Commit | Line | Data |
---|---|---|
920dae64 AT |
1 | =head1 NAME |
2 | X<regular expression> X<regex> X<regexp> | |
3 | ||
4 | perlre - Perl regular expressions | |
5 | ||
6 | =head1 DESCRIPTION | |
7 | ||
8 | This page describes the syntax of regular expressions in Perl. | |
9 | ||
10 | If you haven't used regular expressions before, a quick-start | |
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">. | |
18 | ||
19 | Matching operations can have various modifiers. Modifiers | |
20 | that relate to the interpretation of the regular expression inside | |
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 | |
23 | L<perlop/"Gory details of parsing quoted constructs">. | |
24 | ||
25 | =over 4 | |
26 | ||
27 | =item i | |
28 | X</i> X<regex, case-insensitive> X<regexp, case-insensitive> | |
29 | X<regular expression, case-insensitive> | |
30 | ||
31 | Do case-insensitive pattern matching. | |
32 | ||
33 | If C<use locale> is in effect, the case map is taken from the current | |
34 | locale. See L<perllocale>. | |
35 | ||
36 | =item m | |
37 | X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline> | |
38 | ||
39 | Treat string as multiple lines. That is, change "^" and "$" from matching | |
40 | the start or end of the string to matching the start or end of any | |
41 | line anywhere within the string. | |
42 | ||
43 | =item s | |
44 | X</s> X<regex, single-line> X<regexp, single-line> | |
45 | X<regular expression, single-line> | |
46 | ||
47 | Treat string as single line. That is, change "." to match any character | |
48 | whatsoever, even a newline, which normally it would not match. | |
49 | ||
50 | The C</s> and C</m> modifiers both override the C<$*> setting. That | |
51 | is, no matter what C<$*> contains, C</s> without C</m> will force | |
52 | "^" to match only at the beginning of the string and "$" to match | |
53 | only at the end (or just before a newline at the end) of the string. | |
54 | Together, as /ms, they let the "." match any character whatsoever, | |
55 | while still allowing "^" and "$" to match, respectively, just after | |
56 | and just before newlines within the string. | |
57 | ||
58 | =item x | |
59 | X</x> | |
60 | ||
61 | Extend your pattern's legibility by permitting whitespace and comments. | |
62 | ||
63 | =back | |
64 | ||
65 | These are usually written as "the C</x> modifier", even though the delimiter | |
66 | in question might not really be a slash. Any of these | |
67 | modifiers may also be embedded within the regular expression itself using | |
68 | the C<(?...)> construct. See below. | |
69 | ||
70 | The C</x> modifier itself needs a little more explanation. It tells | |
71 | the regular expression parser to ignore whitespace that is neither | |
72 | backslashed nor within a character class. You can use this to break up | |
73 | your regular expression into (slightly) more readable parts. The C<#> | |
74 | character is also treated as a metacharacter introducing a comment, | |
75 | just as in ordinary Perl code. This also means that if you want real | |
76 | whitespace or C<#> characters in the pattern (outside a character | |
77 | class, where they are unaffected by C</x>), that you'll either have to | |
78 | escape them or encode them using octal or hex escapes. Taken together, | |
79 | these features go a long way towards making Perl's regular expressions | |
80 | more readable. Note that you have to be careful not to include the | |
81 | pattern delimiter in the comment--perl has no way of knowing you did | |
82 | not intend to close the pattern early. See the C-comment deletion code | |
83 | in L<perlop>. | |
84 | X</x> | |
85 | ||
86 | =head2 Regular Expressions | |
87 | ||
88 | The patterns used in Perl pattern matching derive from supplied in | |
89 | the Version 8 regex routines. (The routines are derived | |
90 | (distantly) from Henry Spencer's freely redistributable reimplementation | |
91 | of the V8 routines.) See L<Version 8 Regular Expressions> for | |
92 | details. | |
93 | ||
94 | In particular the following metacharacters have their standard I<egrep>-ish | |
95 | meanings: | |
96 | X<metacharacter> | |
97 | X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]> | |
98 | ||
99 | ||
100 | \ Quote the next metacharacter | |
101 | ^ Match the beginning of the line | |
102 | . Match any character (except newline) | |
103 | $ Match the end of the line (or before newline at the end) | |
104 | | Alternation | |
105 | () Grouping | |
106 | [] Character class | |
107 | ||
108 | By default, the "^" character is guaranteed to match only the | |
109 | beginning of the string, the "$" character only the end (or before the | |
110 | newline at the end), and Perl does certain optimizations with the | |
111 | assumption that the string contains only one line. Embedded newlines | |
112 | will not be matched by "^" or "$". You may, however, wish to treat a | |
113 | string as a multi-line buffer, such that the "^" will match after any | |
114 | newline within the string, and "$" will match before any newline. At the | |
115 | cost of a little more overhead, you can do this by using the /m modifier | |
116 | on the pattern match operator. (Older programs did this by setting C<$*>, | |
117 | but this practice is now deprecated.) | |
118 | X<^> X<$> X</m> | |
119 | ||
120 | To simplify multi-line substitutions, the "." character never matches a | |
121 | newline unless you use the C</s> modifier, which in effect tells Perl to pretend | |
122 | the string is a single line--even if it isn't. The C</s> modifier also | |
123 | overrides the setting of C<$*>, in case you have some (badly behaved) older | |
124 | code that sets it in another module. | |
125 | X<.> X</s> | |
126 | ||
127 | The following standard quantifiers are recognized: | |
128 | X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}> | |
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 | |
138 | as a regular character. In particular, the lower bound | |
139 | is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+" | |
140 | modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited | |
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 | ||
145 | $_ **= $_ , / {$_} / for 2 .. 42; | |
146 | ||
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": | |
152 | X<metacharacter> X<greedy> X<greedyness> | |
153 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?> | |
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 | ||
162 | Because patterns are processed as double quoted strings, the following | |
163 | also work: | |
164 | X<\t> X<\n> X<\r> X<\f> X<\a> X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q> | |
165 | X<\0> X<\c> X<\N> X<\x> | |
166 | ||
167 | \t tab (HT, TAB) | |
168 | \n newline (LF, NL) | |
169 | \r return (CR) | |
170 | \f form feed (FF) | |
171 | \a alarm (bell) (BEL) | |
172 | \e escape (think troff) (ESC) | |
173 | \033 octal char (think of a PDP-11) | |
174 | \x1B hex char | |
175 | \x{263a} wide hex char (Unicode SMILEY) | |
176 | \c[ control char | |
177 | \N{name} named char | |
178 | \l lowercase next char (think vi) | |
179 | \u uppercase next char (think vi) | |
180 | \L lowercase till \E (think vi) | |
181 | \U uppercase till \E (think vi) | |
182 | \E end case modification (think vi) | |
183 | \Q quote (disable) pattern metacharacters till \E | |
184 | ||
185 | If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u> | |
186 | and C<\U> is taken from the current locale. See L<perllocale>. For | |
187 | documentation of C<\N{name}>, see L<charnames>. | |
188 | ||
189 | You cannot include a literal C<$> or C<@> within a C<\Q> sequence. | |
190 | An unescaped C<$> or C<@> interpolates the corresponding variable, | |
191 | while escaping will cause the literal string C<\$> to be matched. | |
192 | You'll need to write something like C<m/\Quser\E\@\Qhost/>. | |
193 | ||
