Commit | Line | Data |
a0d0e21e |
1 | =head1 NAME |
d74e8afc |
2 | X<regular expression> X<regex> X<regexp> |
a0d0e21e |
3 | |
4 | perlre - Perl regular expressions |
5 | |
6 | =head1 DESCRIPTION |
7 | |
91e0c79e |
8 | This page describes the syntax of regular expressions in Perl. |
9 | |
cc46d5f2 |
10 | If you haven't used regular expressions before, a quick-start |
91e0c79e |
11 | introduction is available in L<perlrequick>, and a longer tutorial |
12 | introduction is available in L<perlretut>. |
13 | |
14 | For reference on how regular expressions are used in matching |
15 | operations, plus various examples of the same, see discussions of |
16 | C<m//>, C<s///>, C<qr//> and C<??> in L<perlop/"Regexp Quote-Like |
17 | Operators">. |
cb1a09d0 |
18 | |
19799a22 |
19 | Matching operations can have various modifiers. Modifiers |
5a964f20 |
20 | that relate to the interpretation of the regular expression inside |
19799a22 |
21 | are listed below. Modifiers that alter the way a regular expression |
22 | is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and |
1e66bd83 |
23 | L<perlop/"Gory details of parsing quoted constructs">. |
a0d0e21e |
24 | |
55497cff |
25 | =over 4 |
26 | |
27 | =item i |
d74e8afc |
28 | X</i> X<regex, case-insensitive> X<regexp, case-insensitive> |
29 | X<regular expression, case-insensitive> |
55497cff |
30 | |
31 | Do case-insensitive pattern matching. |
32 | |
a034a98d |
33 | If C<use locale> is in effect, the case map is taken from the current |
34 | locale. See L<perllocale>. |
35 | |
54310121 |
36 | =item m |
d74e8afc |
37 | X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline> |
55497cff |
38 | |
39 | Treat string as multiple lines. That is, change "^" and "$" from matching |
14218588 |
40 | the start or end of the string to matching the start or end of any |
7f761169 |
41 | line anywhere within the string. |
55497cff |
42 | |
54310121 |
43 | =item s |
d74e8afc |
44 | X</s> X<regex, single-line> X<regexp, single-line> |
45 | X<regular expression, single-line> |
55497cff |
46 | |
47 | Treat string as single line. That is, change "." to match any character |
19799a22 |
48 | whatsoever, even a newline, which normally it would not match. |
55497cff |
49 | |
f02c194e |
50 | Used together, as /ms, they let the "." match any character whatsoever, |
fb55449c |
51 | while still allowing "^" and "$" to match, respectively, just after |
19799a22 |
52 | and just before newlines within the string. |
7b8d334a |
53 | |
54310121 |
54 | =item x |
d74e8afc |
55 | X</x> |
55497cff |
56 | |
57 | Extend your pattern's legibility by permitting whitespace and comments. |
58 | |
59 | =back |
a0d0e21e |
60 | |
61 | These are usually written as "the C</x> modifier", even though the delimiter |
14218588 |
62 | in question might not really be a slash. Any of these |
a0d0e21e |
63 | modifiers may also be embedded within the regular expression itself using |
14218588 |
64 | the C<(?...)> construct. See below. |
a0d0e21e |
65 | |
4633a7c4 |
66 | The C</x> modifier itself needs a little more explanation. It tells |
55497cff |
67 | the regular expression parser to ignore whitespace that is neither |
68 | backslashed nor within a character class. You can use this to break up |
4633a7c4 |
69 | your regular expression into (slightly) more readable parts. The C<#> |
54310121 |
70 | character is also treated as a metacharacter introducing a comment, |
55497cff |
71 | just as in ordinary Perl code. This also means that if you want real |
14218588 |
72 | whitespace or C<#> characters in the pattern (outside a character |
f9a3ff1a |
73 | class, where they are unaffected by C</x>), then you'll either have to |
74 | escape them (using backslashes or C<\Q...\E>) or encode them using octal |
8933a740 |
75 | or hex escapes. Taken together, these features go a long way towards |
76 | making Perl's regular expressions more readable. Note that you have to |
77 | be careful not to include the pattern delimiter in the comment--perl has |
78 | no way of knowing you did not intend to close the pattern early. See |
79 | the C-comment deletion code in L<perlop>. Also note that anything inside |
1031e5db |
80 | a C<\Q...\E> stays unaffected by C</x>. |
d74e8afc |
81 | X</x> |
a0d0e21e |
82 | |
83 | =head2 Regular Expressions |
84 | |
19799a22 |
85 | The patterns used in Perl pattern matching derive from supplied in |
14218588 |
86 | the Version 8 regex routines. (The routines are derived |
19799a22 |
87 | (distantly) from Henry Spencer's freely redistributable reimplementation |
88 | of the V8 routines.) See L<Version 8 Regular Expressions> for |
89 | details. |
a0d0e21e |
90 | |
91 | In particular the following metacharacters have their standard I<egrep>-ish |
92 | meanings: |
d74e8afc |
93 | X<metacharacter> |
94 | X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]> |
95 | |
a0d0e21e |
96 | |
54310121 |
97 | \ Quote the next metacharacter |
a0d0e21e |
98 | ^ Match the beginning of the line |
99 | . Match any character (except newline) |
c07a80fd |
100 | $ Match the end of the line (or before newline at the end) |
a0d0e21e |
101 | | Alternation |
102 | () Grouping |
103 | [] Character class |
104 | |
14218588 |
105 | By default, the "^" character is guaranteed to match only the |
106 | beginning of the string, the "$" character only the end (or before the |
107 | newline at the end), and Perl does certain optimizations with the |
a0d0e21e |
108 | assumption that the string contains only one line. Embedded newlines |
109 | will not be matched by "^" or "$". You may, however, wish to treat a |
4a6725af |
110 | string as a multi-line buffer, such that the "^" will match after any |
a0d0e21e |
111 | newline within the string, and "$" will match before any newline. At the |
112 | cost of a little more overhead, you can do this by using the /m modifier |
113 | on the pattern match operator. (Older programs did this by setting C<$*>, |
f02c194e |
114 | but this practice has been removed in perl 5.9.) |
d74e8afc |
115 | X<^> X<$> X</m> |
a0d0e21e |
116 | |
14218588 |
117 | To simplify multi-line substitutions, the "." character never matches a |
55497cff |
118 | newline unless you use the C</s> modifier, which in effect tells Perl to pretend |
f02c194e |
119 | the string is a single line--even if it isn't. |
d74e8afc |
120 | X<.> X</s> |
a0d0e21e |
121 | |
122 | The following standard quantifiers are recognized: |
d74e8afc |
123 | X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}> |
a0d0e21e |
124 | |
125 | * Match 0 or more times |
126 | + Match 1 or more times |
127 | ? Match 1 or 0 times |
128 | {n} Match exactly n times |
129 | {n,} Match at least n times |
130 | {n,m} Match at least n but not more than m times |
131 | |
132 | (If a curly bracket occurs in any other context, it is treated |
b975c076 |
133 | as a regular character. In particular, the lower bound |
134 | is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+" |
25f94b33 |
135 | modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited |
9c79236d |
136 | to integral values less than a preset limit defined when perl is built. |
137 | This is usually 32766 on the most common platforms. The actual limit can |
138 | be seen in the error message generated by code such as this: |
139 | |
820475bd |
140 | $_ **= $_ , / {$_} / for 2 .. 42; |
a0d0e21e |
141 | |
54310121 |
142 | By default, a quantified subpattern is "greedy", that is, it will match as |
143 | many times as possible (given a particular starting location) while still |
144 | allowing the rest of the pattern to match. If you want it to match the |
145 | minimum number of times possible, follow the quantifier with a "?". Note |
146 | that the meanings don't change, just the "greediness": |
d74e8afc |
147 | X<metacharacter> X<greedy> X<greedyness> |
148 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?> |
a0d0e21e |
149 | |
150 | *? Match 0 or more times |
151 | +? Match 1 or more times |
152 | ?? Match 0 or 1 time |
153 | {n}? Match exactly n times |
154 | {n,}? Match at least n times |
155 | {n,m}? Match at least n but not more than m times |
156 | |
5f05dabc |
157 | Because patterns are processed as double quoted strings, the following |
a0d0e21e |
158 | also work: |
d74e8afc |
159 | X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q> |
160 | X<\0> X<\c> X<\N> X<\x> |
a0d0e21e |
161 | |
0f36ee90 |
162 | \t tab (HT, TAB) |
163 | \n newline (LF, NL) |
164 | \r return (CR) |
165 | \f form feed (FF) |
166 | \a alarm (bell) (BEL) |
167 | \e escape (think troff) (ESC) |
cb1a09d0 |
168 | \033 octal char (think of a PDP-11) |
169 | \x1B hex char |
a0ed51b3 |
170 | \x{263a} wide hex char (Unicode SMILEY) |
a0d0e21e |
171 | \c[ control char |
4a2d328f |
172 | \N{name} named char |
cb1a09d0 |
173 | \l lowercase next char (think vi) |
174 | \u uppercase next char (think vi) |
175 | \L lowercase till \E (think vi) |
176 | \U uppercase till \E (think vi) |
177 | \E end case modification (think vi) |
5a964f20 |
178 | \Q quote (disable) pattern metacharacters till \E |
a0d0e21e |
179 | |
a034a98d |
180 | If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u> |
423cee85 |
181 | and C<\U> is taken from the current locale. See L<perllocale>. For |
4a2d328f |
182 | documentation of C<\N{name}>, see L<charnames>. |
a034a98d |
183 | |
1d2dff63 |
184 | You cannot include a literal C<$> or C<@> within a C<\Q> sequence. |
