gives a definition |
1627 | of success. There may be 0, 1, or several different ways that the |
1628 | definition might succeed against a particular string. And if there are |
5a964f20 |
1629 | multiple ways it might succeed, you need to understand backtracking to |
1630 | know which variety of success you will achieve. |
c07a80fd |
1631 | |
19799a22 |
1632 | When using look-ahead assertions and negations, this can all get even |
8b19b778 |
1633 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd |
1634 | followed by "123". You might try to write that as |
1635 | |
871b0233 |
1636 | $_ = "ABC123"; |
1637 | if ( /^\D*(?!123)/ ) { # Wrong! |
1638 | print "Yup, no 123 in $_\n"; |
1639 | } |
c07a80fd |
1640 | |
1641 | But that isn't going to match; at least, not the way you're hoping. It |
1642 | claims that there is no 123 in the string. Here's a clearer picture of |
9b9391b2 |
1643 | why that pattern matches, contrary to popular expectations: |
c07a80fd |
1644 | |
4358a253 |
1645 | $x = 'ABC123'; |
1646 | $y = 'ABC445'; |
c07a80fd |
1647 | |
4358a253 |
1648 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
1649 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; |
c07a80fd |
1650 | |
4358a253 |
1651 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; |
1652 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; |
c07a80fd |
1653 | |
1654 | This prints |
1655 | |
1656 | 2: got ABC |
1657 | 3: got AB |
1658 | 4: got ABC |
1659 | |
5f05dabc |
1660 | You might have expected test 3 to fail because it seems to a more |
c07a80fd |
1661 | general purpose version of test 1. The important difference between |
1662 | them is that test 3 contains a quantifier (C<\D*>) and so can use |
1663 | backtracking, whereas test 1 will not. What's happening is |
1664 | that you've asked "Is it true that at the start of $x, following 0 or more |
5f05dabc |
1665 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd |
1666 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 |
1667 | fail. |
14218588 |
1668 | |
c07a80fd |
1669 | The search engine will initially match C<\D*> with "ABC". Then it will |
14218588 |
1670 | try to match C<(?!123> with "123", which fails. But because |
c07a80fd |
1671 | a quantifier (C<\D*>) has been used in the regular expression, the |
1672 | search engine can backtrack and retry the match differently |
54310121 |
1673 | in the hope of matching the complete regular expression. |
c07a80fd |
1674 | |
5a964f20 |
1675 | The pattern really, I<really> wants to succeed, so it uses the |
1676 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this |
c07a80fd |
1677 | time. Now there's indeed something following "AB" that is not |
14218588 |
1678 | "123". It's "C123", which suffices. |
c07a80fd |
1679 | |
14218588 |
1680 | We can deal with this by using both an assertion and a negation. |
1681 | We'll say that the first part in $1 must be followed both by a digit |
1682 | and by something that's not "123". Remember that the look-aheads |
1683 | are zero-width expressions--they only look, but don't consume any |
1684 | of the string in their match. So rewriting this way produces what |
c07a80fd |
1685 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
1686 | |
4358a253 |
1687 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; |
1688 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; |
c07a80fd |
1689 | |
1690 | 6: got ABC |
1691 | |
5a964f20 |
1692 | In other words, the two zero-width assertions next to each other work as though |
19799a22 |
1693 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd |
1694 | matches only if you're at the beginning of the line AND the end of the |
1695 | line simultaneously. The deeper underlying truth is that juxtaposition in |
1696 | regular expressions always means AND, except when you write an explicit OR |
1697 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", |
1698 | although the attempted matches are made at different positions because "a" |
1699 | is not a zero-width assertion, but a one-width assertion. |
1700 | |
19799a22 |
1701 | B<WARNING>: particularly complicated regular expressions can take |
14218588 |
1702 | exponential time to solve because of the immense number of possible |
9da458fc |
1703 | ways they can use backtracking to try match. For example, without |
1704 | internal optimizations done by the regular expression engine, this will |
1705 | take a painfully long time to run: |
c07a80fd |
1706 | |
e1901655 |
1707 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
1708 | |
1709 | And if you used C<*>'s in the internal groups instead of limiting them |
1710 | to 0 through 5 matches, then it would take forever--or until you ran |
