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 | |
0d017f4d WL |
19 | |
20 | =head2 Modifiers | |
21 | ||
19799a22 | 22 | Matching operations can have various modifiers. Modifiers |
5a964f20 | 23 | that relate to the interpretation of the regular expression inside |
19799a22 | 24 | are listed below. Modifiers that alter the way a regular expression |
5d458dd8 | 25 | is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and |
1e66bd83 | 26 | L<perlop/"Gory details of parsing quoted constructs">. |
a0d0e21e | 27 | |
55497cff | 28 | =over 4 |
29 | ||
54310121 | 30 | =item m |
d74e8afc | 31 | X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline> |
55497cff | 32 | |
33 | Treat string as multiple lines. That is, change "^" and "$" from matching | |
14218588 | 34 | the start or end of the string to matching the start or end of any |
7f761169 | 35 | line anywhere within the string. |
55497cff | 36 | |
54310121 | 37 | =item s |
d74e8afc ITB |
38 | X</s> X<regex, single-line> X<regexp, single-line> |
39 | X<regular expression, single-line> | |
55497cff | 40 | |
41 | Treat string as single line. That is, change "." to match any character | |
19799a22 | 42 | whatsoever, even a newline, which normally it would not match. |
55497cff | 43 | |
34d67d80 | 44 | Used together, as C</ms>, they let the "." match any character whatsoever, |
fb55449c | 45 | while still allowing "^" and "$" to match, respectively, just after |
19799a22 | 46 | and just before newlines within the string. |
7b8d334a | 47 | |
87e95b7f YO |
48 | =item i |
49 | X</i> X<regex, case-insensitive> X<regexp, case-insensitive> | |
50 | X<regular expression, case-insensitive> | |
51 | ||
52 | Do case-insensitive pattern matching. | |
53 | ||
5027a30b KW |
54 | If locale matching rules are in effect, the case map is taken from the |
55 | current | |
17580e7a | 56 | locale for code points less than 255, and from Unicode rules for larger |
ed7efc79 KW |
57 | code points. However, matches that would cross the Unicode |
58 | rules/non-Unicode rules boundary (ords 255/256) will not succeed. See | |
59 | L<perllocale>. | |
60 | ||
61 | There are a number of Unicode characters that match multiple characters | |
62 | under C</i>. For example, C<LATIN SMALL LIGATURE FI> | |
63 | should match the sequence C<fi>. Perl is not | |
64 | currently able to do this when the multiple characters are in the pattern and | |
65 | are split between groupings, or when one or more are quantified. Thus | |
66 | ||
67 | "\N{LATIN SMALL LIGATURE FI}" =~ /fi/i; # Matches | |
68 | "\N{LATIN SMALL LIGATURE FI}" =~ /[fi][fi]/i; # Doesn't match! | |
69 | "\N{LATIN SMALL LIGATURE FI}" =~ /fi*/i; # Doesn't match! | |
70 | ||
71 | # The below doesn't match, and it isn't clear what $1 and $2 would | |
72 | # be even if it did!! | |
73 | "\N{LATIN SMALL LIGATURE FI}" =~ /(f)(i)/i; # Doesn't match! | |
74 | ||
1f59b283 KW |
75 | Perl doesn't match multiple characters in an inverted bracketed |
76 | character class, which otherwise could be highly confusing. See | |
77 | L<perlrecharclass/Negation>. | |
78 | ||
79 | Also, Perl matching doesn't fully conform to the current Unicode C</i> | |
ed7efc79 KW |
80 | recommendations, which ask that the matching be made upon the NFD |
81 | (Normalization Form Decomposed) of the text. However, Unicode is | |
82 | in the process of reconsidering and revising their recommendations. | |
87e95b7f | 83 | |
54310121 | 84 | =item x |
d74e8afc | 85 | X</x> |
55497cff | 86 | |
87 | Extend your pattern's legibility by permitting whitespace and comments. | |
ed7efc79 | 88 | Details in L</"/x"> |
55497cff | 89 | |
87e95b7f YO |
90 | =item p |
91 | X</p> X<regex, preserve> X<regexp, preserve> | |
92 | ||
632a1772 | 93 | Preserve the string matched such that ${^PREMATCH}, ${^MATCH}, and |
87e95b7f YO |
94 | ${^POSTMATCH} are available for use after matching. |
95 | ||
e2e6bec7 DN |
96 | =item g and c |
97 | X</g> X</c> | |
98 | ||
99 | Global matching, and keep the Current position after failed matching. | |
100 | Unlike i, m, s and x, these two flags affect the way the regex is used | |
101 | rather than the regex itself. See | |
102 | L<perlretut/"Using regular expressions in Perl"> for further explanation | |
103 | of the g and c modifiers. | |
104 | ||
b6fa137b FC |
105 | =item a, d, l and u |
106 | X</a> X</d> X</l> X</u> | |
107 | ||
108 | These modifiers, new in 5.14, affect which character-set semantics | |
ed7efc79 KW |
109 | (Unicode, ASCII, etc.) are used, as described below in |
110 | L</Character set modifiers>. | |
b6fa137b | 111 | |
55497cff | 112 | =back |
a0d0e21e LW |
113 | |
114 | These are usually written as "the C</x> modifier", even though the delimiter | |
b6fa137b | 115 | in question might not really be a slash. The modifiers C</imsxadlup> |
ab7bb42d | 116 | may also be embedded within the regular expression itself using |
ed7efc79 | 117 | the C<(?...)> construct, see L</Extended Patterns> below. |
b6fa137b FC |
118 | |
119 | The C</x>, C</l>, C</u>, C</a> and C</d> modifiers need a little more | |
120 | explanation. | |
a0d0e21e | 121 | |
ed7efc79 KW |
122 | =head3 /x |
123 | ||
b6fa137b | 124 | C</x> tells |
7b059540 | 125 | the regular expression parser to ignore most whitespace that is neither |
55497cff | 126 | backslashed nor within a character class. You can use this to break up |
4633a7c4 | 127 | your regular expression into (slightly) more readable parts. The C<#> |
54310121 | 128 | character is also treated as a metacharacter introducing a comment, |
55497cff | 129 | just as in ordinary Perl code. This also means that if you want real |
14218588 | 130 | whitespace or C<#> characters in the pattern (outside a character |
f9a3ff1a | 131 | class, where they are unaffected by C</x>), then you'll either have to |
7b059540 KW |
132 | escape them (using backslashes or C<\Q...\E>) or encode them using octal, |
133 | hex, or C<\N{}> escapes. Taken together, these features go a long way towards | |
8933a740 RGS |
134 | making Perl's regular expressions more readable. Note that you have to |
135 | be careful not to include the pattern delimiter in the comment--perl has | |
136 | no way of knowing you did not intend to close the pattern early. See | |