185 | An unescaped C<$> or C<@> interpolates the corresponding variable, |
186 | while escaping will cause the literal string C<\$> to be matched. |
187 | You'll need to write something like C<m/\Quser\E\@\Qhost/>. |
188 | |
a0d0e21e |
189 | In addition, Perl defines the following: |
d74e8afc |
190 | X<metacharacter> |
191 | X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C> |
192 | X<word> X<whitespace> |
a0d0e21e |
193 | |
81714fb9 |
194 | \w Match a "word" character (alphanumeric plus "_") |
195 | \W Match a non-"word" character |
196 | \s Match a whitespace character |
197 | \S Match a non-whitespace character |
198 | \d Match a digit character |
199 | \D Match a non-digit character |
200 | \pP Match P, named property. Use \p{Prop} for longer names. |
201 | \PP Match non-P |
202 | \X Match eXtended Unicode "combining character sequence", |
203 | equivalent to (?:\PM\pM*) |
204 | \C Match a single C char (octet) even under Unicode. |
205 | NOTE: breaks up characters into their UTF-8 bytes, |
206 | so you may end up with malformed pieces of UTF-8. |
207 | Unsupported in lookbehind. |
208 | \1 Backreference to a a specific group. |
209 | '1' may actually be any positive integer |
210 | \k<name> Named backreference |
211 | \N{name} Named unicode character, or unicode escape. |
212 | \x12 Hexadecimal escape sequence |
213 | \x{1234} Long hexadecimal escape sequence |
a0d0e21e |
214 | |
08ce8fc6 |
215 | A C<\w> matches a single alphanumeric character (an alphabetic |
216 | character, or a decimal digit) or C<_>, not a whole word. Use C<\w+> |
217 | to match a string of Perl-identifier characters (which isn't the same |
218 | as matching an English word). If C<use locale> is in effect, the list |
219 | of alphabetic characters generated by C<\w> is taken from the current |
220 | locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>, |
1209ba90 |
221 | C<\d>, and C<\D> within character classes, but if you try to use them |
08ce8fc6 |
222 | as endpoints of a range, that's not a range, the "-" is understood |
223 | literally. If Unicode is in effect, C<\s> matches also "\x{85}", |
224 | "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about |
491fd90a |
225 | C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general. |
fa11829f |
226 | You can define your own C<\p> and C<\P> properties, see L<perlunicode>. |
d74e8afc |
227 | X<\w> X<\W> X<word> |
a0d0e21e |
228 | |
b8c5462f |
229 | The POSIX character class syntax |
d74e8afc |
230 | X<character class> |
b8c5462f |
231 | |
820475bd |
232 | [:class:] |
b8c5462f |
233 | |
5496314a |
234 | is also available. Note that the C<[> and C<]> braces are I<literal>; |
235 | they must always be used within a character class expression. |
236 | |
237 | # this is correct: |
238 | $string =~ /[[:alpha:]]/; |
239 | |
240 | # this is not, and will generate a warning: |
241 | $string =~ /[:alpha:]/; |
242 | |
243 | The available classes and their backslash equivalents (if available) are |
244 | as follows: |
d74e8afc |
245 | X<character class> |
246 | X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph> |
247 | X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit> |
b8c5462f |
248 | |
249 | alpha |
250 | alnum |
251 | ascii |
aaa51d5e |
252 | blank [1] |
b8c5462f |
253 | cntrl |
254 | digit \d |
255 | graph |
256 | lower |
257 | print |
258 | punct |
aaa51d5e |
259 | space \s [2] |
b8c5462f |
260 | upper |
aaa51d5e |
261 | word \w [3] |
b8c5462f |
262 | xdigit |
263 | |
07698885 |
264 | =over |
265 | |
266 | =item [1] |
267 | |
b432a672 |
268 | A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace". |
07698885 |
269 | |
270 | =item [2] |
271 | |
272 | Not exactly equivalent to C<\s> since the C<[[:space:]]> includes |
b432a672 |
273 | also the (very rare) "vertical tabulator", "\ck", chr(11). |
07698885 |
274 | |
275 | =item [3] |
276 | |
08ce8fc6 |
277 | A Perl extension, see above. |
07698885 |
278 | |
279 | =back |
aaa51d5e |
280 | |
26b44a0a |
281 | For example use C<[:upper:]> to match all the uppercase characters. |
aaa51d5e |
282 | Note that the C<[]> are part of the C<[::]> construct, not part of the |
283 | whole character class. For example: |
b8c5462f |
284 | |
820475bd |
285 | [01[:alpha:]%] |
b8c5462f |
286 | |
593df60c |
287 | matches zero, one, any alphabetic character, and the percentage sign. |
b8c5462f |
288 | |
72ff2908 |
289 | The following equivalences to Unicode \p{} constructs and equivalent |
290 | backslash character classes (if available), will hold: |
d74e8afc |
291 | X<character class> X<\p> X<\p{}> |
72ff2908 |
292 | |
5496314a |
293 | [[:...:]] \p{...} backslash |
b8c5462f |
294 | |
295 | alpha IsAlpha |
296 | alnum IsAlnum |
297 | ascii IsASCII |
b432a672 |
298 | blank IsSpace |
b8c5462f |
299 | cntrl IsCntrl |
3bec3564 |
300 | digit IsDigit \d |
b8c5462f |
301 | graph IsGraph |
302 | lower IsLower |
303 | print IsPrint |
304 | punct IsPunct |
305 | space IsSpace |
3bec3564 |
306 | IsSpacePerl \s |
b8c5462f |
307 | upper IsUpper |
308 | word IsWord |
309 | xdigit IsXDigit |
310 | |
5496314a |
311 | For example C<[[:lower:]]> and C<\p{IsLower}> are equivalent. |
b8c5462f |
312 | |
313 | If the C<utf8> pragma is not used but the C<locale> pragma is, the |
aaa51d5e |
314 | classes correlate with the usual isalpha(3) interface (except for |
b432a672 |
315 | "word" and "blank"). |
b8c5462f |
316 | |
317 | The assumedly non-obviously named classes are: |
318 | |
319 | =over 4 |
320 | |
321 | =item cntrl |
d74e8afc |
322 | X<cntrl> |
b8c5462f |
323 | |
820475bd |
324 | Any control character. Usually characters that don't produce output as |
325 | such but instead control the terminal somehow: for example newline and |
326 | backspace are control characters. All characters with ord() less than |
593df60c |
327 | 32 are most often classified as control characters (assuming ASCII, |
7be5a6cf |
328 | the ISO Latin character sets, and Unicode), as is the character with |
329 | the ord() value of 127 (C<DEL>). |
b8c5462f |
330 | |
331 | =item graph |
d74e8afc |
332 | X<graph> |
b8c5462f |
333 | |
f1cbbd6e |
334 | Any alphanumeric or punctuation (special) character. |
b8c5462f |
335 | |
336 | =item print |
d74e8afc |
337 | X<print> |
b8c5462f |
338 | |
f79b3095 |
339 | Any alphanumeric or punctuation (special) character or the space character. |
b8c5462f |
340 | |
341 | =item punct |
d74e8afc |
342 | X<punct> |
b8c5462f |
343 | |
f1cbbd6e |
344 | Any punctuation (special) character. |
b8c5462f |
345 | |
346 | =item xdigit |
d74e8afc |
347 | X<xdigit> |
b8c5462f |
348 | |
593df60c |
349 | Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would |
820475bd |
350 | work just fine) it is included for completeness. |
b8c5462f |
351 | |
b8c5462f |
352 | =back |
353 | |
354 | You can negate the [::] character classes by prefixing the class name |
355 | with a '^'. This is a Perl extension. For example: |
d74e8afc |
356 | X<character class, negation> |
b8c5462f |
357 | |
5496314a |
358 | POSIX traditional Unicode |
93733859 |
359 | |
5496314a |
360 | [[:^digit:]] \D \P{IsDigit} |
361 | [[:^space:]] \S \P{IsSpace} |
362 | [[:^word:]] \W \P{IsWord} |
b8c5462f |
363 | |
54c18d04 |
364 | Perl respects the POSIX standard in that POSIX character classes are |
365 | only supported within a character class. The POSIX character classes |
366 | [.cc.] and [=cc=] are recognized but B<not> supported and trying to |
367 | use them will cause an error. |
b8c5462f |
368 | |
a0d0e21e |
369 | Perl defines the following zero-width assertions: |
d74e8afc |
370 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> |
371 | X<regexp, zero-width assertion> |
372 | X<regular expression, zero-width assertion> |
373 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> |
a0d0e21e |
374 | |
375 | \b Match a word boundary |
376 | \B Match a non-(word boundary) |
b85d18e9 |
377 | \A Match only at beginning of string |
378 | \Z Match only at end of string, or before newline at the end |
379 | \z Match only at end of string |
9da458fc |
380 | \G Match only at pos() (e.g. at the end-of-match position |
381 | of prior m//g) |
a0d0e21e |
382 | |
14218588 |
383 | A word boundary (C<\b>) is a spot between two characters |
19799a22 |
384 | that has a C<\w> on one side of it and a C<\W> on the other side |
385 | of it (in either order), counting the imaginary characters off the |
386 | beginning and end of the string as matching a C<\W>. (Within |
387 | character classes C<\b> represents backspace rather than a word |
388 | boundary, just as it normally does in any double-quoted string.) |
389 | The C<\A> and C<\Z> are just like "^" and "$", except that they |
390 | won't match multiple times when the C</m> modifier is used, while |
391 | "^" and "$" will match at every internal line boundary. To match |
392 | the actual end of the string and not ignore an optional trailing |
393 | newline, use C<\z>. |
d74e8afc |
394 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 |
395 | |
396 | The C<\G> assertion can be used to chain global matches (using |
397 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. |
398 | It is also useful when writing C<lex>-like scanners, when you have |
399 | several patterns that you want to match against consequent substrings |
400 | of your string, see the previous reference. The actual location |
401 | where C<\G> will match can also be influenced by using C<pos()> as |
25cf8c22 |
402 | an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully |
403 | supported when anchored to the start of the pattern; while it |
404 | is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some |
405 | such uses (C</.\G/g>, for example) currently cause problems, and |
406 | it is recommended that you avoid such usage for now. |
d74e8afc |
407 | X<\G> |
c47ff5f1 |
408 | |
14218588 |
409 | The bracketing construct C<( ... )> creates capture buffers. To |
c47ff5f1 |
410 | refer to the digit'th buffer use \<digit> within the |
14218588 |
411 | match. Outside the match use "$" instead of "\". (The |
81714fb9 |
412 | \<digit> notation works in certain circumstances outside |
14218588 |
413 | the match. See the warning below about \1 vs $1 for details.) |
414 | Referring back to another part of the match is called a |
415 | I<backreference>. |
d74e8afc |
416 | X<regex, capture buffer> X<regexp, capture buffer> |
417 | X<regular expression, capture buffer> X<backreference> |
14218588 |
418 | |
419 | There is no limit to the number of captured substrings that you may |
420 | use. However Perl also uses \10, \11, etc. as aliases for \010, |
fb55449c |
421 | \011, etc. (Recall that 0 means octal, so \011 is the character at |
422 | number 9 in your coded character set; which would be the 10th character, |
81714fb9 |
423 | a horizontal tab under ASCII.) Perl resolves this |
424 | ambiguity by interpreting \10 as a backreference only if at least 10 |
425 | left parentheses have opened before it. Likewise \11 is a |
426 | backreference only if at least 11 left parentheses have opened |
427 | before it. And so on. \1 through \9 are always interpreted as |
fb55449c |
428 | backreferences. |
14218588 |
429 | |
81714fb9 |
430 | Additionally, as of Perl 5.10 you may use named capture buffers and named |
431 | backreferences. The notation is C<< (?<name>...) >> and C<< \k<name> >> |
432 | (you may also use single quotes instead of angle brackets to quote the |
433 | name). The only difference with named capture buffers and unnamed ones is |
434 | that multiple buffers may have the same name and that the contents of |
435 | named capture buffers is available via the C<%+> hash. When multiple |
436 | groups share the same name C<$+{name}> and C<< \k<name> >> refer to the |
437 | leftmost defined group, thus it's possible to do things with named capture |
438 | buffers that would otherwise require C<(??{})> code to accomplish. Named |
439 | capture buffers are numbered just as normal capture buffers are and may be |
440 | referenced via the magic numeric variables or via numeric backreferences |
441 | as well as by name. |
442 | |
14218588 |
443 | Examples: |
a0d0e21e |
444 | |
445 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words |
446 | |
81714fb9 |
447 | /(.)\1/ # find first doubled char |
448 | and print "'$1' is the first doubled character\n"; |
449 | |
450 | /(?<char>.)\k<char>/ # ... a different way |
451 | and print "'$+{char}' is the first doubled character\n"; |
452 | |
453 | /(?<char>.)\1/ # ... mix and match |
454 | and print "'$1' is the first doubled character\n"; |
c47ff5f1 |
455 | |
14218588 |
456 | if (/Time: (..):(..):(..)/) { # parse out values |
a0d0e21e |
457 | $hours = $1; |
458 | $minutes = $2; |
459 | $seconds = $3; |
460 | } |
c47ff5f1 |
461 | |
14218588 |
462 | Several special variables also refer back to portions of the previous |
463 | match. C<$+> returns whatever the last bracket match matched. |
464 | C<$&> returns the entire matched string. (At one point C<$0> did |
465 | also, but now it returns the name of the program.) C<$`> returns |
77ea4f6d |
466 | everything before the matched string. C<$'> returns everything |
467 | after the matched string. And C<$^N> contains whatever was matched by |
468 | the most-recently closed group (submatch). C<$^N> can be used in |
469 | extended patterns (see below), for example to assign a submatch to a |
81714fb9 |
470 | variable. |
d74e8afc |
471 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 |
472 | |
665e98b9 |
473 | The numbered match variables ($1, $2, $3, etc.) and the related punctuation |
77ea4f6d |
474 | set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped |
14218588 |
475 | until the end of the enclosing block or until the next successful |
476 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) |
d74e8afc |
477 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
478 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> |
479 | |
14218588 |
480 | |
665e98b9 |
481 | B<NOTE>: failed matches in Perl do not reset the match variables, |
5146ce24 |
482 | which makes it easier to write code that tests for a series of more |
665e98b9 |
483 | specific cases and remembers the best match. |
484 | |
14218588 |
485 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or |
486 | C<$'> anywhere in the program, it has to provide them for every |
487 | pattern match. This may substantially slow your program. Perl |
488 | uses the same mechanism to produce $1, $2, etc, so you also pay a |
489 | price for each pattern that contains capturing parentheses. (To |
490 | avoid this cost while retaining the grouping behaviour, use the |
491 | extended regular expression C<(?: ... )> instead.) But if you never |
492 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing |
493 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> |
494 | if you can, but if you can't (and some algorithms really appreciate |
495 | them), once you've used them once, use them at will, because you've |
496 | already paid the price. As of 5.005, C<$&> is not so costly as the |
497 | other two. |
d74e8afc |
498 | X<$&> X<$`> X<$'> |
68dc0745 |
499 | |
19799a22 |
500 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
501 | C<\w>, C<\n>. Unlike some other regular expression languages, there |
502 | are no backslashed symbols that aren't alphanumeric. So anything |
c47ff5f1 |
503 | that looks like \\, \(, \), \<, \>, \{, or \} is always |
19799a22 |
504 | interpreted as a literal character, not a metacharacter. This was |
505 | once used in a common idiom to disable or quote the special meanings |
506 | of regular expression metacharacters in a string that you want to |
36bbe248 |
507 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e |
508 | |
509 | $pattern =~ s/(\W)/\\$1/g; |
510 | |
f1cbbd6e |
511 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 |
512 | Today it is more common to use the quotemeta() function or the C<\Q> |
513 | metaquoting escape sequence to disable all metacharacters' special |
514 | meanings like this: |
a0d0e21e |
515 | |
516 | /$unquoted\Q$quoted\E$unquoted/ |
517 | |
9da458fc |
518 | Beware that if you put literal backslashes (those not inside |
519 | interpolated variables) between C<\Q> and C<\E>, double-quotish |
520 | backslash interpolation may lead to confusing results. If you |
521 | I<need> to use literal backslashes within C<\Q...\E>, |
522 | consult L<perlop/"Gory details of parsing quoted constructs">. |
523 | |
19799a22 |
524 | =head2 Extended Patterns |
525 | |
14218588 |
526 | Perl also defines a consistent extension syntax for features not |
527 | found in standard tools like B<awk> and B<lex>. The syntax is a |
528 | pair of parentheses with a question mark as the first thing within |
529 | the parentheses. The character after the question mark indicates |
530 | the extension. |
19799a22 |
531 | |
14218588 |
532 | The stability of these extensions varies widely. Some have been |
533 | part of the core language for many years. Others are experimental |
534 | and may change without warning or be completely removed. Check |
535 | the documentation on an individual feature to verify its current |
536 | status. |
19799a22 |
537 | |
14218588 |
538 | A question mark was chosen for this and for the minimal-matching |
539 | construct because 1) question marks are rare in older regular |
540 | expressions, and 2) whenever you see one, you should stop and |
541 | "question" exactly what is going on. That's psychology... |
a0d0e21e |
542 | |
543 | =over 10 |
544 | |
cc6b7395 |
545 | =item C<(?#text)> |
d74e8afc |
546 | X<(?#)> |
a0d0e21e |
547 | |
14218588 |
548 | A comment. The text is ignored. If the C</x> modifier enables |
19799a22 |
549 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes |
259138e3 |
550 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
551 | C<)> in the comment. |
a0d0e21e |
552 | |
19799a22 |
553 | =item C<(?imsx-imsx)> |
d74e8afc |
554 | X<(?)> |
19799a22 |
555 | |
0b6d1084 |
556 | One or more embedded pattern-match modifiers, to be turned on (or |
557 | turned off, if preceded by C<->) for the remainder of the pattern or |
558 | the remainder of the enclosing pattern group (if any). This is |
559 | particularly useful for dynamic patterns, such as those read in from a |
560 | configuration file, read in as an argument, are specified in a table |
561 | somewhere, etc. Consider the case that some of which want to be case |
562 | sensitive and some do not. The case insensitive ones need to include |
563 | merely C<(?i)> at the front of the pattern. For example: |
19799a22 |
564 | |
565 | $pattern = "foobar"; |
566 | if ( /$pattern/i ) { } |
567 | |
568 | # more flexible: |
569 | |
570 | $pattern = "(?i)foobar"; |
571 | if ( /$pattern/ ) { } |
572 | |
0b6d1084 |
573 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 |
574 | |
575 | ( (?i) blah ) \s+ \1 |
576 | |