194 | In addition, Perl defines the following: | |
195 | X<metacharacter> | |
196 | X<\w> X<\W> X<\s> X<\S> X<\d> X<\D> X<\X> X<\p> X<\P> X<\C> | |
197 | X<word> X<whitespace> | |
198 | ||
199 | \w Match a "word" character (alphanumeric plus "_") | |
200 | \W Match a non-"word" character | |
201 | \s Match a whitespace character | |
202 | \S Match a non-whitespace character | |
203 | \d Match a digit character | |
204 | \D Match a non-digit character | |
205 | \pP Match P, named property. Use \p{Prop} for longer names. | |
206 | \PP Match non-P | |
207 | \X Match eXtended Unicode "combining character sequence", | |
208 | equivalent to (?:\PM\pM*) | |
209 | \C Match a single C char (octet) even under Unicode. | |
210 | NOTE: breaks up characters into their UTF-8 bytes, | |
211 | so you may end up with malformed pieces of UTF-8. | |
212 | Unsupported in lookbehind. | |
213 | ||
214 | A C<\w> matches a single alphanumeric character (an alphabetic | |
215 | character, or a decimal digit) or C<_>, not a whole word. Use C<\w+> | |
216 | to match a string of Perl-identifier characters (which isn't the same | |
217 | as matching an English word). If C<use locale> is in effect, the list | |
218 | of alphabetic characters generated by C<\w> is taken from the current | |
219 | locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>, | |
220 | C<\d>, and C<\D> within character classes, but if you try to use them | |
221 | as endpoints of a range, that's not a range, the "-" is understood | |
222 | literally. If Unicode is in effect, C<\s> matches also "\x{85}", | |
223 | "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about | |
224 | C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general. | |
225 | You can define your own C<\p> and C<\P> properties, see L<perlunicode>. | |
226 | X<\w> X<\W> X<word> | |
227 | ||
228 | The POSIX character class syntax | |
229 | X<character class> | |
230 | ||
231 | [:class:] | |
232 | ||
233 | is also available. The available classes and their backslash | |
234 | equivalents (if available) are as follows: | |
235 | X<character class> | |
236 | X<alpha> X<alnum> X<ascii> X<blank> X<cntrl> X<digit> X<graph> | |
237 | X<lower> X<print> X<punct> X<space> X<upper> X<word> X<xdigit> | |
238 | ||
239 | alpha | |
240 | alnum | |
241 | ascii | |
242 | blank [1] | |
243 | cntrl | |
244 | digit \d | |
245 | graph | |
246 | lower | |
247 | ||
248 | punct | |
249 | space \s [2] | |
250 | upper | |
251 | word \w [3] | |
252 | xdigit | |
253 | ||
254 | =over | |
255 | ||
256 | =item [1] | |
257 | ||
258 | A GNU extension equivalent to C<[ \t]>, "all horizontal whitespace". | |
259 | ||
260 | =item [2] | |
261 | ||
262 | Not exactly equivalent to C<\s> since the C<[[:space:]]> includes | |
263 | also the (very rare) "vertical tabulator", "\ck", chr(11). | |
264 | ||
265 | =item [3] | |
266 | ||
267 | A Perl extension, see above. | |
268 | ||
269 | =back | |
270 | ||
271 | For example use C<[:upper:]> to match all the uppercase characters. | |
272 | Note that the C<[]> are part of the C<[::]> construct, not part of the | |
273 | whole character class. For example: | |
274 | ||
275 | [01[:alpha:]%] | |
276 | ||
277 | matches zero, one, any alphabetic character, and the percentage sign. | |
278 | ||
279 | The following equivalences to Unicode \p{} constructs and equivalent | |
280 | backslash character classes (if available), will hold: | |
281 | X<character class> X<\p> X<\p{}> | |
282 | ||
283 | [:...:] \p{...} backslash | |
284 | ||
285 | alpha IsAlpha | |
286 | alnum IsAlnum | |
287 | ascii IsASCII | |
288 | blank IsSpace | |
289 | cntrl IsCntrl | |
290 | digit IsDigit \d | |
291 | graph IsGraph | |
292 | lower IsLower | |
293 | print IsPrint | |
294 | punct IsPunct | |
295 | space IsSpace | |
296 | IsSpacePerl \s | |
297 | upper IsUpper | |
298 | word IsWord | |
299 | xdigit IsXDigit | |
300 | ||
301 | For example C<[:lower:]> and C<\p{IsLower}> are equivalent. | |
302 | ||
303 | If the C<utf8> pragma is not used but the C<locale> pragma is, the | |
304 | classes correlate with the usual isalpha(3) interface (except for | |
305 | "word" and "blank"). | |
306 | ||
307 | The assumedly non-obviously named classes are: | |
308 | ||
309 | =over 4 | |
310 | ||
311 | =item cntrl | |
312 | X<cntrl> | |
313 | ||
314 | Any control character. Usually characters that don't produce output as | |
315 | such but instead control the terminal somehow: for example newline and | |
316 | backspace are control characters. All characters with ord() less than | |
317 | 32 are most often classified as control characters (assuming ASCII, | |
318 | the ISO Latin character sets, and Unicode), as is the character with | |
319 | the ord() value of 127 (C<DEL>). | |
320 | ||
321 | =item graph | |
322 | X<graph> | |
323 | ||
324 | Any alphanumeric or punctuation (special) character. | |
325 | ||
326 | =item print | |
327 | X<print> | |
328 | ||
329 | Any alphanumeric or punctuation (special) character or the space character. | |
330 | ||
331 | =item punct | |
332 | X<punct> | |
333 | ||
334 | Any punctuation (special) character. | |
335 | ||
336 | =item xdigit | |
337 | X<xdigit> | |
338 | ||
339 | Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would | |
340 | work just fine) it is included for completeness. | |
341 | ||
342 | =back | |
343 | ||
344 | You can negate the [::] character classes by prefixing the class name | |
345 | with a '^'. This is a Perl extension. For example: | |
346 | X<character class, negation> | |
347 | ||
348 | POSIX traditional Unicode | |
349 | ||
350 | [:^digit:] \D \P{IsDigit} | |
351 | [:^space:] \S \P{IsSpace} | |
352 | [:^word:] \W \P{IsWord} | |
353 | ||
354 | Perl respects the POSIX standard in that POSIX character classes are | |
355 | only supported within a character class. The POSIX character classes | |
356 | [.cc.] and [=cc=] are recognized but B<not> supported and trying to | |
357 | use them will cause an error. | |
358 | ||
359 | Perl defines the following zero-width assertions: | |
360 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> | |
361 | X<regexp, zero-width assertion> | |
362 | X<regular expression, zero-width assertion> | |
363 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> | |
364 | ||
365 | \b Match a word boundary | |
366 | \B Match a non-(word boundary) | |
367 | \A Match only at beginning of string | |
368 | \Z Match only at end of string, or before newline at the end | |
369 | \z Match only at end of string | |
370 | \G Match only at pos() (e.g. at the end-of-match position | |
371 | of prior m//g) | |
372 | ||
373 | A word boundary (C<\b>) is a spot between two characters | |
374 | that has a C<\w> on one side of it and a C<\W> on the other side | |
375 | of it (in either order), counting the imaginary characters off the | |
376 | beginning and end of the string as matching a C<\W>. (Within | |
377 | character classes C<\b> represents backspace rather than a word | |
378 | boundary, just as it normally does in any double-quoted string.) | |
379 | The C<\A> and C<\Z> are just like "^" and "$", except that they | |
380 | won't match multiple times when the C</m> modifier is used, while | |
381 | "^" and "$" will match at every internal line boundary. To match | |
382 | the actual end of the string and not ignore an optional trailing | |
383 | newline, use C<\z>. | |
384 | X<\b> X<\A> X<\Z> X<\z> X</m> | |
385 | ||
386 | The C<\G> assertion can be used to chain global matches (using | |
387 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. | |
388 | It is also useful when writing C<lex>-like scanners, when you have | |
389 | several patterns that you want to match against consequent substrings | |
390 | of your string, see the previous reference. The actual location | |
391 | where C<\G> will match can also be influenced by using C<pos()> as | |
392 | an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully | |
393 | supported when anchored to the start of the pattern; while it | |
394 | is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some | |
395 | such uses (C</.\G/g>, for example) currently cause problems, and | |
396 | it is recommended that you avoid such usage for now. | |
397 | X<\G> | |
398 | ||
399 | The bracketing construct C<( ... )> creates capture buffers. To | |
400 | refer to the digit'th buffer use \<digit> within the | |
401 | match. Outside the match use "$" instead of "\". (The | |
402 | \<digit> notation works in certain circumstances outside | |
403 | the match. See the warning below about \1 vs $1 for details.) | |
404 | Referring back to another part of the match is called a | |
405 | I<backreference>. | |
406 | X<regex, capture buffer> X<regexp, capture buffer> | |
407 | X<regular expression, capture buffer> X<backreference> | |
408 | ||
409 | There is no limit to the number of captured substrings that you may | |
410 | use. However Perl also uses \10, \11, etc. as aliases for \010, | |
411 | \011, etc. (Recall that 0 means octal, so \011 is the character at | |
412 | number 9 in your coded character set; which would be the 10th character, | |
413 | a horizontal tab under ASCII.) Perl resolves this | |
414 | ambiguity by interpreting \10 as a backreference only if at least 10 | |
415 | left parentheses have opened before it. Likewise \11 is a | |
416 | backreference only if at least 11 left parentheses have opened | |
417 | before it. And so on. \1 through \9 are always interpreted as | |
418 | backreferences. | |
419 | ||
420 | Examples: | |
421 | ||
422 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
423 | ||
424 | if (/(.)\1/) { # find first doubled char | |
425 | print "'$1' is the first doubled character\n"; | |
426 | } | |
427 | ||
428 | if (/Time: (..):(..):(..)/) { # parse out values | |
429 | $hours = $1; | |
430 | $minutes = $2; | |
431 | $seconds = $3; | |
432 | } | |
433 | ||
434 | Several special variables also refer back to portions of the previous | |
435 | match. C<$+> returns whatever the last bracket match matched. | |
436 | C<$&> returns the entire matched string. (At one point C<$0> did | |
437 | also, but now it returns the name of the program.) C<$`> returns | |
438 | everything before the matched string. C<$'> returns everything | |
439 | after the matched string. And C<$^N> contains whatever was matched by | |
440 | the most-recently closed group (submatch). C<$^N> can be used in | |
441 | extended patterns (see below), for example to assign a submatch to a | |
442 | variable. | |
443 | X<$+> X<$^N> X<$&> X<$`> X<$'> | |
444 | ||
445 | The numbered match variables ($1, $2, $3, etc.) and the related punctuation | |
446 | set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped | |
447 | until the end of the enclosing block or until the next successful | |
448 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
449 | X<$+> X<$^N> X<$&> X<$`> X<$'> | |
450 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> | |
451 | ||
452 | ||
453 | B<NOTE>: failed matches in Perl do not reset the match variables, | |
454 | which makes it easier to write code that tests for a series of more | |
455 | specific cases and remembers the best match. | |
456 | ||
457 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or | |
458 | C<$'> anywhere in the program, it has to provide them for every | |
459 | pattern match. This may substantially slow your program. Perl | |
460 | uses the same mechanism to produce $1, $2, etc, so you also pay a | |
461 | price for each pattern that contains capturing parentheses. (To | |
462 | avoid this cost while retaining the grouping behaviour, use the | |
463 | extended regular expression C<(?: ... )> instead.) But if you never | |
464 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
465 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
466 | if you can, but if you can't (and some algorithms really appreciate | |
467 | them), once you've used them once, use them at will, because you've | |
468 | already paid the price. As of 5.005, C<$&> is not so costly as the | |
469 | other two. | |
470 | X<$&> X<$`> X<$'> | |
471 | ||
472 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, | |
473 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
474 | are no backslashed symbols that aren't alphanumeric. So anything | |
475 | that looks like \\, \(, \), \<, \>, \{, or \} is always | |
476 | interpreted as a literal character, not a metacharacter. This was | |
477 | once used in a common idiom to disable or quote the special meanings | |
478 | of regular expression metacharacters in a string that you want to | |
479 | use for a pattern. Simply quote all non-"word" characters: | |
480 | ||
481 | $pattern =~ s/(\W)/\\$1/g; | |
482 | ||
483 | (If C<use locale> is set, then this depends on the current locale.) | |
484 | Today it is more common to use the quotemeta() function or the C<\Q> | |
485 | metaquoting escape sequence to disable all metacharacters' special | |
486 | meanings like this: | |
487 | ||
488 | /$unquoted\Q$quoted\E$unquoted/ | |
489 | ||
490 | Beware that if you put literal backslashes (those not inside | |
491 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
492 | backslash interpolation may lead to confusing results. If you | |
493 | I<need> to use literal backslashes within C<\Q...\E>, | |
494 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
495 | ||
496 | =head2 Extended Patterns | |
497 | ||
498 | Perl also defines a consistent extension syntax for features not | |
499 | found in standard tools like B<awk> and B<lex>. The syntax is a | |
500 | pair of parentheses with a question mark as the first thing within | |
501 | the parentheses. The character after the question mark indicates | |
502 | the extension. | |
503 | ||
504 | The stability of these extensions varies widely. Some have been | |
505 | part of the core language for many years. Others are experimental | |
506 | and may change without warning or be completely removed. Check | |
507 | the documentation on an individual feature to verify its current | |
508 | status. | |
509 | ||
510 | A question mark was chosen for this and for the minimal-matching | |
511 | construct because 1) question marks are rare in older regular | |
512 | expressions, and 2) whenever you see one, you should stop and | |
513 | "question" exactly what is going on. That's psychology... | |
514 | ||
515 | =over 10 | |
516 | ||
517 | =item C<(?#text)> | |
518 | X<(?#)> | |
519 | ||
520 | A comment. The text is ignored. If the C</x> modifier enables | |
521 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes | |
522 | the comment as soon as it sees a C<)>, so there is no way to put a literal | |
523 | C<)> in the comment. | |
524 | ||
525 | =item C<(?imsx-imsx)> | |
526 | X<(?)> | |
527 | ||
528 | One or more embedded pattern-match modifiers, to be turned on (or | |
529 | turned off, if preceded by C<->) for the remainder of the pattern or | |
530 | the remainder of the enclosing pattern group (if any). This is | |
531 | particularly useful for dynamic patterns, such as those read in from a | |
532 | configuration file, read in as an argument, are specified in a table | |
533 | somewhere, etc. Consider the case that some of which want to be case | |
534 | sensitive and some do not. The case insensitive ones need to include | |
535 | merely C<(?i)> at the front of the pattern. For example: | |
536 | ||
537 | $pattern = "foobar"; | |
538 | if ( /$pattern/i ) { } | |
539 | ||
540 | # more flexible: | |
541 | ||
542 | $pattern = "(?i)foobar"; | |
543 | if ( /$pattern/ ) { } | |
544 | ||
545 | These modifiers are restored at the end of the enclosing group. For example, | |
546 | ||
547 | ( (?i) blah ) \s+ \1 | |
548 | ||
549 | will match a repeated (I<including the case>!) word C<blah> in any | |