137 | the C-comment deletion code in L<perlop>. Also note that anything inside | |
7651b971 | 138 | a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect |
0b928c2f | 139 | space interpretation within a single multi-character construct. For |
7651b971 | 140 | example in C<\x{...}>, regardless of the C</x> modifier, there can be no |
9bb1f947 | 141 | spaces. Same for a L<quantifier|/Quantifiers> such as C<{3}> or |
f9e949fd KW |
142 | C<{5,}>. Similarly, C<(?:...)> can't have a space between the C<?> and C<:>, |
143 | but can between the C<(> and C<?>. Within any delimiters for such a | |
144 | construct, allowed spaces are not affected by C</x>, and depend on the | |
145 | construct. For example, C<\x{...}> can't have spaces because hexadecimal | |
146 | numbers don't have spaces in them. But, Unicode properties can have spaces, so | |
0b928c2f | 147 | in C<\p{...}> there can be spaces that follow the Unicode rules, for which see |
9bb1f947 | 148 | L<perluniprops/Properties accessible through \p{} and \P{}>. |
d74e8afc | 149 | X</x> |
a0d0e21e | 150 | |
ed7efc79 KW |
151 | =head3 Character set modifiers |
152 | ||
153 | C</d>, C</u>, C</a>, and C</l>, available starting in 5.14, are called | |
154 | the character set modifiers; they affect the character set semantics | |
155 | used for the regular expression. | |
156 | ||
157 | At any given time, exactly one of these modifiers is in effect. Once | |
158 | compiled, the behavior doesn't change regardless of what rules are in | |
159 | effect when the regular expression is executed. And if a regular | |
160 | expression is interpolated into a larger one, the original's rules | |
161 | continue to apply to it, and only it. | |
162 | ||
ca9560b2 KW |
163 | Note that the modifiers affect only pattern matching, and do not extend |
164 | to any replacement done. For example, | |
165 | ||
166 | s/foo/\Ubar/l | |
167 | ||
168 | will uppercase "bar", but the C</l> does not affect how the C<\U> | |
169 | operates. If C<use locale> is in effect, the C<\U> will use locale | |
170 | rules; if C<use feature 'unicode_strings'> is in effect, it will | |
171 | use Unicode rules, etc. | |
172 | ||
ed7efc79 KW |
173 | =head4 /l |
174 | ||
175 | means to use the current locale's rules (see L<perllocale>) when pattern | |
176 | matching. For example, C<\w> will match the "word" characters of that | |
177 | locale, and C<"/i"> case-insensitive matching will match according to | |
178 | the locale's case folding rules. The locale used will be the one in | |
179 | effect at the time of execution of the pattern match. This may not be | |
180 | the same as the compilation-time locale, and can differ from one match | |
181 | to another if there is an intervening call of the | |
b6fa137b | 182 | L<setlocale() function|perllocale/The setlocale function>. |
ed7efc79 KW |
183 | |
184 | Perl only supports single-byte locales. This means that code points | |
185 | above 255 are treated as Unicode no matter what locale is in effect. | |
186 | Under Unicode rules, there are a few case-insensitive matches that cross | |
187 | the 255/256 boundary. These are disallowed under C</l>. For example, | |
188 | 0xFF does not caselessly match the character at 0x178, C<LATIN CAPITAL | |
189 | LETTER Y WITH DIAERESIS>, because 0xFF may not be C<LATIN SMALL LETTER Y | |
190 | WITH DIAERESIS> in the current locale, and Perl has no way of knowing if | |
191 | that character even exists in the locale, much less what code point it | |
192 | is. | |
193 | ||
194 | This modifier may be specified to be the default by C<use locale>, but | |
195 | see L</Which character set modifier is in effect?>. | |
b6fa137b FC |
196 | X</l> |
197 | ||
ed7efc79 KW |
198 | =head4 /u |
199 | ||
200 | means to use Unicode rules when pattern matching. On ASCII platforms, | |
201 | this means that the code points between 128 and 255 take on their | |
b6fa137b FC |
202 | Latin-1 (ISO-8859-1) meanings (which are the same as Unicode's), whereas |
203 | in strict ASCII their meanings are undefined. Thus the platform | |
ed7efc79 KW |
204 | effectively becomes a Unicode platform, hence, for example, C<\w> will |
205 | match any of the more than 100_000 word characters in Unicode. | |
206 | ||
207 | Unlike most locales, which are specific to a language and country pair, | |
208 | Unicode classifies all the characters that are letters I<somewhere> as | |
209 | C<\w>. For example, your locale might not think that C<LATIN SMALL | |
210 | LETTER ETH> is a letter (unless you happen to speak Icelandic), but | |
211 | Unicode does. Similarly, all the characters that are decimal digits | |
212 | somewhere in the world will match C<\d>; this is hundreds, not 10, | |
213 | possible matches. And some of those digits look like some of the 10 | |
214 | ASCII digits, but mean a different number, so a human could easily think | |
215 | a number is a different quantity than it really is. For example, | |
216 | C<BENGALI DIGIT FOUR> (U+09EA) looks very much like an | |
217 | C<ASCII DIGIT EIGHT> (U+0038). And, C<\d+>, may match strings of digits | |
218 | that are a mixture from different writing systems, creating a security | |
916cec3f | 219 | issue. L<Unicode::UCDE<sol>num()|Unicode::UCD/num> can be used to sort this out. |
ed7efc79 KW |
220 | |
221 | Also, case-insensitive matching works on the full set of Unicode | |
222 | characters. The C<KELVIN SIGN>, for example matches the letters "k" and | |
223 | "K"; and C<LATIN SMALL LIGATURE FF> matches the sequence "ff", which, | |
224 | if you're not prepared, might make it look like a hexadecimal constant, | |
225 | presenting another potential security issue. See | |
226 | L<http://unicode.org/reports/tr36> for a detailed discussion of Unicode | |
227 | security issues. | |
228 | ||
6368643f KW |
229 | On the EBCDIC platforms that Perl handles, the native character set is |
230 | equivalent to Latin-1. Thus this modifier changes behavior only when | |
ed7efc79 KW |
231 | the C<"/i"> modifier is also specified, and it turns out it affects only |
232 | two characters, giving them full Unicode semantics: the C<MICRO SIGN> | |
6368643f | 233 | will match the Greek capital and small letters C<MU>, otherwise not; and |
ed7efc79 KW |