577 | will match a repeated (I<including the case>!) word C<blah> in any |
14218588 |
578 | case, assuming C<x> modifier, and no C<i> modifier outside this |
19799a22 |
579 | group. |
580 | |
5a964f20 |
581 | =item C<(?:pattern)> |
d74e8afc |
582 | X<(?:)> |
a0d0e21e |
583 | |
ca9dfc88 |
584 | =item C<(?imsx-imsx:pattern)> |
585 | |
5a964f20 |
586 | This is for clustering, not capturing; it groups subexpressions like |
587 | "()", but doesn't make backreferences as "()" does. So |
a0d0e21e |
588 | |
5a964f20 |
589 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e |
590 | |
591 | is like |
592 | |
5a964f20 |
593 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e |
594 | |
19799a22 |
595 | but doesn't spit out extra fields. It's also cheaper not to capture |
596 | characters if you don't need to. |
a0d0e21e |
597 | |
19799a22 |
598 | Any letters between C<?> and C<:> act as flags modifiers as with |
599 | C<(?imsx-imsx)>. For example, |
ca9dfc88 |
600 | |
601 | /(?s-i:more.*than).*million/i |
602 | |
14218588 |
603 | is equivalent to the more verbose |
ca9dfc88 |
604 | |
605 | /(?:(?s-i)more.*than).*million/i |
606 | |
5a964f20 |
607 | =item C<(?=pattern)> |
d74e8afc |
608 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> |
a0d0e21e |
609 | |
19799a22 |
610 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e |
611 | matches a word followed by a tab, without including the tab in C<$&>. |
612 | |
5a964f20 |
613 | =item C<(?!pattern)> |
d74e8afc |
614 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> |
a0d0e21e |
615 | |
19799a22 |
616 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e |
617 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 |
618 | however that look-ahead and look-behind are NOT the same thing. You cannot |
619 | use this for look-behind. |
7b8d334a |
620 | |
5a964f20 |
621 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a |
622 | will not do what you want. That's because the C<(?!foo)> is just saying that |
623 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will |
624 | match. You would have to do something like C</(?!foo)...bar/> for that. We |
625 | say "like" because there's the case of your "bar" not having three characters |
626 | before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>. |
627 | Sometimes it's still easier just to say: |
a0d0e21e |
628 | |
a3cb178b |
629 | if (/bar/ && $` !~ /foo$/) |
a0d0e21e |
630 | |
19799a22 |
631 | For look-behind see below. |
c277df42 |
632 | |
c47ff5f1 |
633 | =item C<(?<=pattern)> |
d74e8afc |
634 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> |
c277df42 |
635 | |
c47ff5f1 |
636 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 |
637 | matches a word that follows a tab, without including the tab in C<$&>. |
638 | Works only for fixed-width look-behind. |
c277df42 |
639 | |
5a964f20 |
640 | =item C<(?<!pattern)> |
d74e8afc |
641 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> |
c277df42 |
642 | |
19799a22 |
643 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
644 | matches any occurrence of "foo" that does not follow "bar". Works |
645 | only for fixed-width look-behind. |
c277df42 |
646 | |
81714fb9 |
647 | =item C<(?'NAME'pattern)> |
648 | |
649 | =item C<< (?<NAME>pattern) >> |
650 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> |
651 | |
652 | A named capture buffer. Identical in every respect to normal capturing |
653 | parens C<()> but for the additional fact that C<%+> may be used after |
654 | a succesful match to refer to a named buffer. See C<perlvar> for more |
655 | details on the C<%+> hash. |
656 | |
657 | If multiple distinct capture buffers have the same name then the |
658 | $+{NAME} will refer to the leftmost defined buffer in the match. |
659 | |
660 | The forms C<(?'NAME'pattern)> and C<(?<NAME>pattern)> are equivalent. |
661 | |
662 | B<NOTE:> While the notation of this construct is the same as the similar |
663 | function in .NET regexes, the behavior is not, in Perl the buffers are |
664 | numbered sequentially regardless of being named or not. Thus in the |
665 | pattern |
666 | |
667 | /(x)(?<foo>y)(z)/ |
668 | |
669 | $+{foo} will be the same as $2, and $3 will contain 'z' instead of |
670 | the opposite which is what a .NET regex hacker might expect. |
671 | |
672 | Currently NAME is restricted to word chars only. In other words, it |
673 | must match C</^\w+$/>. |
674 | |
675 | =item C<< \k<name> >> |
676 | |
677 | =item C<< \k'name' >> |
678 | |
679 | Named backreference. Similar to numeric backreferences, except that |
680 | the group is designated by name and not number. If multiple groups |
681 | have the same name then it refers to the leftmost defined group in |
682 | the current match. |
683 | |
684 | It is an error to refer to a name not defined by a C<(?<NAME>)> |
685 | earlier in the pattern. |
686 | |
687 | Both forms are equivalent. |
688 | |
cc6b7395 |
689 | =item C<(?{ code })> |
d74e8afc |
690 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 |
691 | |
19799a22 |
692 | B<WARNING>: This extended regular expression feature is considered |
693 | highly experimental, and may be changed or deleted without notice. |
c277df42 |
694 | |
cc46d5f2 |
695 | This zero-width assertion evaluates any embedded Perl code. It |
19799a22 |
696 | always succeeds, and its C<code> is not interpolated. Currently, |
697 | the rules to determine where the C<code> ends are somewhat convoluted. |
698 | |
77ea4f6d |
699 | This feature can be used together with the special variable C<$^N> to |
700 | capture the results of submatches in variables without having to keep |
701 | track of the number of nested parentheses. For example: |
702 | |
703 | $_ = "The brown fox jumps over the lazy dog"; |
704 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; |
705 | print "color = $color, animal = $animal\n"; |
706 | |
754091cb |
707 | Inside the C<(?{...})> block, C<$_> refers to the string the regular |
708 | expression is matching against. You can also use C<pos()> to know what is |
fa11829f |
709 | the current position of matching within this string. |
754091cb |
710 | |
19799a22 |
711 | The C<code> is properly scoped in the following sense: If the assertion |
712 | is backtracked (compare L<"Backtracking">), all changes introduced after |
713 | C<local>ization are undone, so that |
b9ac3b5b |
714 | |
715 | $_ = 'a' x 8; |
716 | m< |
717 | (?{ $cnt = 0 }) # Initialize $cnt. |
718 | ( |
719 | a |
720 | (?{ |
721 | local $cnt = $cnt + 1; # Update $cnt, backtracking-safe. |
722 | }) |
723 | )* |
724 | aaaa |
725 | (?{ $res = $cnt }) # On success copy to non-localized |
726 | # location. |
727 | >x; |
728 | |
19799a22 |
729 | will set C<$res = 4>. Note that after the match, $cnt returns to the globally |
14218588 |
730 | introduced value, because the scopes that restrict C<local> operators |
b9ac3b5b |
731 | are unwound. |
732 | |
19799a22 |
733 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> |
734 | switch. If I<not> used in this way, the result of evaluation of |
735 | C<code> is put into the special variable C<$^R>. This happens |
736 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions |
737 | inside the same regular expression. |
b9ac3b5b |
738 | |
19799a22 |
739 | The assignment to C<$^R> above is properly localized, so the old |
740 | value of C<$^R> is restored if the assertion is backtracked; compare |
741 | L<"Backtracking">. |
b9ac3b5b |
742 | |
61528107 |
743 | Due to an unfortunate implementation issue, the Perl code contained in these |
744 | blocks is treated as a compile time closure that can have seemingly bizarre |
6bda09f9 |
745 | consequences when used with lexically scoped variables inside of subroutines |
61528107 |
746 | or loops. There are various workarounds for this, including simply using |
747 | global variables instead. If you are using this construct and strange results |
6bda09f9 |
748 | occur then check for the use of lexically scoped variables. |
749 | |
19799a22 |
750 | For reasons of security, this construct is forbidden if the regular |
751 | expression involves run-time interpolation of variables, unless the |
752 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the |
753 | variables contain results of C<qr//> operator (see |
754 | L<perlop/"qr/STRING/imosx">). |
871b0233 |
755 | |
14218588 |
756 | This restriction is because of the wide-spread and remarkably convenient |
19799a22 |
757 | custom of using run-time determined strings as patterns. For example: |
871b0233 |
758 | |
759 | $re = <>; |
760 | chomp $re; |
761 | $string =~ /$re/; |
762 | |
14218588 |
763 | Before Perl knew how to execute interpolated code within a pattern, |
764 | this operation was completely safe from a security point of view, |
765 | although it could raise an exception from an illegal pattern. If |
766 | you turn on the C<use re 'eval'>, though, it is no longer secure, |
767 | so you should only do so if you are also using taint checking. |
768 | Better yet, use the carefully constrained evaluation within a Safe |
cc46d5f2 |
769 | compartment. See L<perlsec> for details about both these mechanisms. |
871b0233 |
770 | |
8988a1bb |
771 | Because perl's regex engine is not currently re-entrant, interpolated |
772 | code may not invoke the regex engine either directly with C<m//> or C<s///>), |
773 | or indirectly with functions such as C<split>. |
774 | |
14455d6c |
775 | =item C<(??{ code })> |
d74e8afc |
776 | X<(??{})> |