1711 | out of stack space. Moreover, these internal optimizations are not |
1712 | always applicable. For example, if you put C<{0,5}> instead of C<*> |
1713 | on the external group, no current optimization is applicable, and the |
1714 | match takes a long time to finish. |
c07a80fd |
1715 | |
9da458fc |
1716 | A powerful tool for optimizing such beasts is what is known as an |
1717 | "independent group", |
c47ff5f1 |
1718 | which does not backtrack (see L<C<< (?>pattern) >>>). Note also that |
9da458fc |
1719 | zero-length look-ahead/look-behind assertions will not backtrack to make |
5d458dd8 |
1720 | the tail match, since they are in "logical" context: only |
14218588 |
1721 | whether they match is considered relevant. For an example |
9da458fc |
1722 | where side-effects of look-ahead I<might> have influenced the |
c47ff5f1 |
1723 | following match, see L<C<< (?>pattern) >>>. |
c277df42 |
1724 | |
a0d0e21e |
1725 | =head2 Version 8 Regular Expressions |
d74e8afc |
1726 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> |
a0d0e21e |
1727 | |
5a964f20 |
1728 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e |
1729 | routines, here are the pattern-matching rules not described above. |
1730 | |
54310121 |
1731 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e |
1732 | with a special meaning described here or above. You can cause |
5a964f20 |
1733 | characters that normally function as metacharacters to be interpreted |
5f05dabc |
1734 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
a0d0e21e |
1735 | character; "\\" matches a "\"). A series of characters matches that |
1736 | series of characters in the target string, so the pattern C<blurfl> |
1737 | would match "blurfl" in the target string. |
1738 | |
1739 | You can specify a character class, by enclosing a list of characters |
5d458dd8 |
1740 | in C<[]>, which will match any character from the list. If the |
a0d0e21e |
1741 | first character after the "[" is "^", the class matches any character not |
14218588 |
1742 | in the list. Within a list, the "-" character specifies a |
5a964f20 |
1743 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 |
1744 | inclusive. If you want either "-" or "]" itself to be a member of a |
1745 | class, put it at the start of the list (possibly after a "^"), or |
1746 | escape it with a backslash. "-" is also taken literally when it is |
1747 | at the end of the list, just before the closing "]". (The |
84850974 |
1748 | following all specify the same class of three characters: C<[-az]>, |
1749 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which |
5d458dd8 |
1750 | specifies a class containing twenty-six characters, even on EBCDIC-based |
1751 | character sets.) Also, if you try to use the character |
1752 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of |
1753 | a range, the "-" is understood literally. |
a0d0e21e |
1754 | |
8ada0baa |
1755 | Note also that the whole range idea is rather unportable between |
1756 | character sets--and even within character sets they may cause results |
1757 | you probably didn't expect. A sound principle is to use only ranges |
1758 | that begin from and end at either alphabets of equal case ([a-e], |
1759 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, |
1760 | spell out the character sets in full. |
1761 | |
54310121 |
1762 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e |
1763 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
1764 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string |
5d458dd8 |
1765 | of octal digits, matches the character whose coded character set value |
1766 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, |
1767 | matches the character whose numeric value is I<nn>. The expression \cI<x> |
1768 | matches the character control-I<x>. Finally, the "." metacharacter |
fb55449c |
1769 | matches any character except "\n" (unless you use C</s>). |
a0d0e21e |
1770 | |
1771 | You can specify a series of alternatives for a pattern using "|" to |
1772 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", |
5a964f20 |
1773 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e |
1774 | first alternative includes everything from the last pattern delimiter |
1775 | ("(", "[", or the beginning of the pattern) up to the first "|", and |
1776 | the last alternative contains everything from the last "|" to the next |
14218588 |
1777 | pattern delimiter. That's why it's common practice to include |
1778 | alternatives in parentheses: to minimize confusion about where they |
a3cb178b |
1779 | start and end. |
1780 | |
5a964f20 |
1781 | Alternatives are tried from left to right, so the first |
a3cb178b |
1782 | alternative found for which the entire expression matches, is the one that |