234 | the C<LATIN CAPITAL LETTER SHARP S> will match any of C<SS>, C<Ss>, |
235 | C<sS>, and C<ss>, otherwise not. | |
236 | ||
237 | This modifier may be specified to be the default by C<use feature | |
238 | 'unicode_strings>, but see | |
239 | L</Which character set modifier is in effect?>. | |
b6fa137b FC |
240 | X</u> |
241 | ||
ed7efc79 KW |
242 | =head4 /a |
243 | ||
244 | is the same as C</u>, except that C<\d>, C<\s>, C<\w>, and the | |
b6fa137b FC |
245 | Posix character classes are restricted to matching in the ASCII range |
246 | only. That is, with this modifier, C<\d> always means precisely the | |
247 | digits C<"0"> to C<"9">; C<\s> means the five characters C<[ \f\n\r\t]>; | |
248 | C<\w> means the 63 characters C<[A-Za-z0-9_]>; and likewise, all the | |
249 | Posix classes such as C<[[:print:]]> match only the appropriate | |
ed7efc79 KW |
250 | ASCII-range characters. |
251 | ||
252 | This modifier is useful for people who only incidentally use Unicode. | |
253 | With it, one can write C<\d> with confidence that it will only match | |
254 | ASCII characters, and should the need arise to match beyond ASCII, you | |
255 | can use C<\p{Digit}>, or C<\p{Word}> for C<\w>. There are similar | |
256 | C<\p{...}> constructs that can match white space and Posix classes | |
6368643f | 257 | beyond ASCII. See L<perlrecharclass/POSIX Character Classes>. |
ed7efc79 KW |
258 | |
259 | As you would expect, this modifier causes, for example, C<\D> to mean | |
260 | the same thing as C<[^0-9]>; in fact, all non-ASCII characters match | |
261 | C<\D>, C<\S>, and C<\W>. C<\b> still means to match at the boundary | |
262 | between C<\w> and C<\W>, using the C</a> definitions of them (similarly | |
263 | for C<\B>). | |
264 | ||
265 | Otherwise, C</a> behaves like the C</u> modifier, in that | |
266 | case-insensitive matching uses Unicode semantics; for example, "k" will | |
267 | match the Unicode C<\N{KELVIN SIGN}> under C</i> matching, and code | |
268 | points in the Latin1 range, above ASCII will have Unicode rules when it | |
269 | comes to case-insensitive matching. | |
270 | ||
271 | To forbid ASCII/non-ASCII matches (like "k" with C<\N{KELVIN SIGN}>), | |
272 | specify the "a" twice, for example C</aai> or C</aia> | |
273 | ||
274 | To reiterate, this modifier provides protection for applications that | |
275 | don't wish to be exposed to all of Unicode. Specifying it twice | |
276 | gives added protection. | |
277 | ||
278 | This modifier may be specified to be the default by C<use re '/a'> | |
279 | or C<use re '/aa'>, but see | |
280 | L</Which character set modifier is in effect?>. | |
b6fa137b | 281 | X</a> |
ed7efc79 KW |
282 | X</aa> |
283 | ||
284 | =head4 /d | |
285 | ||
286 | This modifier means to use the "Default" native rules of the platform | |
287 | except when there is cause to use Unicode rules instead, as follows: | |
288 | ||
289 | =over 4 | |
290 | ||
291 | =item 1 | |
292 | ||
293 | the target string is encoded in UTF-8; or | |
294 | ||
295 | =item 2 | |
296 | ||
297 | the pattern is encoded in UTF-8; or | |
298 | ||
299 | =item 3 | |
300 | ||
301 | the pattern explicitly mentions a code point that is above 255 (say by | |
302 | C<\x{100}>); or | |
303 | ||
304 | =item 4 | |
b6fa137b | 305 | |
ed7efc79 KW |
306 | the pattern uses a Unicode name (C<\N{...}>); or |
307 | ||
308 | =item 5 | |
309 | ||
310 | the pattern uses a Unicode property (C<\p{...}>) | |
311 | ||
312 | =back | |
313 | ||
314 | Another mnemonic for this modifier is "Depends", as the rules actually | |
315 | used depend on various things, and as a result you can get unexpected | |
316 | results. See L<perlunicode/The "Unicode Bug">. | |
317 | ||
318 | On ASCII platforms, the native rules are ASCII, and on EBCDIC platforms | |
319 | (at least the ones that Perl handles), they are Latin-1. | |
320 | ||
321 | Here are some examples of how that works on an ASCII platform: | |
322 | ||
323 | $str = "\xDF"; # $str is not in UTF-8 format. | |
324 | $str =~ /^\w/; # No match, as $str isn't in UTF-8 format. | |
325 | $str .= "\x{0e0b}"; # Now $str is in UTF-8 format. | |
326 | $str =~ /^\w/; # Match! $str is now in UTF-8 format. | |
327 | chop $str; | |
328 | $str =~ /^\w/; # Still a match! $str remains in UTF-8 format. | |
329 | ||
330 | =head4 Which character set modifier is in effect? | |
331 | ||
332 | Which of these modifiers is in effect at any given point in a regular | |
333 | expression depends on a fairly complex set of interactions. As | |
334 | explained below in L</Extended Patterns> it is possible to explicitly | |
335 | specify modifiers that apply only to portions of a regular expression. | |
336 | The innermost always has priority over any outer ones, and one applying | |
6368643f KW |
337 | to the whole expression has priority over any of the default settings that are |
338 | described in the remainder of this section. | |
ed7efc79 | 339 | |
916cec3f | 340 | The C<L<use re 'E<sol>foo'|re/"'/flags' mode">> pragma can be used to set |
ed7efc79 KW |
341 | default modifiers (including these) for regular expressions compiled |
342 | within its scope. This pragma has precedence over the other pragmas | |
6368643f | 343 | listed below that change the defaults. |
ed7efc79 KW |
344 | |
345 | Otherwise, C<L<use locale|perllocale>> sets the default modifier to C</l>; | |
346 | and C<L<use feature 'unicode_strings|feature>> or | |
347 | C<L<use 5.012|perlfunc/use VERSION>> (or higher) set the default to | |
348 | C</u> when not in the same scope as either C<L<use locale|perllocale>> | |
6368643f KW |
349 | or C<L<use bytes|bytes>>. Unlike the mechanisms mentioned above, these |
350 | affect operations besides regular expressions pattern matching, and so | |
351 | give more consistent results with other operators, including using | |
352 | C<\U>, C<\l>, etc. in substitution replacements. | |
ed7efc79 KW |
353 | |
354 | If none of the above apply, for backwards compatibility reasons, the | |
355 | C</d> modifier is the one in effect by default. As this can lead to | |
356 | unexpected results, it is best to specify which other rule set should be | |
357 | used. | |
358 | ||
359 | =head4 Character set modifier behavior prior to Perl 5.14 | |
360 | ||