777 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> |
0f5d15d6 |
778 | |
19799a22 |
779 | B<WARNING>: This extended regular expression feature is considered |
780 | highly experimental, and may be changed or deleted without notice. |
9da458fc |
781 | A simplified version of the syntax may be introduced for commonly |
782 | used idioms. |
0f5d15d6 |
783 | |
19799a22 |
784 | This is a "postponed" regular subexpression. The C<code> is evaluated |
785 | at run time, at the moment this subexpression may match. The result |
786 | of evaluation is considered as a regular expression and matched as |
61528107 |
787 | if it were inserted instead of this construct. Note that this means |
6bda09f9 |
788 | that the contents of capture buffers defined inside an eval'ed pattern |
789 | are not available outside of the pattern, and vice versa, there is no |
790 | way for the inner pattern to refer to a capture buffer defined outside. |
791 | Thus, |
792 | |
793 | ('a' x 100)=~/(??{'(.)' x 100})/ |
794 | |
81714fb9 |
795 | B<will> match, it will B<not> set $1. |
0f5d15d6 |
796 | |
428594d9 |
797 | The C<code> is not interpolated. As before, the rules to determine |
19799a22 |
798 | where the C<code> ends are currently somewhat convoluted. |
799 | |
800 | The following pattern matches a parenthesized group: |
0f5d15d6 |
801 | |
802 | $re = qr{ |
803 | \( |
804 | (?: |
805 | (?> [^()]+ ) # Non-parens without backtracking |
806 | | |
14455d6c |
807 | (??{ $re }) # Group with matching parens |
0f5d15d6 |
808 | )* |
809 | \) |
810 | }x; |
811 | |
6bda09f9 |
812 | See also C<(?PARNO)> for a different, more efficient way to accomplish |
813 | the same task. |
814 | |
8988a1bb |
815 | Because perl's regex engine is not currently re-entrant, delayed |
816 | code may not invoke the regex engine either directly with C<m//> or C<s///>), |
817 | or indirectly with functions such as C<split>. |
818 | |
6bda09f9 |
819 | Recursing deeper than 50 times without consuming any input string will |
61528107 |
820 | result in a fatal error. The maximum depth is compiled into perl, so |
6bda09f9 |
821 | changing it requires a custom build. |
822 | |
823 | =item C<(?PARNO)> C<(?R)> |
824 | |
825 | X<(?PARNO)> X<(?1)> |
826 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
827 | |
61528107 |
828 | B<WARNING>: This extended regular expression feature is considered |
6bda09f9 |
829 | highly experimental, and may be changed or deleted without notice. |
830 | |
81714fb9 |
831 | Similar to C<(??{ code })> except it does not involve compiling any code, |
832 | instead it treats the contents of a capture buffer as an independent |
61528107 |
833 | pattern that must match at the current position. Capture buffers |
81714fb9 |
834 | contained by the pattern will have the value as determined by the |
6bda09f9 |
835 | outermost recursion. |
836 | |
837 | PARNO is a sequence of digits not starting with 0 whose value |
81714fb9 |
838 | reflects the paren-number of the capture buffer to recurse to. |
6bda09f9 |
839 | C<(?R)> curses to the beginning of the pattern. |
840 | |
81714fb9 |
841 | The following pattern matches a function foo() which may contain |
842 | balanced parenthesis as the argument. |
6bda09f9 |
843 | |
844 | $re = qr{ ( # paren group 1 (full function) |
81714fb9 |
845 | foo |
6bda09f9 |
846 | ( # paren group 2 (parens) |
847 | \( |
848 | ( # paren group 3 (contents of parens) |
849 | (?: |
850 | (?> [^()]+ ) # Non-parens without backtracking |
851 | | |
852 | (?2) # Recurse to start of paren group 2 |
853 | )* |
854 | ) |
855 | \) |
856 | ) |
857 | ) |
858 | }x; |
859 | |
860 | If the pattern was used as follows |
861 | |
862 | 'foo(bar(baz)+baz(bop))'=~/$re/ |
863 | and print "\$1 = $1\n", |
864 | "\$2 = $2\n", |
865 | "\$3 = $3\n"; |
866 | |
867 | the output produced should be the following: |
868 | |
869 | $1 = foo(bar(baz)+baz(bop)) |
870 | $2 = (bar(baz)+baz(bop)) |
81714fb9 |
871 | $3 = bar(baz)+baz(bop) |
6bda09f9 |
872 | |
81714fb9 |
873 | If there is no corresponding capture buffer defined, then it is a |
61528107 |
874 | fatal error. Recursing deeper than 50 times without consuming any input |
81714fb9 |
875 | string will also result in a fatal error. The maximum depth is compiled |
6bda09f9 |
876 | into perl, so changing it requires a custom build. |
877 | |
81714fb9 |
878 | B<Note> that this pattern does not behave the same way as the equivalent |
6bda09f9 |
879 | PCRE or Python construct of the same form. In perl you can backtrack into |
880 | a recursed group, in PCRE and Python the recursed into group is treated |
81714fb9 |
881 | as atomic. Also, constructs like (?i:(?1)) or (?:(?i)(?1)) do not affect |
882 | the pattern being recursed into. |
6bda09f9 |
883 | |
c47ff5f1 |
884 | =item C<< (?>pattern) >> |
6bda09f9 |
885 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 |
886 | |
19799a22 |
887 | B<WARNING>: This extended regular expression feature is considered |
888 | highly experimental, and may be changed or deleted without notice. |
889 | |
890 | An "independent" subexpression, one which matches the substring |
891 | that a I<standalone> C<pattern> would match if anchored at the given |
9da458fc |
892 | position, and it matches I<nothing other than this substring>. This |
19799a22 |
893 | construct is useful for optimizations of what would otherwise be |
894 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). |
9da458fc |
895 | It may also be useful in places where the "grab all you can, and do not |
896 | give anything back" semantic is desirable. |
19799a22 |
897 | |
c47ff5f1 |
898 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 |
899 | (anchored at the beginning of string, as above) will match I<all> |
900 | characters C<a> at the beginning of string, leaving no C<a> for |
901 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, |
902 | since the match of the subgroup C<a*> is influenced by the following |
903 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside |
904 | C<a*ab> will match fewer characters than a standalone C<a*>, since |
905 | this makes the tail match. |
906 | |
c47ff5f1 |
907 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
19799a22 |
908 | C<(?=(pattern))\1>. This matches the same substring as a standalone |
909 | C<a+>, and the following C<\1> eats the matched string; it therefore |
c47ff5f1 |
910 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 |
911 | (The difference between these two constructs is that the second one |
912 | uses a capturing group, thus shifting ordinals of backreferences |
913 | in the rest of a regular expression.) |
914 | |
915 | Consider this pattern: |
c277df42 |
916 | |
871b0233 |
917 | m{ \( |
918 | ( |
9da458fc |
919 | [^()]+ # x+ |
871b0233 |
920 | | |
921 | \( [^()]* \) |
922 | )+ |
923 | \) |
924 | }x |
5a964f20 |
925 | |
19799a22 |
926 | That will efficiently match a nonempty group with matching parentheses |
927 | two levels deep or less. However, if there is no such group, it |
928 | will take virtually forever on a long string. That's because there |
929 | are so many different ways to split a long string into several |
930 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar |
931 | to a subpattern of the above pattern. Consider how the pattern |
932 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several |
933 | seconds, but that each extra letter doubles this time. This |
934 | exponential performance will make it appear that your program has |
14218588 |
935 | hung. However, a tiny change to this pattern |
5a964f20 |
936 | |
871b0233 |
937 | m{ \( |
938 | ( |
9da458fc |
939 | (?> [^()]+ ) # change x+ above to (?> x+ ) |
871b0233 |
940 | | |
941 | \( [^()]* \) |
942 | )+ |
943 | \) |
944 | }x |
c277df42 |
945 | |
c47ff5f1 |
946 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 |
947 | this yourself would be a productive exercise), but finishes in a fourth |
948 | the time when used on a similar string with 1000000 C<a>s. Be aware, |
949 | however, that this pattern currently triggers a warning message under |
9f1b1f2d |
950 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b |
951 | C<"matches null string many times in regex">. |
c277df42 |
952 | |
c47ff5f1 |
953 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 |
954 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 |
955 | This was only 4 times slower on a string with 1000000 C<a>s. |
956 | |
9da458fc |
957 | The "grab all you can, and do not give anything back" semantic is desirable |
958 | in many situations where on the first sight a simple C<()*> looks like |
959 | the correct solution. Suppose we parse text with comments being delimited |
960 | by C<#> followed by some optional (horizontal) whitespace. Contrary to |
4375e838 |
961 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc |
962 | the comment delimiter, because it may "give up" some whitespace if |
963 | the remainder of the pattern can be made to match that way. The correct |
964 | answer is either one of these: |
965 | |
966 | (?>#[ \t]*) |
967 | #[ \t]*(?![ \t]) |
968 | |
969 | For example, to grab non-empty comments into $1, one should use either |
970 | one of these: |
971 | |
972 | / (?> \# [ \t]* ) ( .+ ) /x; |
973 | / \# [ \t]* ( [^ \t] .* ) /x; |
974 | |
975 | Which one you pick depends on which of these expressions better reflects |
976 | the above specification of comments. |
977 | |
6bda09f9 |
978 | In some literature this construct is called "atomic matching" or |
