Commit | Line | Data |
---|---|---|
a0d0e21e | 1 | =head1 NAME |
d74e8afc | 2 | X<regular expression> X<regex> X<regexp> |
a0d0e21e LW |
3 | |
4 | perlre - Perl regular expressions | |
5 | ||
6 | =head1 DESCRIPTION | |
7 | ||
5d458dd8 | 8 | This page describes the syntax of regular expressions in Perl. |
91e0c79e | 9 | |
cc46d5f2 | 10 | If you haven't used regular expressions before, a quick-start |
91e0c79e MJD |
11 | introduction is available in L<perlrequick>, and a longer tutorial |
12 | introduction is available in L<perlretut>. | |
13 | ||
14 | For reference on how regular expressions are used in matching | |
15 | operations, plus various examples of the same, see discussions of | |
16 | C<m//>, C<s///>, C<qr//> and C<??> in L<perlop/"Regexp Quote-Like | |
17 | Operators">. | |
cb1a09d0 | 18 | |
67cdf558 KW |
19 | New in v5.22, L<C<use re 'strict'>|re/'strict' mode> applies stricter |
20 | rules than otherwise when compiling regular expression patterns. It can | |
21 | find things that, while legal, may not be what you intended. | |
0d017f4d WL |
22 | |
23 | =head2 Modifiers | |
24 | ||
4cb6b395 KW |
25 | =head3 Overview |
26 | ||
19799a22 | 27 | Matching operations can have various modifiers. Modifiers |
5a964f20 | 28 | that relate to the interpretation of the regular expression inside |
19799a22 | 29 | are listed below. Modifiers that alter the way a regular expression |
5d458dd8 | 30 | is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and |
1e66bd83 | 31 | L<perlop/"Gory details of parsing quoted constructs">. |
a0d0e21e | 32 | |
55497cff | 33 | =over 4 |
34 | ||
54310121 | 35 | =item m |
d74e8afc | 36 | X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline> |
55497cff | 37 | |
38 | Treat string as multiple lines. That is, change "^" and "$" from matching | |
1ca12bda RS |
39 | the start of the string's first line and the end of its last line to |
40 | matching the start and end of each line within the string. | |
55497cff | 41 | |
54310121 | 42 | =item s |
d74e8afc ITB |
43 | X</s> X<regex, single-line> X<regexp, single-line> |
44 | X<regular expression, single-line> | |
55497cff | 45 | |
46 | Treat string as single line. That is, change "." to match any character | |
19799a22 | 47 | whatsoever, even a newline, which normally it would not match. |
55497cff | 48 | |
34d67d80 | 49 | Used together, as C</ms>, they let the "." match any character whatsoever, |
fb55449c | 50 | while still allowing "^" and "$" to match, respectively, just after |
19799a22 | 51 | and just before newlines within the string. |
7b8d334a | 52 | |
87e95b7f YO |
53 | =item i |
54 | X</i> X<regex, case-insensitive> X<regexp, case-insensitive> | |
55 | X<regular expression, case-insensitive> | |
56 | ||
57 | Do case-insensitive pattern matching. | |
58 | ||
5027a30b KW |
59 | If locale matching rules are in effect, the case map is taken from the |
60 | current | |
17580e7a | 61 | locale for code points less than 255, and from Unicode rules for larger |
ed7efc79 KW |
62 | code points. However, matches that would cross the Unicode |
63 | rules/non-Unicode rules boundary (ords 255/256) will not succeed. See | |
64 | L<perllocale>. | |
65 | ||
66 | There are a number of Unicode characters that match multiple characters | |
67 | under C</i>. For example, C<LATIN SMALL LIGATURE FI> | |
68 | should match the sequence C<fi>. Perl is not | |
69 | currently able to do this when the multiple characters are in the pattern and | |
70 | are split between groupings, or when one or more are quantified. Thus | |
71 | ||
72 | "\N{LATIN SMALL LIGATURE FI}" =~ /fi/i; # Matches | |
73 | "\N{LATIN SMALL LIGATURE FI}" =~ /[fi][fi]/i; # Doesn't match! | |
74 | "\N{LATIN SMALL LIGATURE FI}" =~ /fi*/i; # Doesn't match! | |
75 | ||
76 | # The below doesn't match, and it isn't clear what $1 and $2 would | |
77 | # be even if it did!! | |
78 | "\N{LATIN SMALL LIGATURE FI}" =~ /(f)(i)/i; # Doesn't match! | |
79 | ||
9d53c457 KW |
80 | Perl doesn't match multiple characters in a bracketed |
81 | character class unless the character that maps to them is explicitly | |
82 | mentioned, and it doesn't match them at all if the character class is | |
83 | inverted, which otherwise could be highly confusing. See | |
84 | L<perlrecharclass/Bracketed Character Classes>, and | |
1f59b283 KW |
85 | L<perlrecharclass/Negation>. |
86 | ||
54310121 | 87 | =item x |
d74e8afc | 88 | X</x> |
55497cff | 89 | |
90 | Extend your pattern's legibility by permitting whitespace and comments. | |
ed7efc79 | 91 | Details in L</"/x"> |
55497cff | 92 | |
87e95b7f YO |
93 | =item p |
94 | X</p> X<regex, preserve> X<regexp, preserve> | |
95 | ||
632a1772 | 96 | Preserve the string matched such that ${^PREMATCH}, ${^MATCH}, and |
87e95b7f YO |
97 | ${^POSTMATCH} are available for use after matching. |
98 | ||
13b0f67d DM |
99 | In Perl 5.20 and higher this is ignored. Due to a new copy-on-write |
100 | mechanism, ${^PREMATCH}, ${^MATCH}, and ${^POSTMATCH} will be available | |
101 | after the match regardless of the modifier. | |
102 | ||
b6fa137b FC |
103 | =item a, d, l and u |
104 | X</a> X</d> X</l> X</u> | |
105 | ||
850b7ec9 | 106 | These modifiers, all new in 5.14, affect which character-set rules |
516074bb | 107 | (Unicode, etc.) are used, as described below in |
ed7efc79 | 108 | L</Character set modifiers>. |
b6fa137b | 109 | |
33be4c61 MH |
110 | =item n |
111 | X</n> X<regex, non-capture> X<regexp, non-capture> | |
112 | X<regular expression, non-capture> | |
113 | ||
114 | Prevent the grouping metacharacters C<()> from capturing. This modifier, | |
115 | new in 5.22, will stop C<$1>, C<$2>, etc... from being filled in. | |
116 | ||
117 | "hello" =~ /(hi|hello)/; # $1 is "hello" | |
118 | "hello" =~ /(hi|hello)/n; # $1 is undef | |
119 | ||
25941dca | 120 | This is equivalent to putting C<?:> at the beginning of every capturing group: |
33be4c61 MH |
121 | |
122 | "hello" =~ /(?:hi|hello)/; # $1 is undef | |
123 | ||
124 | C</n> can be negated on a per-group basis. Alternatively, named captures | |
125 | may still be used. | |
126 | ||
127 | "hello" =~ /(?-n:(hi|hello))/n; # $1 is "hello" | |
bdafd784 KW |
128 | "hello" =~ /(?<greet>hi|hello)/n; # $1 is "hello", $+{greet} is |
129 | # "hello" | |
33be4c61 | 130 | |
7cf040c1 RS |
131 | =item Other Modifiers |
132 | ||
133 | There are a number of flags that can be found at the end of regular | |
134 | expression constructs that are I<not> generic regular expression flags, but | |
135 | apply to the operation being performed, like matching or substitution (C<m//> | |
136 | or C<s///> respectively). | |
137 | ||
138 | Flags described further in | |
139 | L<perlretut/"Using regular expressions in Perl"> are: | |
140 | ||
141 | c - keep the current position during repeated matching | |
142 | g - globally match the pattern repeatedly in the string | |
143 | ||
144 | Substitution-specific modifiers described in | |
145 | ||
33be4c61 | 146 | L<perlop/"s/PATTERN/REPLACEMENT/msixpodualngcer"> are: |
171e7319 | 147 | |
7cf040c1 RS |
148 | e - evaluate the right-hand side as an expression |
149 | ee - evaluate the right side as a string then eval the result | |
150 | o - pretend to optimize your code, but actually introduce bugs | |
151 | r - perform non-destructive substitution and return the new value | |
171e7319 | 152 | |
55497cff | 153 | =back |
a0d0e21e | 154 | |
516074bb KW |
155 | Regular expression modifiers are usually written in documentation |
156 | as e.g., "the C</x> modifier", even though the delimiter | |
4cb6b395 | 157 | in question might not really be a slash. The modifiers C</imnsxadlup> |
ab7bb42d | 158 | may also be embedded within the regular expression itself using |
ed7efc79 | 159 | the C<(?...)> construct, see L</Extended Patterns> below. |
b6fa137b | 160 | |
4cb6b395 KW |
161 | =head3 Details on some modifiers |
162 | ||
163 | Some of the modifiers require more explanation than given in the | |
164 | L</Overview> above. | |
165 | ||
166 | =head4 /x | |
ed7efc79 | 167 | |
b6fa137b | 168 | C</x> tells |
7b059540 | 169 | the regular expression parser to ignore most whitespace that is neither |
7c688e65 KW |
170 | backslashed nor within a bracketed character class. You can use this to |
171 | break up your regular expression into (slightly) more readable parts. | |
172 | Also, the C<#> character is treated as a metacharacter introducing a | |
173 | comment that runs up to the pattern's closing delimiter, or to the end | |
174 | of the current line if the pattern extends onto the next line. Hence, | |
175 | this is very much like an ordinary Perl code comment. (You can include | |
176 | the closing delimiter within the comment only if you precede it with a | |
177 | backslash, so be careful!) | |
178 | ||
179 | Use of C</x> means that if you want real | |
180 | whitespace or C<#> characters in the pattern (outside a bracketed character | |
181 | class, which is unaffected by C</x>), then you'll either have to | |
7b059540 | 182 | escape them (using backslashes or C<\Q...\E>) or encode them using octal, |
7c688e65 | 183 | hex, or C<\N{}> escapes. |
8be3c4ca KW |
184 | It is ineffective to try to continue a comment onto the next line by |
185 | escaping the C<\n> with a backslash or C<\Q>. | |
7c688e65 KW |
186 | |
187 | You can use L</(?#text)> to create a comment that ends earlier than the | |
188 | end of the current line, but C<text> also can't contain the closing | |
189 | delimiter unless escaped with a backslash. | |
190 | ||
191 | Taken together, these features go a long way towards | |
192 | making Perl's regular expressions more readable. Here's an example: | |
193 | ||
194 | # Delete (most) C comments. | |
195 | $program =~ s { | |
196 | /\* # Match the opening delimiter. | |
197 | .*? # Match a minimal number of characters. | |
198 | \*/ # Match the closing delimiter. | |
199 | } []gsx; | |
200 | ||
201 | Note that anything inside | |
7651b971 | 202 | a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect |
0b928c2f | 203 | space interpretation within a single multi-character construct. For |
7651b971 | 204 | example in C<\x{...}>, regardless of the C</x> modifier, there can be no |
9bb1f947 | 205 | spaces. Same for a L<quantifier|/Quantifiers> such as C<{3}> or |
000947ad KW |
206 | C<{5,}>. Similarly, C<(?:...)> can't have a space between the C<(>, |
207 | C<?>, and C<:>. Within any delimiters for such a | |
f9e949fd KW |
208 | construct, allowed spaces are not affected by C</x>, and depend on the |
209 | construct. For example, C<\x{...}> can't have spaces because hexadecimal | |
210 | numbers don't have spaces in them. But, Unicode properties can have spaces, so | |
0b928c2f | 211 | in C<\p{...}> there can be spaces that follow the Unicode rules, for which see |
9bb1f947 | 212 | L<perluniprops/Properties accessible through \p{} and \P{}>. |
d74e8afc | 213 | X</x> |
a0d0e21e | 214 | |
8373491a KW |
215 | The set of characters that are deemed whitespace are those that Unicode |
216 | calls "Pattern White Space", namely: | |
217 | ||
218 | U+0009 CHARACTER TABULATION | |
219 | U+000A LINE FEED | |
220 | U+000B LINE TABULATION | |
221 | U+000C FORM FEED | |
222 | U+000D CARRIAGE RETURN | |
223 | U+0020 SPACE | |
224 | U+0085 NEXT LINE | |
225 | U+200E LEFT-TO-RIGHT MARK | |
226 | U+200F RIGHT-TO-LEFT MARK | |
227 | U+2028 LINE SEPARATOR | |
228 | U+2029 PARAGRAPH SEPARATOR | |
229 | ||
4cb6b395 | 230 | =head4 Character set modifiers |
ed7efc79 KW |
231 | |
232 | C</d>, C</u>, C</a>, and C</l>, available starting in 5.14, are called | |
850b7ec9 | 233 | the character set modifiers; they affect the character set rules |
ed7efc79 KW |
234 | used for the regular expression. |
235 | ||
808432af KW |
236 | The C</d>, C</u>, and C</l> modifiers are not likely to be of much use |
237 | to you, and so you need not worry about them very much. They exist for | |
238 | Perl's internal use, so that complex regular expression data structures | |
239 | can be automatically serialized and later exactly reconstituted, | |
240 | including all their nuances. But, since Perl can't keep a secret, and | |
241 | there may be rare instances where they are useful, they are documented | |
242 | here. | |
ed7efc79 | 243 | |
808432af KW |
244 | The C</a> modifier, on the other hand, may be useful. Its purpose is to |
245 | allow code that is to work mostly on ASCII data to not have to concern | |
246 | itself with Unicode. | |
ca9560b2 | 247 | |
808432af KW |
248 | Briefly, C</l> sets the character set to that of whatever B<L>ocale is in |
249 | effect at the time of the execution of the pattern match. | |
ca9560b2 | 250 | |
808432af KW |
251 | C</u> sets the character set to B<U>nicode. |
252 | ||
253 | C</a> also sets the character set to Unicode, BUT adds several | |
254 | restrictions for B<A>SCII-safe matching. | |
255 | ||
256 | C</d> is the old, problematic, pre-5.14 B<D>efault character set | |
257 | behavior. Its only use is to force that old behavior. | |
258 | ||
259 | At any given time, exactly one of these modifiers is in effect. Their | |
260 | existence allows Perl to keep the originally compiled behavior of a | |
261 | regular expression, regardless of what rules are in effect when it is | |
262 | actually executed. And if it is interpolated into a larger regex, the | |
263 | original's rules continue to apply to it, and only it. | |
264 | ||
265 | The C</l> and C</u> modifiers are automatically selected for | |
266 | regular expressions compiled within the scope of various pragmas, | |
267 | and we recommend that in general, you use those pragmas instead of | |
268 | specifying these modifiers explicitly. For one thing, the modifiers | |
269 | affect only pattern matching, and do not extend to even any replacement | |
270 | done, whereas using the pragmas give consistent results for all | |
271 | appropriate operations within their scopes. For example, | |
272 | ||
273 | s/foo/\Ubar/il | |
274 | ||
275 | will match "foo" using the locale's rules for case-insensitive matching, | |
276 | but the C</l> does not affect how the C<\U> operates. Most likely you | |
277 | want both of them to use locale rules. To do this, instead compile the | |
278 | regular expression within the scope of C<use locale>. This both | |
ce4fe27b KW |
279 | implicitly adds the C</l>, and applies locale rules to the C<\U>. The |
280 | lesson is to C<use locale>, and not C</l> explicitly. | |
808432af KW |
281 | |
282 | Similarly, it would be better to use C<use feature 'unicode_strings'> | |
283 | instead of, | |
284 | ||
285 | s/foo/\Lbar/iu | |
286 | ||
287 | to get Unicode rules, as the C<\L> in the former (but not necessarily | |
288 | the latter) would also use Unicode rules. | |
289 | ||
290 | More detail on each of the modifiers follows. Most likely you don't | |
291 | need to know this detail for C</l>, C</u>, and C</d>, and can skip ahead | |
292 | to L<E<sol>a|/E<sol>a (and E<sol>aa)>. | |
ca9560b2 | 293 | |
ed7efc79 KW |
294 | =head4 /l |
295 | ||
296 | means to use the current locale's rules (see L<perllocale>) when pattern | |
297 | matching. For example, C<\w> will match the "word" characters of that | |
298 | locale, and C<"/i"> case-insensitive matching will match according to | |
299 | the locale's case folding rules. The locale used will be the one in | |
300 | effect at the time of execution of the pattern match. This may not be | |
301 | the same as the compilation-time locale, and can differ from one match | |
302 | to another if there is an intervening call of the | |
b6fa137b | 303 | L<setlocale() function|perllocale/The setlocale function>. |
ed7efc79 | 304 | |
31f05a37 KW |
305 | The only non-single-byte locale Perl supports is (starting in v5.20) |
306 | UTF-8. This means that code points above 255 are treated as Unicode no | |
307 | matter what locale is in effect (since UTF-8 implies Unicode). | |
308 | ||
ed7efc79 | 309 | Under Unicode rules, there are a few case-insensitive matches that cross |
31f05a37 KW |
310 | the 255/256 boundary. Except for UTF-8 locales in Perls v5.20 and |
311 | later, these are disallowed under C</l>. For example, 0xFF (on ASCII | |
312 | platforms) does not caselessly match the character at 0x178, C<LATIN | |
313 | CAPITAL LETTER Y WITH DIAERESIS>, because 0xFF may not be C<LATIN SMALL | |
314 | LETTER Y WITH DIAERESIS> in the current locale, and Perl has no way of | |
315 | knowing if that character even exists in the locale, much less what code | |
316 | point it is. | |
317 | ||
318 | In a UTF-8 locale in v5.20 and later, the only visible difference | |
319 | between locale and non-locale in regular expressions should be tainting | |
320 | (see L<perlsec>). | |
ed7efc79 KW |
321 | |
322 | This modifier may be specified to be the default by C<use locale>, but | |
323 | see L</Which character set modifier is in effect?>. | |
b6fa137b FC |
324 | X</l> |
325 | ||
ed7efc79 KW |
326 | =head4 /u |
327 | ||
328 | means to use Unicode rules when pattern matching. On ASCII platforms, | |
329 | this means that the code points between 128 and 255 take on their | |
808432af KW |
330 | Latin-1 (ISO-8859-1) meanings (which are the same as Unicode's). |
331 | (Otherwise Perl considers their meanings to be undefined.) Thus, | |
332 | under this modifier, the ASCII platform effectively becomes a Unicode | |
333 | platform; and hence, for example, C<\w> will match any of the more than | |
334 | 100_000 word characters in Unicode. | |
ed7efc79 KW |
335 | |
336 | Unlike most locales, which are specific to a language and country pair, | |
516074bb KW |
337 | Unicode classifies all the characters that are letters I<somewhere> in |
338 | the world as | |
ed7efc79 KW |
339 | C<\w>. For example, your locale might not think that C<LATIN SMALL |