361 | Prior to 5.14, there were no explicit modifiers, but C</l> was implied | |
362 | for regexes compiled within the scope of C<use locale>, and C</d> was | |
363 | implied otherwise. However, interpolating a regex into a larger regex | |
364 | would ignore the original compilation in favor of whatever was in effect | |
365 | at the time of the second compilation. There were a number of | |
366 | inconsistencies (bugs) with the C</d> modifier, where Unicode rules | |
367 | would be used when inappropriate, and vice versa. C<\p{}> did not imply | |
368 | Unicode rules, and neither did all occurrences of C<\N{}>, until 5.12. | |
b6fa137b | 369 | |
a0d0e21e LW |
370 | =head2 Regular Expressions |
371 | ||
04838cea RGS |
372 | =head3 Metacharacters |
373 | ||
384f06ae | 374 | The patterns used in Perl pattern matching evolved from those supplied in |
14218588 | 375 | the Version 8 regex routines. (The routines are derived |
19799a22 GS |
376 | (distantly) from Henry Spencer's freely redistributable reimplementation |
377 | of the V8 routines.) See L<Version 8 Regular Expressions> for | |
378 | details. | |
a0d0e21e LW |
379 | |
380 | In particular the following metacharacters have their standard I<egrep>-ish | |
381 | meanings: | |
d74e8afc ITB |
382 | X<metacharacter> |
383 | X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]> | |
384 | ||
a0d0e21e | 385 | |
f793d64a KW |
386 | \ Quote the next metacharacter |
387 | ^ Match the beginning of the line | |
388 | . Match any character (except newline) | |
389 | $ Match the end of the line (or before newline at the end) | |
390 | | Alternation | |
391 | () Grouping | |
392 | [] Bracketed Character class | |
a0d0e21e | 393 | |
14218588 GS |
394 | By default, the "^" character is guaranteed to match only the |
395 | beginning of the string, the "$" character only the end (or before the | |
396 | newline at the end), and Perl does certain optimizations with the | |
a0d0e21e LW |
397 | assumption that the string contains only one line. Embedded newlines |
398 | will not be matched by "^" or "$". You may, however, wish to treat a | |
4a6725af | 399 | string as a multi-line buffer, such that the "^" will match after any |
0d520e8e YO |
400 | newline within the string (except if the newline is the last character in |
401 | the string), and "$" will match before any newline. At the | |
a0d0e21e LW |
402 | cost of a little more overhead, you can do this by using the /m modifier |
403 | on the pattern match operator. (Older programs did this by setting C<$*>, | |
0b928c2f | 404 | but this option was removed in perl 5.9.) |
d74e8afc | 405 | X<^> X<$> X</m> |
a0d0e21e | 406 | |
14218588 | 407 | To simplify multi-line substitutions, the "." character never matches a |
55497cff | 408 | newline unless you use the C</s> modifier, which in effect tells Perl to pretend |
f02c194e | 409 | the string is a single line--even if it isn't. |
d74e8afc | 410 | X<.> X</s> |
a0d0e21e | 411 | |
04838cea RGS |
412 | =head3 Quantifiers |
413 | ||
a0d0e21e | 414 | The following standard quantifiers are recognized: |
d74e8afc | 415 | X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}> |
a0d0e21e | 416 | |
f793d64a KW |
417 | * Match 0 or more times |
418 | + Match 1 or more times | |
419 | ? Match 1 or 0 times | |
420 | {n} Match exactly n times | |
421 | {n,} Match at least n times | |
422 | {n,m} Match at least n but not more than m times | |
a0d0e21e | 423 | |
0b928c2f FC |
424 | (If a curly bracket occurs in any other context and does not form part of |
425 | a backslashed sequence like C<\x{...}>, it is treated | |
b975c076 | 426 | as a regular character. In particular, the lower bound |
527e91da BB |
427 | is not optional.) The "*" quantifier is equivalent to C<{0,}>, the "+" |
428 | quantifier to C<{1,}>, and the "?" quantifier to C<{0,1}>. n and m are limited | |
d0b16107 | 429 | to non-negative integral values less than a preset limit defined when perl is built. |
9c79236d GS |
430 | This is usually 32766 on the most common platforms. The actual limit can |
431 | be seen in the error message generated by code such as this: | |
432 | ||
820475bd | 433 | $_ **= $_ , / {$_} / for 2 .. 42; |
a0d0e21e | 434 | |
54310121 | 435 | By default, a quantified subpattern is "greedy", that is, it will match as |
436 | many times as possible (given a particular starting location) while still | |
437 | allowing the rest of the pattern to match. If you want it to match the | |
438 | minimum number of times possible, follow the quantifier with a "?". Note | |
439 | that the meanings don't change, just the "greediness": | |
0d017f4d | 440 | X<metacharacter> X<greedy> X<greediness> |
d74e8afc | 441 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?> |
a0d0e21e | 442 | |
f793d64a KW |
443 | *? Match 0 or more times, not greedily |
444 | +? Match 1 or more times, not greedily | |
445 | ?? Match 0 or 1 time, not greedily | |
0b928c2f | 446 | {n}? Match exactly n times, not greedily (redundant) |
f793d64a KW |
447 | {n,}? Match at least n times, not greedily |
448 | {n,m}? Match at least n but not more than m times, not greedily | |
a0d0e21e | 449 | |
b9b4dddf YO |
450 | By default, when a quantified subpattern does not allow the rest of the |
451 | overall pattern to match, Perl will backtrack. However, this behaviour is | |
0d017f4d | 452 | sometimes undesirable. Thus Perl provides the "possessive" quantifier form |
b9b4dddf YO |
453 | as well. |
454 | ||
f793d64a KW |
455 | *+ Match 0 or more times and give nothing back |
456 | ++ Match 1 or more times and give nothing back | |
457 | ?+ Match 0 or 1 time and give nothing back | |
458 | {n}+ Match exactly n times and give nothing back (redundant) | |
459 | {n,}+ Match at least n times and give nothing back | |
460 | {n,m}+ Match at least n but not more than m times and give nothing back | |
b9b4dddf YO |
461 | |
462 | For instance, | |
463 | ||
464 | 'aaaa' =~ /a++a/ | |
465 | ||
466 | will never match, as the C<a++> will gobble up all the C<a>'s in the | |
467 | string and won't leave any for the remaining part of the pattern. This | |
468 | feature can be extremely useful to give perl hints about where it | |
469 | shouldn't backtrack. For instance, the typical "match a double-quoted | |