1783 | is chosen. This means that alternatives are not necessarily greedy. For |
628afcb5 |
1784 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b |
1785 | part will match, as that is the first alternative tried, and it successfully |
1786 | matches the target string. (This might not seem important, but it is |
1787 | important when you are capturing matched text using parentheses.) |
1788 | |
5a964f20 |
1789 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b |
1790 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e |
1791 | |
14218588 |
1792 | Within a pattern, you may designate subpatterns for later reference |
1793 | by enclosing them in parentheses, and you may refer back to the |
1794 | I<n>th subpattern later in the pattern using the metacharacter |
1795 | \I<n>. Subpatterns are numbered based on the left to right order |
1796 | of their opening parenthesis. A backreference matches whatever |
1797 | actually matched the subpattern in the string being examined, not |
1798 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will |
1799 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern |
1800 | 1 matched "0x", even though the rule C<0|0x> could potentially match |
1801 | the leading 0 in the second number. |
cb1a09d0 |
1802 | |
19799a22 |
1803 | =head2 Warning on \1 vs $1 |
cb1a09d0 |
1804 | |
5a964f20 |
1805 | Some people get too used to writing things like: |
cb1a09d0 |
1806 | |
1807 | $pattern =~ s/(\W)/\\\1/g; |
1808 | |
1809 | This is grandfathered for the RHS of a substitute to avoid shocking the |
1810 | B<sed> addicts, but it's a dirty habit to get into. That's because in |
d1be9408 |
1811 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 |
1812 | the usual double-quoted string means a control-A. The customary Unix |
1813 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit |
1814 | of doing that, you get yourself into trouble if you then add an C</e> |
1815 | modifier. |
1816 | |
5a964f20 |
1817 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 |
1818 | |
1819 | Or if you try to do |
1820 | |
1821 | s/(\d+)/\1000/; |
1822 | |
1823 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with |
14218588 |
1824 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 |
1825 | with the operation of matching a backreference. Certainly they mean two |
1826 | different things on the I<left> side of the C<s///>. |
9fa51da4 |
1827 | |
c84d73f1 |
1828 | =head2 Repeated patterns matching zero-length substring |
1829 | |
19799a22 |
1830 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 |
1831 | |
1832 | Regular expressions provide a terse and powerful programming language. As |
1833 | with most other power tools, power comes together with the ability |
1834 | to wreak havoc. |
1835 | |
1836 | A common abuse of this power stems from the ability to make infinite |
628afcb5 |
1837 | loops using regular expressions, with something as innocuous as: |
c84d73f1 |
1838 | |
1839 | 'foo' =~ m{ ( o? )* }x; |
1840 | |
1841 | The C<o?> can match at the beginning of C<'foo'>, and since the position |
1842 | in the string is not moved by the match, C<o?> would match again and again |
14218588 |
1843 | because of the C<*> modifier. Another common way to create a similar cycle |
c84d73f1 |
1844 | is with the looping modifier C<//g>: |
1845 | |
1846 | @matches = ( 'foo' =~ m{ o? }xg ); |
1847 | |
1848 | or |
1849 | |
1850 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; |
1851 | |
1852 | or the loop implied by split(). |
1853 | |
1854 | However, long experience has shown that many programming tasks may |
14218588 |
1855 | be significantly simplified by using repeated subexpressions that |
1856 | may match zero-length substrings. Here's a simple example being: |
c84d73f1 |
1857 | |
1858 | @chars = split //, $string; # // is not magic in split |
1859 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / |
1860 | |
9da458fc |
1861 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 |
1862 | the infinite loop>. The rules for this are different for lower-level |
1863 | loops given by the greedy modifiers C<*+{}>, and for higher-level |
1864 | ones like the C</g> modifier or split() operator. |
1865 | |
19799a22 |
1866 | The lower-level loops are I<interrupted> (that is, the loop is |
1867 | broken) when Perl detects that a repeated expression matched a |
1868 | zero-length substring. Thus |
c84d73f1 |
1869 | |
1870 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; |
1871 | |
5d458dd8 |
1872 | is made equivalent to |
c84d73f1 |
1873 | |
5d458dd8 |
1874 | m{ (?: NON_ZERO_LENGTH )* |
1875 | | |
1876 | (?: ZERO_LENGTH )? |
c84d73f1 |
1877 | }x; |
1878 | |