340 | LETTER ETH> is a letter (unless you happen to speak Icelandic), but | |
341 | Unicode does. Similarly, all the characters that are decimal digits | |
342 | somewhere in the world will match C<\d>; this is hundreds, not 10, | |
343 | possible matches. And some of those digits look like some of the 10 | |
344 | ASCII digits, but mean a different number, so a human could easily think | |
345 | a number is a different quantity than it really is. For example, | |
346 | C<BENGALI DIGIT FOUR> (U+09EA) looks very much like an | |
347 | C<ASCII DIGIT EIGHT> (U+0038). And, C<\d+>, may match strings of digits | |
348 | that are a mixture from different writing systems, creating a security | |
67592e11 | 349 | issue. L<Unicode::UCD/num()> can be used to sort |
516074bb KW |
350 | this out. Or the C</a> modifier can be used to force C<\d> to match |
351 | just the ASCII 0 through 9. | |
ed7efc79 | 352 | |
516074bb KW |
353 | Also, under this modifier, case-insensitive matching works on the full |
354 | set of Unicode | |
ed7efc79 KW |
355 | characters. The C<KELVIN SIGN>, for example matches the letters "k" and |
356 | "K"; and C<LATIN SMALL LIGATURE FF> matches the sequence "ff", which, | |
357 | if you're not prepared, might make it look like a hexadecimal constant, | |
358 | presenting another potential security issue. See | |
359 | L<http://unicode.org/reports/tr36> for a detailed discussion of Unicode | |
360 | security issues. | |
361 | ||
ed7efc79 | 362 | This modifier may be specified to be the default by C<use feature |
66cbab2c KW |
363 | 'unicode_strings>, C<use locale ':not_characters'>, or |
364 | C<L<use 5.012|perlfunc/use VERSION>> (or higher), | |
808432af | 365 | but see L</Which character set modifier is in effect?>. |
b6fa137b FC |
366 | X</u> |
367 | ||
ed7efc79 KW |
368 | =head4 /d |
369 | ||
370 | This modifier means to use the "Default" native rules of the platform | |
371 | except when there is cause to use Unicode rules instead, as follows: | |
372 | ||
373 | =over 4 | |
374 | ||
375 | =item 1 | |
376 | ||
377 | the target string is encoded in UTF-8; or | |
378 | ||
379 | =item 2 | |
380 | ||
381 | the pattern is encoded in UTF-8; or | |
382 | ||
383 | =item 3 | |
384 | ||
385 | the pattern explicitly mentions a code point that is above 255 (say by | |
386 | C<\x{100}>); or | |
387 | ||
388 | =item 4 | |
b6fa137b | 389 | |
ed7efc79 KW |
390 | the pattern uses a Unicode name (C<\N{...}>); or |
391 | ||
392 | =item 5 | |
393 | ||
ce4fe27b | 394 | the pattern uses a Unicode property (C<\p{...}> or C<\P{...}>); or |
9d1a5160 KW |
395 | |
396 | =item 6 | |
397 | ||
64935bc6 KW |
398 | the pattern uses a Unicode break (C<\b{...}> or C<\B{...}>); or |
399 | ||
400 | =item 7 | |
401 | ||
9d1a5160 | 402 | the pattern uses L</C<(?[ ])>> |
ed7efc79 KW |
403 | |
404 | =back | |
405 | ||
406 | Another mnemonic for this modifier is "Depends", as the rules actually | |
407 | used depend on various things, and as a result you can get unexpected | |
808432af KW |
408 | results. See L<perlunicode/The "Unicode Bug">. The Unicode Bug has |
409 | become rather infamous, leading to yet another (printable) name for this | |
410 | modifier, "Dodgy". | |
ed7efc79 | 411 | |
4b9734bf KW |
412 | Unless the pattern or string are encoded in UTF-8, only ASCII characters |
413 | can match positively. | |
ed7efc79 KW |
414 | |
415 | Here are some examples of how that works on an ASCII platform: | |
416 | ||
417 | $str = "\xDF"; # $str is not in UTF-8 format. | |
418 | $str =~ /^\w/; # No match, as $str isn't in UTF-8 format. | |
419 | $str .= "\x{0e0b}"; # Now $str is in UTF-8 format. | |
420 | $str =~ /^\w/; # Match! $str is now in UTF-8 format. | |
421 | chop $str; | |
422 | $str =~ /^\w/; # Still a match! $str remains in UTF-8 format. | |
423 | ||
808432af KW |
424 | This modifier is automatically selected by default when none of the |
425 | others are, so yet another name for it is "Default". | |
426 | ||
427 | Because of the unexpected behaviors associated with this modifier, you | |
428 | probably should only use it to maintain weird backward compatibilities. | |
429 | ||
430 | =head4 /a (and /aa) | |
431 | ||
432 | This modifier stands for ASCII-restrict (or ASCII-safe). This modifier, | |
433 | unlike the others, may be doubled-up to increase its effect. | |
434 | ||
435 | When it appears singly, it causes the sequences C<\d>, C<\s>, C<\w>, and | |
436 | the Posix character classes to match only in the ASCII range. They thus | |
437 | revert to their pre-5.6, pre-Unicode meanings. Under C</a>, C<\d> | |
438 | always means precisely the digits C<"0"> to C<"9">; C<\s> means the five | |
779cf272 KW |
439 | characters C<[ \f\n\r\t]>, and starting in Perl v5.18, the vertical tab; |
440 | C<\w> means the 63 characters | |
808432af KW |
441 | C<[A-Za-z0-9_]>; and likewise, all the Posix classes such as |
442 | C<[[:print:]]> match only the appropriate ASCII-range characters. | |
443 | ||
444 | This modifier is useful for people who only incidentally use Unicode, | |
445 | and who do not wish to be burdened with its complexities and security | |
446 | concerns. | |
447 | ||
448 | With C</a>, one can write C<\d> with confidence that it will only match | |
449 | ASCII characters, and should the need arise to match beyond ASCII, you | |
450 | can instead use C<\p{Digit}> (or C<\p{Word}> for C<\w>). There are | |
451 | similar C<\p{...}> constructs that can match beyond ASCII both white | |
452 | space (see L<perlrecharclass/Whitespace>), and Posix classes (see | |
453 | L<perlrecharclass/POSIX Character Classes>). Thus, this modifier | |
454 | doesn't mean you can't use Unicode, it means that to get Unicode | |
455 | matching you must explicitly use a construct (C<\p{}>, C<\P{}>) that | |
456 | signals Unicode. | |
457 | ||
458 | As you would expect, this modifier causes, for example, C<\D> to mean | |
459 | the same thing as C<[^0-9]>; in fact, all non-ASCII characters match | |
460 | C<\D>, C<\S>, and C<\W>. C<\b> still means to match at the boundary | |
461 | between C<\w> and C<\W>, using the C</a> definitions of them (similarly | |
462 | for C<\B>). | |
463 | ||
464 | Otherwise, C</a> behaves like the C</u> modifier, in that | |
850b7ec9 | 465 | case-insensitive matching uses Unicode rules; for example, "k" will |
808432af KW |
466 | match the Unicode C<\N{KELVIN SIGN}> under C</i> matching, and code |
467 | points in the Latin1 range, above ASCII will have Unicode rules when it | |
468 | comes to case-insensitive matching. | |
469 | ||
470 | To forbid ASCII/non-ASCII matches (like "k" with C<\N{KELVIN SIGN}>), | |
471 | specify the "a" twice, for example C</aai> or C</aia>. (The first | |
472 | occurrence of "a" restricts the C<\d>, etc., and the second occurrence | |
473 | adds the C</i> restrictions.) But, note that code points outside the | |
474 | ASCII range will use Unicode rules for C</i> matching, so the modifier | |
475 | doesn't really restrict things to just ASCII; it just forbids the | |
476 | intermixing of ASCII and non-ASCII. | |
477 | ||
478 | To summarize, this modifier provides protection for applications that | |
479 | don't wish to be exposed to all of Unicode. Specifying it twice | |
480 | gives added protection. | |
481 | ||
482 | This modifier may be specified to be the default by C<use re '/a'> | |
483 | or C<use re '/aa'>. If you do so, you may actually have occasion to use | |
31dc26d6 | 484 | the C</u> modifier explicitly if there are a few regular expressions |
808432af KW |
485 | where you do want full Unicode rules (but even here, it's best if |
486 | everything were under feature C<"unicode_strings">, along with the | |
487 | C<use re '/aa'>). Also see L</Which character set modifier is in | |
488 | effect?>. | |
489 | X</a> | |
490 | X</aa> | |
491 | ||
ed7efc79 KW |
492 | =head4 Which character set modifier is in effect? |
493 | ||
494 | Which of these modifiers is in effect at any given point in a regular | |
808432af KW |
495 | expression depends on a fairly complex set of interactions. These have |
496 | been designed so that in general you don't have to worry about it, but | |
497 | this section gives the gory details. As | |
ed7efc79 KW |
498 | explained below in L</Extended Patterns> it is possible to explicitly |
499 | specify modifiers that apply only to portions of a regular expression. | |
500 | The innermost always has priority over any outer ones, and one applying | |
6368643f KW |
501 | to the whole expression has priority over any of the default settings that are |
502 | described in the remainder of this section. | |
ed7efc79 | 503 | |
916cec3f | 504 | The C<L<use re 'E<sol>foo'|re/"'/flags' mode">> pragma can be used to set |
ed7efc79 KW |
505 | default modifiers (including these) for regular expressions compiled |
506 | within its scope. This pragma has precedence over the other pragmas | |
516074bb | 507 | listed below that also change the defaults. |
ed7efc79 KW |
508 | |
509 | Otherwise, C<L<use locale|perllocale>> sets the default modifier to C</l>; | |
66cbab2c | 510 | and C<L<use feature 'unicode_strings|feature>>, or |
ed7efc79 KW |
511 | C<L<use 5.012|perlfunc/use VERSION>> (or higher) set the default to |
512 | C</u> when not in the same scope as either C<L<use locale|perllocale>> | |
66cbab2c KW |
513 | or C<L<use bytes|bytes>>. |
514 | (C<L<use locale ':not_characters'|perllocale/Unicode and UTF-8>> also | |
515 | sets the default to C</u>, overriding any plain C<use locale>.) | |
516 | Unlike the mechanisms mentioned above, these | |
6368643f KW |
517 | affect operations besides regular expressions pattern matching, and so |
518 | give more consistent results with other operators, including using | |
519 | C<\U>, C<\l>, etc. in substitution replacements. | |
ed7efc79 KW |
520 | |
521 | If none of the above apply, for backwards compatibility reasons, the | |
522 | C</d> modifier is the one in effect by default. As this can lead to | |
523 | unexpected results, it is best to specify which other rule set should be | |
524 | used. | |
525 | ||
526 | =head4 Character set modifier behavior prior to Perl 5.14 | |
527 | ||
528 | Prior to 5.14, there were no explicit modifiers, but C</l> was implied | |
529 | for regexes compiled within the scope of C<use locale>, and C</d> was | |
530 | implied otherwise. However, interpolating a regex into a larger regex | |
531 | would ignore the original compilation in favor of whatever was in effect | |
532 | at the time of the second compilation. There were a number of | |
533 | inconsistencies (bugs) with the C</d> modifier, where Unicode rules | |
534 | would be used when inappropriate, and vice versa. C<\p{}> did not imply | |
535 | Unicode rules, and neither did all occurrences of C<\N{}>, until 5.12. | |
b6fa137b | 536 | |
a0d0e21e LW |
537 | =head2 Regular Expressions |
538 | ||
04838cea RGS |
539 | =head3 Metacharacters |
540 | ||
384f06ae | 541 | The patterns used in Perl pattern matching evolved from those supplied in |
14218588 | 542 | the Version 8 regex routines. (The routines are derived |
19799a22 GS |
543 | (distantly) from Henry Spencer's freely redistributable reimplementation |
544 | of the V8 routines.) See L<Version 8 Regular Expressions> for | |
545 | details. | |
a0d0e21e LW |
546 | |
547 | In particular the following metacharacters have their standard I<egrep>-ish | |
548 | meanings: | |
d74e8afc ITB |
549 | X<metacharacter> |
550 | X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]> | |
551 | ||
a0d0e21e | 552 | |
f793d64a KW |
553 | \ Quote the next metacharacter |
554 | ^ Match the beginning of the line | |
555 | . Match any character (except newline) | |
363e3e5a RS |
556 | $ Match the end of the string (or before newline at the end |
557 | of the string) | |
f793d64a KW |
558 | | Alternation |
559 | () Grouping | |
560 | [] Bracketed Character class | |
a0d0e21e | 561 | |
14218588 GS |
562 | By default, the "^" character is guaranteed to match only the |
563 | beginning of the string, the "$" character only the end (or before the | |
564 | newline at the end), and Perl does certain optimizations with the | |
a0d0e21e LW |
565 | assumption that the string contains only one line. Embedded newlines |
566 | will not be matched by "^" or "$". You may, however, wish to treat a | |
4a6725af | 567 | string as a multi-line buffer, such that the "^" will match after any |
0d520e8e YO |
568 | newline within the string (except if the newline is the last character in |
569 | the string), and "$" will match before any newline. At the | |
a0d0e21e LW |
570 | cost of a little more overhead, you can do this by using the /m modifier |
571 | on the pattern match operator. (Older programs did this by setting C<$*>, | |
db7cd43a | 572 | but this option was removed in perl 5.10.) |
d74e8afc | 573 | X<^> X<$> X</m> |
a0d0e21e | 574 | |
14218588 | 575 | To simplify multi-line substitutions, the "." character never matches a |
55497cff | 576 | newline unless you use the C</s> modifier, which in effect tells Perl to pretend |
f02c194e | 577 | the string is a single line--even if it isn't. |
d74e8afc | 578 | X<.> X</s> |
a0d0e21e | 579 | |
04838cea RGS |
580 | =head3 Quantifiers |
581 | ||
a0d0e21e | 582 | The following standard quantifiers are recognized: |
d74e8afc | 583 | X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}> |
a0d0e21e | 584 | |
f793d64a KW |
585 | * Match 0 or more times |
586 | + Match 1 or more times | |
587 | ? Match 1 or 0 times | |
588 | {n} Match exactly n times | |
589 | {n,} Match at least n times | |
590 | {n,m} Match at least n but not more than m times | |
a0d0e21e | 591 | |
0b928c2f | 592 | (If a curly bracket occurs in any other context and does not form part of |
4d68ffa0 | 593 | a backslashed sequence like C<\x{...}>, it is treated as a regular |
412f55bb KW |
594 | character. However, a deprecation warning is raised for all such |
595 | occurrences, and in Perl v5.26, literal uses of a curly bracket will be | |
596 | required to be escaped, say by preceding them with a backslash (C<"\{">) | |
597 | or enclosing them within square brackets (C<"[{]">). This change will | |
598 | allow for future syntax extensions (like making the lower bound of a | |
599 | quantifier optional), and better error checking of quantifiers.) | |
9af81bfe KW |
600 | |
601 | The "*" quantifier is equivalent to C<{0,}>, the "+" | |
527e91da | 602 | quantifier to C<{1,}>, and the "?" quantifier to C<{0,1}>. n and m are limited |
d0b16107 | 603 | to non-negative integral values less than a preset limit defined when perl is built. |
9c79236d GS |
604 | This is usually 32766 on the most common platforms. The actual limit can |
605 | be seen in the error message generated by code such as this: | |
606 | ||
820475bd | 607 | $_ **= $_ , / {$_} / for 2 .. 42; |
a0d0e21e | 608 | |
54310121 | 609 | By default, a quantified subpattern is "greedy", that is, it will match as |
610 | many times as possible (given a particular starting location) while still | |
611 | allowing the rest of the pattern to match. If you want it to match the | |
612 | minimum number of times possible, follow the quantifier with a "?". Note | |
613 | that the meanings don't change, just the "greediness": | |
0d017f4d | 614 | X<metacharacter> X<greedy> X<greediness> |
d74e8afc | 615 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?> |
a0d0e21e | 616 | |
f793d64a KW |
617 | *? Match 0 or more times, not greedily |
618 | +? Match 1 or more times, not greedily | |
619 | ?? Match 0 or 1 time, not greedily | |
0b928c2f | 620 | {n}? Match exactly n times, not greedily (redundant) |
f793d64a KW |
621 | {n,}? Match at least n times, not greedily |
622 | {n,m}? Match at least n but not more than m times, not greedily | |
a0d0e21e | 623 | |
5f3789aa | 624 | Normally when a quantified subpattern does not allow the rest of the |
b9b4dddf | 625 | overall pattern to match, Perl will backtrack. However, this behaviour is |
0d017f4d | 626 | sometimes undesirable. Thus Perl provides the "possessive" quantifier form |
b9b4dddf YO |
627 | as well. |
628 | ||
f793d64a KW |
629 | *+ Match 0 or more times and give nothing back |
630 | ++ Match 1 or more times and give nothing back | |
631 | ?+ Match 0 or 1 time and give nothing back | |
632 | {n}+ Match exactly n times and give nothing back (redundant) | |
633 | {n,}+ Match at least n times and give nothing back | |
634 | {n,m}+ Match at least n but not more than m times and give nothing back | |
b9b4dddf YO |
635 | |
636 | For instance, | |
637 | ||
638 | 'aaaa' =~ /a++a/ | |
639 | ||
640 | will never match, as the C<a++> will gobble up all the C<a>'s in the | |
641 | string and won't leave any for the remaining part of the pattern. This | |
642 | feature can be extremely useful to give perl hints about where it | |
643 | shouldn't backtrack. For instance, the typical "match a double-quoted | |
644 | string" problem can be most efficiently performed when written as: | |
645 | ||
646 | /"(?:[^"\\]++|\\.)*+"/ | |
647 | ||
0d017f4d | 648 | as we know that if the final quote does not match, backtracking will not |
0b928c2f FC |
649 | help. See the independent subexpression |
650 | L</C<< (?>pattern) >>> for more details; | |
b9b4dddf YO |
651 | possessive quantifiers are just syntactic sugar for that construct. For |
652 | instance the above example could also be written as follows: | |
653 | ||
654 | /"(?>(?:(?>[^"\\]+)|\\.)*)"/ | |
655 | ||
5f3789aa YO |
656 | Note that the possessive quantifier modifier can not be be combined |
657 | with the non-greedy modifier. This is because it would make no sense. | |
658 | Consider the follow equivalency table: | |
659 | ||
660 | Illegal Legal | |
661 | ------------ ------ | |
662 | X??+ X{0} | |
663 | X+?+ X{1} | |
664 | X{min,max}?+ X{min} | |
665 | ||
04838cea RGS |