979 | "possessive matching". |
980 | |
5a964f20 |
981 | =item C<(?(condition)yes-pattern|no-pattern)> |
d74e8afc |
982 | X<(?()> |
c277df42 |
983 | |
5a964f20 |
984 | =item C<(?(condition)yes-pattern)> |
c277df42 |
985 | |
19799a22 |
986 | B<WARNING>: This extended regular expression feature is considered |
987 | highly experimental, and may be changed or deleted without notice. |
988 | |
c277df42 |
989 | Conditional expression. C<(condition)> should be either an integer in |
990 | parentheses (which is valid if the corresponding pair of parentheses |
19799a22 |
991 | matched), or look-ahead/look-behind/evaluate zero-width assertion. |
c277df42 |
992 | |
19799a22 |
993 | For example: |
c277df42 |
994 | |
5a964f20 |
995 | m{ ( \( )? |
871b0233 |
996 | [^()]+ |
5a964f20 |
997 | (?(1) \) ) |
871b0233 |
998 | }x |
c277df42 |
999 | |
1000 | matches a chunk of non-parentheses, possibly included in parentheses |
1001 | themselves. |
a0d0e21e |
1002 | |
a0d0e21e |
1003 | =back |
1004 | |
c07a80fd |
1005 | =head2 Backtracking |
d74e8afc |
1006 | X<backtrack> X<backtracking> |
c07a80fd |
1007 | |
35a734be |
1008 | NOTE: This section presents an abstract approximation of regular |
1009 | expression behavior. For a more rigorous (and complicated) view of |
1010 | the rules involved in selecting a match among possible alternatives, |
1011 | see L<Combining pieces together>. |
1012 | |
c277df42 |
1013 | A fundamental feature of regular expression matching involves the |
5a964f20 |
1014 | notion called I<backtracking>, which is currently used (when needed) |
c277df42 |
1015 | by all regular expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc |
1016 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
1017 | internally, but the general principle outlined here is valid. |
c07a80fd |
1018 | |
1019 | For a regular expression to match, the I<entire> regular expression must |
1020 | match, not just part of it. So if the beginning of a pattern containing a |
1021 | quantifier succeeds in a way that causes later parts in the pattern to |
1022 | fail, the matching engine backs up and recalculates the beginning |
1023 | part--that's why it's called backtracking. |
1024 | |
1025 | Here is an example of backtracking: Let's say you want to find the |
1026 | word following "foo" in the string "Food is on the foo table.": |
1027 | |
1028 | $_ = "Food is on the foo table."; |
1029 | if ( /\b(foo)\s+(\w+)/i ) { |
1030 | print "$2 follows $1.\n"; |
1031 | } |
1032 | |
1033 | When the match runs, the first part of the regular expression (C<\b(foo)>) |
1034 | finds a possible match right at the beginning of the string, and loads up |
1035 | $1 with "Foo". However, as soon as the matching engine sees that there's |
1036 | no whitespace following the "Foo" that it had saved in $1, it realizes its |
68dc0745 |
1037 | mistake and starts over again one character after where it had the |
c07a80fd |
1038 | tentative match. This time it goes all the way until the next occurrence |
1039 | of "foo". The complete regular expression matches this time, and you get |
1040 | the expected output of "table follows foo." |
1041 | |
1042 | Sometimes minimal matching can help a lot. Imagine you'd like to match |
1043 | everything between "foo" and "bar". Initially, you write something |
1044 | like this: |
1045 | |
1046 | $_ = "The food is under the bar in the barn."; |
1047 | if ( /foo(.*)bar/ ) { |
1048 | print "got <$1>\n"; |
1049 | } |
1050 | |
1051 | Which perhaps unexpectedly yields: |
1052 | |
1053 | got <d is under the bar in the > |
1054 | |
1055 | That's because C<.*> was greedy, so you get everything between the |
14218588 |
1056 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd |
1057 | to use minimal matching to make sure you get the text between a "foo" |
1058 | and the first "bar" thereafter. |
1059 | |
1060 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } |
1061 | got <d is under the > |
1062 | |
1063 | Here's another example: let's say you'd like to match a number at the end |
b6e13d97 |
1064 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd |
1065 | So you write this: |
1066 | |
1067 | $_ = "I have 2 numbers: 53147"; |
1068 | if ( /(.*)(\d*)/ ) { # Wrong! |
1069 | print "Beginning is <$1>, number is <$2>.\n"; |
1070 | } |
1071 | |
1072 | That won't work at all, because C<.*> was greedy and gobbled up the |
1073 | whole string. As C<\d*> can match on an empty string the complete |
1074 | regular expression matched successfully. |
1075 | |
8e1088bc |
1076 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd |
1077 | |
1078 | Here are some variants, most of which don't work: |
1079 | |
1080 | $_ = "I have 2 numbers: 53147"; |
1081 | @pats = qw{ |
1082 | (.*)(\d*) |
1083 | (.*)(\d+) |
1084 | (.*?)(\d*) |
1085 | (.*?)(\d+) |
1086 | (.*)(\d+)$ |
1087 | (.*?)(\d+)$ |
1088 | (.*)\b(\d+)$ |
1089 | (.*\D)(\d+)$ |
1090 | }; |
1091 | |
1092 | for $pat (@pats) { |
1093 | printf "%-12s ", $pat; |
1094 | if ( /$pat/ ) { |
1095 | print "<$1> <$2>\n"; |
1096 | } else { |
1097 | print "FAIL\n"; |
1098 | } |
1099 | } |
1100 | |
1101 | That will print out: |
1102 | |
1103 | (.*)(\d*) <I have 2 numbers: 53147> <> |
1104 | (.*)(\d+) <I have 2 numbers: 5314> <7> |
1105 | (.*?)(\d*) <> <> |
1106 | (.*?)(\d+) <I have > <2> |
1107 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> |
1108 | (.*?)(\d+)$ <I have 2 numbers: > <53147> |
1109 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> |
1110 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> |
1111 | |
1112 | As you see, this can be a bit tricky. It's important to realize that a |
1113 | regular expression is merely a set of assertions that gives a definition |
1114 | of success. There may be 0, 1, or several different ways that the |
1115 | definition might succeed against a particular string. And if there are |
5a964f20 |
1116 | multiple ways it might succeed, you need to understand backtracking to |
1117 | know which variety of success you will achieve. |
c07a80fd |
1118 | |
19799a22 |
1119 | When using look-ahead assertions and negations, this can all get even |
8b19b778 |
1120 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd |
1121 | followed by "123". You might try to write that as |
1122 | |
871b0233 |
1123 | $_ = "ABC123"; |
1124 | if ( /^\D*(?!123)/ ) { # Wrong! |
1125 | print "Yup, no 123 in $_\n"; |
1126 | } |
c07a80fd |
1127 | |
1128 | But that isn't going to match; at least, not the way you're hoping. It |
1129 | claims that there is no 123 in the string. Here's a clearer picture of |
9b9391b2 |
1130 | why that pattern matches, contrary to popular expectations: |
c07a80fd |
1131 | |
4358a253 |
1132 | $x = 'ABC123'; |
1133 | $y = 'ABC445'; |
c07a80fd |
1134 | |
4358a253 |
1135 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
1136 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; |
c07a80fd |
1137 | |
4358a253 |
1138 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; |
1139 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; |
c07a80fd |
1140 | |
1141 | This prints |
1142 | |
1143 | 2: got ABC |
1144 | 3: got AB |
1145 | 4: got ABC |
1146 | |
5f05dabc |
1147 | You might have expected test 3 to fail because it seems to a more |
c07a80fd |
1148 | general purpose version of test 1. The important difference between |
1149 | them is that test 3 contains a quantifier (C<\D*>) and so can use |
1150 | backtracking, whereas test 1 will not. What's happening is |
1151 | that you've asked "Is it true that at the start of $x, following 0 or more |
5f05dabc |
1152 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd |
1153 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 |
1154 | fail. |
14218588 |
1155 | |
c07a80fd |
1156 | The search engine will initially match C<\D*> with "ABC". Then it will |
14218588 |
1157 | try to match C<(?!123> with "123", which fails. But because |
c07a80fd |
1158 | a quantifier (C<\D*>) has been used in the regular expression, the |
1159 | search engine can backtrack and retry the match differently |
54310121 |
1160 | in the hope of matching the complete regular expression. |
c07a80fd |
1161 | |
5a964f20 |
1162 | The pattern really, I<really> wants to succeed, so it uses the |
1163 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this |
c07a80fd |
1164 | time. Now there's indeed something following "AB" that is not |
14218588 |
1165 | "123". It's "C123", which suffices. |
c07a80fd |
1166 | |
14218588 |
1167 | We can deal with this by using both an assertion and a negation. |
1168 | We'll say that the first part in $1 must be followed both by a digit |
1169 | and by something that's not "123". Remember that the look-aheads |
1170 | are zero-width expressions--they only look, but don't consume any |
1171 | of the string in their match. So rewriting this way produces what |
c07a80fd |
1172 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
1173 | |
4358a253 |
1174 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; |
1175 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; |
c07a80fd |
1176 | |
1177 | 6: got ABC |
1178 | |
5a964f20 |
1179 | In other words, the two zero-width assertions next to each other work as though |
19799a22 |
1180 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd |
1181 | matches only if you're at the beginning of the line AND the end of the |
1182 | line simultaneously. The deeper underlying truth is that juxtaposition in |