550 | case, assuming C<x> modifier, and no C<i> modifier outside this | |
551 | group. | |
552 | ||
553 | =item C<(?:pattern)> | |
554 | X<(?:)> | |
555 | ||
556 | =item C<(?imsx-imsx:pattern)> | |
557 | ||
558 | This is for clustering, not capturing; it groups subexpressions like | |
559 | "()", but doesn't make backreferences as "()" does. So | |
560 | ||
561 | @fields = split(/\b(?:a|b|c)\b/) | |
562 | ||
563 | is like | |
564 | ||
565 | @fields = split(/\b(a|b|c)\b/) | |
566 | ||
567 | but doesn't spit out extra fields. It's also cheaper not to capture | |
568 | characters if you don't need to. | |
569 | ||
570 | Any letters between C<?> and C<:> act as flags modifiers as with | |
571 | C<(?imsx-imsx)>. For example, | |
572 | ||
573 | /(?s-i:more.*than).*million/i | |
574 | ||
575 | is equivalent to the more verbose | |
576 | ||
577 | /(?:(?s-i)more.*than).*million/i | |
578 | ||
579 | =item C<(?=pattern)> | |
580 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> | |
581 | ||
582 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> | |
583 | matches a word followed by a tab, without including the tab in C<$&>. | |
584 | ||
585 | =item C<(?!pattern)> | |
586 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> | |
587 | ||
588 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> | |
589 | matches any occurrence of "foo" that isn't followed by "bar". Note | |
590 | however that look-ahead and look-behind are NOT the same thing. You cannot | |
591 | use this for look-behind. | |
592 | ||
593 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> | |
594 | will not do what you want. That's because the C<(?!foo)> is just saying that | |
595 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
596 | match. You would have to do something like C</(?!foo)...bar/> for that. We | |
597 | say "like" because there's the case of your "bar" not having three characters | |
598 | before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>. | |
599 | Sometimes it's still easier just to say: | |
600 | ||
601 | if (/bar/ && $` !~ /foo$/) | |
602 | ||
603 | For look-behind see below. | |
604 | ||
605 | =item C<(?<=pattern)> | |
606 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> | |
607 | ||
608 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> | |
609 | matches a word that follows a tab, without including the tab in C<$&>. | |
610 | Works only for fixed-width look-behind. | |
611 | ||
612 | =item C<(?<!pattern)> | |
613 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> | |
614 | ||
615 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> | |
616 | matches any occurrence of "foo" that does not follow "bar". Works | |
617 | only for fixed-width look-behind. | |
618 | ||
619 | =item C<(?{ code })> | |
620 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> | |
621 | ||
622 | B<WARNING>: This extended regular expression feature is considered | |
623 | highly experimental, and may be changed or deleted without notice. | |
624 | ||
625 | This zero-width assertion evaluates any embedded Perl code. It | |
626 | always succeeds, and its C<code> is not interpolated. Currently, | |
627 | the rules to determine where the C<code> ends are somewhat convoluted. | |
628 | ||
629 | This feature can be used together with the special variable C<$^N> to | |
630 | capture the results of submatches in variables without having to keep | |
631 | track of the number of nested parentheses. For example: | |
632 | ||
633 | $_ = "The brown fox jumps over the lazy dog"; | |
634 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
635 | print "color = $color, animal = $animal\n"; | |
636 | ||
637 | Inside the C<(?{...})> block, C<$_> refers to the string the regular | |
638 | expression is matching against. You can also use C<pos()> to know what is | |
639 | the current position of matching within this string. | |
640 | ||
641 | The C<code> is properly scoped in the following sense: If the assertion | |
642 | is backtracked (compare L<"Backtracking">), all changes introduced after | |
643 | C<local>ization are undone, so that | |
644 | ||
645 | $_ = 'a' x 8; | |
646 | m< | |
647 | (?{ $cnt = 0 }) # Initialize $cnt. | |
648 | ( | |
649 | a | |
650 | (?{ | |
651 | local $cnt = $cnt + 1; # Update $cnt, backtracking-safe. | |
652 | }) | |
653 | )* | |
654 | aaaa | |
655 | (?{ $res = $cnt }) # On success copy to non-localized | |
656 | # location. | |
657 | >x; | |
658 | ||
659 | will set C<$res = 4>. Note that after the match, $cnt returns to the globally | |
660 | introduced value, because the scopes that restrict C<local> operators | |
661 | are unwound. | |
662 | ||
663 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> | |
664 | switch. If I<not> used in this way, the result of evaluation of | |
665 | C<code> is put into the special variable C<$^R>. This happens | |
666 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions | |
667 | inside the same regular expression. | |
668 | ||
669 | The assignment to C<$^R> above is properly localized, so the old | |
670 | value of C<$^R> is restored if the assertion is backtracked; compare | |
671 | L<"Backtracking">. | |
672 | ||
673 | For reasons of security, this construct is forbidden if the regular | |
674 | expression involves run-time interpolation of variables, unless the | |
675 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
676 | variables contain results of C<qr//> operator (see | |
677 | L<perlop/"qr/STRING/imosx">). | |
678 | ||
679 | This restriction is because of the wide-spread and remarkably convenient | |
680 | custom of using run-time determined strings as patterns. For example: | |
681 | ||
682 | $re = <>; | |
683 | chomp $re; | |
684 | $string =~ /$re/; | |
685 | ||
686 | Before Perl knew how to execute interpolated code within a pattern, | |
687 | this operation was completely safe from a security point of view, | |
688 | although it could raise an exception from an illegal pattern. If | |
689 | you turn on the C<use re 'eval'>, though, it is no longer secure, | |
690 | so you should only do so if you are also using taint checking. | |
691 | Better yet, use the carefully constrained evaluation within a Safe | |
692 | compartment. See L<perlsec> for details about both these mechanisms. | |
693 | ||
694 | =item C<(??{ code })> | |
695 | X<(??{})> | |
696 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> | |
697 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> | |
698 | ||
699 | B<WARNING>: This extended regular expression feature is considered | |
700 | highly experimental, and may be changed or deleted without notice. | |
701 | A simplified version of the syntax may be introduced for commonly | |
702 | used idioms. | |
703 | ||
704 | This is a "postponed" regular subexpression. The C<code> is evaluated | |
705 | at run time, at the moment this subexpression may match. The result | |
706 | of evaluation is considered as a regular expression and matched as | |
707 | if it were inserted instead of this construct. | |
708 | ||
709 | The C<code> is not interpolated. As before, the rules to determine | |
710 | where the C<code> ends are currently somewhat convoluted. | |
711 | ||
712 | The following pattern matches a parenthesized group: | |
713 | ||
714 | $re = qr{ | |
715 | \( | |
716 | (?: | |
717 | (?> [^()]+ ) # Non-parens without backtracking | |
718 | | | |
719 | (??{ $re }) # Group with matching parens | |
720 | )* | |
721 | \) | |
722 | }x; | |
723 | ||
724 | =item C<< (?>pattern) >> | |
725 | X<backtrack> X<backtracking> | |
726 | ||
727 | B<WARNING>: This extended regular expression feature is considered | |
728 | highly experimental, and may be changed or deleted without notice. | |
729 | ||
730 | An "independent" subexpression, one which matches the substring | |
731 | that a I<standalone> C<pattern> would match if anchored at the given | |