666 | =head3 Escape sequences |
667 | ||
0b928c2f | 668 | Because patterns are processed as double-quoted strings, the following |
a0d0e21e LW |
669 | also work: |
670 | ||
f793d64a KW |
671 | \t tab (HT, TAB) |
672 | \n newline (LF, NL) | |
673 | \r return (CR) | |
674 | \f form feed (FF) | |
675 | \a alarm (bell) (BEL) | |
676 | \e escape (think troff) (ESC) | |
f793d64a | 677 | \cK control char (example: VT) |
dc0d9c48 | 678 | \x{}, \x00 character whose ordinal is the given hexadecimal number |
fb121860 | 679 | \N{name} named Unicode character or character sequence |
f793d64a | 680 | \N{U+263D} Unicode character (example: FIRST QUARTER MOON) |
f0a2b745 | 681 | \o{}, \000 character whose ordinal is the given octal number |
f793d64a KW |
682 | \l lowercase next char (think vi) |
683 | \u uppercase next char (think vi) | |
ad81d09f KE |
684 | \L lowercase until \E (think vi) |
685 | \U uppercase until \E (think vi) | |
686 | \Q quote (disable) pattern metacharacters until \E | |
f793d64a | 687 | \E end either case modification or quoted section, think vi |
a0d0e21e | 688 | |
9bb1f947 | 689 | Details are in L<perlop/Quote and Quote-like Operators>. |
1d2dff63 | 690 | |
e1d1eefb | 691 | =head3 Character Classes and other Special Escapes |
04838cea | 692 | |
a0d0e21e | 693 | In addition, Perl defines the following: |
d0b16107 | 694 | X<\g> X<\k> X<\K> X<backreference> |
a0d0e21e | 695 | |
f793d64a KW |
696 | Sequence Note Description |
697 | [...] [1] Match a character according to the rules of the | |
698 | bracketed character class defined by the "...". | |
699 | Example: [a-z] matches "a" or "b" or "c" ... or "z" | |
700 | [[:...:]] [2] Match a character according to the rules of the POSIX | |
701 | character class "..." within the outer bracketed | |
702 | character class. Example: [[:upper:]] matches any | |
703 | uppercase character. | |
572224ce | 704 | (?[...]) [8] Extended bracketed character class |
d35dd6c6 KW |
705 | \w [3] Match a "word" character (alphanumeric plus "_", plus |
706 | other connector punctuation chars plus Unicode | |
0b928c2f | 707 | marks) |
f793d64a KW |
708 | \W [3] Match a non-"word" character |
709 | \s [3] Match a whitespace character | |
710 | \S [3] Match a non-whitespace character | |
711 | \d [3] Match a decimal digit character | |
712 | \D [3] Match a non-digit character | |
713 | \pP [3] Match P, named property. Use \p{Prop} for longer names | |
714 | \PP [3] Match non-P | |
715 | \X [4] Match Unicode "eXtended grapheme cluster" | |
c27a5cfe | 716 | \1 [5] Backreference to a specific capture group or buffer. |
f793d64a KW |
717 | '1' may actually be any positive integer. |
718 | \g1 [5] Backreference to a specific or previous group, | |
719 | \g{-1} [5] The number may be negative indicating a relative | |
c27a5cfe | 720 | previous group and may optionally be wrapped in |
f793d64a KW |
721 | curly brackets for safer parsing. |
722 | \g{name} [5] Named backreference | |
723 | \k<name> [5] Named backreference | |
724 | \K [6] Keep the stuff left of the \K, don't include it in $& | |
2171640d | 725 | \N [7] Any character but \n. Not affected by /s modifier |
f793d64a KW |
726 | \v [3] Vertical whitespace |
727 | \V [3] Not vertical whitespace | |
728 | \h [3] Horizontal whitespace | |
729 | \H [3] Not horizontal whitespace | |
730 | \R [4] Linebreak | |
e1d1eefb | 731 | |
9bb1f947 KW |
732 | =over 4 |
733 | ||
734 | =item [1] | |
735 | ||
736 | See L<perlrecharclass/Bracketed Character Classes> for details. | |
df225385 | 737 | |
9bb1f947 | 738 | =item [2] |
b8c5462f | 739 | |
9bb1f947 | 740 | See L<perlrecharclass/POSIX Character Classes> for details. |
b8c5462f | 741 | |
9bb1f947 | 742 | =item [3] |
5496314a | 743 | |
9bb1f947 | 744 | See L<perlrecharclass/Backslash sequences> for details. |
5496314a | 745 | |
9bb1f947 | 746 | =item [4] |
5496314a | 747 | |
9bb1f947 | 748 | See L<perlrebackslash/Misc> for details. |
d0b16107 | 749 | |
9bb1f947 | 750 | =item [5] |
b8c5462f | 751 | |
c27a5cfe | 752 | See L</Capture groups> below for details. |
93733859 | 753 | |
9bb1f947 | 754 | =item [6] |
b8c5462f | 755 | |
9bb1f947 KW |
756 | See L</Extended Patterns> below for details. |
757 | ||
758 | =item [7] | |
759 | ||
760 | Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the | |
fb121860 KW |
761 | character or character sequence whose name is C<NAME>; and similarly |
762 | when of the form C<\N{U+I<hex>}>, it matches the character whose Unicode | |
763 | code point is I<hex>. Otherwise it matches any character but C<\n>. | |
9bb1f947 | 764 | |
572224ce KW |
765 | =item [8] |
766 | ||
767 | See L<perlrecharclass/Extended Bracketed Character Classes> for details. | |
768 | ||
9bb1f947 | 769 | =back |
d0b16107 | 770 | |
04838cea RGS |
771 | =head3 Assertions |
772 | ||
a0d0e21e | 773 | Perl defines the following zero-width assertions: |
d74e8afc ITB |
774 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> |
775 | X<regexp, zero-width assertion> | |
776 | X<regular expression, zero-width assertion> | |
777 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> | |
a0d0e21e | 778 | |
64935bc6 KW |
779 | \b{} Match at Unicode boundary of specified type |
780 | \B{} Match where corresponding \b{} doesn't match | |
9bb1f947 KW |
781 | \b Match a word boundary |
782 | \B Match except at a word boundary | |
783 | \A Match only at beginning of string | |
784 | \Z Match only at end of string, or before newline at the end | |
785 | \z Match only at end of string | |
786 | \G Match only at pos() (e.g. at the end-of-match position | |
9da458fc | 787 | of prior m//g) |
a0d0e21e | 788 | |
64935bc6 KW |
789 | A Unicode boundary (C<\b{}>), available starting in v5.22, is a spot |
790 | between two characters, or before the first character in the string, or | |
791 | after the final character in the string where certain criteria defined | |
792 | by Unicode are met. See L<perlrebackslash/\b{}, \b, \B{}, \B> for | |
793 | details. | |
794 | ||
14218588 | 795 | A word boundary (C<\b>) is a spot between two characters |
19799a22 GS |
796 | that has a C<\w> on one side of it and a C<\W> on the other side |
797 | of it (in either order), counting the imaginary characters off the | |
798 | beginning and end of the string as matching a C<\W>. (Within | |
799 | character classes C<\b> represents backspace rather than a word | |
800 | boundary, just as it normally does in any double-quoted string.) | |
801 | The C<\A> and C<\Z> are just like "^" and "$", except that they | |
802 | won't match multiple times when the C</m> modifier is used, while | |
803 | "^" and "$" will match at every internal line boundary. To match | |
804 | the actual end of the string and not ignore an optional trailing | |
805 | newline, use C<\z>. | |
d74e8afc | 806 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 GS |
807 | |
808 | The C<\G> assertion can be used to chain global matches (using | |
809 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. | |
810 | It is also useful when writing C<lex>-like scanners, when you have | |
811 | several patterns that you want to match against consequent substrings | |
0b928c2f | 812 | of your string; see the previous reference. The actual location |
19799a22 | 813 | where C<\G> will match can also be influenced by using C<pos()> as |
58e23c8d | 814 | an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length |
0b928c2f FC |
815 | matches (see L</"Repeated Patterns Matching a Zero-length Substring">) |
816 | is modified somewhat, in that contents to the left of C<\G> are | |
58e23c8d YO |
817 | not counted when determining the length of the match. Thus the following |
818 | will not match forever: | |
d74e8afc | 819 | X<\G> |
c47ff5f1 | 820 | |
e761bb84 CO |
821 | my $string = 'ABC'; |
822 | pos($string) = 1; | |
823 | while ($string =~ /(.\G)/g) { | |
824 | print $1; | |
825 | } | |
58e23c8d YO |
826 | |
827 | It will print 'A' and then terminate, as it considers the match to | |
828 | be zero-width, and thus will not match at the same position twice in a | |
829 | row. | |
830 | ||
831 | It is worth noting that C<\G> improperly used can result in an infinite | |
832 | loop. Take care when using patterns that include C<\G> in an alternation. | |
833 | ||
d5e7783a DM |
834 | Note also that C<s///> will refuse to overwrite part of a substitution |
835 | that has already been replaced; so for example this will stop after the | |
836 | first iteration, rather than iterating its way backwards through the | |
837 | string: | |
838 | ||
839 | $_ = "123456789"; | |
840 | pos = 6; | |
841 | s/.(?=.\G)/X/g; | |
842 | print; # prints 1234X6789, not XXXXX6789 | |
843 | ||
844 | ||
c27a5cfe | 845 | =head3 Capture groups |
04838cea | 846 | |
c27a5cfe KW |
847 | The bracketing construct C<( ... )> creates capture groups (also referred to as |
848 | capture buffers). To refer to the current contents of a group later on, within | |
d8b950dc KW |
849 | the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>) |
850 | for the second, and so on. | |
851 | This is called a I<backreference>. | |
d74e8afc | 852 | X<regex, capture buffer> X<regexp, capture buffer> |
c27a5cfe | 853 | X<regex, capture group> X<regexp, capture group> |
d74e8afc | 854 | X<regular expression, capture buffer> X<backreference> |
c27a5cfe | 855 | X<regular expression, capture group> X<backreference> |
1f1031fe | 856 | X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference> |
d8b950dc KW |
857 | X<named capture buffer> X<regular expression, named capture buffer> |
858 | X<named capture group> X<regular expression, named capture group> | |
859 | X<%+> X<$+{name}> X<< \k<name> >> | |
860 | There is no limit to the number of captured substrings that you may use. | |
861 | Groups are numbered with the leftmost open parenthesis being number 1, etc. If | |
862 | a group did not match, the associated backreference won't match either. (This | |
863 | can happen if the group is optional, or in a different branch of an | |
864 | alternation.) | |
865 | You can omit the C<"g">, and write C<"\1">, etc, but there are some issues with | |
866 | this form, described below. | |
867 | ||
868 | You can also refer to capture groups relatively, by using a negative number, so | |
869 | that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture | |
870 | group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For | |
871 | example: | |
5624f11d YO |
872 | |
873 | / | |
c27a5cfe KW |
874 | (Y) # group 1 |
875 | ( # group 2 | |
876 | (X) # group 3 | |
877 | \g{-1} # backref to group 3 | |
878 | \g{-3} # backref to group 1 | |
5624f11d YO |
879 | ) |
880 | /x | |
881 | ||
d8b950dc KW |
882 | would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to |
883 | interpolate regexes into larger regexes and not have to worry about the | |
884 | capture groups being renumbered. | |
885 | ||
886 | You can dispense with numbers altogether and create named capture groups. | |
887 | The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to | |
888 | reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may | |
889 | also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.) | |
890 | I<name> must not begin with a number, nor contain hyphens. | |
891 | When different groups within the same pattern have the same name, any reference | |
892 | to that name assumes the leftmost defined group. Named groups count in | |
893 | absolute and relative numbering, and so can also be referred to by those | |
894 | numbers. | |
895 | (It's possible to do things with named capture groups that would otherwise | |
896 | require C<(??{})>.) | |
897 | ||
898 | Capture group contents are dynamically scoped and available to you outside the | |
899 | pattern until the end of the enclosing block or until the next successful | |
900 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
901 | You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">, | |
902 | etc); or by name via the C<%+> hash, using C<"$+{I<name>}">. | |
903 | ||
904 | Braces are required in referring to named capture groups, but are optional for | |
905 | absolute or relative numbered ones. Braces are safer when creating a regex by | |
906 | concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a> | |
907 | contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which | |
908 | is probably not what you intended. | |
909 | ||
910 | The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that | |
911 | there were no named nor relative numbered capture groups. Absolute numbered | |
0b928c2f FC |
912 | groups were referred to using C<\1>, |
913 | C<\2>, etc., and this notation is still | |
d8b950dc KW |
914 | accepted (and likely always will be). But it leads to some ambiguities if |
915 | there are more than 9 capture groups, as C<\10> could mean either the tenth | |
916 | capture group, or the character whose ordinal in octal is 010 (a backspace in | |
917 | ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference | |
918 | only if at least 10 left parentheses have opened before it. Likewise C<\11> is | |
919 | a backreference only if at least 11 left parentheses have opened before it. | |
e1f120a9 KW |
920 | And so on. C<\1> through C<\9> are always interpreted as backreferences. |
921 | There are several examples below that illustrate these perils. You can avoid | |
922 | the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups; | |
923 | and for octal constants always using C<\o{}>, or for C<\077> and below, using 3 | |
924 | digits padded with leading zeros, since a leading zero implies an octal | |
925 | constant. | |
d8b950dc KW |
926 | |
927 | The C<\I<digit>> notation also works in certain circumstances outside | |
ed7efc79 | 928 | the pattern. See L</Warning on \1 Instead of $1> below for details. |
81714fb9 | 929 | |
14218588 | 930 | Examples: |
a0d0e21e LW |
931 | |
932 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
933 | ||
d8b950dc | 934 | /(.)\g1/ # find first doubled char |
81714fb9 YO |
935 | and print "'$1' is the first doubled character\n"; |
936 | ||
937 | /(?<char>.)\k<char>/ # ... a different way | |
938 | and print "'$+{char}' is the first doubled character\n"; | |
939 | ||
d8b950dc | 940 | /(?'char'.)\g1/ # ... mix and match |
81714fb9 | 941 | and print "'$1' is the first doubled character\n"; |
c47ff5f1 | 942 | |
14218588 | 943 | if (/Time: (..):(..):(..)/) { # parse out values |
f793d64a KW |
944 | $hours = $1; |
945 | $minutes = $2; | |
946 | $seconds = $3; | |
a0d0e21e | 947 | } |
c47ff5f1 | 948 | |
9d860678 KW |
949 | /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference |
950 | /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal | |
951 | /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference | |
952 | /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal | |
953 | ||
954 | $a = '(.)\1'; # Creates problems when concatenated. | |
955 | $b = '(.)\g{1}'; # Avoids the problems. | |
956 | "aa" =~ /${a}/; # True | |
957 | "aa" =~ /${b}/; # True | |
958 | "aa0" =~ /${a}0/; # False! | |
959 | "aa0" =~ /${b}0/; # True | |
dc0d9c48 KW |
960 | "aa\x08" =~ /${a}0/; # True! |
961 | "aa\x08" =~ /${b}0/; # False | |
9d860678 | 962 | |
14218588 GS |
963 | Several special variables also refer back to portions of the previous |
964 | match. C<$+> returns whatever the last bracket match matched. | |
965 | C<$&> returns the entire matched string. (At one point C<$0> did | |
966 | also, but now it returns the name of the program.) C<$`> returns | |
77ea4f6d JV |
967 | everything before the matched string. C<$'> returns everything |
968 | after the matched string. And C<$^N> contains whatever was matched by | |
969 | the most-recently closed group (submatch). C<$^N> can be used in | |
970 | extended patterns (see below), for example to assign a submatch to a | |
81714fb9 | 971 | variable. |
d74e8afc | 972 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 | 973 | |
d8b950dc KW |
974 | These special variables, like the C<%+> hash and the numbered match variables |
975 | (C<$1>, C<$2>, C<$3>, etc.) are dynamically scoped | |
14218588 GS |
976 | until the end of the enclosing block or until the next successful |
977 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
d74e8afc ITB |
978 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
979 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> | |
980 | ||
0d017f4d | 981 | B<NOTE>: Failed matches in Perl do not reset the match variables, |
5146ce24 | 982 | which makes it easier to write code that tests for a series of more |
665e98b9 JH |
983 | specific cases and remembers the best match. |
984 | ||
13b0f67d DM |
985 | B<WARNING>: If your code is to run on Perl 5.16 or earlier, |
986 | beware that once Perl sees that you need one of C<$&>, C<$`>, or | |
14218588 | 987 | C<$'> anywhere in the program, it has to provide them for every |
13b0f67d DM |
988 | pattern match. This may substantially slow your program. |
989 | ||
990 | Perl uses the same mechanism to produce C<$1>, C<$2>, etc, so you also | |
991 | pay a price for each pattern that contains capturing parentheses. | |
992 | (To avoid this cost while retaining the grouping behaviour, use the | |
14218588 GS |
993 | extended regular expression C<(?: ... )> instead.) But if you never |
994 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
995 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
996 | if you can, but if you can't (and some algorithms really appreciate | |
997 | them), once you've used them once, use them at will, because you've | |
13b0f67d | 998 | already paid the price. |
d74e8afc | 999 | X<$&> X<$`> X<$'> |
68dc0745 | 1000 | |
13b0f67d DM |
1001 | Perl 5.16 introduced a slightly more efficient mechanism that notes |
1002 | separately whether each of C<$`>, C<$&>, and C<$'> have been seen, and | |