1879 | The higher level-loops preserve an additional state between iterations: |
5d458dd8 |
1880 | whether the last match was zero-length. To break the loop, the following |
c84d73f1 |
1881 | match after a zero-length match is prohibited to have a length of zero. |
5d458dd8 |
1882 | This prohibition interacts with backtracking (see L<"Backtracking">), |
c84d73f1 |
1883 | and so the I<second best> match is chosen if the I<best> match is of |
1884 | zero length. |
1885 | |
19799a22 |
1886 | For example: |
c84d73f1 |
1887 | |
1888 | $_ = 'bar'; |
1889 | s/\w??/<$&>/g; |
1890 | |
20fb949f |
1891 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
5d458dd8 |
1892 | match given by non-greedy C<??> is the zero-length match, and the I<second |
c84d73f1 |
1893 | best> match is what is matched by C<\w>. Thus zero-length matches |
1894 | alternate with one-character-long matches. |
1895 | |
5d458dd8 |
1896 | Similarly, for repeated C<m/()/g> the second-best match is the match at the |
c84d73f1 |
1897 | position one notch further in the string. |
1898 | |
19799a22 |
1899 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 |
1900 | the matched string, and is reset by each assignment to pos(). |
9da458fc |
1901 | Zero-length matches at the end of the previous match are ignored |
1902 | during C<split>. |
c84d73f1 |
1903 | |
35a734be |
1904 | =head2 Combining pieces together |
1905 | |
1906 | Each of the elementary pieces of regular expressions which were described |
1907 | before (such as C<ab> or C<\Z>) could match at most one substring |
1908 | at the given position of the input string. However, in a typical regular |
1909 | expression these elementary pieces are combined into more complicated |
1910 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc |
1911 | (in these examples C<S> and C<T> are regular subexpressions). |
1912 | |
1913 | Such combinations can include alternatives, leading to a problem of choice: |
1914 | if we match a regular expression C<a|ab> against C<"abc">, will it match |
1915 | substring C<"a"> or C<"ab">? One way to describe which substring is |
1916 | actually matched is the concept of backtracking (see L<"Backtracking">). |
1917 | However, this description is too low-level and makes you think |
1918 | in terms of a particular implementation. |
1919 | |
1920 | Another description starts with notions of "better"/"worse". All the |
1921 | substrings which may be matched by the given regular expression can be |
1922 | sorted from the "best" match to the "worst" match, and it is the "best" |
1923 | match which is chosen. This substitutes the question of "what is chosen?" |
1924 | by the question of "which matches are better, and which are worse?". |
1925 | |
1926 | Again, for elementary pieces there is no such question, since at most |
1927 | one match at a given position is possible. This section describes the |
1928 | notion of better/worse for combining operators. In the description |
1929 | below C<S> and C<T> are regular subexpressions. |
1930 | |
13a2d996 |
1931 | =over 4 |
35a734be |
1932 | |
1933 | =item C<ST> |
1934 | |
1935 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are |
1936 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings |
5d458dd8 |
1937 | which can be matched by C<T>. |
35a734be |
1938 | |
1939 | If C<A> is better match for C<S> than C<A'>, C<AB> is a better |
1940 | match than C<A'B'>. |
1941 | |
1942 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if |
1943 | C<B> is better match for C<T> than C<B'>. |
1944 | |
1945 | =item C<S|T> |
1946 | |
1947 | When C<S> can match, it is a better match than when only C<T> can match. |
1948 | |
1949 | Ordering of two matches for C<S> is the same as for C<S>. Similar for |
1950 | two matches for C<T>. |
1951 | |
1952 | =item C<S{REPEAT_COUNT}> |
1953 | |
1954 | Matches as C<SSS...S> (repeated as many times as necessary). |
1955 | |
1956 | =item C<S{min,max}> |
1957 | |
1958 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. |
1959 | |
1960 | =item C<S{min,max}?> |
1961 | |
1962 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. |
1963 | |
1964 | =item C<S?>, C<S*>, C<S+> |
1965 | |
1966 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. |
1967 | |
1968 | =item C<S??>, C<S*?>, C<S+?> |
1969 | |
1970 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. |
1971 | |
c47ff5f1 |
1972 | =item C<< (?>S) >> |
35a734be |
1973 | |
1974 | Matches the best match for C<S> and only that. |
1975 | |
1976 | =item C<(?=S)>, C<(?<=S)> |
1977 | |
1978 | Only the best match for C<S> is considered. (This is important only if |
1979 | C<S> has capturing parentheses, and backreferences are used somewhere |
1980 | else in the whole regular expression.) |
1981 | |
1982 | =item C<(?!S)>, C<(?<!S)> |