732 | position, and it matches I<nothing other than this substring>. This | |
733 | construct is useful for optimizations of what would otherwise be | |
734 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). | |
735 | It may also be useful in places where the "grab all you can, and do not | |
736 | give anything back" semantic is desirable. | |
737 | ||
738 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> | |
739 | (anchored at the beginning of string, as above) will match I<all> | |
740 | characters C<a> at the beginning of string, leaving no C<a> for | |
741 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, | |
742 | since the match of the subgroup C<a*> is influenced by the following | |
743 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside | |
744 | C<a*ab> will match fewer characters than a standalone C<a*>, since | |
745 | this makes the tail match. | |
746 | ||
747 | An effect similar to C<< (?>pattern) >> may be achieved by writing | |
748 | C<(?=(pattern))\1>. This matches the same substring as a standalone | |
749 | C<a+>, and the following C<\1> eats the matched string; it therefore | |
750 | makes a zero-length assertion into an analogue of C<< (?>...) >>. | |
751 | (The difference between these two constructs is that the second one | |
752 | uses a capturing group, thus shifting ordinals of backreferences | |
753 | in the rest of a regular expression.) | |
754 | ||
755 | Consider this pattern: | |
756 | ||
757 | m{ \( | |
758 | ( | |
759 | [^()]+ # x+ | |
760 | | | |
761 | \( [^()]* \) | |
762 | )+ | |
763 | \) | |
764 | }x | |
765 | ||
766 | That will efficiently match a nonempty group with matching parentheses | |
767 | two levels deep or less. However, if there is no such group, it | |
768 | will take virtually forever on a long string. That's because there | |
769 | are so many different ways to split a long string into several | |
770 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
771 | to a subpattern of the above pattern. Consider how the pattern | |
772 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
773 | seconds, but that each extra letter doubles this time. This | |
774 | exponential performance will make it appear that your program has | |
775 | hung. However, a tiny change to this pattern | |
776 | ||
777 | m{ \( | |
778 | ( | |
779 | (?> [^()]+ ) # change x+ above to (?> x+ ) | |
780 | | | |
781 | \( [^()]* \) | |
782 | )+ | |
783 | \) | |
784 | }x | |
785 | ||
786 | which uses C<< (?>...) >> matches exactly when the one above does (verifying | |
787 | this yourself would be a productive exercise), but finishes in a fourth | |
788 | the time when used on a similar string with 1000000 C<a>s. Be aware, | |
789 | however, that this pattern currently triggers a warning message under | |
790 | the C<use warnings> pragma or B<-w> switch saying it | |
791 | C<"matches null string many times in regex">. | |
792 | ||
793 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable | |
794 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. | |
795 | This was only 4 times slower on a string with 1000000 C<a>s. | |
796 | ||
797 | The "grab all you can, and do not give anything back" semantic is desirable | |
798 | in many situations where on the first sight a simple C<()*> looks like | |
799 | the correct solution. Suppose we parse text with comments being delimited | |
800 | by C<#> followed by some optional (horizontal) whitespace. Contrary to | |
801 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match | |
802 | the comment delimiter, because it may "give up" some whitespace if | |
803 | the remainder of the pattern can be made to match that way. The correct | |
804 | answer is either one of these: | |
805 | ||
806 | (?>#[ \t]*) | |
807 | #[ \t]*(?![ \t]) | |
808 | ||
809 | For example, to grab non-empty comments into $1, one should use either | |
810 | one of these: | |
811 | ||
812 | / (?> \# [ \t]* ) ( .+ ) /x; | |
813 | / \# [ \t]* ( [^ \t] .* ) /x; | |
814 | ||
815 | Which one you pick depends on which of these expressions better reflects | |
816 | the above specification of comments. | |
817 | ||
818 | =item C<(?(condition)yes-pattern|no-pattern)> | |
819 | X<(?()> | |
820 | ||
821 | =item C<(?(condition)yes-pattern)> | |
822 | ||
823 | B<WARNING>: This extended regular expression feature is considered | |
824 | highly experimental, and may be changed or deleted without notice. | |
825 | ||
826 | Conditional expression. C<(condition)> should be either an integer in | |
827 | parentheses (which is valid if the corresponding pair of parentheses | |
828 | matched), or look-ahead/look-behind/evaluate zero-width assertion. | |
829 | ||
830 | For example: | |
831 | ||
832 | m{ ( \( )? | |
833 | [^()]+ | |
834 | (?(1) \) ) | |
835 | }x | |
836 | ||
837 | matches a chunk of non-parentheses, possibly included in parentheses | |
838 | themselves. | |
839 | ||
840 | =back | |
841 | ||
842 | =head2 Backtracking | |
843 | X<backtrack> X<backtracking> | |
844 | ||
845 | NOTE: This section presents an abstract approximation of regular | |
846 | expression behavior. For a more rigorous (and complicated) view of | |
847 | the rules involved in selecting a match among possible alternatives, | |
848 | see L<Combining pieces together>. | |
849 | ||
850 | A fundamental feature of regular expression matching involves the | |
851 | notion called I<backtracking>, which is currently used (when needed) | |
852 | by all regular expression quantifiers, namely C<*>, C<*?>, C<+>, | |
853 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized | |
854 | internally, but the general principle outlined here is valid. | |
855 | ||
856 | For a regular expression to match, the I<entire> regular expression must | |
857 | match, not just part of it. So if the beginning of a pattern containing a | |
858 | quantifier succeeds in a way that causes later parts in the pattern to | |
859 | fail, the matching engine backs up and recalculates the beginning | |
860 | part--that's why it's called backtracking. | |
861 | ||
862 | Here is an example of backtracking: Let's say you want to find the | |
863 | word following "foo" in the string "Food is on the foo table.": | |
864 | ||
865 | $_ = "Food is on the foo table."; | |
866 | if ( /\b(foo)\s+(\w+)/i ) { | |
867 | print "$2 follows $1.\n"; | |
868 | } | |
869 | ||
870 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
871 | finds a possible match right at the beginning of the string, and loads up | |
872 | $1 with "Foo". However, as soon as the matching engine sees that there's | |
873 | no whitespace following the "Foo" that it had saved in $1, it realizes its | |
874 | mistake and starts over again one character after where it had the | |
875 | tentative match. This time it goes all the way until the next occurrence | |
876 | of "foo". The complete regular expression matches this time, and you get | |
877 | the expected output of "table follows foo." | |
878 | ||
879 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
880 | everything between "foo" and "bar". Initially, you write something | |
881 | like this: | |
882 | ||
883 | $_ = "The food is under the bar in the barn."; | |
884 | if ( /foo(.*)bar/ ) { | |
885 | print "got <$1>\n"; | |
886 | } | |
887 | ||
888 | Which perhaps unexpectedly yields: | |
889 | ||
890 | got <d is under the bar in the > | |
891 | ||
892 | That's because C<.*> was greedy, so you get everything between the | |
893 | I<first> "foo" and the I<last> "bar". Here it's more effective | |
894 | to use minimal matching to make sure you get the text between a "foo" | |
895 | and the first "bar" thereafter. | |
896 | ||
897 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
898 | got <d is under the > | |
899 | ||
900 | Here's another example: let's say you'd like to match a number at the end | |
901 | of a string, and you also want to keep the preceding part of the match. | |