1183 | regular expressions always means AND, except when you write an explicit OR |
1184 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", |
1185 | although the attempted matches are made at different positions because "a" |
1186 | is not a zero-width assertion, but a one-width assertion. |
1187 | |
19799a22 |
1188 | B<WARNING>: particularly complicated regular expressions can take |
14218588 |
1189 | exponential time to solve because of the immense number of possible |
9da458fc |
1190 | ways they can use backtracking to try match. For example, without |
1191 | internal optimizations done by the regular expression engine, this will |
1192 | take a painfully long time to run: |
c07a80fd |
1193 | |
e1901655 |
1194 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
1195 | |
1196 | And if you used C<*>'s in the internal groups instead of limiting them |
1197 | to 0 through 5 matches, then it would take forever--or until you ran |
1198 | out of stack space. Moreover, these internal optimizations are not |
1199 | always applicable. For example, if you put C<{0,5}> instead of C<*> |
1200 | on the external group, no current optimization is applicable, and the |
1201 | match takes a long time to finish. |
c07a80fd |
1202 | |
9da458fc |
1203 | A powerful tool for optimizing such beasts is what is known as an |
1204 | "independent group", |
c47ff5f1 |
1205 | which does not backtrack (see L<C<< (?>pattern) >>>). Note also that |
9da458fc |
1206 | zero-length look-ahead/look-behind assertions will not backtrack to make |
14218588 |
1207 | the tail match, since they are in "logical" context: only |
1208 | whether they match is considered relevant. For an example |
9da458fc |
1209 | where side-effects of look-ahead I<might> have influenced the |
c47ff5f1 |
1210 | following match, see L<C<< (?>pattern) >>>. |
c277df42 |
1211 | |
a0d0e21e |
1212 | =head2 Version 8 Regular Expressions |
d74e8afc |
1213 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> |
a0d0e21e |
1214 | |
5a964f20 |
1215 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e |
1216 | routines, here are the pattern-matching rules not described above. |
1217 | |
54310121 |
1218 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e |
1219 | with a special meaning described here or above. You can cause |
5a964f20 |
1220 | characters that normally function as metacharacters to be interpreted |
5f05dabc |
1221 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
a0d0e21e |
1222 | character; "\\" matches a "\"). A series of characters matches that |
1223 | series of characters in the target string, so the pattern C<blurfl> |
1224 | would match "blurfl" in the target string. |
1225 | |
1226 | You can specify a character class, by enclosing a list of characters |
5a964f20 |
1227 | in C<[]>, which will match any one character from the list. If the |
a0d0e21e |
1228 | first character after the "[" is "^", the class matches any character not |
14218588 |
1229 | in the list. Within a list, the "-" character specifies a |
5a964f20 |
1230 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 |
1231 | inclusive. If you want either "-" or "]" itself to be a member of a |
1232 | class, put it at the start of the list (possibly after a "^"), or |
1233 | escape it with a backslash. "-" is also taken literally when it is |
1234 | at the end of the list, just before the closing "]". (The |
84850974 |
1235 | following all specify the same class of three characters: C<[-az]>, |
1236 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which |
fb55449c |
1237 | specifies a class containing twenty-six characters, even on EBCDIC |
1238 | based coded character sets.) Also, if you try to use the character |
1239 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of |
1240 | a range, that's not a range, the "-" is understood literally. |
a0d0e21e |
1241 | |
8ada0baa |
1242 | Note also that the whole range idea is rather unportable between |
1243 | character sets--and even within character sets they may cause results |
1244 | you probably didn't expect. A sound principle is to use only ranges |
1245 | that begin from and end at either alphabets of equal case ([a-e], |
1246 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, |
1247 | spell out the character sets in full. |
1248 | |
54310121 |
1249 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e |
1250 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
1251 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string |
fb55449c |
1252 | of octal digits, matches the character whose coded character set value |
1253 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, |
1254 | matches the character whose numeric value is I<nn>. The expression \cI<x> |
1255 | matches the character control-I<x>. Finally, the "." metacharacter |
1256 | matches any character except "\n" (unless you use C</s>). |
a0d0e21e |
1257 | |
1258 | You can specify a series of alternatives for a pattern using "|" to |
1259 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", |
5a964f20 |
1260 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e |
1261 | first alternative includes everything from the last pattern delimiter |
1262 | ("(", "[", or the beginning of the pattern) up to the first "|", and |
1263 | the last alternative contains everything from the last "|" to the next |
14218588 |
1264 | pattern delimiter. That's why it's common practice to include |
1265 | alternatives in parentheses: to minimize confusion about where they |
a3cb178b |
1266 | start and end. |
1267 | |
5a964f20 |
1268 | Alternatives are tried from left to right, so the first |
a3cb178b |
1269 | alternative found for which the entire expression matches, is the one that |
1270 | is chosen. This means that alternatives are not necessarily greedy. For |
628afcb5 |
1271 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b |
1272 | part will match, as that is the first alternative tried, and it successfully |
1273 | matches the target string. (This might not seem important, but it is |
1274 | important when you are capturing matched text using parentheses.) |
1275 | |
5a964f20 |
1276 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b |
1277 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e |
1278 | |
14218588 |
1279 | Within a pattern, you may designate subpatterns for later reference |
1280 | by enclosing them in parentheses, and you may refer back to the |
1281 | I<n>th subpattern later in the pattern using the metacharacter |
1282 | \I<n>. Subpatterns are numbered based on the left to right order |
1283 | of their opening parenthesis. A backreference matches whatever |
1284 | actually matched the subpattern in the string being examined, not |
1285 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will |
1286 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern |
1287 | 1 matched "0x", even though the rule C<0|0x> could potentially match |
1288 | the leading 0 in the second number. |
cb1a09d0 |
1289 | |
19799a22 |
1290 | =head2 Warning on \1 vs $1 |
cb1a09d0 |
1291 | |
5a964f20 |
1292 | Some people get too used to writing things like: |
cb1a09d0 |
1293 | |
1294 | $pattern =~ s/(\W)/\\\1/g; |
1295 | |
1296 | This is grandfathered for the RHS of a substitute to avoid shocking the |
1297 | B<sed> addicts, but it's a dirty habit to get into. That's because in |
d1be9408 |
1298 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 |
1299 | the usual double-quoted string means a control-A. The customary Unix |
1300 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit |
1301 | of doing that, you get yourself into trouble if you then add an C</e> |
1302 | modifier. |
1303 | |
5a964f20 |
1304 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 |
1305 | |
1306 | Or if you try to do |
1307 | |
1308 | s/(\d+)/\1000/; |
1309 | |
1310 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with |
14218588 |
1311 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 |
1312 | with the operation of matching a backreference. Certainly they mean two |
1313 | different things on the I<left> side of the C<s///>. |
9fa51da4 |
1314 | |
c84d73f1 |
1315 | =head2 Repeated patterns matching zero-length substring |
1316 | |
19799a22 |
1317 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 |
1318 | |
1319 | Regular expressions provide a terse and powerful programming language. As |
1320 | with most other power tools, power comes together with the ability |
1321 | to wreak havoc. |
1322 | |
1323 | A common abuse of this power stems from the ability to make infinite |
628afcb5 |
1324 | loops using regular expressions, with something as innocuous as: |
c84d73f1 |
1325 | |
1326 | 'foo' =~ m{ ( o? )* }x; |
1327 | |
1328 | The C<o?> can match at the beginning of C<'foo'>, and since the position |
1329 | in the string is not moved by the match, C<o?> would match again and again |
14218588 |
1330 | because of the C<*> modifier. Another common way to create a similar cycle |
c84d73f1 |
1331 | is with the looping modifier C<//g>: |
1332 | |
1333 | @matches = ( 'foo' =~ m{ o? }xg ); |
1334 | |
1335 | or |
1336 | |
1337 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; |
1338 | |
1339 | or the loop implied by split(). |
1340 | |
1341 | However, long experience has shown that many programming tasks may |
14218588 |
1342 | be significantly simplified by using repeated subexpressions that |