470 | string" problem can be most efficiently performed when written as: | |
471 | ||
472 | /"(?:[^"\\]++|\\.)*+"/ | |
473 | ||
0d017f4d | 474 | as we know that if the final quote does not match, backtracking will not |
0b928c2f FC |
475 | help. See the independent subexpression |
476 | L</C<< (?>pattern) >>> for more details; | |
b9b4dddf YO |
477 | possessive quantifiers are just syntactic sugar for that construct. For |
478 | instance the above example could also be written as follows: | |
479 | ||
480 | /"(?>(?:(?>[^"\\]+)|\\.)*)"/ | |
481 | ||
04838cea RGS |
482 | =head3 Escape sequences |
483 | ||
0b928c2f | 484 | Because patterns are processed as double-quoted strings, the following |
a0d0e21e LW |
485 | also work: |
486 | ||
f793d64a KW |
487 | \t tab (HT, TAB) |
488 | \n newline (LF, NL) | |
489 | \r return (CR) | |
490 | \f form feed (FF) | |
491 | \a alarm (bell) (BEL) | |
492 | \e escape (think troff) (ESC) | |
f793d64a | 493 | \cK control char (example: VT) |
dc0d9c48 | 494 | \x{}, \x00 character whose ordinal is the given hexadecimal number |
fb121860 | 495 | \N{name} named Unicode character or character sequence |
f793d64a | 496 | \N{U+263D} Unicode character (example: FIRST QUARTER MOON) |
f0a2b745 | 497 | \o{}, \000 character whose ordinal is the given octal number |
f793d64a KW |
498 | \l lowercase next char (think vi) |
499 | \u uppercase next char (think vi) | |
500 | \L lowercase till \E (think vi) | |
501 | \U uppercase till \E (think vi) | |
502 | \Q quote (disable) pattern metacharacters till \E | |
503 | \E end either case modification or quoted section, think vi | |
a0d0e21e | 504 | |
9bb1f947 | 505 | Details are in L<perlop/Quote and Quote-like Operators>. |
1d2dff63 | 506 | |
e1d1eefb | 507 | =head3 Character Classes and other Special Escapes |
04838cea | 508 | |
a0d0e21e | 509 | In addition, Perl defines the following: |
d0b16107 | 510 | X<\g> X<\k> X<\K> X<backreference> |
a0d0e21e | 511 | |
f793d64a KW |
512 | Sequence Note Description |
513 | [...] [1] Match a character according to the rules of the | |
514 | bracketed character class defined by the "...". | |
515 | Example: [a-z] matches "a" or "b" or "c" ... or "z" | |
516 | [[:...:]] [2] Match a character according to the rules of the POSIX | |
517 | character class "..." within the outer bracketed | |
518 | character class. Example: [[:upper:]] matches any | |
519 | uppercase character. | |
d35dd6c6 KW |
520 | \w [3] Match a "word" character (alphanumeric plus "_", plus |
521 | other connector punctuation chars plus Unicode | |
0b928c2f | 522 | marks) |
f793d64a KW |
523 | \W [3] Match a non-"word" character |
524 | \s [3] Match a whitespace character | |
525 | \S [3] Match a non-whitespace character | |
526 | \d [3] Match a decimal digit character | |
527 | \D [3] Match a non-digit character | |
528 | \pP [3] Match P, named property. Use \p{Prop} for longer names | |
529 | \PP [3] Match non-P | |
530 | \X [4] Match Unicode "eXtended grapheme cluster" | |
531 | \C Match a single C-language char (octet) even if that is | |
532 | part of a larger UTF-8 character. Thus it breaks up | |
533 | characters into their UTF-8 bytes, so you may end up | |
534 | with malformed pieces of UTF-8. Unsupported in | |
535 | lookbehind. | |
c27a5cfe | 536 | \1 [5] Backreference to a specific capture group or buffer. |
f793d64a KW |
537 | '1' may actually be any positive integer. |
538 | \g1 [5] Backreference to a specific or previous group, | |
539 | \g{-1} [5] The number may be negative indicating a relative | |
c27a5cfe | 540 | previous group and may optionally be wrapped in |
f793d64a KW |
541 | curly brackets for safer parsing. |
542 | \g{name} [5] Named backreference | |
543 | \k<name> [5] Named backreference | |
544 | \K [6] Keep the stuff left of the \K, don't include it in $& | |
545 | \N [7] Any character but \n (experimental). Not affected by | |
546 | /s modifier | |
547 | \v [3] Vertical whitespace | |
548 | \V [3] Not vertical whitespace | |
549 | \h [3] Horizontal whitespace | |
550 | \H [3] Not horizontal whitespace | |
551 | \R [4] Linebreak | |
e1d1eefb | 552 | |
9bb1f947 KW |
553 | =over 4 |
554 | ||
555 | =item [1] | |
556 | ||
557 | See L<perlrecharclass/Bracketed Character Classes> for details. | |
df225385 | 558 | |
9bb1f947 | 559 | =item [2] |
b8c5462f | 560 | |
9bb1f947 | 561 | See L<perlrecharclass/POSIX Character Classes> for details. |
b8c5462f | 562 | |
9bb1f947 | 563 | =item [3] |
5496314a | 564 | |
9bb1f947 | 565 | See L<perlrecharclass/Backslash sequences> for details. |
5496314a | 566 | |
9bb1f947 | 567 | =item [4] |
5496314a | 568 | |
9bb1f947 | 569 | See L<perlrebackslash/Misc> for details. |
d0b16107 | 570 | |
9bb1f947 | 571 | =item [5] |
b8c5462f | 572 | |
c27a5cfe | 573 | See L</Capture groups> below for details. |
93733859 | 574 | |
9bb1f947 | 575 | =item [6] |
b8c5462f | 576 | |
9bb1f947 KW |
577 | See L</Extended Patterns> below for details. |
578 | ||
579 | =item [7] | |
580 | ||
581 | Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the | |
fb121860 KW |
582 | character or character sequence whose name is C<NAME>; and similarly |
583 | when of the form C<\N{U+I<hex>}>, it matches the character whose Unicode | |
584 | code point is I<hex>. Otherwise it matches any character but C<\n>. | |
9bb1f947 KW |
585 | |
586 | =back | |
d0b16107 | 587 | |
04838cea RGS |
588 | =head3 Assertions |
589 | ||
a0d0e21e | 590 | Perl defines the following zero-width assertions: |
d74e8afc ITB |
591 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> |
592 | X<regexp, zero-width assertion> | |
593 | X<regular expression, zero-width assertion> | |
594 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> | |
a0d0e21e | 595 | |
9bb1f947 KW |
596 | \b Match a word boundary |
597 | \B Match except at a word boundary | |
598 | \A Match only at beginning of string | |
599 | \Z Match only at end of string, or before newline at the end | |
600 | \z Match only at end of string | |
601 | \G Match only at pos() (e.g. at the end-of-match position | |
9da458fc | 602 | of prior m//g) |
a0d0e21e | 603 | |
14218588 | 604 | A word boundary (C<\b>) is a spot between two characters |
19799a22 GS |
605 | that has a C<\w> on one side of it and a C<\W> on the other side |