902 | So you write this: | |
903 | ||
904 | $_ = "I have 2 numbers: 53147"; | |
905 | if ( /(.*)(\d*)/ ) { # Wrong! | |
906 | print "Beginning is <$1>, number is <$2>.\n"; | |
907 | } | |
908 | ||
909 | That won't work at all, because C<.*> was greedy and gobbled up the | |
910 | whole string. As C<\d*> can match on an empty string the complete | |
911 | regular expression matched successfully. | |
912 | ||
913 | Beginning is <I have 2 numbers: 53147>, number is <>. | |
914 | ||
915 | Here are some variants, most of which don't work: | |
916 | ||
917 | $_ = "I have 2 numbers: 53147"; | |
918 | @pats = qw{ | |
919 | (.*)(\d*) | |
920 | (.*)(\d+) | |
921 | (.*?)(\d*) | |
922 | (.*?)(\d+) | |
923 | (.*)(\d+)$ | |
924 | (.*?)(\d+)$ | |
925 | (.*)\b(\d+)$ | |
926 | (.*\D)(\d+)$ | |
927 | }; | |
928 | ||
929 | for $pat (@pats) { | |
930 | printf "%-12s ", $pat; | |
931 | if ( /$pat/ ) { | |
932 | print "<$1> <$2>\n"; | |
933 | } else { | |
934 | print "FAIL\n"; | |
935 | } | |
936 | } | |
937 | ||
938 | That will print out: | |
939 | ||
940 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
941 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
942 | (.*?)(\d*) <> <> | |
943 | (.*?)(\d+) <I have > <2> | |
944 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
945 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
946 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
947 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
948 | ||
949 | As you see, this can be a bit tricky. It's important to realize that a | |
950 | regular expression is merely a set of assertions that gives a definition | |
951 | of success. There may be 0, 1, or several different ways that the | |
952 | definition might succeed against a particular string. And if there are | |
953 | multiple ways it might succeed, you need to understand backtracking to | |
954 | know which variety of success you will achieve. | |
955 | ||
956 | When using look-ahead assertions and negations, this can all get even | |
957 | trickier. Imagine you'd like to find a sequence of non-digits not | |
958 | followed by "123". You might try to write that as | |
959 | ||
960 | $_ = "ABC123"; | |
961 | if ( /^\D*(?!123)/ ) { # Wrong! | |
962 | print "Yup, no 123 in $_\n"; | |
963 | } | |
964 | ||
965 | But that isn't going to match; at least, not the way you're hoping. It | |
966 | claims that there is no 123 in the string. Here's a clearer picture of | |
967 | why that pattern matches, contrary to popular expectations: | |
968 | ||
969 | $x = 'ABC123'; | |
970 | $y = 'ABC445'; | |
971 | ||
972 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; | |
973 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; | |
974 | ||
975 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; | |
976 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; | |
977 | ||
978 | This prints | |
979 | ||
980 | 2: got ABC | |
981 | 3: got AB | |
982 | 4: got ABC | |
983 | ||
984 | You might have expected test 3 to fail because it seems to a more | |
985 | general purpose version of test 1. The important difference between | |
986 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
987 | backtracking, whereas test 1 will not. What's happening is | |
988 | that you've asked "Is it true that at the start of $x, following 0 or more | |
989 | non-digits, you have something that's not 123?" If the pattern matcher had | |
990 | let C<\D*> expand to "ABC", this would have caused the whole pattern to | |
991 | fail. | |
992 | ||
993 | The search engine will initially match C<\D*> with "ABC". Then it will | |
994 | try to match C<(?!123> with "123", which fails. But because | |
995 | a quantifier (C<\D*>) has been used in the regular expression, the | |
996 | search engine can backtrack and retry the match differently | |
997 | in the hope of matching the complete regular expression. | |
998 | ||
999 | The pattern really, I<really> wants to succeed, so it uses the | |
1000 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
1001 | time. Now there's indeed something following "AB" that is not | |
1002 | "123". It's "C123", which suffices. | |
1003 | ||
1004 | We can deal with this by using both an assertion and a negation. | |
1005 | We'll say that the first part in $1 must be followed both by a digit | |
1006 | and by something that's not "123". Remember that the look-aheads | |
1007 | are zero-width expressions--they only look, but don't consume any | |
1008 | of the string in their match. So rewriting this way produces what | |
1009 | you'd expect; that is, case 5 will fail, but case 6 succeeds: | |
1010 | ||
1011 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; | |
1012 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; | |
1013 | ||
1014 | 6: got ABC | |
1015 | ||
1016 | In other words, the two zero-width assertions next to each other work as though | |
1017 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> | |
1018 | matches only if you're at the beginning of the line AND the end of the | |
1019 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
1020 | regular expressions always means AND, except when you write an explicit OR | |
1021 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
1022 | although the attempted matches are made at different positions because "a" | |
1023 | is not a zero-width assertion, but a one-width assertion. | |
1024 | ||
1025 | B<WARNING>: particularly complicated regular expressions can take | |
1026 | exponential time to solve because of the immense number of possible | |
1027 | ways they can use backtracking to try match. For example, without | |
1028 | internal optimizations done by the regular expression engine, this will | |
1029 | take a painfully long time to run: | |
1030 | ||
1031 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ | |
1032 | ||
1033 | And if you used C<*>'s in the internal groups instead of limiting them | |
1034 | to 0 through 5 matches, then it would take forever--or until you ran | |
1035 | out of stack space. Moreover, these internal optimizations are not | |
1036 | always applicable. For example, if you put C<{0,5}> instead of C<*> | |
1037 | on the external group, no current optimization is applicable, and the | |
1038 | match takes a long time to finish. | |
1039 | ||
1040 | A powerful tool for optimizing such beasts is what is known as an | |
1041 | "independent group", | |
1042 | which does not backtrack (see L<C<< (?>pattern) >>>). Note also that | |
1043 | zero-length look-ahead/look-behind assertions will not backtrack to make | |
1044 | the tail match, since they are in "logical" context: only | |
1045 | whether they match is considered relevant. For an example | |
1046 | where side-effects of look-ahead I<might> have influenced the | |
1047 | following match, see L<C<< (?>pattern) >>>. | |
1048 | ||
1049 | =head2 Version 8 Regular Expressions | |
1050 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> | |
1051 | ||
1052 | In case you're not familiar with the "regular" Version 8 regex | |
1053 | routines, here are the pattern-matching rules not described above. | |
1054 | ||
1055 | Any single character matches itself, unless it is a I<metacharacter> | |
1056 | with a special meaning described here or above. You can cause | |
1057 | characters that normally function as metacharacters to be interpreted | |
1058 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any | |
1059 | character; "\\" matches a "\"). A series of characters matches that | |
1060 | series of characters in the target string, so the pattern C<blurfl> | |
1061 | would match "blurfl" in the target string. | |
1062 | ||
1063 | You can specify a character class, by enclosing a list of characters | |
1064 | in C<[]>, which will match any one character from the list. If the | |
1065 | first character after the "[" is "^", the class matches any character not | |