1003 | thus may only need to copy part of the string. Perl 5.20 introduced a | |
1004 | much more efficient copy-on-write mechanism which eliminates any slowdown. | |
1005 | ||
1006 | As another workaround for this problem, Perl 5.10.0 introduced C<${^PREMATCH}>, | |
cde0cee5 YO |
1007 | C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&> |
1008 | and C<$'>, B<except> that they are only guaranteed to be defined after a | |
87e95b7f | 1009 | successful match that was executed with the C</p> (preserve) modifier. |
cde0cee5 YO |
1010 | The use of these variables incurs no global performance penalty, unlike |
1011 | their punctuation char equivalents, however at the trade-off that you | |
13b0f67d DM |
1012 | have to tell perl when you want to use them. As of Perl 5.20, these three |
1013 | variables are equivalent to C<$`>, C<$&> and C<$'>, and C</p> is ignored. | |
87e95b7f | 1014 | X</p> X<p modifier> |
cde0cee5 | 1015 | |
9d727203 KW |
1016 | =head2 Quoting metacharacters |
1017 | ||
19799a22 GS |
1018 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
1019 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
1020 | are no backslashed symbols that aren't alphanumeric. So anything | |
0f264506 | 1021 | that looks like \\, \(, \), \[, \], \{, or \} is always |
19799a22 GS |
1022 | interpreted as a literal character, not a metacharacter. This was |
1023 | once used in a common idiom to disable or quote the special meanings | |
1024 | of regular expression metacharacters in a string that you want to | |
36bbe248 | 1025 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e LW |
1026 | |
1027 | $pattern =~ s/(\W)/\\$1/g; | |
1028 | ||
f1cbbd6e | 1029 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 GS |
1030 | Today it is more common to use the quotemeta() function or the C<\Q> |
1031 | metaquoting escape sequence to disable all metacharacters' special | |
1032 | meanings like this: | |
a0d0e21e LW |
1033 | |
1034 | /$unquoted\Q$quoted\E$unquoted/ | |
1035 | ||
9da458fc IZ |
1036 | Beware that if you put literal backslashes (those not inside |
1037 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
1038 | backslash interpolation may lead to confusing results. If you | |
1039 | I<need> to use literal backslashes within C<\Q...\E>, | |
1040 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
1041 | ||
736fe711 KW |
1042 | C<quotemeta()> and C<\Q> are fully described in L<perlfunc/quotemeta>. |
1043 | ||
19799a22 GS |
1044 | =head2 Extended Patterns |
1045 | ||
14218588 | 1046 | Perl also defines a consistent extension syntax for features not |
0b928c2f FC |
1047 | found in standard tools like B<awk> and |
1048 | B<lex>. The syntax for most of these is a | |
14218588 GS |
1049 | pair of parentheses with a question mark as the first thing within |
1050 | the parentheses. The character after the question mark indicates | |
1051 | the extension. | |
19799a22 | 1052 | |
14218588 GS |
1053 | The stability of these extensions varies widely. Some have been |
1054 | part of the core language for many years. Others are experimental | |
1055 | and may change without warning or be completely removed. Check | |
1056 | the documentation on an individual feature to verify its current | |
1057 | status. | |
19799a22 | 1058 | |
14218588 GS |
1059 | A question mark was chosen for this and for the minimal-matching |
1060 | construct because 1) question marks are rare in older regular | |
1061 | expressions, and 2) whenever you see one, you should stop and | |
0b928c2f | 1062 | "question" exactly what is going on. That's psychology.... |
a0d0e21e | 1063 | |
70ca8714 | 1064 | =over 4 |
a0d0e21e | 1065 | |
cc6b7395 | 1066 | =item C<(?#text)> |
d74e8afc | 1067 | X<(?#)> |
a0d0e21e | 1068 | |
7c688e65 KW |
1069 | A comment. The text is ignored. |
1070 | Note that Perl closes | |
259138e3 | 1071 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
7c688e65 KW |
1072 | C<)> in the comment. The pattern's closing delimiter must be escaped by |
1073 | a backslash if it appears in the comment. | |
1074 | ||
1075 | See L</E<sol>x> for another way to have comments in patterns. | |
a0d0e21e | 1076 | |
4cb6b395 | 1077 | =item C<(?adlupimnsx-imnsx)> |
fb85c044 | 1078 | |
4cb6b395 | 1079 | =item C<(?^alupimnsx)> |
fb85c044 | 1080 | X<(?)> X<(?^)> |
19799a22 | 1081 | |
0b6d1084 JH |
1082 | One or more embedded pattern-match modifiers, to be turned on (or |
1083 | turned off, if preceded by C<->) for the remainder of the pattern or | |
fb85c044 KW |
1084 | the remainder of the enclosing pattern group (if any). |
1085 | ||
fb85c044 | 1086 | This is particularly useful for dynamic patterns, such as those read in from a |
0d017f4d | 1087 | configuration file, taken from an argument, or specified in a table |
0b928c2f FC |
1088 | somewhere. Consider the case where some patterns want to be |
1089 | case-sensitive and some do not: The case-insensitive ones merely need to | |
0d017f4d | 1090 | include C<(?i)> at the front of the pattern. For example: |
19799a22 GS |
1091 | |
1092 | $pattern = "foobar"; | |
5d458dd8 | 1093 | if ( /$pattern/i ) { } |
19799a22 GS |
1094 | |
1095 | # more flexible: | |
1096 | ||
1097 | $pattern = "(?i)foobar"; | |
5d458dd8 | 1098 | if ( /$pattern/ ) { } |
19799a22 | 1099 | |
0b6d1084 | 1100 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 | 1101 | |
d8b950dc | 1102 | ( (?i) blah ) \s+ \g1 |
19799a22 | 1103 | |
0d017f4d WL |
1104 | will match C<blah> in any case, some spaces, and an exact (I<including the case>!) |
1105 | repetition of the previous word, assuming the C</x> modifier, and no C</i> | |
1106 | modifier outside this group. | |
19799a22 | 1107 | |
8eb5594e | 1108 | These modifiers do not carry over into named subpatterns called in the |
dd72e27b | 1109 | enclosing group. In other words, a pattern such as C<((?i)(?&NAME))> does not |
8eb5594e DR |
1110 | change the case-sensitivity of the "NAME" pattern. |
1111 | ||
dc925305 KW |
1112 | Any of these modifiers can be set to apply globally to all regular |
1113 | expressions compiled within the scope of a C<use re>. See | |
a0bbd6ff | 1114 | L<re/"'/flags' mode">. |
dc925305 | 1115 | |
9de15fec | 1116 | Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately |
4cb6b395 | 1117 | after the C<"?"> is a shorthand equivalent to C<d-imnsx>. Flags (except |
9de15fec KW |
1118 | C<"d">) may follow the caret to override it. |
1119 | But a minus sign is not legal with it. | |
1120 | ||
dc925305 | 1121 | Note that the C<a>, C<d>, C<l>, C<p>, and C<u> modifiers are special in |
e1d8d8ac | 1122 | that they can only be enabled, not disabled, and the C<a>, C<d>, C<l>, and |
dc925305 | 1123 | C<u> modifiers are mutually exclusive: specifying one de-specifies the |
ed7efc79 KW |
1124 | others, and a maximum of one (or two C<a>'s) may appear in the |
1125 | construct. Thus, for | |
0b928c2f | 1126 | example, C<(?-p)> will warn when compiled under C<use warnings>; |
b6fa137b | 1127 | C<(?-d:...)> and C<(?dl:...)> are fatal errors. |
9de15fec KW |
1128 | |
1129 | Note also that the C<p> modifier is special in that its presence | |
1130 | anywhere in a pattern has a global effect. | |
cde0cee5 | 1131 | |
5a964f20 | 1132 | =item C<(?:pattern)> |
d74e8afc | 1133 | X<(?:)> |
a0d0e21e | 1134 | |
4cb6b395 | 1135 | =item C<(?adluimnsx-imnsx:pattern)> |
ca9dfc88 | 1136 | |
4cb6b395 | 1137 | =item C<(?^aluimnsx:pattern)> |
fb85c044 KW |
1138 | X<(?^:)> |
1139 | ||
5a964f20 TC |
1140 | This is for clustering, not capturing; it groups subexpressions like |
1141 | "()", but doesn't make backreferences as "()" does. So | |
a0d0e21e | 1142 | |
5a964f20 | 1143 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e LW |
1144 | |
1145 | is like | |
1146 | ||
5a964f20 | 1147 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e | 1148 | |
19799a22 GS |
1149 | but doesn't spit out extra fields. It's also cheaper not to capture |
1150 | characters if you don't need to. | |
a0d0e21e | 1151 | |
19799a22 | 1152 | Any letters between C<?> and C<:> act as flags modifiers as with |
4cb6b395 | 1153 | C<(?adluimnsx-imnsx)>. For example, |
ca9dfc88 IZ |
1154 | |
1155 | /(?s-i:more.*than).*million/i | |
1156 | ||
14218588 | 1157 | is equivalent to the more verbose |
ca9dfc88 IZ |
1158 | |
1159 | /(?:(?s-i)more.*than).*million/i | |
1160 | ||
4cb6b395 KW |
1161 | Note that any C<(...)> constructs enclosed within this one will still |
1162 | capture unless the C</n> modifier is in effect. | |
1163 | ||
fb85c044 | 1164 | Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately |
4cb6b395 | 1165 | after the C<"?"> is a shorthand equivalent to C<d-imnsx>. Any positive |
9de15fec | 1166 | flags (except C<"d">) may follow the caret, so |
fb85c044 KW |
1167 | |
1168 | (?^x:foo) | |
1169 | ||
1170 | is equivalent to | |
1171 | ||
4cb6b395 | 1172 | (?x-imns:foo) |
fb85c044 KW |
1173 | |
1174 | The caret tells Perl that this cluster doesn't inherit the flags of any | |
4cb6b395 | 1175 | surrounding pattern, but uses the system defaults (C<d-imnsx>), |
fb85c044 KW |
1176 | modified by any flags specified. |
1177 | ||
1178 | The caret allows for simpler stringification of compiled regular | |
1179 | expressions. These look like | |
1180 | ||
1181 | (?^:pattern) | |
1182 | ||
1183 | with any non-default flags appearing between the caret and the colon. | |
1184 | A test that looks at such stringification thus doesn't need to have the | |
1185 | system default flags hard-coded in it, just the caret. If new flags are | |
1186 | added to Perl, the meaning of the caret's expansion will change to include | |
1187 | the default for those flags, so the test will still work, unchanged. | |
1188 | ||
1189 | Specifying a negative flag after the caret is an error, as the flag is | |
1190 | redundant. | |
1191 | ||
1192 | Mnemonic for C<(?^...)>: A fresh beginning since the usual use of a caret is | |
1193 | to match at the beginning. | |
1194 | ||
594d7033 YO |
1195 | =item C<(?|pattern)> |
1196 | X<(?|)> X<Branch reset> | |
1197 | ||
1198 | This is the "branch reset" pattern, which has the special property | |
c27a5cfe | 1199 | that the capture groups are numbered from the same starting point |
99d59c4d | 1200 | in each alternation branch. It is available starting from perl 5.10.0. |
4deaaa80 | 1201 | |
c27a5cfe | 1202 | Capture groups are numbered from left to right, but inside this |
693596a8 | 1203 | construct the numbering is restarted for each branch. |
4deaaa80 | 1204 | |
c27a5cfe | 1205 | The numbering within each branch will be as normal, and any groups |
4deaaa80 PJ |
1206 | following this construct will be numbered as though the construct |
1207 | contained only one branch, that being the one with the most capture | |
c27a5cfe | 1208 | groups in it. |
4deaaa80 | 1209 | |
0b928c2f | 1210 | This construct is useful when you want to capture one of a |
4deaaa80 PJ |
1211 | number of alternative matches. |
1212 | ||
1213 | Consider the following pattern. The numbers underneath show in | |
c27a5cfe | 1214 | which group the captured content will be stored. |
594d7033 YO |
1215 | |
1216 | ||
1217 | # before ---------------branch-reset----------- after | |
1218 | / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x | |
1219 | # 1 2 2 3 2 3 4 | |
1220 | ||
ab106183 A |
1221 | Be careful when using the branch reset pattern in combination with |
1222 | named captures. Named captures are implemented as being aliases to | |
c27a5cfe | 1223 | numbered groups holding the captures, and that interferes with the |
ab106183 A |
1224 | implementation of the branch reset pattern. If you are using named |
1225 | captures in a branch reset pattern, it's best to use the same names, | |
1226 | in the same order, in each of the alternations: | |
1227 | ||
1228 | /(?| (?<a> x ) (?<b> y ) | |
1229 | | (?<a> z ) (?<b> w )) /x | |
1230 | ||
1231 | Not doing so may lead to surprises: | |
1232 | ||
1233 | "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x; | |
1234 | say $+ {a}; # Prints '12' | |
1235 | say $+ {b}; # *Also* prints '12'. | |
1236 | ||
c27a5cfe KW |
1237 | The problem here is that both the group named C<< a >> and the group |
1238 | named C<< b >> are aliases for the group belonging to C<< $1 >>. | |
90a18110 | 1239 | |
ee9b8eae YO |
1240 | =item Look-Around Assertions |
1241 | X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround> | |
1242 | ||
0b928c2f | 1243 | Look-around assertions are zero-width patterns which match a specific |
ee9b8eae YO |
1244 | pattern without including it in C<$&>. Positive assertions match when |
1245 | their subpattern matches, negative assertions match when their subpattern | |
1246 | fails. Look-behind matches text up to the current match position, | |
1247 | look-ahead matches text following the current match position. | |
1248 | ||
1249 | =over 4 | |
1250 | ||
5a964f20 | 1251 | =item C<(?=pattern)> |
d74e8afc | 1252 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> |
a0d0e21e | 1253 | |
19799a22 | 1254 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e LW |
1255 | matches a word followed by a tab, without including the tab in C<$&>. |
1256 | ||
5a964f20 | 1257 | =item C<(?!pattern)> |
d74e8afc | 1258 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> |
a0d0e21e | 1259 | |
19799a22 | 1260 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e | 1261 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 GS |
1262 | however that look-ahead and look-behind are NOT the same thing. You cannot |
1263 | use this for look-behind. | |
7b8d334a | 1264 | |
5a964f20 | 1265 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a GS |
1266 | will not do what you want. That's because the C<(?!foo)> is just saying that |
1267 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
0b928c2f | 1268 | match. Use look-behind instead (see below). |
c277df42 | 1269 | |
ee9b8eae YO |
1270 | =item C<(?<=pattern)> C<\K> |
1271 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K> | |
c277df42 | 1272 | |
c47ff5f1 | 1273 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 GS |
1274 | matches a word that follows a tab, without including the tab in C<$&>. |
1275 | Works only for fixed-width look-behind. | |
c277df42 | 1276 | |
3d9df1a7 KE |
1277 | There is a special form of this construct, called C<\K> (available since |
1278 | Perl 5.10.0), which causes the | |
ee9b8eae | 1279 | regex engine to "keep" everything it had matched prior to the C<\K> and |
0b928c2f | 1280 | not include it in C<$&>. This effectively provides variable-length |
ee9b8eae YO |
1281 | look-behind. The use of C<\K> inside of another look-around assertion |
1282 | is allowed, but the behaviour is currently not well defined. | |
1283 | ||
c62285ac | 1284 | For various reasons C<\K> may be significantly more efficient than the |
ee9b8eae YO |
1285 | equivalent C<< (?<=...) >> construct, and it is especially useful in |
1286 | situations where you want to efficiently remove something following | |
1287 | something else in a string. For instance | |
1288 | ||
1289 | s/(foo)bar/$1/g; | |
1290 | ||
1291 | can be rewritten as the much more efficient | |
1292 | ||
1293 | s/foo\Kbar//g; | |
1294 | ||
5a964f20 | 1295 | =item C<(?<!pattern)> |
d74e8afc | 1296 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> |
c277df42 | 1297 | |
19799a22 GS |
1298 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
1299 | matches any occurrence of "foo" that does not follow "bar". Works | |
1300 | only for fixed-width look-behind. | |
c277df42 | 1301 | |
ee9b8eae YO |
1302 | =back |
1303 | ||
81714fb9 YO |
1304 | =item C<(?'NAME'pattern)> |
1305 | ||
1306 | =item C<< (?<NAME>pattern) >> | |
1307 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> | |
1308 | ||
c27a5cfe | 1309 | A named capture group. Identical in every respect to normal capturing |
0b928c2f FC |
1310 | parentheses C<()> but for the additional fact that the group |
1311 | can be referred to by name in various regular expression | |
1312 | constructs (like C<\g{NAME}>) and can be accessed by name | |
1313 | after a successful match via C<%+> or C<%->. See L<perlvar> | |
90a18110 | 1314 | for more details on the C<%+> and C<%-> hashes. |
81714fb9 | 1315 | |
c27a5cfe KW |
1316 | If multiple distinct capture groups have the same name then the |
1317 | $+{NAME} will refer to the leftmost defined group in the match. | |
81714fb9 | 1318 | |
0d017f4d | 1319 | The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent. |
81714fb9 YO |
1320 | |
1321 | B<NOTE:> While the notation of this construct is the same as the similar | |
c27a5cfe | 1322 | function in .NET regexes, the behavior is not. In Perl the groups are |
81714fb9 YO |
1323 | numbered sequentially regardless of being named or not. Thus in the |
1324 | pattern | |
1325 | ||
1326 | /(x)(?<foo>y)(z)/ | |
1327 | ||
1328 | $+{foo} will be the same as $2, and $3 will contain 'z' instead of | |
1329 | the opposite which is what a .NET regex hacker might expect. | |
1330 | ||
1f1031fe YO |
1331 | Currently NAME is restricted to simple identifiers only. |
1332 | In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or | |
1333 | its Unicode extension (see L<utf8>), | |
1334 | though it isn't extended by the locale (see L<perllocale>). | |
81714fb9 | 1335 | |
1f1031fe | 1336 | B<NOTE:> In order to make things easier for programmers with experience |
ae5648b3 | 1337 | with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >> |
0d017f4d | 1338 | may be used instead of C<< (?<NAME>pattern) >>; however this form does not |
64c5a566 | 1339 | support the use of single quotes as a delimiter for the name. |
81714fb9 | 1340 | |
1f1031fe YO |
1341 | =item C<< \k<NAME> >> |
1342 | ||
1343 | =item C<< \k'NAME' >> | |
81714fb9 YO |
1344 | |
1345 | Named backreference. Similar to numeric backreferences, except that | |