606 | of it (in either order), counting the imaginary characters off the | |
607 | beginning and end of the string as matching a C<\W>. (Within | |
608 | character classes C<\b> represents backspace rather than a word | |
609 | boundary, just as it normally does in any double-quoted string.) | |
610 | The C<\A> and C<\Z> are just like "^" and "$", except that they | |
611 | won't match multiple times when the C</m> modifier is used, while | |
612 | "^" and "$" will match at every internal line boundary. To match | |
613 | the actual end of the string and not ignore an optional trailing | |
614 | newline, use C<\z>. | |
d74e8afc | 615 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 GS |
616 | |
617 | The C<\G> assertion can be used to chain global matches (using | |
618 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. | |
619 | It is also useful when writing C<lex>-like scanners, when you have | |
620 | several patterns that you want to match against consequent substrings | |
0b928c2f | 621 | of your string; see the previous reference. The actual location |
19799a22 | 622 | where C<\G> will match can also be influenced by using C<pos()> as |
58e23c8d | 623 | an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length |
0b928c2f FC |
624 | matches (see L</"Repeated Patterns Matching a Zero-length Substring">) |
625 | is modified somewhat, in that contents to the left of C<\G> are | |
58e23c8d YO |
626 | not counted when determining the length of the match. Thus the following |
627 | will not match forever: | |
d74e8afc | 628 | X<\G> |
c47ff5f1 | 629 | |
e761bb84 CO |
630 | my $string = 'ABC'; |
631 | pos($string) = 1; | |
632 | while ($string =~ /(.\G)/g) { | |
633 | print $1; | |
634 | } | |
58e23c8d YO |
635 | |
636 | It will print 'A' and then terminate, as it considers the match to | |
637 | be zero-width, and thus will not match at the same position twice in a | |
638 | row. | |
639 | ||
640 | It is worth noting that C<\G> improperly used can result in an infinite | |
641 | loop. Take care when using patterns that include C<\G> in an alternation. | |
642 | ||
c27a5cfe | 643 | =head3 Capture groups |
04838cea | 644 | |
c27a5cfe KW |
645 | The bracketing construct C<( ... )> creates capture groups (also referred to as |
646 | capture buffers). To refer to the current contents of a group later on, within | |
d8b950dc KW |
647 | the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>) |
648 | for the second, and so on. | |
649 | This is called a I<backreference>. | |
d74e8afc | 650 | X<regex, capture buffer> X<regexp, capture buffer> |
c27a5cfe | 651 | X<regex, capture group> X<regexp, capture group> |
d74e8afc | 652 | X<regular expression, capture buffer> X<backreference> |
c27a5cfe | 653 | X<regular expression, capture group> X<backreference> |
1f1031fe | 654 | X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference> |
d8b950dc KW |
655 | X<named capture buffer> X<regular expression, named capture buffer> |
656 | X<named capture group> X<regular expression, named capture group> | |
657 | X<%+> X<$+{name}> X<< \k<name> >> | |
658 | There is no limit to the number of captured substrings that you may use. | |
659 | Groups are numbered with the leftmost open parenthesis being number 1, etc. If | |
660 | a group did not match, the associated backreference won't match either. (This | |
661 | can happen if the group is optional, or in a different branch of an | |
662 | alternation.) | |
663 | You can omit the C<"g">, and write C<"\1">, etc, but there are some issues with | |
664 | this form, described below. | |
665 | ||
666 | You can also refer to capture groups relatively, by using a negative number, so | |
667 | that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture | |
668 | group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For | |
669 | example: | |
5624f11d YO |
670 | |
671 | / | |
c27a5cfe KW |
672 | (Y) # group 1 |
673 | ( # group 2 | |
674 | (X) # group 3 | |
675 | \g{-1} # backref to group 3 | |
676 | \g{-3} # backref to group 1 | |
5624f11d YO |
677 | ) |
678 | /x | |
679 | ||
d8b950dc KW |
680 | would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to |
681 | interpolate regexes into larger regexes and not have to worry about the | |
682 | capture groups being renumbered. | |
683 | ||
684 | You can dispense with numbers altogether and create named capture groups. | |
685 | The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to | |
686 | reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may | |
687 | also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.) | |
688 | I<name> must not begin with a number, nor contain hyphens. | |
689 | When different groups within the same pattern have the same name, any reference | |
690 | to that name assumes the leftmost defined group. Named groups count in | |
691 | absolute and relative numbering, and so can also be referred to by those | |
692 | numbers. | |
693 | (It's possible to do things with named capture groups that would otherwise | |
694 | require C<(??{})>.) | |
695 | ||
696 | Capture group contents are dynamically scoped and available to you outside the | |
697 | pattern until the end of the enclosing block or until the next successful | |
698 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
699 | You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">, | |
700 | etc); or by name via the C<%+> hash, using C<"$+{I<name>}">. | |
701 | ||
702 | Braces are required in referring to named capture groups, but are optional for | |
703 | absolute or relative numbered ones. Braces are safer when creating a regex by | |
704 | concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a> | |
705 | contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which | |
706 | is probably not what you intended. | |
707 | ||
708 | The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that | |
709 | there were no named nor relative numbered capture groups. Absolute numbered | |
0b928c2f FC |
710 | groups were referred to using C<\1>, |
711 | C<\2>, etc., and this notation is still | |
d8b950dc KW |
712 | accepted (and likely always will be). But it leads to some ambiguities if |
713 | there are more than 9 capture groups, as C<\10> could mean either the tenth | |