1983 | |
1984 | For this grouping operator there is no need to describe the ordering, since |
1985 | only whether or not C<S> can match is important. |
1986 | |
6bda09f9 |
1987 | =item C<(??{ EXPR })>, C<(?PARNO)> |
35a734be |
1988 | |
1989 | The ordering is the same as for the regular expression which is |
6bda09f9 |
1990 | the result of EXPR, or the pattern contained by capture buffer PARNO. |
35a734be |
1991 | |
1992 | =item C<(?(condition)yes-pattern|no-pattern)> |
1993 | |
1994 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is |
1995 | already determined. The ordering of the matches is the same as for the |
1996 | chosen subexpression. |
1997 | |
1998 | =back |
1999 | |
2000 | The above recipes describe the ordering of matches I<at a given position>. |
2001 | One more rule is needed to understand how a match is determined for the |
2002 | whole regular expression: a match at an earlier position is always better |
2003 | than a match at a later position. |
2004 | |
c84d73f1 |
2005 | =head2 Creating custom RE engines |
2006 | |
2007 | Overloaded constants (see L<overload>) provide a simple way to extend |
2008 | the functionality of the RE engine. |
2009 | |
2010 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which |
6b0ac556 |
2011 | matches at boundary between whitespace characters and non-whitespace |
c84d73f1 |
2012 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
2013 | at these positions, so we want to have each C<\Y|> in the place of the |
2014 | more complicated version. We can create a module C<customre> to do |
2015 | this: |
2016 | |
2017 | package customre; |
2018 | use overload; |
2019 | |
2020 | sub import { |
2021 | shift; |
2022 | die "No argument to customre::import allowed" if @_; |
2023 | overload::constant 'qr' => \&convert; |
2024 | } |
2025 | |
2026 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} |
2027 | |
580a9fe1 |
2028 | # We must also take care of not escaping the legitimate \\Y| |
2029 | # sequence, hence the presence of '\\' in the conversion rules. |
5d458dd8 |
2030 | my %rules = ( '\\' => '\\\\', |
c84d73f1 |
2031 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
2032 | sub convert { |
2033 | my $re = shift; |
5d458dd8 |
2034 | $re =~ s{ |
c84d73f1 |
2035 | \\ ( \\ | Y . ) |
2036 | } |
5d458dd8 |
2037 | { $rules{$1} or invalid($re,$1) }sgex; |
c84d73f1 |
2038 | return $re; |
2039 | } |
2040 | |
2041 | Now C<use customre> enables the new escape in constant regular |
2042 | expressions, i.e., those without any runtime variable interpolations. |
2043 | As documented in L<overload>, this conversion will work only over |
2044 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable |
2045 | part of this regular expression needs to be converted explicitly |
2046 | (but only if the special meaning of C<\Y|> should be enabled inside $re): |
2047 | |
2048 | use customre; |
2049 | $re = <>; |
2050 | chomp $re; |
2051 | $re = customre::convert $re; |
2052 | /\Y|$re\Y|/; |
2053 | |
1f1031fe |
2054 | =head1 PCRE/Python Support |
2055 | |
2056 | As of Perl 5.10 Perl supports several Python/PCRE specific extensions |
2057 | to the regex syntax. While Perl programmers are encouraged to use the |
2058 | Perl specific syntax, the following are legal in Perl 5.10: |
2059 | |
2060 | =over 4 |
2061 | |
ee9b8eae |
2062 | =item C<< (?PE<lt>NAMEE<gt>pattern) >> |
1f1031fe |
2063 | |
2064 | Define a named capture buffer. Equivalent to C<< (?<NAME>pattern) >>. |
2065 | |
2066 | =item C<< (?P=NAME) >> |
2067 | |
2068 | Backreference to a named capture buffer. Equivalent to C<< \g{NAME} >>. |
2069 | |
2070 | =item C<< (?P>NAME) >> |
2071 | |
2072 | Subroutine call to a named capture buffer. Equivalent to C<< (?&NAME) >>. |
2073 | |
ee9b8eae |
2074 | =back |
1f1031fe |
2075 | |
19799a22 |
2076 | =head1 BUGS |
2077 | |
9da458fc |
2078 | This document varies from difficult to understand to completely |
2079 | and utterly opaque. The wandering prose riddled with jargon is |
2080 | hard to fathom in several places. |
2081 | |
2082 | This document needs a rewrite that separates the tutorial content |
2083 | from the reference content. |
19799a22 |
2084 | |
2085 | =head1 SEE ALSO |
9fa51da4 |
2086 | |
91e0c79e |
2087 | L<perlrequick>. |
2088 | |
2089 | L<perlretut>. |
2090 | |
9b599b2a |
2091 | L<perlop/"Regexp Quote-Like Operators">. |
2092 | |
1e66bd83 |
2093 | L<perlop/"Gory details of parsing quoted constructs">. |
2094 | |
14218588 |
2095 | L<perlfaq6>. |
2096 | |
9b599b2a |
2097 | L<perlfunc/pos>. |
2098 | |
2099 | L<perllocale>. |
2100 | |
fb55449c |
2101 | L<perlebcdic>. |
2102 | |
14218588 |
2103 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
2104 | by O'Reilly and Associates. |