1346 | the group is designated by name and not number. If multiple groups | |
1347 | have the same name then it refers to the leftmost defined group in | |
1348 | the current match. | |
1349 | ||
0d017f4d | 1350 | It is an error to refer to a name not defined by a C<< (?<NAME>) >> |
81714fb9 YO |
1351 | earlier in the pattern. |
1352 | ||
1353 | Both forms are equivalent. | |
1354 | ||
1f1031fe | 1355 | B<NOTE:> In order to make things easier for programmers with experience |
0d017f4d | 1356 | with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >> |
64c5a566 | 1357 | may be used instead of C<< \k<NAME> >>. |
1f1031fe | 1358 | |
cc6b7395 | 1359 | =item C<(?{ code })> |
d74e8afc | 1360 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 | 1361 | |
83f32aba RS |
1362 | B<WARNING>: Using this feature safely requires that you understand its |
1363 | limitations. Code executed that has side effects may not perform identically | |
1364 | from version to version due to the effect of future optimisations in the regex | |
1365 | engine. For more information on this, see L</Embedded Code Execution | |
1366 | Frequency>. | |
c277df42 | 1367 | |
e128ab2c DM |
1368 | This zero-width assertion executes any embedded Perl code. It always |
1369 | succeeds, and its return value is set as C<$^R>. | |
19799a22 | 1370 | |
e128ab2c DM |
1371 | In literal patterns, the code is parsed at the same time as the |
1372 | surrounding code. While within the pattern, control is passed temporarily | |
1373 | back to the perl parser, until the logically-balancing closing brace is | |
1374 | encountered. This is similar to the way that an array index expression in | |
1375 | a literal string is handled, for example | |
77ea4f6d | 1376 | |
e128ab2c DM |
1377 | "abc$array[ 1 + f('[') + g()]def" |
1378 | ||
1379 | In particular, braces do not need to be balanced: | |
1380 | ||
576fa024 | 1381 | s/abc(?{ f('{'); })/def/ |
e128ab2c DM |
1382 | |
1383 | Even in a pattern that is interpolated and compiled at run-time, literal | |
1384 | code blocks will be compiled once, at perl compile time; the following | |
1385 | prints "ABCD": | |
1386 | ||
1387 | print "D"; | |
1388 | my $qr = qr/(?{ BEGIN { print "A" } })/; | |
1389 | my $foo = "foo"; | |
1390 | /$foo$qr(?{ BEGIN { print "B" } })/; | |
1391 | BEGIN { print "C" } | |
1392 | ||
1393 | In patterns where the text of the code is derived from run-time | |
1394 | information rather than appearing literally in a source code /pattern/, | |
1395 | the code is compiled at the same time that the pattern is compiled, and | |
5771dda0 | 1396 | for reasons of security, C<use re 'eval'> must be in scope. This is to |
e128ab2c DM |
1397 | stop user-supplied patterns containing code snippets from being |
1398 | executable. | |
1399 | ||
5771dda0 | 1400 | In situations where you need to enable this with C<use re 'eval'>, you should |
e128ab2c DM |
1401 | also have taint checking enabled. Better yet, use the carefully |
1402 | constrained evaluation within a Safe compartment. See L<perlsec> for | |
1403 | details about both these mechanisms. | |
1404 | ||
1405 | From the viewpoint of parsing, lexical variable scope and closures, | |
1406 | ||
1407 | /AAA(?{ BBB })CCC/ | |
1408 | ||
1409 | behaves approximately like | |
1410 | ||
1411 | /AAA/ && do { BBB } && /CCC/ | |
1412 | ||
1413 | Similarly, | |
1414 | ||
1415 | qr/AAA(?{ BBB })CCC/ | |
1416 | ||
1417 | behaves approximately like | |
77ea4f6d | 1418 | |
e128ab2c DM |
1419 | sub { /AAA/ && do { BBB } && /CCC/ } |
1420 | ||
1421 | In particular: | |
1422 | ||
1423 | { my $i = 1; $r = qr/(?{ print $i })/ } | |
1424 | my $i = 2; | |
1425 | /$r/; # prints "1" | |
1426 | ||
1427 | Inside a C<(?{...})> block, C<$_> refers to the string the regular | |
754091cb | 1428 | expression is matching against. You can also use C<pos()> to know what is |
fa11829f | 1429 | the current position of matching within this string. |
754091cb | 1430 | |
e128ab2c DM |
1431 | The code block introduces a new scope from the perspective of lexical |
1432 | variable declarations, but B<not> from the perspective of C<local> and | |
1433 | similar localizing behaviours. So later code blocks within the same | |
1434 | pattern will still see the values which were localized in earlier blocks. | |
1435 | These accumulated localizations are undone either at the end of a | |
1436 | successful match, or if the assertion is backtracked (compare | |
1437 | L<"Backtracking">). For example, | |
b9ac3b5b GS |
1438 | |
1439 | $_ = 'a' x 8; | |
5d458dd8 | 1440 | m< |
d1fbf752 | 1441 | (?{ $cnt = 0 }) # Initialize $cnt. |
b9ac3b5b | 1442 | ( |
5d458dd8 | 1443 | a |
b9ac3b5b | 1444 | (?{ |
d1fbf752 KW |
1445 | local $cnt = $cnt + 1; # Update $cnt, |
1446 | # backtracking-safe. | |
b9ac3b5b | 1447 | }) |
5d458dd8 | 1448 | )* |
b9ac3b5b | 1449 | aaaa |
d1fbf752 KW |
1450 | (?{ $res = $cnt }) # On success copy to |
1451 | # non-localized location. | |
b9ac3b5b GS |
1452 | >x; |
1453 | ||
e128ab2c DM |
1454 | will initially increment C<$cnt> up to 8; then during backtracking, its |
1455 | value will be unwound back to 4, which is the value assigned to C<$res>. | |
1456 | At the end of the regex execution, $cnt will be wound back to its initial | |
1457 | value of 0. | |
1458 | ||
1459 | This assertion may be used as the condition in a | |
b9ac3b5b | 1460 | |
e128ab2c DM |
1461 | (?(condition)yes-pattern|no-pattern) |
1462 | ||
1463 | switch. If I<not> used in this way, the result of evaluation of C<code> | |
1464 | is put into the special variable C<$^R>. This happens immediately, so | |
1465 | C<$^R> can be used from other C<(?{ code })> assertions inside the same | |
1466 | regular expression. | |
b9ac3b5b | 1467 | |
19799a22 GS |
1468 | The assignment to C<$^R> above is properly localized, so the old |
1469 | value of C<$^R> is restored if the assertion is backtracked; compare | |
1470 | L<"Backtracking">. | |
b9ac3b5b | 1471 | |
e128ab2c DM |
1472 | Note that the special variable C<$^N> is particularly useful with code |
1473 | blocks to capture the results of submatches in variables without having to | |
1474 | keep track of the number of nested parentheses. For example: | |
1475 | ||
1476 | $_ = "The brown fox jumps over the lazy dog"; | |
1477 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
1478 | print "color = $color, animal = $animal\n"; | |
1479 | ||
8988a1bb | 1480 | |
14455d6c | 1481 | =item C<(??{ code })> |
d74e8afc ITB |
1482 | X<(??{})> |
1483 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> | |
0f5d15d6 | 1484 | |
83f32aba RS |
1485 | B<WARNING>: Using this feature safely requires that you understand its |
1486 | limitations. Code executed that has side effects may not perform | |
1487 | identically from version to version due to the effect of future | |
1488 | optimisations in the regex engine. For more information on this, see | |
1489 | L</Embedded Code Execution Frequency>. | |
0f5d15d6 | 1490 | |
e128ab2c DM |
1491 | This is a "postponed" regular subexpression. It behaves in I<exactly> the |
1492 | same way as a C<(?{ code })> code block as described above, except that | |
1493 | its return value, rather than being assigned to C<$^R>, is treated as a | |
1494 | pattern, compiled if it's a string (or used as-is if its a qr// object), | |
1495 | then matched as if it were inserted instead of this construct. | |
6bda09f9 | 1496 | |
e128ab2c DM |
1497 | During the matching of this sub-pattern, it has its own set of |
1498 | captures which are valid during the sub-match, but are discarded once | |
1499 | control returns to the main pattern. For example, the following matches, | |
1500 | with the inner pattern capturing "B" and matching "BB", while the outer | |
1501 | pattern captures "A"; | |
1502 | ||
1503 | my $inner = '(.)\1'; | |
1504 | "ABBA" =~ /^(.)(??{ $inner })\1/; | |
1505 | print $1; # prints "A"; | |
6bda09f9 | 1506 | |
e128ab2c DM |
1507 | Note that this means that there is no way for the inner pattern to refer |
1508 | to a capture group defined outside. (The code block itself can use C<$1>, | |
1509 | etc., to refer to the enclosing pattern's capture groups.) Thus, although | |
0f5d15d6 | 1510 | |
e128ab2c DM |
1511 | ('a' x 100)=~/(??{'(.)' x 100})/ |
1512 | ||
1513 | I<will> match, it will I<not> set $1 on exit. | |
19799a22 GS |
1514 | |
1515 | The following pattern matches a parenthesized group: | |
0f5d15d6 | 1516 | |
d1fbf752 KW |
1517 | $re = qr{ |
1518 | \( | |
1519 | (?: | |
1520 | (?> [^()]+ ) # Non-parens without backtracking | |
1521 | | | |
1522 | (??{ $re }) # Group with matching parens | |
1523 | )* | |
1524 | \) | |
1525 | }x; | |
0f5d15d6 | 1526 | |
93f313ef KW |
1527 | See also |
1528 | L<C<(?I<PARNO>)>|/(?PARNO) (?-PARNO) (?+PARNO) (?R) (?0)> | |
1529 | for a different, more efficient way to accomplish | |
6bda09f9 YO |
1530 | the same task. |
1531 | ||
e128ab2c DM |
1532 | Executing a postponed regular expression 50 times without consuming any |
1533 | input string will result in a fatal error. The maximum depth is compiled | |
1534 | into perl, so changing it requires a custom build. | |
6bda09f9 | 1535 | |
93f313ef | 1536 | =item C<(?I<PARNO>)> C<(?-I<PARNO>)> C<(?+I<PARNO>)> C<(?R)> C<(?0)> |
542fa716 | 1537 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)> |
6bda09f9 | 1538 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
d1b2014a YO |
1539 | X<regex, relative recursion> X<GOSUB> X<GOSTART> |
1540 | ||
1541 | Recursive subpattern. Treat the contents of a given capture buffer in the | |
1542 | current pattern as an independent subpattern and attempt to match it at | |
1543 | the current position in the string. Information about capture state from | |
1544 | the caller for things like backreferences is available to the subpattern, | |
1545 | but capture buffers set by the subpattern are not visible to the caller. | |
6bda09f9 | 1546 | |
e128ab2c DM |
1547 | Similar to C<(??{ code })> except that it does not involve executing any |
1548 | code or potentially compiling a returned pattern string; instead it treats | |
1549 | the part of the current pattern contained within a specified capture group | |
d1b2014a YO |
1550 | as an independent pattern that must match at the current position. Also |
1551 | different is the treatment of capture buffers, unlike C<(??{ code })> | |
1552 | recursive patterns have access to their callers match state, so one can | |
1553 | use backreferences safely. | |
6bda09f9 | 1554 | |
93f313ef | 1555 | I<PARNO> is a sequence of digits (not starting with 0) whose value reflects |
c27a5cfe | 1556 | the paren-number of the capture group to recurse to. C<(?R)> recurses to |
894be9b7 | 1557 | the beginning of the whole pattern. C<(?0)> is an alternate syntax for |
93f313ef | 1558 | C<(?R)>. If I<PARNO> is preceded by a plus or minus sign then it is assumed |
c27a5cfe | 1559 | to be relative, with negative numbers indicating preceding capture groups |
542fa716 | 1560 | and positive ones following. Thus C<(?-1)> refers to the most recently |
c27a5cfe | 1561 | declared group, and C<(?+1)> indicates the next group to be declared. |
c74340f9 | 1562 | Note that the counting for relative recursion differs from that of |
c27a5cfe | 1563 | relative backreferences, in that with recursion unclosed groups B<are> |
c74340f9 | 1564 | included. |
6bda09f9 | 1565 | |
81714fb9 | 1566 | The following pattern matches a function foo() which may contain |
f145b7e9 | 1567 | balanced parentheses as the argument. |
6bda09f9 | 1568 | |
d1fbf752 | 1569 | $re = qr{ ( # paren group 1 (full function) |
81714fb9 | 1570 | foo |
d1fbf752 | 1571 | ( # paren group 2 (parens) |
6bda09f9 | 1572 | \( |
d1fbf752 | 1573 | ( # paren group 3 (contents of parens) |
6bda09f9 | 1574 | (?: |
d1fbf752 | 1575 | (?> [^()]+ ) # Non-parens without backtracking |
6bda09f9 | 1576 | | |
d1fbf752 | 1577 | (?2) # Recurse to start of paren group 2 |
6bda09f9 YO |
1578 | )* |
1579 | ) | |
1580 | \) | |
1581 | ) | |
1582 | ) | |
1583 | }x; | |
1584 | ||
1585 | If the pattern was used as follows | |
1586 | ||
1587 | 'foo(bar(baz)+baz(bop))'=~/$re/ | |
1588 | and print "\$1 = $1\n", | |
1589 | "\$2 = $2\n", | |
1590 | "\$3 = $3\n"; | |
1591 | ||
1592 | the output produced should be the following: | |
1593 | ||
1594 | $1 = foo(bar(baz)+baz(bop)) | |
1595 | $2 = (bar(baz)+baz(bop)) | |
81714fb9 | 1596 | $3 = bar(baz)+baz(bop) |
6bda09f9 | 1597 | |
c27a5cfe | 1598 | If there is no corresponding capture group defined, then it is a |
61528107 | 1599 | fatal error. Recursing deeper than 50 times without consuming any input |
81714fb9 | 1600 | string will also result in a fatal error. The maximum depth is compiled |
6bda09f9 YO |
1601 | into perl, so changing it requires a custom build. |
1602 | ||
542fa716 YO |
1603 | The following shows how using negative indexing can make it |
1604 | easier to embed recursive patterns inside of a C<qr//> construct | |
1605 | for later use: | |
1606 | ||
1607 | my $parens = qr/(\((?:[^()]++|(?-1))*+\))/; | |
c77257ed | 1608 | if (/foo $parens \s+ \+ \s+ bar $parens/x) { |
542fa716 YO |
1609 | # do something here... |
1610 | } | |
1611 | ||
81714fb9 | 1612 | B<Note> that this pattern does not behave the same way as the equivalent |
0d017f4d | 1613 | PCRE or Python construct of the same form. In Perl you can backtrack into |
6bda09f9 | 1614 | a recursed group, in PCRE and Python the recursed into group is treated |
542fa716 YO |
1615 | as atomic. Also, modifiers are resolved at compile time, so constructs |
1616 | like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will | |
1617 | be processed. | |
6bda09f9 | 1618 | |
894be9b7 YO |
1619 | =item C<(?&NAME)> |
1620 | X<(?&NAME)> | |
1621 | ||
93f313ef | 1622 | Recurse to a named subpattern. Identical to C<(?I<PARNO>)> except that the |
0d017f4d | 1623 | parenthesis to recurse to is determined by name. If multiple parentheses have |
894be9b7 YO |
1624 | the same name, then it recurses to the leftmost. |
1625 | ||
1626 | It is an error to refer to a name that is not declared somewhere in the | |
1627 | pattern. | |
1628 | ||
1f1031fe YO |
1629 | B<NOTE:> In order to make things easier for programmers with experience |
1630 | with the Python or PCRE regex engines the pattern C<< (?P>NAME) >> | |
64c5a566 | 1631 | may be used instead of C<< (?&NAME) >>. |
1f1031fe | 1632 | |
e2e6a0f1 YO |
1633 | =item C<(?(condition)yes-pattern|no-pattern)> |
1634 | X<(?()> | |
286f584a | 1635 | |
e2e6a0f1 | 1636 | =item C<(?(condition)yes-pattern)> |
286f584a | 1637 | |
41ef34de ML |
1638 | Conditional expression. Matches C<yes-pattern> if C<condition> yields |
1639 | a true value, matches C<no-pattern> otherwise. A missing pattern always | |
1640 | matches. | |
1641 | ||
25e26d77 | 1642 | C<(condition)> should be one of: 1) an integer in |
e2e6a0f1 | 1643 | parentheses (which is valid if the corresponding pair of parentheses |
25e26d77 | 1644 | matched); 2) a look-ahead/look-behind/evaluate zero-width assertion; 3) a |
c27a5cfe | 1645 | name in angle brackets or single quotes (which is valid if a group |
25e26d77 | 1646 | with the given name matched); or 4) the special symbol (R) (true when |
e2e6a0f1 YO |
1647 | evaluated inside of recursion or eval). Additionally the R may be |
1648 | followed by a number, (which will be true when evaluated when recursing | |
1649 | inside of the appropriate group), or by C<&NAME>, in which case it will | |
1650 | be true only when evaluated during recursion in the named group. | |
1651 | ||
1652 | Here's a summary of the possible predicates: | |
1653 | ||
1654 | =over 4 | |
1655 | ||
1656 | =item (1) (2) ... | |
1657 | ||
c27a5cfe | 1658 | Checks if the numbered capturing group has matched something. |
e2e6a0f1 YO |
1659 | |
1660 | =item (<NAME>) ('NAME') | |
1661 | ||
c27a5cfe | 1662 | Checks if a group with the given name has matched something. |
e2e6a0f1 | 1663 | |
f01cd190 FC |
1664 | =item (?=...) (?!...) (?<=...) (?<!...) |
1665 | ||
1666 | Checks whether the pattern matches (or does not match, for the '!' | |
1667 | variants). | |
1668 | ||
e2e6a0f1 YO |
1669 | =item (?{ CODE }) |
1670 | ||
f01cd190 | 1671 | Treats the return value of the code block as the condition. |
e2e6a0f1 YO |
1672 | |
1673 | =item (R) | |
1674 | ||
1675 | Checks if the expression has been evaluated inside of recursion. | |
1676 | ||
1677 | =item (R1) (R2) ... | |
1678 | ||
1679 | Checks if the expression has been evaluated while executing directly | |
1680 | inside of the n-th capture group. This check is the regex equivalent of | |
1681 | ||
1682 | if ((caller(0))[3] eq 'subname') { ... } | |
1683 | ||
1684 | In other words, it does not check the full recursion stack. | |
1685 | ||
1686 | =item (R&NAME) | |
1687 | ||
1688 | Similar to C<(R1)>, this predicate checks to see if we're executing | |
1689 | directly inside of the leftmost group with a given name (this is the same | |
1690 | logic used by C<(?&NAME)> to disambiguate). It does not check the full | |
1691 | stack, but only the name of the innermost active recursion. | |
1692 | ||
1693 | =item (DEFINE) | |
1694 | ||
1695 | In this case, the yes-pattern is never directly executed, and no | |
1696 | no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient. | |
1697 | See below for details. | |
1698 | ||
1699 | =back | |
1700 | ||
1701 | For example: | |
1702 | ||
1703 | m{ ( \( )? | |
1704 | [^()]+ | |
1705 | (?(1) \) ) | |
1706 | }x | |
1707 | ||
1708 | matches a chunk of non-parentheses, possibly included in parentheses | |
1709 | themselves. | |
1710 | ||
0b928c2f FC |
1711 | A special form is the C<(DEFINE)> predicate, which never executes its |
1712 | yes-pattern directly, and does not allow a no-pattern. This allows one to | |
1713 | define subpatterns which will be executed only by the recursion mechanism. | |