714 | capture group, or the character whose ordinal in octal is 010 (a backspace in | |
715 | ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference | |
716 | only if at least 10 left parentheses have opened before it. Likewise C<\11> is | |
717 | a backreference only if at least 11 left parentheses have opened before it. | |
e1f120a9 KW |
718 | And so on. C<\1> through C<\9> are always interpreted as backreferences. |
719 | There are several examples below that illustrate these perils. You can avoid | |
720 | the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups; | |
721 | and for octal constants always using C<\o{}>, or for C<\077> and below, using 3 | |
722 | digits padded with leading zeros, since a leading zero implies an octal | |
723 | constant. | |
d8b950dc KW |
724 | |
725 | The C<\I<digit>> notation also works in certain circumstances outside | |
ed7efc79 | 726 | the pattern. See L</Warning on \1 Instead of $1> below for details. |
81714fb9 | 727 | |
14218588 | 728 | Examples: |
a0d0e21e LW |
729 | |
730 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
731 | ||
d8b950dc | 732 | /(.)\g1/ # find first doubled char |
81714fb9 YO |
733 | and print "'$1' is the first doubled character\n"; |
734 | ||
735 | /(?<char>.)\k<char>/ # ... a different way | |
736 | and print "'$+{char}' is the first doubled character\n"; | |
737 | ||
d8b950dc | 738 | /(?'char'.)\g1/ # ... mix and match |
81714fb9 | 739 | and print "'$1' is the first doubled character\n"; |
c47ff5f1 | 740 | |
14218588 | 741 | if (/Time: (..):(..):(..)/) { # parse out values |
f793d64a KW |
742 | $hours = $1; |
743 | $minutes = $2; | |
744 | $seconds = $3; | |
a0d0e21e | 745 | } |
c47ff5f1 | 746 | |
9d860678 KW |
747 | /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference |
748 | /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal | |
749 | /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference | |
750 | /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal | |
751 | ||
752 | $a = '(.)\1'; # Creates problems when concatenated. | |
753 | $b = '(.)\g{1}'; # Avoids the problems. | |
754 | "aa" =~ /${a}/; # True | |
755 | "aa" =~ /${b}/; # True | |
756 | "aa0" =~ /${a}0/; # False! | |
757 | "aa0" =~ /${b}0/; # True | |
dc0d9c48 KW |
758 | "aa\x08" =~ /${a}0/; # True! |
759 | "aa\x08" =~ /${b}0/; # False | |
9d860678 | 760 | |
14218588 GS |
761 | Several special variables also refer back to portions of the previous |
762 | match. C<$+> returns whatever the last bracket match matched. | |
763 | C<$&> returns the entire matched string. (At one point C<$0> did | |
764 | also, but now it returns the name of the program.) C<$`> returns | |
77ea4f6d JV |
765 | everything before the matched string. C<$'> returns everything |
766 | after the matched string. And C<$^N> contains whatever was matched by | |
767 | the most-recently closed group (submatch). C<$^N> can be used in | |
768 | extended patterns (see below), for example to assign a submatch to a | |
81714fb9 | 769 | variable. |
d74e8afc | 770 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 | 771 | |
d8b950dc KW |
772 | These special variables, like the C<%+> hash and the numbered match variables |
773 | (C<$1>, C<$2>, C<$3>, etc.) are dynamically scoped | |
14218588 GS |
774 | until the end of the enclosing block or until the next successful |
775 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
d74e8afc ITB |
776 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
777 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> | |
778 | ||
0d017f4d | 779 | B<NOTE>: Failed matches in Perl do not reset the match variables, |
5146ce24 | 780 | which makes it easier to write code that tests for a series of more |
665e98b9 JH |
781 | specific cases and remembers the best match. |
782 | ||
14218588 GS |
783 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or |
784 | C<$'> anywhere in the program, it has to provide them for every | |
785 | pattern match. This may substantially slow your program. Perl | |
d8b950dc | 786 | uses the same mechanism to produce C<$1>, C<$2>, etc, so you also pay a |
14218588 GS |
787 | price for each pattern that contains capturing parentheses. (To |
788 | avoid this cost while retaining the grouping behaviour, use the | |
789 | extended regular expression C<(?: ... )> instead.) But if you never | |
790 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
791 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
792 | if you can, but if you can't (and some algorithms really appreciate | |
793 | them), once you've used them once, use them at will, because you've | |
794 | already paid the price. As of 5.005, C<$&> is not so costly as the | |
795 | other two. | |
d74e8afc | 796 | X<$&> X<$`> X<$'> |
68dc0745 | 797 | |
99d59c4d | 798 | As a workaround for this problem, Perl 5.10.0 introduces C<${^PREMATCH}>, |
cde0cee5 YO |
799 | C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&> |
800 | and C<$'>, B<except> that they are only guaranteed to be defined after a | |
87e95b7f | 801 | successful match that was executed with the C</p> (preserve) modifier. |
cde0cee5 YO |
802 | The use of these variables incurs no global performance penalty, unlike |
803 | their punctuation char equivalents, however at the trade-off that you | |
804 | have to tell perl when you want to use them. | |
87e95b7f | 805 | X</p> X<p modifier> |
cde0cee5 | 806 | |
9d727203 KW |
807 | =head2 Quoting metacharacters |
808 | ||
19799a22 GS |
809 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
810 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
811 | are no backslashed symbols that aren't alphanumeric. So anything | |
c47ff5f1 | 812 | that looks like \\, \(, \), \<, \>, \{, or \} is always |
19799a22 GS |
813 | interpreted as a literal character, not a metacharacter. This was |
814 | once used in a common idiom to disable or quote the special meanings | |
815 | of regular expression metacharacters in a string that you want to | |
36bbe248 | 816 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e LW |
817 | |
818 | $pattern =~ s/(\W)/\\$1/g; | |
819 | ||
f1cbbd6e | 820 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 GS |