1066 | in the list. Within a list, the "-" character specifies a | |
1067 | range, so that C<a-z> represents all characters between "a" and "z", | |
1068 | inclusive. If you want either "-" or "]" itself to be a member of a | |
1069 | class, put it at the start of the list (possibly after a "^"), or | |
1070 | escape it with a backslash. "-" is also taken literally when it is | |
1071 | at the end of the list, just before the closing "]". (The | |
1072 | following all specify the same class of three characters: C<[-az]>, | |
1073 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which | |
1074 | specifies a class containing twenty-six characters, even on EBCDIC | |
1075 | based coded character sets.) Also, if you try to use the character | |
1076 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of | |
1077 | a range, that's not a range, the "-" is understood literally. | |
1078 | ||
1079 | Note also that the whole range idea is rather unportable between | |
1080 | character sets--and even within character sets they may cause results | |
1081 | you probably didn't expect. A sound principle is to use only ranges | |
1082 | that begin from and end at either alphabets of equal case ([a-e], | |
1083 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, | |
1084 | spell out the character sets in full. | |
1085 | ||
1086 | Characters may be specified using a metacharacter syntax much like that | |
1087 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, | |
1088 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string | |
1089 | of octal digits, matches the character whose coded character set value | |
1090 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, | |
1091 | matches the character whose numeric value is I<nn>. The expression \cI<x> | |
1092 | matches the character control-I<x>. Finally, the "." metacharacter | |
1093 | matches any character except "\n" (unless you use C</s>). | |
1094 | ||
1095 | You can specify a series of alternatives for a pattern using "|" to | |
1096 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", | |
1097 | or "foe" in the target string (as would C<f(e|i|o)e>). The | |
1098 | first alternative includes everything from the last pattern delimiter | |
1099 | ("(", "[", or the beginning of the pattern) up to the first "|", and | |
1100 | the last alternative contains everything from the last "|" to the next | |
1101 | pattern delimiter. That's why it's common practice to include | |
1102 | alternatives in parentheses: to minimize confusion about where they | |
1103 | start and end. | |
1104 | ||
1105 | Alternatives are tried from left to right, so the first | |
1106 | alternative found for which the entire expression matches, is the one that | |
1107 | is chosen. This means that alternatives are not necessarily greedy. For | |
1108 | example: when matching C<foo|foot> against "barefoot", only the "foo" | |
1109 | part will match, as that is the first alternative tried, and it successfully | |
1110 | matches the target string. (This might not seem important, but it is | |
1111 | important when you are capturing matched text using parentheses.) | |
1112 | ||
1113 | Also remember that "|" is interpreted as a literal within square brackets, | |
1114 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. | |
1115 | ||
1116 | Within a pattern, you may designate subpatterns for later reference | |
1117 | by enclosing them in parentheses, and you may refer back to the | |
1118 | I<n>th subpattern later in the pattern using the metacharacter | |
1119 | \I<n>. Subpatterns are numbered based on the left to right order | |
1120 | of their opening parenthesis. A backreference matches whatever | |
1121 | actually matched the subpattern in the string being examined, not | |
1122 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will | |
1123 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern | |
1124 | 1 matched "0x", even though the rule C<0|0x> could potentially match | |
1125 | the leading 0 in the second number. | |
1126 | ||
1127 | =head2 Warning on \1 vs $1 | |
1128 | ||
1129 | Some people get too used to writing things like: | |
1130 | ||
1131 | $pattern =~ s/(\W)/\\\1/g; | |
1132 | ||
1133 | This is grandfathered for the RHS of a substitute to avoid shocking the | |
1134 | B<sed> addicts, but it's a dirty habit to get into. That's because in | |
1135 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in | |
1136 | the usual double-quoted string means a control-A. The customary Unix | |
1137 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
1138 | of doing that, you get yourself into trouble if you then add an C</e> | |
1139 | modifier. | |
1140 | ||
1141 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w | |
1142 | ||
1143 | Or if you try to do | |
1144 | ||
1145 | s/(\d+)/\1000/; | |
1146 | ||
1147 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
1148 | C<${1}000>. The operation of interpolation should not be confused | |
1149 | with the operation of matching a backreference. Certainly they mean two | |
1150 | different things on the I<left> side of the C<s///>. | |
1151 | ||
1152 | =head2 Repeated patterns matching zero-length substring | |
1153 | ||
1154 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. | |
1155 | ||
1156 | Regular expressions provide a terse and powerful programming language. As | |
1157 | with most other power tools, power comes together with the ability | |
1158 | to wreak havoc. | |
1159 | ||
1160 | A common abuse of this power stems from the ability to make infinite | |
1161 | loops using regular expressions, with something as innocuous as: | |
1162 | ||
1163 | 'foo' =~ m{ ( o? )* }x; | |
1164 | ||
1165 | The C<o?> can match at the beginning of C<'foo'>, and since the position | |
1166 | in the string is not moved by the match, C<o?> would match again and again | |
1167 | because of the C<*> modifier. Another common way to create a similar cycle | |
1168 | is with the looping modifier C<//g>: | |
1169 | ||
1170 | @matches = ( 'foo' =~ m{ o? }xg ); | |
1171 | ||
1172 | or | |
1173 | ||
1174 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
1175 | ||
1176 | or the loop implied by split(). | |
1177 | ||
1178 | However, long experience has shown that many programming tasks may | |
1179 | be significantly simplified by using repeated subexpressions that | |
1180 | may match zero-length substrings. Here's a simple example being: | |
1181 | ||
1182 | @chars = split //, $string; # // is not magic in split | |
1183 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / | |
1184 | ||
1185 | Thus Perl allows such constructs, by I<forcefully breaking | |
1186 | the infinite loop>. The rules for this are different for lower-level | |
1187 | loops given by the greedy modifiers C<*+{}>, and for higher-level | |
1188 | ones like the C</g> modifier or split() operator. | |
1189 | ||
1190 | The lower-level loops are I<interrupted> (that is, the loop is | |
1191 | broken) when Perl detects that a repeated expression matched a | |
1192 | zero-length substring. Thus | |
1193 | ||
1194 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
1195 | ||
1196 | is made equivalent to | |
1197 | ||
1198 | m{ (?: NON_ZERO_LENGTH )* | |
1199 | | | |
1200 | (?: ZERO_LENGTH )? | |
1201 | }x; | |
1202 | ||
1203 | The higher level-loops preserve an additional state between iterations: | |
1204 | whether the last match was zero-length. To break the loop, the following | |
1205 | match after a zero-length match is prohibited to have a length of zero. | |
1206 | This prohibition interacts with backtracking (see L<"Backtracking">), | |
1207 | and so the I<second best> match is chosen if the I<best> match is of | |
1208 | zero length. | |
1209 | ||
1210 | For example: | |
1211 | ||
1212 | $_ = 'bar'; | |
1213 | s/\w??/<$&>/g; | |
1214 | ||