1343 | may match zero-length substrings. Here's a simple example being: |
c84d73f1 |
1344 | |
1345 | @chars = split //, $string; # // is not magic in split |
1346 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / |
1347 | |
9da458fc |
1348 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 |
1349 | the infinite loop>. The rules for this are different for lower-level |
1350 | loops given by the greedy modifiers C<*+{}>, and for higher-level |
1351 | ones like the C</g> modifier or split() operator. |
1352 | |
19799a22 |
1353 | The lower-level loops are I<interrupted> (that is, the loop is |
1354 | broken) when Perl detects that a repeated expression matched a |
1355 | zero-length substring. Thus |
c84d73f1 |
1356 | |
1357 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; |
1358 | |
1359 | is made equivalent to |
1360 | |
1361 | m{ (?: NON_ZERO_LENGTH )* |
1362 | | |
1363 | (?: ZERO_LENGTH )? |
1364 | }x; |
1365 | |
1366 | The higher level-loops preserve an additional state between iterations: |
1367 | whether the last match was zero-length. To break the loop, the following |
1368 | match after a zero-length match is prohibited to have a length of zero. |
1369 | This prohibition interacts with backtracking (see L<"Backtracking">), |
1370 | and so the I<second best> match is chosen if the I<best> match is of |
1371 | zero length. |
1372 | |
19799a22 |
1373 | For example: |
c84d73f1 |
1374 | |
1375 | $_ = 'bar'; |
1376 | s/\w??/<$&>/g; |
1377 | |
20fb949f |
1378 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
c84d73f1 |
1379 | match given by non-greedy C<??> is the zero-length match, and the I<second |
1380 | best> match is what is matched by C<\w>. Thus zero-length matches |
1381 | alternate with one-character-long matches. |
1382 | |
1383 | Similarly, for repeated C<m/()/g> the second-best match is the match at the |
1384 | position one notch further in the string. |
1385 | |
19799a22 |
1386 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 |
1387 | the matched string, and is reset by each assignment to pos(). |
9da458fc |
1388 | Zero-length matches at the end of the previous match are ignored |
1389 | during C<split>. |
c84d73f1 |
1390 | |
35a734be |
1391 | =head2 Combining pieces together |
1392 | |
1393 | Each of the elementary pieces of regular expressions which were described |
1394 | before (such as C<ab> or C<\Z>) could match at most one substring |
1395 | at the given position of the input string. However, in a typical regular |
1396 | expression these elementary pieces are combined into more complicated |
1397 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc |
1398 | (in these examples C<S> and C<T> are regular subexpressions). |
1399 | |
1400 | Such combinations can include alternatives, leading to a problem of choice: |
1401 | if we match a regular expression C<a|ab> against C<"abc">, will it match |
1402 | substring C<"a"> or C<"ab">? One way to describe which substring is |
1403 | actually matched is the concept of backtracking (see L<"Backtracking">). |
1404 | However, this description is too low-level and makes you think |
1405 | in terms of a particular implementation. |
1406 | |
1407 | Another description starts with notions of "better"/"worse". All the |
1408 | substrings which may be matched by the given regular expression can be |
1409 | sorted from the "best" match to the "worst" match, and it is the "best" |
1410 | match which is chosen. This substitutes the question of "what is chosen?" |
1411 | by the question of "which matches are better, and which are worse?". |
1412 | |
1413 | Again, for elementary pieces there is no such question, since at most |
1414 | one match at a given position is possible. This section describes the |
1415 | notion of better/worse for combining operators. In the description |
1416 | below C<S> and C<T> are regular subexpressions. |
1417 | |
13a2d996 |
1418 | =over 4 |
35a734be |
1419 | |
1420 | =item C<ST> |
1421 | |
1422 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are |
1423 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings |
1424 | which can be matched by C<T>. |
1425 | |
1426 | If C<A> is better match for C<S> than C<A'>, C<AB> is a better |
1427 | match than C<A'B'>. |
1428 | |
1429 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if |
1430 | C<B> is better match for C<T> than C<B'>. |
1431 | |
1432 | =item C<S|T> |
1433 | |
1434 | When C<S> can match, it is a better match than when only C<T> can match. |
1435 | |
1436 | Ordering of two matches for C<S> is the same as for C<S>. Similar for |
1437 | two matches for C<T>. |
1438 | |
1439 | =item C<S{REPEAT_COUNT}> |
1440 | |
1441 | Matches as C<SSS...S> (repeated as many times as necessary). |
1442 | |
1443 | =item C<S{min,max}> |
1444 | |
1445 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. |
1446 | |
1447 | =item C<S{min,max}?> |
1448 | |
1449 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. |
1450 | |
1451 | =item C<S?>, C<S*>, C<S+> |
1452 | |
1453 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. |
1454 | |
1455 | =item C<S??>, C<S*?>, C<S+?> |
1456 | |
1457 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. |
1458 | |
c47ff5f1 |
1459 | =item C<< (?>S) >> |
35a734be |
1460 | |
1461 | Matches the best match for C<S> and only that. |
1462 | |
1463 | =item C<(?=S)>, C<(?<=S)> |
1464 | |
1465 | Only the best match for C<S> is considered. (This is important only if |
1466 | C<S> has capturing parentheses, and backreferences are used somewhere |
1467 | else in the whole regular expression.) |
1468 | |
1469 | =item C<(?!S)>, C<(?<!S)> |
1470 | |
1471 | For this grouping operator there is no need to describe the ordering, since |
1472 | only whether or not C<S> can match is important. |
1473 | |
6bda09f9 |
1474 | =item C<(??{ EXPR })>, C<(?PARNO)> |
35a734be |
1475 | |
1476 | The ordering is the same as for the regular expression which is |
6bda09f9 |
1477 | the result of EXPR, or the pattern contained by capture buffer PARNO. |
35a734be |
1478 | |
1479 | =item C<(?(condition)yes-pattern|no-pattern)> |
1480 | |
1481 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is |
1482 | already determined. The ordering of the matches is the same as for the |
1483 | chosen subexpression. |
1484 | |
1485 | =back |
1486 | |
1487 | The above recipes describe the ordering of matches I<at a given position>. |
1488 | One more rule is needed to understand how a match is determined for the |
1489 | whole regular expression: a match at an earlier position is always better |
1490 | than a match at a later position. |
1491 | |
c84d73f1 |
1492 | =head2 Creating custom RE engines |
1493 | |
1494 | Overloaded constants (see L<overload>) provide a simple way to extend |
1495 | the functionality of the RE engine. |
1496 | |
1497 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which |
6b0ac556 |
1498 | matches at boundary between whitespace characters and non-whitespace |
c84d73f1 |
1499 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
1500 | at these positions, so we want to have each C<\Y|> in the place of the |
1501 | more complicated version. We can create a module C<customre> to do |
1502 | this: |
1503 | |
1504 | package customre; |
1505 | use overload; |
1506 | |
1507 | sub import { |
1508 | shift; |
1509 | die "No argument to customre::import allowed" if @_; |
1510 | overload::constant 'qr' => \&convert; |
1511 | } |
1512 | |
1513 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} |
1514 | |
580a9fe1 |
1515 | # We must also take care of not escaping the legitimate \\Y| |
1516 | # sequence, hence the presence of '\\' in the conversion rules. |
141db969 |
1517 | my %rules = ( '\\' => '\\\\', |
c84d73f1 |
1518 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
1519 | sub convert { |
1520 | my $re = shift; |
1521 | $re =~ s{ |
1522 | \\ ( \\ | Y . ) |
1523 | } |
1524 | { $rules{$1} or invalid($re,$1) }sgex; |
1525 | return $re; |
1526 | } |
1527 | |
1528 | Now C<use customre> enables the new escape in constant regular |
1529 | expressions, i.e., those without any runtime variable interpolations. |
1530 | As documented in L<overload>, this conversion will work only over |
1531 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable |
1532 | part of this regular expression needs to be converted explicitly |
1533 | (but only if the special meaning of C<\Y|> should be enabled inside $re): |
1534 | |
1535 | use customre; |
1536 | $re = <>; |
1537 | chomp $re; |
1538 | $re = customre::convert $re; |
1539 | /\Y|$re\Y|/; |
1540 | |
19799a22 |
1541 | =head1 BUGS |
1542 | |
9da458fc |
1543 | This document varies from difficult to understand to completely |
1544 | and utterly opaque. The wandering prose riddled with jargon is |
1545 | hard to fathom in several places. |
1546 | |
1547 | This document needs a rewrite that separates the tutorial content |
1548 | from the reference content. |
19799a22 |
1549 | |
1550 | =head1 SEE ALSO |
9fa51da4 |
1551 | |
91e0c79e |
1552 | L<perlrequick>. |
1553 | |
1554 | L<perlretut>. |
1555 | |
9b599b2a |
1556 | L<perlop/"Regexp Quote-Like Operators">. |
1557 | |
1e66bd83 |
1558 | L<perlop/"Gory details of parsing quoted constructs">. |
1559 | |
14218588 |
1560 | L<perlfaq6>. |
1561 | |
9b599b2a |
1562 | L<perlfunc/pos>. |
1563 | |
1564 | L<perllocale>. |
1565 | |
fb55449c |
1566 | L<perlebcdic>. |
1567 | |
14218588 |
1568 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
1569 | by O'Reilly and Associates. |