e2e6a0f1 YO |
1714 | This way, you can define a set of regular expression rules that can be |
1715 | bundled into any pattern you choose. | |
1716 | ||
1717 | It is recommended that for this usage you put the DEFINE block at the | |
1718 | end of the pattern, and that you name any subpatterns defined within it. | |
1719 | ||
1720 | Also, it's worth noting that patterns defined this way probably will | |
31dc26d6 | 1721 | not be as efficient, as the optimizer is not very clever about |
e2e6a0f1 YO |
1722 | handling them. |
1723 | ||
1724 | An example of how this might be used is as follows: | |
1725 | ||
2bf803e2 | 1726 | /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT)) |
e2e6a0f1 | 1727 | (?(DEFINE) |
2bf803e2 | 1728 | (?<NAME_PAT>....) |
22dc6719 | 1729 | (?<ADDRESS_PAT>....) |
e2e6a0f1 YO |
1730 | )/x |
1731 | ||
c27a5cfe KW |
1732 | Note that capture groups matched inside of recursion are not accessible |
1733 | after the recursion returns, so the extra layer of capturing groups is | |
e2e6a0f1 YO |
1734 | necessary. Thus C<$+{NAME_PAT}> would not be defined even though |
1735 | C<$+{NAME}> would be. | |
286f584a | 1736 | |
51a1303c BF |
1737 | Finally, keep in mind that subpatterns created inside a DEFINE block |
1738 | count towards the absolute and relative number of captures, so this: | |
1739 | ||
1740 | my @captures = "a" =~ /(.) # First capture | |
1741 | (?(DEFINE) | |
1742 | (?<EXAMPLE> 1 ) # Second capture | |
1743 | )/x; | |
1744 | say scalar @captures; | |
1745 | ||
1746 | Will output 2, not 1. This is particularly important if you intend to | |
1747 | compile the definitions with the C<qr//> operator, and later | |
1748 | interpolate them in another pattern. | |
1749 | ||
c47ff5f1 | 1750 | =item C<< (?>pattern) >> |
6bda09f9 | 1751 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 | 1752 | |
19799a22 GS |
1753 | An "independent" subexpression, one which matches the substring |
1754 | that a I<standalone> C<pattern> would match if anchored at the given | |
9da458fc | 1755 | position, and it matches I<nothing other than this substring>. This |
19799a22 GS |
1756 | construct is useful for optimizations of what would otherwise be |
1757 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). | |
9da458fc IZ |
1758 | It may also be useful in places where the "grab all you can, and do not |
1759 | give anything back" semantic is desirable. | |
19799a22 | 1760 | |
c47ff5f1 | 1761 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 GS |
1762 | (anchored at the beginning of string, as above) will match I<all> |
1763 | characters C<a> at the beginning of string, leaving no C<a> for | |
1764 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, | |
1765 | since the match of the subgroup C<a*> is influenced by the following | |
1766 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside | |
1767 | C<a*ab> will match fewer characters than a standalone C<a*>, since | |
1768 | this makes the tail match. | |
1769 | ||
0b928c2f FC |
1770 | C<< (?>pattern) >> does not disable backtracking altogether once it has |
1771 | matched. It is still possible to backtrack past the construct, but not | |
1772 | into it. So C<< ((?>a*)|(?>b*))ar >> will still match "bar". | |
1773 | ||
c47ff5f1 | 1774 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
0b928c2f FC |
1775 | C<(?=(pattern))\g{-1}>. This matches the same substring as a standalone |
1776 | C<a+>, and the following C<\g{-1}> eats the matched string; it therefore | |
c47ff5f1 | 1777 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 GS |
1778 | (The difference between these two constructs is that the second one |
1779 | uses a capturing group, thus shifting ordinals of backreferences | |
1780 | in the rest of a regular expression.) | |
1781 | ||
1782 | Consider this pattern: | |
c277df42 | 1783 | |
871b0233 | 1784 | m{ \( |
e2e6a0f1 | 1785 | ( |
f793d64a | 1786 | [^()]+ # x+ |
e2e6a0f1 | 1787 | | |
871b0233 IZ |
1788 | \( [^()]* \) |
1789 | )+ | |
e2e6a0f1 | 1790 | \) |
871b0233 | 1791 | }x |
5a964f20 | 1792 | |
19799a22 GS |
1793 | That will efficiently match a nonempty group with matching parentheses |
1794 | two levels deep or less. However, if there is no such group, it | |
1795 | will take virtually forever on a long string. That's because there | |
1796 | are so many different ways to split a long string into several | |
1797 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
1798 | to a subpattern of the above pattern. Consider how the pattern | |
1799 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
1800 | seconds, but that each extra letter doubles this time. This | |
1801 | exponential performance will make it appear that your program has | |
14218588 | 1802 | hung. However, a tiny change to this pattern |
5a964f20 | 1803 | |
e2e6a0f1 YO |
1804 | m{ \( |
1805 | ( | |
f793d64a | 1806 | (?> [^()]+ ) # change x+ above to (?> x+ ) |
e2e6a0f1 | 1807 | | |
871b0233 IZ |
1808 | \( [^()]* \) |
1809 | )+ | |
e2e6a0f1 | 1810 | \) |
871b0233 | 1811 | }x |
c277df42 | 1812 | |
c47ff5f1 | 1813 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 TC |
1814 | this yourself would be a productive exercise), but finishes in a fourth |
1815 | the time when used on a similar string with 1000000 C<a>s. Be aware, | |
0b928c2f FC |
1816 | however, that, when this construct is followed by a |
1817 | quantifier, it currently triggers a warning message under | |
9f1b1f2d | 1818 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b | 1819 | C<"matches null string many times in regex">. |
c277df42 | 1820 | |
c47ff5f1 | 1821 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 | 1822 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 IZ |
1823 | This was only 4 times slower on a string with 1000000 C<a>s. |
1824 | ||
9da458fc IZ |
1825 | The "grab all you can, and do not give anything back" semantic is desirable |
1826 | in many situations where on the first sight a simple C<()*> looks like | |
1827 | the correct solution. Suppose we parse text with comments being delimited | |
1828 | by C<#> followed by some optional (horizontal) whitespace. Contrary to | |
4375e838 | 1829 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc IZ |
1830 | the comment delimiter, because it may "give up" some whitespace if |
1831 | the remainder of the pattern can be made to match that way. The correct | |
1832 | answer is either one of these: | |
1833 | ||
1834 | (?>#[ \t]*) | |
1835 | #[ \t]*(?![ \t]) | |
1836 | ||
1837 | For example, to grab non-empty comments into $1, one should use either | |
1838 | one of these: | |
1839 | ||
1840 | / (?> \# [ \t]* ) ( .+ ) /x; | |
1841 | / \# [ \t]* ( [^ \t] .* ) /x; | |
1842 | ||
1843 | Which one you pick depends on which of these expressions better reflects | |
1844 | the above specification of comments. | |
1845 | ||
6bda09f9 YO |
1846 | In some literature this construct is called "atomic matching" or |
1847 | "possessive matching". | |
1848 | ||
b9b4dddf YO |
1849 | Possessive quantifiers are equivalent to putting the item they are applied |
1850 | to inside of one of these constructs. The following equivalences apply: | |
1851 | ||
1852 | Quantifier Form Bracketing Form | |
1853 | --------------- --------------- | |
1854 | PAT*+ (?>PAT*) | |
1855 | PAT++ (?>PAT+) | |
1856 | PAT?+ (?>PAT?) | |
1857 | PAT{min,max}+ (?>PAT{min,max}) | |
1858 | ||
9d1a5160 | 1859 | =item C<(?[ ])> |
f4f5fe57 | 1860 | |
572224ce | 1861 | See L<perlrecharclass/Extended Bracketed Character Classes>. |
9d1a5160 | 1862 | |
e2e6a0f1 YO |
1863 | =back |
1864 | ||
1865 | =head2 Special Backtracking Control Verbs | |
1866 | ||
e2e6a0f1 YO |
1867 | These special patterns are generally of the form C<(*VERB:ARG)>. Unless |
1868 | otherwise stated the ARG argument is optional; in some cases, it is | |
fee50582 | 1869 | mandatory. |
e2e6a0f1 YO |
1870 | |
1871 | Any pattern containing a special backtracking verb that allows an argument | |
e1020413 | 1872 | has the special behaviour that when executed it sets the current package's |
5d458dd8 YO |
1873 | C<$REGERROR> and C<$REGMARK> variables. When doing so the following |
1874 | rules apply: | |
e2e6a0f1 | 1875 | |
5d458dd8 YO |
1876 | On failure, the C<$REGERROR> variable will be set to the ARG value of the |
1877 | verb pattern, if the verb was involved in the failure of the match. If the | |
1878 | ARG part of the pattern was omitted, then C<$REGERROR> will be set to the | |
1879 | name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was | |
1880 | none. Also, the C<$REGMARK> variable will be set to FALSE. | |
e2e6a0f1 | 1881 | |
5d458dd8 YO |
1882 | On a successful match, the C<$REGERROR> variable will be set to FALSE, and |
1883 | the C<$REGMARK> variable will be set to the name of the last | |
1884 | C<(*MARK:NAME)> pattern executed. See the explanation for the | |
1885 | C<(*MARK:NAME)> verb below for more details. | |
e2e6a0f1 | 1886 | |
5d458dd8 | 1887 | B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1> |
0b928c2f | 1888 | and most other regex-related variables. They are not local to a scope, nor |
5d458dd8 YO |
1889 | readonly, but instead are volatile package variables similar to C<$AUTOLOAD>. |
1890 | Use C<local> to localize changes to them to a specific scope if necessary. | |
e2e6a0f1 YO |
1891 | |
1892 | If a pattern does not contain a special backtracking verb that allows an | |
5d458dd8 | 1893 | argument, then C<$REGERROR> and C<$REGMARK> are not touched at all. |
e2e6a0f1 | 1894 | |
70ca8714 | 1895 | =over 3 |
e2e6a0f1 | 1896 | |
fee50582 | 1897 | =item Verbs |
e2e6a0f1 YO |
1898 | |
1899 | =over 4 | |
1900 | ||
5d458dd8 | 1901 | =item C<(*PRUNE)> C<(*PRUNE:NAME)> |
f7819f85 | 1902 | X<(*PRUNE)> X<(*PRUNE:NAME)> |
54612592 | 1903 | |
5d458dd8 YO |
1904 | This zero-width pattern prunes the backtracking tree at the current point |
1905 | when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>, | |
1906 | where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached, | |
1907 | A may backtrack as necessary to match. Once it is reached, matching | |
1908 | continues in B, which may also backtrack as necessary; however, should B | |
1909 | not match, then no further backtracking will take place, and the pattern | |
1910 | will fail outright at the current starting position. | |
54612592 YO |
1911 | |
1912 | The following example counts all the possible matching strings in a | |
1913 | pattern (without actually matching any of them). | |
1914 | ||
e2e6a0f1 | 1915 | 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1916 | print "Count=$count\n"; |
1917 | ||
1918 | which produces: | |
1919 | ||
1920 | aaab | |
1921 | aaa | |
1922 | aa | |
1923 | a | |
1924 | aab | |
1925 | aa | |
1926 | a | |
1927 | ab | |
1928 | a | |
1929 | Count=9 | |
1930 | ||
5d458dd8 | 1931 | If we add a C<(*PRUNE)> before the count like the following |
54612592 | 1932 | |
5d458dd8 | 1933 | 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1934 | print "Count=$count\n"; |
1935 | ||
0b928c2f | 1936 | we prevent backtracking and find the count of the longest matching string |
353c6505 | 1937 | at each matching starting point like so: |
54612592 YO |
1938 | |
1939 | aaab | |
1940 | aab | |
1941 | ab | |
1942 | Count=3 | |
1943 | ||
5d458dd8 | 1944 | Any number of C<(*PRUNE)> assertions may be used in a pattern. |
54612592 | 1945 | |
5d458dd8 YO |
1946 | See also C<< (?>pattern) >> and possessive quantifiers for other ways to |
1947 | control backtracking. In some cases, the use of C<(*PRUNE)> can be | |
1948 | replaced with a C<< (?>pattern) >> with no functional difference; however, | |
1949 | C<(*PRUNE)> can be used to handle cases that cannot be expressed using a | |
1950 | C<< (?>pattern) >> alone. | |
54612592 | 1951 | |
5d458dd8 YO |
1952 | =item C<(*SKIP)> C<(*SKIP:NAME)> |
1953 | X<(*SKIP)> | |
e2e6a0f1 | 1954 | |
5d458dd8 | 1955 | This zero-width pattern is similar to C<(*PRUNE)>, except that on |
e2e6a0f1 | 1956 | failure it also signifies that whatever text that was matched leading up |
5d458dd8 YO |
1957 | to the C<(*SKIP)> pattern being executed cannot be part of I<any> match |
1958 | of this pattern. This effectively means that the regex engine "skips" forward | |
1959 | to this position on failure and tries to match again, (assuming that | |
1960 | there is sufficient room to match). | |
1961 | ||
1962 | The name of the C<(*SKIP:NAME)> pattern has special significance. If a | |
1963 | C<(*MARK:NAME)> was encountered while matching, then it is that position | |
1964 | which is used as the "skip point". If no C<(*MARK)> of that name was | |
1965 | encountered, then the C<(*SKIP)> operator has no effect. When used | |
1966 | without a name the "skip point" is where the match point was when | |
1967 | executing the (*SKIP) pattern. | |
1968 | ||
0b928c2f | 1969 | Compare the following to the examples in C<(*PRUNE)>; note the string |
24b23f37 YO |
1970 | is twice as long: |
1971 | ||
d1fbf752 KW |
1972 | 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/; |
1973 | print "Count=$count\n"; | |
24b23f37 YO |
1974 | |
1975 | outputs | |
1976 | ||
1977 | aaab | |
1978 | aaab | |
1979 | Count=2 | |
1980 | ||
5d458dd8 | 1981 | Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)> |
353c6505 | 1982 | executed, the next starting point will be where the cursor was when the |
5d458dd8 YO |
1983 | C<(*SKIP)> was executed. |
1984 | ||
5d458dd8 | 1985 | =item C<(*MARK:NAME)> C<(*:NAME)> |
b16db30f | 1986 | X<(*MARK)> X<(*MARK:NAME)> X<(*:NAME)> |
5d458dd8 YO |
1987 | |
1988 | This zero-width pattern can be used to mark the point reached in a string | |
1989 | when a certain part of the pattern has been successfully matched. This | |
1990 | mark may be given a name. A later C<(*SKIP)> pattern will then skip | |
1991 | forward to that point if backtracked into on failure. Any number of | |
b4222fa9 | 1992 | C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated. |
5d458dd8 YO |
1993 | |
1994 | In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)> | |
1995 | can be used to "label" a pattern branch, so that after matching, the | |
1996 | program can determine which branches of the pattern were involved in the | |
1997 | match. | |
1998 | ||
1999 | When a match is successful, the C<$REGMARK> variable will be set to the | |
2000 | name of the most recently executed C<(*MARK:NAME)> that was involved | |
2001 | in the match. | |
2002 | ||
2003 | This can be used to determine which branch of a pattern was matched | |
c27a5cfe | 2004 | without using a separate capture group for each branch, which in turn |
5d458dd8 YO |
2005 | can result in a performance improvement, as perl cannot optimize |
2006 | C</(?:(x)|(y)|(z))/> as efficiently as something like | |
2007 | C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>. | |
2008 | ||
2009 | When a match has failed, and unless another verb has been involved in | |
2010 | failing the match and has provided its own name to use, the C<$REGERROR> | |
2011 | variable will be set to the name of the most recently executed | |
2012 | C<(*MARK:NAME)>. | |
2013 | ||
42ac7c82 | 2014 | See L</(*SKIP)> for more details. |
5d458dd8 | 2015 | |
b62d2d15 YO |
2016 | As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>. |
2017 | ||
5d458dd8 YO |
2018 | =item C<(*THEN)> C<(*THEN:NAME)> |
2019 | ||
ac9d8485 | 2020 | This is similar to the "cut group" operator C<::> from Perl 6. Like |
5d458dd8 YO |
2021 | C<(*PRUNE)>, this verb always matches, and when backtracked into on |
2022 | failure, it causes the regex engine to try the next alternation in the | |
ac9d8485 FC |
2023 | innermost enclosing group (capturing or otherwise) that has alternations. |
2024 | The two branches of a C<(?(condition)yes-pattern|no-pattern)> do not | |
2025 | count as an alternation, as far as C<(*THEN)> is concerned. | |
5d458dd8 YO |
2026 | |
2027 | Its name comes from the observation that this operation combined with the | |
2028 | alternation operator (C<|>) can be used to create what is essentially a | |
2029 | pattern-based if/then/else block: | |
2030 | ||
2031 | ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) | |
2032 | ||
2033 | Note that if this operator is used and NOT inside of an alternation then | |
2034 | it acts exactly like the C<(*PRUNE)> operator. | |
2035 | ||
2036 | / A (*PRUNE) B / | |
2037 | ||
2038 | is the same as | |
2039 | ||
2040 | / A (*THEN) B / | |
2041 | ||
2042 | but | |
2043 | ||
25e26d77 | 2044 | / ( A (*THEN) B | C ) / |
5d458dd8 YO |
2045 | |
2046 | is not the same as | |
2047 | ||
25e26d77 | 2048 | / ( A (*PRUNE) B | C ) / |
5d458dd8 YO |
2049 | |
2050 | as after matching the A but failing on the B the C<(*THEN)> verb will | |
2051 | backtrack and try C; but the C<(*PRUNE)> verb will simply fail. | |
24b23f37 | 2052 | |
fee50582 | 2053 | =item C<(*COMMIT)> C<(*COMMIT:args)> |
e2e6a0f1 | 2054 | X<(*COMMIT)> |
24b23f37 | 2055 | |
241e7389 | 2056 | This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a |
5d458dd8 YO |
2057 | zero-width pattern similar to C<(*SKIP)>, except that when backtracked |
2058 | into on failure it causes the match to fail outright. No further attempts | |
2059 | to find a valid match by advancing the start pointer will occur again. | |
2060 | For example, | |
24b23f37 | 2061 | |
d1fbf752 KW |
2062 | 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/; |
2063 | print "Count=$count\n"; | |
24b23f37 YO |
2064 | |
2065 | outputs | |
2066 | ||
2067 | aaab | |
2068 | Count=1 | |
2069 | ||
e2e6a0f1 YO |
2070 | In other words, once the C<(*COMMIT)> has been entered, and if the pattern |
2071 | does not match, the regex engine will not try any further matching on the | |
2072 | rest of the string. | |
c277df42 | 2073 | |
fee50582 | 2074 | =item C<(*FAIL)> C<(*F)> C<(*FAIL:arg)> |