1215 | results in C<< <><b><><a><><r><> >>. At each position of the string the best | |
1216 | match given by non-greedy C<??> is the zero-length match, and the I<second | |
1217 | best> match is what is matched by C<\w>. Thus zero-length matches | |
1218 | alternate with one-character-long matches. | |
1219 | ||
1220 | Similarly, for repeated C<m/()/g> the second-best match is the match at the | |
1221 | position one notch further in the string. | |
1222 | ||
1223 | The additional state of being I<matched with zero-length> is associated with | |
1224 | the matched string, and is reset by each assignment to pos(). | |
1225 | Zero-length matches at the end of the previous match are ignored | |
1226 | during C<split>. | |
1227 | ||
1228 | =head2 Combining pieces together | |
1229 | ||
1230 | Each of the elementary pieces of regular expressions which were described | |
1231 | before (such as C<ab> or C<\Z>) could match at most one substring | |
1232 | at the given position of the input string. However, in a typical regular | |
1233 | expression these elementary pieces are combined into more complicated | |
1234 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc | |
1235 | (in these examples C<S> and C<T> are regular subexpressions). | |
1236 | ||
1237 | Such combinations can include alternatives, leading to a problem of choice: | |
1238 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
1239 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
1240 | actually matched is the concept of backtracking (see L<"Backtracking">). | |
1241 | However, this description is too low-level and makes you think | |
1242 | in terms of a particular implementation. | |
1243 | ||
1244 | Another description starts with notions of "better"/"worse". All the | |
1245 | substrings which may be matched by the given regular expression can be | |
1246 | sorted from the "best" match to the "worst" match, and it is the "best" | |
1247 | match which is chosen. This substitutes the question of "what is chosen?" | |
1248 | by the question of "which matches are better, and which are worse?". | |
1249 | ||
1250 | Again, for elementary pieces there is no such question, since at most | |
1251 | one match at a given position is possible. This section describes the | |
1252 | notion of better/worse for combining operators. In the description | |
1253 | below C<S> and C<T> are regular subexpressions. | |
1254 | ||
1255 | =over 4 | |
1256 | ||
1257 | =item C<ST> | |
1258 | ||
1259 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are | |
1260 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings | |
1261 | which can be matched by C<T>. | |
1262 | ||
1263 | If C<A> is better match for C<S> than C<A'>, C<AB> is a better | |
1264 | match than C<A'B'>. | |
1265 | ||
1266 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if | |
1267 | C<B> is better match for C<T> than C<B'>. | |
1268 | ||
1269 | =item C<S|T> | |
1270 | ||
1271 | When C<S> can match, it is a better match than when only C<T> can match. | |
1272 | ||
1273 | Ordering of two matches for C<S> is the same as for C<S>. Similar for | |
1274 | two matches for C<T>. | |
1275 | ||
1276 | =item C<S{REPEAT_COUNT}> | |
1277 | ||
1278 | Matches as C<SSS...S> (repeated as many times as necessary). | |
1279 | ||
1280 | =item C<S{min,max}> | |
1281 | ||
1282 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
1283 | ||
1284 | =item C<S{min,max}?> | |
1285 | ||
1286 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
1287 | ||
1288 | =item C<S?>, C<S*>, C<S+> | |
1289 | ||
1290 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
1291 | ||
1292 | =item C<S??>, C<S*?>, C<S+?> | |
1293 | ||
1294 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
1295 | ||
1296 | =item C<< (?>S) >> | |
1297 | ||
1298 | Matches the best match for C<S> and only that. | |
1299 | ||
1300 | =item C<(?=S)>, C<(?<=S)> | |
1301 | ||
1302 | Only the best match for C<S> is considered. (This is important only if | |
1303 | C<S> has capturing parentheses, and backreferences are used somewhere | |
1304 | else in the whole regular expression.) | |
1305 | ||
1306 | =item C<(?!S)>, C<(?<!S)> | |
1307 | ||
1308 | For this grouping operator there is no need to describe the ordering, since | |
1309 | only whether or not C<S> can match is important. | |
1310 | ||
1311 | =item C<(??{ EXPR })> | |
1312 | ||
1313 | The ordering is the same as for the regular expression which is | |
1314 | the result of EXPR. | |
1315 | ||
1316 | =item C<(?(condition)yes-pattern|no-pattern)> | |
1317 | ||
1318 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is | |
1319 | already determined. The ordering of the matches is the same as for the | |
1320 | chosen subexpression. | |
1321 | ||
1322 | =back | |
1323 | ||
1324 | The above recipes describe the ordering of matches I<at a given position>. | |
1325 | One more rule is needed to understand how a match is determined for the | |
1326 | whole regular expression: a match at an earlier position is always better | |
1327 | than a match at a later position. | |
1328 | ||
1329 | =head2 Creating custom RE engines | |
1330 | ||
1331 | Overloaded constants (see L<overload>) provide a simple way to extend | |
1332 | the functionality of the RE engine. | |
1333 | ||
1334 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
1335 | matches at boundary between whitespace characters and non-whitespace | |
1336 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly | |
1337 | at these positions, so we want to have each C<\Y|> in the place of the | |
1338 | more complicated version. We can create a module C<customre> to do | |
1339 | this: | |
1340 | ||
1341 | package customre; | |
1342 | use overload; | |
1343 | ||
1344 | sub import { | |
1345 | shift; | |
1346 | die "No argument to customre::import allowed" if @_; | |
1347 | overload::constant 'qr' => \&convert; | |
1348 | } | |
1349 | ||
1350 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
1351 | ||
1352 | # We must also take care of not escaping the legitimate \\Y| | |
1353 | # sequence, hence the presence of '\\' in the conversion rules. | |
1354 | my %rules = ( '\\' => '\\\\', | |
1355 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); | |
1356 | sub convert { | |
1357 | my $re = shift; | |
1358 | $re =~ s{ | |
1359 | \\ ( \\ | Y . ) | |
1360 | } | |
1361 | { $rules{$1} or invalid($re,$1) }sgex; | |
1362 | return $re; | |
1363 | } | |
1364 | ||
1365 | Now C<use customre> enables the new escape in constant regular | |
1366 | expressions, i.e., those without any runtime variable interpolations. | |
1367 | As documented in L<overload>, this conversion will work only over | |
1368 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
1369 | part of this regular expression needs to be converted explicitly | |
1370 | (but only if the special meaning of C<\Y|> should be enabled inside $re): | |
1371 | ||
1372 | use customre; | |
1373 | $re = <>; | |
1374 | chomp $re; | |
1375 | $re = customre::convert $re; | |
1376 | /\Y|$re\Y|/; | |
1377 | ||
1378 | =head1 BUGS | |
1379 | ||
1380 | This document varies from difficult to understand to completely | |
1381 | and utterly opaque. The wandering prose riddled with jargon is | |
1382 | hard to fathom in several places. | |
1383 | ||
1384 | This document needs a rewrite that separates the tutorial content | |
1385 | from the reference content. | |
1386 | ||
1387 | =head1 SEE ALSO | |
1388 | ||
1389 | L<perlrequick>. | |
1390 | ||
1391 | L<perlretut>. | |
1392 | ||
1393 | L<perlop/"Regexp Quote-Like Operators">. | |
1394 | ||
1395 | L<perlop/"Gory details of parsing quoted constructs">. | |
1396 | ||
1397 | L<perlfaq6>. | |
1398 | ||
1399 | L<perlfunc/pos>. | |
1400 | ||
1401 | L<perllocale>. | |
1402 | ||
1403 | L<perlebcdic>. | |
1404 | ||
1405 | I<Mastering Regular Expressions> by Jeffrey Friedl, published | |
1406 | by O'Reilly and Associates. |