e2e6a0f1 | 2075 | X<(*FAIL)> X<(*F)> |
9af228c6 | 2076 | |
e2e6a0f1 YO |
2077 | This pattern matches nothing and always fails. It can be used to force the |
2078 | engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In | |
fee50582 YO |
2079 | fact, C<(?!)> gets optimised into C<(*FAIL)> internally. You can provide |
2080 | an argument so that if the match fails because of this FAIL directive | |
2081 | the argument can be obtained from $REGERROR. | |
9af228c6 | 2082 | |
e2e6a0f1 | 2083 | It is probably useful only when combined with C<(?{})> or C<(??{})>. |
9af228c6 | 2084 | |
fee50582 | 2085 | =item C<(*ACCEPT)> C<(*ACCEPT:arg)> |
e2e6a0f1 | 2086 | X<(*ACCEPT)> |
9af228c6 | 2087 | |
e2e6a0f1 YO |
2088 | This pattern matches nothing and causes the end of successful matching at |
2089 | the point at which the C<(*ACCEPT)> pattern was encountered, regardless of | |
2090 | whether there is actually more to match in the string. When inside of a | |
0d017f4d | 2091 | nested pattern, such as recursion, or in a subpattern dynamically generated |
e2e6a0f1 | 2092 | via C<(??{})>, only the innermost pattern is ended immediately. |
9af228c6 | 2093 | |
c27a5cfe | 2094 | If the C<(*ACCEPT)> is inside of capturing groups then the groups are |
e2e6a0f1 YO |
2095 | marked as ended at the point at which the C<(*ACCEPT)> was encountered. |
2096 | For instance: | |
9af228c6 | 2097 | |
e2e6a0f1 | 2098 | 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x; |
9af228c6 | 2099 | |
e2e6a0f1 | 2100 | will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not |
0b928c2f | 2101 | be set. If another branch in the inner parentheses was matched, such as in the |
e2e6a0f1 | 2102 | string 'ACDE', then the C<D> and C<E> would have to be matched as well. |
9af228c6 | 2103 | |
fee50582 YO |
2104 | You can provide an argument, which will be available in the var $REGMARK |
2105 | after the match completes. | |
2106 | ||
9af228c6 | 2107 | =back |
c277df42 | 2108 | |
a0d0e21e LW |
2109 | =back |
2110 | ||
c07a80fd | 2111 | =head2 Backtracking |
d74e8afc | 2112 | X<backtrack> X<backtracking> |
c07a80fd | 2113 | |
35a734be IZ |
2114 | NOTE: This section presents an abstract approximation of regular |
2115 | expression behavior. For a more rigorous (and complicated) view of | |
2116 | the rules involved in selecting a match among possible alternatives, | |
0d017f4d | 2117 | see L<Combining RE Pieces>. |
35a734be | 2118 | |
c277df42 | 2119 | A fundamental feature of regular expression matching involves the |
5a964f20 | 2120 | notion called I<backtracking>, which is currently used (when needed) |
0d017f4d | 2121 | by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc IZ |
2122 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
2123 | internally, but the general principle outlined here is valid. | |
c07a80fd | 2124 | |
2125 | For a regular expression to match, the I<entire> regular expression must | |
2126 | match, not just part of it. So if the beginning of a pattern containing a | |
2127 | quantifier succeeds in a way that causes later parts in the pattern to | |
2128 | fail, the matching engine backs up and recalculates the beginning | |
2129 | part--that's why it's called backtracking. | |
2130 | ||
2131 | Here is an example of backtracking: Let's say you want to find the | |
2132 | word following "foo" in the string "Food is on the foo table.": | |
2133 | ||
2134 | $_ = "Food is on the foo table."; | |
2135 | if ( /\b(foo)\s+(\w+)/i ) { | |
f793d64a | 2136 | print "$2 follows $1.\n"; |
c07a80fd | 2137 | } |
2138 | ||
2139 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
2140 | finds a possible match right at the beginning of the string, and loads up | |
2141 | $1 with "Foo". However, as soon as the matching engine sees that there's | |
2142 | no whitespace following the "Foo" that it had saved in $1, it realizes its | |
68dc0745 | 2143 | mistake and starts over again one character after where it had the |
c07a80fd | 2144 | tentative match. This time it goes all the way until the next occurrence |
2145 | of "foo". The complete regular expression matches this time, and you get | |
2146 | the expected output of "table follows foo." | |
2147 | ||
2148 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
2149 | everything between "foo" and "bar". Initially, you write something | |
2150 | like this: | |
2151 | ||
2152 | $_ = "The food is under the bar in the barn."; | |
2153 | if ( /foo(.*)bar/ ) { | |
f793d64a | 2154 | print "got <$1>\n"; |
c07a80fd | 2155 | } |
2156 | ||
2157 | Which perhaps unexpectedly yields: | |
2158 | ||
2159 | got <d is under the bar in the > | |
2160 | ||
2161 | That's because C<.*> was greedy, so you get everything between the | |
14218588 | 2162 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd | 2163 | to use minimal matching to make sure you get the text between a "foo" |
2164 | and the first "bar" thereafter. | |
2165 | ||
2166 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
2167 | got <d is under the > | |
2168 | ||
0d017f4d | 2169 | Here's another example. Let's say you'd like to match a number at the end |
b6e13d97 | 2170 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd | 2171 | So you write this: |
2172 | ||
2173 | $_ = "I have 2 numbers: 53147"; | |
f793d64a KW |
2174 | if ( /(.*)(\d*)/ ) { # Wrong! |
2175 | print "Beginning is <$1>, number is <$2>.\n"; | |
c07a80fd | 2176 | } |
2177 | ||
2178 | That won't work at all, because C<.*> was greedy and gobbled up the | |
2179 | whole string. As C<\d*> can match on an empty string the complete | |
2180 | regular expression matched successfully. | |
2181 | ||
8e1088bc | 2182 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd | 2183 | |
2184 | Here are some variants, most of which don't work: | |
2185 | ||
2186 | $_ = "I have 2 numbers: 53147"; | |
2187 | @pats = qw{ | |
f793d64a KW |
2188 | (.*)(\d*) |
2189 | (.*)(\d+) | |
2190 | (.*?)(\d*) | |
2191 | (.*?)(\d+) | |
2192 | (.*)(\d+)$ | |
2193 | (.*?)(\d+)$ | |
2194 | (.*)\b(\d+)$ | |
2195 | (.*\D)(\d+)$ | |
c07a80fd | 2196 | }; |
2197 | ||
2198 | for $pat (@pats) { | |
f793d64a KW |
2199 | printf "%-12s ", $pat; |
2200 | if ( /$pat/ ) { | |
2201 | print "<$1> <$2>\n"; | |
2202 | } else { | |
2203 | print "FAIL\n"; | |
2204 | } | |
c07a80fd | 2205 | } |
2206 | ||
2207 | That will print out: | |
2208 | ||
2209 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
2210 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
2211 | (.*?)(\d*) <> <> | |
2212 | (.*?)(\d+) <I have > <2> | |
2213 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
2214 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
2215 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
2216 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
2217 | ||
2218 | As you see, this can be a bit tricky. It's important to realize that a | |
2219 | regular expression is merely a set of assertions that gives a definition | |
2220 | of success. There may be 0, 1, or several different ways that the | |
2221 | definition might succeed against a particular string. And if there are | |
5a964f20 TC |
2222 | multiple ways it might succeed, you need to understand backtracking to |
2223 | know which variety of success you will achieve. | |
c07a80fd | 2224 | |
19799a22 | 2225 | When using look-ahead assertions and negations, this can all get even |
8b19b778 | 2226 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd | 2227 | followed by "123". You might try to write that as |
2228 | ||
871b0233 | 2229 | $_ = "ABC123"; |
f793d64a KW |
2230 | if ( /^\D*(?!123)/ ) { # Wrong! |
2231 | print "Yup, no 123 in $_\n"; | |
871b0233 | 2232 | } |
c07a80fd | 2233 | |
2234 | But that isn't going to match; at least, not the way you're hoping. It | |
2235 | claims that there is no 123 in the string. Here's a clearer picture of | |
9b9391b2 | 2236 | why that pattern matches, contrary to popular expectations: |
c07a80fd | 2237 | |
4358a253 SS |
2238 | $x = 'ABC123'; |
2239 | $y = 'ABC445'; | |
c07a80fd | 2240 | |
4358a253 SS |
2241 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
2242 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; | |
c07a80fd | 2243 | |
4358a253 SS |
2244 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; |
2245 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; | |
c07a80fd | 2246 | |
2247 | This prints | |
2248 | ||
2249 | 2: got ABC | |
2250 | 3: got AB | |
2251 | 4: got ABC | |
2252 | ||
5f05dabc | 2253 | You might have expected test 3 to fail because it seems to a more |
c07a80fd | 2254 | general purpose version of test 1. The important difference between |
2255 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
2256 | backtracking, whereas test 1 will not. What's happening is | |
2257 | that you've asked "Is it true that at the start of $x, following 0 or more | |
5f05dabc | 2258 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd | 2259 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 | 2260 | fail. |
14218588 | 2261 | |
c07a80fd | 2262 | The search engine will initially match C<\D*> with "ABC". Then it will |
0b928c2f | 2263 | try to match C<(?!123)> with "123", which fails. But because |
c07a80fd | 2264 | a quantifier (C<\D*>) has been used in the regular expression, the |
2265 | search engine can backtrack and retry the match differently | |
54310121 | 2266 | in the hope of matching the complete regular expression. |
c07a80fd | 2267 | |
5a964f20 TC |
2268 | The pattern really, I<really> wants to succeed, so it uses the |
2269 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
c07a80fd | 2270 | time. Now there's indeed something following "AB" that is not |
14218588 | 2271 | "123". It's "C123", which suffices. |
c07a80fd | 2272 | |
14218588 GS |
2273 | We can deal with this by using both an assertion and a negation. |
2274 | We'll say that the first part in $1 must be followed both by a digit | |
2275 | and by something that's not "123". Remember that the look-aheads | |
2276 | are zero-width expressions--they only look, but don't consume any | |
2277 | of the string in their match. So rewriting this way produces what | |
c07a80fd | 2278 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
2279 | ||
4358a253 SS |
2280 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; |
2281 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; | |
c07a80fd | 2282 | |
2283 | 6: got ABC | |
2284 | ||
5a964f20 | 2285 | In other words, the two zero-width assertions next to each other work as though |
19799a22 | 2286 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd | 2287 | matches only if you're at the beginning of the line AND the end of the |
2288 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
2289 | regular expressions always means AND, except when you write an explicit OR | |
2290 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
2291 | although the attempted matches are made at different positions because "a" | |
2292 | is not a zero-width assertion, but a one-width assertion. | |
2293 | ||
0d017f4d | 2294 | B<WARNING>: Particularly complicated regular expressions can take |
14218588 | 2295 | exponential time to solve because of the immense number of possible |
0d017f4d | 2296 | ways they can use backtracking to try for a match. For example, without |
9da458fc IZ |
2297 | internal optimizations done by the regular expression engine, this will |
2298 | take a painfully long time to run: | |
c07a80fd | 2299 | |
e1901655 IZ |
2300 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
2301 | ||
2302 | And if you used C<*>'s in the internal groups instead of limiting them | |
2303 | to 0 through 5 matches, then it would take forever--or until you ran | |
2304 | out of stack space. Moreover, these internal optimizations are not | |
2305 | always applicable. For example, if you put C<{0,5}> instead of C<*> | |
2306 | on the external group, no current optimization is applicable, and the | |
2307 | match takes a long time to finish. | |
c07a80fd | 2308 | |
9da458fc IZ |
2309 | A powerful tool for optimizing such beasts is what is known as an |
2310 | "independent group", | |
96090e4f | 2311 | which does not backtrack (see L</C<< (?>pattern) >>>). Note also that |
9da458fc | 2312 | zero-length look-ahead/look-behind assertions will not backtrack to make |
5d458dd8 | 2313 | the tail match, since they are in "logical" context: only |
14218588 | 2314 | whether they match is considered relevant. For an example |
9da458fc | 2315 | where side-effects of look-ahead I<might> have influenced the |
96090e4f | 2316 | following match, see L</C<< (?>pattern) >>>. |
c277df42 | 2317 | |
a0d0e21e | 2318 | =head2 Version 8 Regular Expressions |
d74e8afc | 2319 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> |
a0d0e21e | 2320 | |
5a964f20 | 2321 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e LW |
2322 | routines, here are the pattern-matching rules not described above. |
2323 | ||
54310121 | 2324 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e | 2325 | with a special meaning described here or above. You can cause |
5a964f20 | 2326 | characters that normally function as metacharacters to be interpreted |
5f05dabc | 2327 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
0d017f4d WL |
2328 | character; "\\" matches a "\"). This escape mechanism is also required |
2329 | for the character used as the pattern delimiter. | |
2330 | ||
2331 | A series of characters matches that series of characters in the target | |
0b928c2f | 2332 | string, so the pattern C<blurfl> would match "blurfl" in the target |
0d017f4d | 2333 | string. |
a0d0e21e LW |
2334 | |
2335 | You can specify a character class, by enclosing a list of characters | |
5d458dd8 | 2336 | in C<[]>, which will match any character from the list. If the |
a0d0e21e | 2337 | first character after the "[" is "^", the class matches any character not |
14218588 | 2338 | in the list. Within a list, the "-" character specifies a |
5a964f20 | 2339 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 GS |
2340 | inclusive. If you want either "-" or "]" itself to be a member of a |
2341 | class, put it at the start of the list (possibly after a "^"), or | |
2342 | escape it with a backslash. "-" is also taken literally when it is | |
2343 | at the end of the list, just before the closing "]". (The | |
84850974 DD |
2344 | following all specify the same class of three characters: C<[-az]>, |
2345 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which | |
5d458dd8 YO |
2346 | specifies a class containing twenty-six characters, even on EBCDIC-based |
2347 | character sets.) Also, if you try to use the character | |
2348 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of | |
2349 | a range, the "-" is understood literally. | |
a0d0e21e | 2350 | |
8ada0baa | 2351 | Note also that the whole range idea is rather unportable between |
b927b7e9 KW |
2352 | character sets, except for four situations that Perl handles specially. |
2353 | Any subset of the ranges C<[A-Z]>, C<[a-z]>, and C<[0-9]> are guaranteed | |
2354 | to match the expected subset of ASCII characters, no matter what | |
2355 | character set the platform is running. The fourth portable way to | |
2356 | specify ranges is to use the C<\N{...}> syntax to specify either end | |
2357 | point of the range. For example, C<[\N{U+04}-\N{U+07}]> means to match | |
2358 | the Unicode code points C<\N{U+04}>, C<\N{U+05}>, C<\N{U+06}>, and | |
2359 | C<\N{U+07}>, whatever their native values may be on the platform. Under | |
2360 | L<use re 'strict'|re/'strict' mode> or within a L</C<(?[ ])>>, a warning | |
2361 | is raised, if enabled, and the other end point of a range which has a | |
2362 | C<\N{...}> endpoint is not portably specified. For example, | |
2363 | ||
2364 | [\N{U+00}-\x06] # Warning under "use re 'strict'". | |
2365 | ||
2366 | It is hard to understand without digging what exactly matches ranges | |
2367 | other than subsets of C<[A-Z]>, C<[a-z]>, and C<[0-9]>. A sound | |
2368 | principle is to use only ranges that begin from and end at either | |
2369 | alphabetics of equal case ([a-e], [A-E]), or digits ([0-9]). Anything | |
2370 | else is unsafe or unclear. If in doubt, spell out the range in full. | |
8ada0baa | 2371 | |
54310121 | 2372 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e LW |
2373 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
2374 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string | |
dc0d9c48 | 2375 | of three octal digits, matches the character whose coded character set value |
5d458dd8 | 2376 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, |
dc0d9c48 | 2377 | matches the character whose ordinal is I<nn>. The expression \cI<x> |
5d458dd8 | 2378 | matches the character control-I<x>. Finally, the "." metacharacter |
fb55449c | 2379 | matches any character except "\n" (unless you use C</s>). |
a0d0e21e LW |
2380 | |
2381 | You can specify a series of alternatives for a pattern using "|" to | |
2382 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", | |
5a964f20 | 2383 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e | 2384 | first alternative includes everything from the last pattern delimiter |
0b928c2f | 2385 | ("(", "(?:", etc. or the beginning of the pattern) up to the first "|", and |
a0d0e21e | 2386 | the last alternative contains everything from the last "|" to the next |
0b928c2f | 2387 | closing pattern delimiter. That's why it's common practice to include |
14218588 | 2388 | alternatives in parentheses: to minimize confusion about where they |
a3cb178b GS |
2389 | start and end. |
2390 | ||
5a964f20 | 2391 | Alternatives are tried from left to right, so the first |
a3cb178b GS |
2392 | alternative found for which the entire expression matches, is the one that |
2393 | is chosen. This means that alternatives are not necessarily greedy. For | |
628afcb5 | 2394 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b GS |
2395 | part will match, as that is the first alternative tried, and it successfully |
2396 | matches the target string. (This might not seem important, but it is | |
2397 | important when you are capturing matched text using parentheses.) | |
2398 | ||
5a964f20 | 2399 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b | 2400 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e | 2401 | |
14218588 GS |
2402 | Within a pattern, you may designate subpatterns for later reference |
2403 | by enclosing them in parentheses, and you may refer back to the | |
2404 | I<n>th subpattern later in the pattern using the metacharacter | |
0b928c2f | 2405 | \I<n> or \gI<n>. Subpatterns are numbered based on the left to right order |
14218588 GS |
2406 | of their opening parenthesis. A backreference matches whatever |
2407 | actually matched the subpattern in the string being examined, not | |
d8b950dc | 2408 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\g1\d*> will |
14218588 GS |
2409 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern |
2410 | 1 matched "0x", even though the rule C<0|0x> could potentially match | |
2411 | the leading 0 in the second number. | |
cb1a09d0 | 2412 | |
0d017f4d | 2413 | =head2 Warning on \1 Instead of $1 |
cb1a09d0 | 2414 | |
5a964f20 | 2415 | Some people get too used to writing things like: |
cb1a09d0 AD |
2416 | |
2417 | $pattern =~ s/(\W)/\\\1/g; | |
2418 | ||
3ff1c45a KW |
2419 | This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid |
2420 | shocking the | |
cb1a09d0 | 2421 | B<sed> addicts, but it's a dirty habit to get into. That's because in |
d1be9408 | 2422 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 AD |
2423 | the usual double-quoted string means a control-A. The customary Unix |
2424 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
2425 | of doing that, you get yourself into trouble if you then add an C</e> | |
2426 | modifier. | |
2427 | ||
f793d64a | 2428 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 AD |
2429 | |
2430 | Or if you try to do | |
2431 | ||
2432 | s/(\d+)/\1000/; | |
2433 | ||
2434 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
14218588 | 2435 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 AD |
2436 | with the operation of matching a backreference. Certainly they mean two |
2437 | different things on the I<left> side of the C<s///>. | |
9fa51da4 | 2438 | |
0d017f4d | 2439 | =head2 Repeated Patterns Matching a Zero-length Substring |
c84d73f1 | 2440 | |
19799a22 | 2441 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 IZ |
2442 | |
2443 | Regular expressions provide a terse and powerful programming language. As | |
2444 | with most other power tools, power comes together with the ability | |
2445 | to wreak havoc. | |
2446 | ||
2447 | A common abuse of this power stems from the ability to make infinite | |
628afcb5 | 2448 | loops using regular expressions, with something as innocuous as: |
c84d73f1 IZ |
2449 | |
2450 | 'foo' =~ m{ ( o? )* }x; | |
2451 | ||
0d017f4d | 2452 | The C<o?> matches at the beginning of C<'foo'>, and since the position |
c84d73f1 | 2453 | in the string is not moved by the match, C<o?> would match again and again |
527e91da | 2454 | because of the C<*> quantifier. Another common way to create a similar cycle |
c84d73f1 IZ |
2455 | is with the looping modifier C<//g>: |
2456 | ||
2457 | @matches = ( 'foo' =~ m{ o? }xg ); | |
2458 | ||
2459 | or | |
2460 | ||
2461 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
2462 | ||
2463 | or the loop implied by split(). | |
2464 | ||
2465 | However, long experience has shown that many programming tasks may | |
14218588 GS |
2466 | be significantly simplified by using repeated subexpressions that |
2467 | may match zero-length substrings. Here's a simple example being: | |
c84d73f1 | 2468 | |
d1fbf752 | 2469 | @chars = split //, $string; # // is not magic in split |
c84d73f1 IZ |
2470 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / |
2471 | ||
9da458fc | 2472 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 | 2473 | the infinite loop>. The rules for this are different for lower-level |
527e91da | 2474 | loops given by the greedy quantifiers C<*+{}>, and for higher-level |
c84d73f1 IZ |
2475 | ones like the C</g> modifier or split() operator. |
2476 | ||
19799a22 GS |
2477 | The lower-level loops are I<interrupted> (that is, the loop is |
2478 | broken) when Perl detects that a repeated expression matched a | |
2479 | zero-length substring. Thus | |
c84d73f1 IZ |
2480 | |
2481 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
2482 | ||
5d458dd8 | 2483 | is made equivalent to |
c84d73f1 | 2484 | |
0b928c2f FC |
2485 | m{ (?: NON_ZERO_LENGTH )* (?: ZERO_LENGTH )? }x; |
2486 | ||
2487 | For example, this program | |
2488 | ||
2489 | #!perl -l | |
2490 | "aaaaab" =~ / | |
2491 | (?: | |
2492 | a # non-zero | |
2493 | | # or | |
2494 | (?{print "hello"}) # print hello whenever this | |
2495 | # branch is tried | |
2496 | (?=(b)) # zero-width assertion | |
2497 | )* # any number of times | |
2498 | /x; | |
2499 | print $&; | |
2500 | print $1; | |
c84d73f1 | 2501 | |
0b928c2f FC |
2502 | prints |
2503 | ||
2504 | hello | |
2505 | aaaaa | |
2506 | b | |
2507 | ||
2508 | Notice that "hello" is only printed once, as when Perl sees that the sixth | |
2509 | iteration of the outermost C<(?:)*> matches a zero-length string, it stops | |
2510 | the C<*>. | |
2511 | ||
2512 | The higher-level loops preserve an additional state between iterations: | |
5d458dd8 | 2513 | whether the last match was zero-length. To break the loop, the following |
c84d73f1 | 2514 | match after a zero-length match is prohibited to have a length of zero. |
5d458dd8 | 2515 | This prohibition interacts with backtracking (see L<"Backtracking">), |
c84d73f1 IZ |
2516 | and so the I<second best> match is chosen if the I<best> match is of |
2517 | zero length. | |
2518 | ||
19799a22 | 2519 | For example: |
c84d73f1 IZ |
2520 | |
2521 | $_ = 'bar'; | |
2522 | s/\w??/<$&>/g; | |
2523 | ||
20fb949f | 2524 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
5d458dd8 | 2525 | match given by non-greedy C<??> is the zero-length match, and the I<second |
c84d73f1 IZ |
2526 | best> match is what is matched by C<\w>. Thus zero-length matches |
2527 | alternate with one-character-long matches. | |
2528 | ||
5d458dd8 | 2529 | Similarly, for repeated C<m/()/g> the second-best match is the match at the |
c84d73f1 IZ |
2530 | position one notch further in the string. |
2531 | ||
19799a22 | 2532 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 | 2533 | the matched string, and is reset by each assignment to pos(). |
9da458fc IZ |
2534 | Zero-length matches at the end of the previous match are ignored |
2535 | during C<split>. | |
c84d73f1 | 2536 | |
0d017f4d | 2537 | =head2 Combining RE Pieces |
35a734be IZ |
2538 | |
2539 | Each of the elementary pieces of regular expressions which were described | |
2540 | before (such as C<ab> or C<\Z>) could match at most one substring | |
2541 | at the given position of the input string. However, in a typical regular | |
2542 | expression these elementary pieces are combined into more complicated | |
0b928c2f | 2543 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc. |
35a734be IZ |
2544 | (in these examples C<S> and C<T> are regular subexpressions). |
2545 | ||
2546 | Such combinations can include alternatives, leading to a problem of choice: | |
2547 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
2548 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
2549 | actually matched is the concept of backtracking (see L<"Backtracking">). | |
2550 | However, this description is too low-level and makes you think | |
2551 | in terms of a particular implementation. | |
2552 | ||
2553 | Another description starts with notions of "better"/"worse". All the | |
2554 | substrings which may be matched by the given regular expression can be | |
2555 | sorted from the "best" match to the "worst" match, and it is the "best" | |
2556 | match which is chosen. This substitutes the question of "what is chosen?" | |
2557 | by the question of "which matches are better, and which are worse?". | |
2558 | ||
2559 | Again, for elementary pieces there is no such question, since at most | |
2560 | one match at a given position is possible. This section describes the | |
2561 | notion of better/worse for combining operators. In the description | |
2562 | below C<S> and C<T> are regular subexpressions. | |
2563 | ||
13a2d996 | 2564 | =over 4 |
35a734be IZ |
2565 | |
2566 | =item C<ST> | |
2567 | ||
2568 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are | |
2569 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings | |
5d458dd8 | 2570 | which can be matched by C<T>. |
35a734be | 2571 | |
0b928c2f | 2572 | If C<A> is a better match for C<S> than C<A'>, C<AB> is a better |
35a734be IZ |
2573 | match than C<A'B'>. |
2574 | ||
2575 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if | |
0b928c2f | 2576 | C<B> is a better match for C<T> than C<B'>. |
35a734be IZ |
2577 | |
2578 | =item C<S|T> | |
2579 | ||
2580 | When C<S> can match, it is a better match than when only C<T> can match. | |
2581 | ||
2582 | Ordering of two matches for C<S> is the same as for C<S>. Similar for | |
2583 | two matches for C<T>. | |
2584 | ||
2585 | =item C<S{REPEAT_COUNT}> | |
2586 | ||
2587 | Matches as C<SSS...S> (repeated as many times as necessary). | |
2588 | ||
2589 | =item C<S{min,max}> | |
2590 | ||
2591 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
2592 | ||
2593 | =item C<S{min,max}?> | |
2594 | ||
2595 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
2596 | ||
2597 | =item C<S?>, C<S*>, C<S+> | |
2598 | ||
2599 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
2600 | ||
2601 | =item C<S??>, C<S*?>, C<S+?> | |
2602 | ||
2603 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
2604 | ||
c47ff5f1 | 2605 | =item C<< (?>S) >> |
35a734be IZ |
2606 | |
2607 | Matches the best match for C<S> and only that. | |
2608 | ||
2609 | =item C<(?=S)>, C<(?<=S)> | |
2610 | ||
2611 | Only the best match for C<S> is considered. (This is important only if | |
2612 | C<S> has capturing parentheses, and backreferences are used somewhere | |
2613 | else in the whole regular expression.) | |
2614 | ||
2615 | =item C<(?!S)>, C<(?<!S)> | |
2616 | ||
2617 | For this grouping operator there is no need to describe the ordering, since | |
2618 | only whether or not C<S> can match is important. | |
2619 | ||
93f313ef | 2620 | =item C<(??{ EXPR })>, C<(?I<PARNO>)> |
35a734be IZ |
2621 | |
2622 | The ordering is the same as for the regular expression which is | |
93f313ef | 2623 | the result of EXPR, or the pattern contained by capture group I<PARNO>. |
35a734be IZ |
2624 | |
2625 | =item C<(?(condition)yes-pattern|no-pattern)> | |
2626 | ||
2627 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is | |
2628 | already determined. The ordering of the matches is the same as for the | |
2629 | chosen subexpression. | |
2630 | ||
2631 | =back | |
2632 | ||
2633 | The above recipes describe the ordering of matches I<at a given position>. | |
2634 | One more rule is needed to understand how a match is determined for the | |
2635 | whole regular expression: a match at an earlier position is always better | |
2636 | than a match at a later position. | |
2637 | ||
0d017f4d | 2638 | =head2 Creating Custom RE Engines |
c84d73f1 | 2639 | |
0b928c2f FC |
2640 | As of Perl 5.10.0, one can create custom regular expression engines. This |
2641 | is not for the faint of heart, as they have to plug in at the C level. See | |
2642 | L<perlreapi> for more details. | |
2643 | ||
2644 | As an alternative, overloaded constants (see L<overload>) provide a simple | |
2645 | way to extend the functionality of the RE engine, by substituting one | |
2646 | pattern for another. | |
c84d73f1 IZ |
2647 | |
2648 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
0d017f4d | 2649 | matches at a boundary between whitespace characters and non-whitespace |
c84d73f1 IZ |
2650 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
2651 | at these positions, so we want to have each C<\Y|> in the place of the | |
2652 | more complicated version. We can create a module C<customre> to do | |
2653 | this: | |
2654 | ||
2655 | package customre; | |
2656 | use overload; | |
2657 | ||
2658 | sub import { | |
2659 | shift; | |
2660 | die "No argument to customre::import allowed" if @_; | |
2661 | overload::constant 'qr' => \&convert; | |
2662 | } | |
2663 | ||
2664 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
2665 | ||
580a9fe1 RGS |
2666 | # We must also take care of not escaping the legitimate \\Y| |
2667 | # sequence, hence the presence of '\\' in the conversion rules. | |
5d458dd8 | 2668 | my %rules = ( '\\' => '\\\\', |
f793d64a | 2669 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
c84d73f1 IZ |
2670 | sub convert { |
2671 | my $re = shift; | |
5d458dd8 | 2672 | $re =~ s{ |
c84d73f1 IZ |
2673 | \\ ( \\ | Y . ) |
2674 | } | |
5d458dd8 | 2675 | { $rules{$1} or invalid($re,$1) }sgex; |
c84d73f1 IZ |
2676 | return $re; |
2677 | } | |
2678 | ||
2679 | Now C<use customre> enables the new escape in constant regular | |
2680 | expressions, i.e., those without any runtime variable interpolations. | |
2681 | As documented in L<overload>, this conversion will work only over | |
2682 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
2683 | part of this regular expression needs to be converted explicitly | |
2684 | (but only if the special meaning of C<\Y|> should be enabled inside $re): | |
2685 | ||
2686 | use customre; | |
2687 | $re = <>; | |
2688 | chomp $re; | |
2689 | $re = customre::convert $re; | |
2690 | /\Y|$re\Y|/; | |
2691 | ||
83f32aba RS |
2692 | =head2 Embedded Code Execution Frequency |
2693 | ||
2694 | The exact rules for how often (??{}) and (?{}) are executed in a pattern | |
2695 | are unspecified. In the case of a successful match you can assume that | |
2696 | they DWIM and will be executed in left to right order the appropriate | |
2697 | number of times in the accepting path of the pattern as would any other | |
2698 | meta-pattern. How non-accepting pathways and match failures affect the | |
2699 | number of times a pattern is executed is specifically unspecified and | |
2700 | may vary depending on what optimizations can be applied to the pattern | |
2701 | and is likely to change from version to version. | |
2702 | ||
2703 | For instance in | |
2704 | ||
2705 | "aaabcdeeeee"=~/a(?{print "a"})b(?{print "b"})cde/; | |
2706 | ||
2707 | the exact number of times "a" or "b" are printed out is unspecified for | |
2708 | failure, but you may assume they will be printed at least once during | |
2709 | a successful match, additionally you may assume that if "b" is printed, | |
2710 | it will be preceded by at least one "a". | |
2711 | ||
2712 | In the case of branching constructs like the following: | |
2713 | ||
2714 | /a(b|(?{ print "a" }))c(?{ print "c" })/; | |
2715 | ||
2716 | you can assume that the input "ac" will output "ac", and that "abc" | |
2717 | will output only "c". | |
2718 | ||
2719 | When embedded code is quantified, successful matches will call the | |
2720 | code once for each matched iteration of the quantifier. For | |
2721 | example: | |
2722 | ||
2723 | "good" =~ /g(?:o(?{print "o"}))*d/; | |
2724 | ||
2725 | will output "o" twice. | |
2726 | ||
0b928c2f | 2727 | =head2 PCRE/Python Support |
1f1031fe | 2728 | |
0b928c2f | 2729 | As of Perl 5.10.0, Perl supports several Python/PCRE-specific extensions |
1f1031fe | 2730 | to the regex syntax. While Perl programmers are encouraged to use the |
0b928c2f | 2731 | Perl-specific syntax, the following are also accepted: |
1f1031fe YO |
2732 | |
2733 | =over 4 | |
2734 | ||
ae5648b3 | 2735 | =item C<< (?PE<lt>NAMEE<gt>pattern) >> |
1f1031fe | 2736 | |
c27a5cfe | 2737 | Define a named capture group. Equivalent to C<< (?<NAME>pattern) >>. |
1f1031fe YO |
2738 | |
2739 | =item C<< (?P=NAME) >> | |
2740 | ||
c27a5cfe | 2741 | Backreference to a named capture group. Equivalent to C<< \g{NAME} >>. |
1f1031fe YO |
2742 | |
2743 | =item C<< (?P>NAME) >> | |
2744 | ||
c27a5cfe | 2745 | Subroutine call to a named capture group. Equivalent to C<< (?&NAME) >>. |
1f1031fe | 2746 | |
ee9b8eae | 2747 | =back |
1f1031fe | 2748 | |
19799a22 GS |
2749 | =head1 BUGS |
2750 | ||
88c9975e KW |
2751 | Many regular expression constructs don't work on EBCDIC platforms. |
2752 | ||
ed7efc79 KW |
2753 | There are a number of issues with regard to case-insensitive matching |
2754 | in Unicode rules. See C<i> under L</Modifiers> above. | |
2755 | ||
9da458fc IZ |
2756 | This document varies from difficult to understand to completely |
2757 | and utterly opaque. The wandering prose riddled with jargon is | |
2758 | hard to fathom in several places. | |
2759 | ||
2760 | This document needs a rewrite that separates the tutorial content | |
2761 | from the reference content. | |
19799a22 GS |
2762 | |
2763 | =head1 SEE ALSO | |
9fa51da4 | 2764 | |
91e0c79e MJD |
2765 | L<perlrequick>. |
2766 | ||
2767 | L<perlretut>. | |
2768 | ||
9b599b2a GS |
2769 | L<perlop/"Regexp Quote-Like Operators">. |
2770 | ||
1e66bd83 PP |
2771 | L<perlop/"Gory details of parsing quoted constructs">. |
2772 | ||
14218588 GS |
2773 | L<perlfaq6>. |
2774 | ||
9b599b2a GS |
2775 | L<perlfunc/pos>. |
2776 | ||
2777 | L<perllocale>. | |
2778 | ||
fb55449c JH |
2779 | L<perlebcdic>. |
2780 | ||
14218588 GS |
2781 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
2782 | by O'Reilly and Associates. |