821 | Today it is more common to use the quotemeta() function or the C<\Q> |
822 | metaquoting escape sequence to disable all metacharacters' special | |
823 | meanings like this: | |
a0d0e21e LW |
824 | |
825 | /$unquoted\Q$quoted\E$unquoted/ | |
826 | ||
9da458fc IZ |
827 | Beware that if you put literal backslashes (those not inside |
828 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
829 | backslash interpolation may lead to confusing results. If you | |
830 | I<need> to use literal backslashes within C<\Q...\E>, | |
831 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
832 | ||
19799a22 GS |
833 | =head2 Extended Patterns |
834 | ||
14218588 | 835 | Perl also defines a consistent extension syntax for features not |
0b928c2f FC |
836 | found in standard tools like B<awk> and |
837 | B<lex>. The syntax for most of these is a | |
14218588 GS |
838 | pair of parentheses with a question mark as the first thing within |
839 | the parentheses. The character after the question mark indicates | |
840 | the extension. | |
19799a22 | 841 | |
14218588 GS |
842 | The stability of these extensions varies widely. Some have been |
843 | part of the core language for many years. Others are experimental | |
844 | and may change without warning or be completely removed. Check | |
845 | the documentation on an individual feature to verify its current | |
846 | status. | |
19799a22 | 847 | |
14218588 GS |
848 | A question mark was chosen for this and for the minimal-matching |
849 | construct because 1) question marks are rare in older regular | |
850 | expressions, and 2) whenever you see one, you should stop and | |
0b928c2f | 851 | "question" exactly what is going on. That's psychology.... |
a0d0e21e | 852 | |
70ca8714 | 853 | =over 4 |
a0d0e21e | 854 | |
cc6b7395 | 855 | =item C<(?#text)> |
d74e8afc | 856 | X<(?#)> |
a0d0e21e | 857 | |
14218588 | 858 | A comment. The text is ignored. If the C</x> modifier enables |
19799a22 | 859 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes |
259138e3 GS |
860 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
861 | C<)> in the comment. | |
a0d0e21e | 862 | |
cfaf538b | 863 | =item C<(?adlupimsx-imsx)> |
fb85c044 | 864 | |
cfaf538b | 865 | =item C<(?^alupimsx)> |
fb85c044 | 866 | X<(?)> X<(?^)> |
19799a22 | 867 | |
0b6d1084 JH |
868 | One or more embedded pattern-match modifiers, to be turned on (or |
869 | turned off, if preceded by C<->) for the remainder of the pattern or | |
fb85c044 KW |
870 | the remainder of the enclosing pattern group (if any). |
871 | ||
fb85c044 | 872 | This is particularly useful for dynamic patterns, such as those read in from a |
0d017f4d | 873 | configuration file, taken from an argument, or specified in a table |
0b928c2f FC |
874 | somewhere. Consider the case where some patterns want to be |
875 | case-sensitive and some do not: The case-insensitive ones merely need to | |
0d017f4d | 876 | include C<(?i)> at the front of the pattern. For example: |
19799a22 GS |
877 | |
878 | $pattern = "foobar"; | |
5d458dd8 | 879 | if ( /$pattern/i ) { } |
19799a22 GS |
880 | |
881 | # more flexible: | |
882 | ||
883 | $pattern = "(?i)foobar"; | |
5d458dd8 | 884 | if ( /$pattern/ ) { } |
19799a22 | 885 | |
0b6d1084 | 886 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 | 887 | |
d8b950dc | 888 | ( (?i) blah ) \s+ \g1 |
19799a22 | 889 | |
0d017f4d WL |
890 | will match C<blah> in any case, some spaces, and an exact (I<including the case>!) |
891 | repetition of the previous word, assuming the C</x> modifier, and no C</i> | |
892 | modifier outside this group. | |
19799a22 | 893 | |
8eb5594e DR |
894 | These modifiers do not carry over into named subpatterns called in the |
895 | enclosing group. In other words, a pattern such as C<((?i)(&NAME))> does not | |
896 | change the case-sensitivity of the "NAME" pattern. | |
897 | ||
dc925305 KW |
898 | Any of these modifiers can be set to apply globally to all regular |
899 | expressions compiled within the scope of a C<use re>. See | |
a0bbd6ff | 900 | L<re/"'/flags' mode">. |
dc925305 | 901 | |
9de15fec KW |
902 | Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately |
903 | after the C<"?"> is a shorthand equivalent to C<d-imsx>. Flags (except | |
904 | C<"d">) may follow the caret to override it. | |
905 | But a minus sign is not legal with it. | |
906 | ||
dc925305 | 907 | Note that the C<a>, C<d>, C<l>, C<p>, and C<u> modifiers are special in |
e1d8d8ac | 908 | that they can only be enabled, not disabled, and the C<a>, C<d>, C<l>, and |
dc925305 | 909 | C<u> modifiers are mutually exclusive: specifying one de-specifies the |
ed7efc79 KW |
910 | others, and a maximum of one (or two C<a>'s) may appear in the |
911 | construct. Thus, for | |
0b928c2f | 912 | example, C<(?-p)> will warn when compiled under C<use warnings>; |
b6fa137b | 913 | C<(?-d:...)> and C<(?dl:...)> are fatal errors. |
9de15fec KW |
914 | |
915 | Note also that the C<p> modifier is special in that its presence | |
916 | anywhere in a pattern has a global effect. | |
cde0cee5 | 917 | |
5a964f20 | 918 | =item C<(?:pattern)> |
d74e8afc | 919 | X<(?:)> |
a0d0e21e | 920 | |
cfaf538b | 921 | =item C<(?adluimsx-imsx:pattern)> |
ca9dfc88 | 922 | |
cfaf538b | 923 | =item C<(?^aluimsx:pattern)> |
fb85c044 KW |
924 | X<(?^:)> |
925 | ||
5a964f20 TC |
926 | This is for clustering, not capturing; it groups subexpressions like |
927 | "()", but doesn't make backreferences as "()" does. So | |
a0d0e21e | 928 | |
5a964f20 | 929 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e LW |
930 | |
931 | is like | |
932 | ||
5a964f20 | 933 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e | 934 | |
19799a22 GS |
935 | but doesn't spit out extra fields. It's also cheaper not to capture |
936 | characters if you don't need to. | |
a0d0e21e | 937 | |
19799a22 | 938 | Any letters between C<?> and C<:> act as flags modifiers as with |
cfaf538b | 939 | C<(?adluimsx-imsx)>. For example, |
ca9dfc88 IZ |
940 | |
941 | /(?s-i:more.*than).*million/i | |
942 | ||
14218588 | 943 | is equivalent to the more verbose |
ca9dfc88 IZ |
944 | |
945 | /(?:(?s-i)more.*than).*million/i | |
946 | ||
fb85c044 | 947 | Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately |
9de15fec KW |
948 | after the C<"?"> is a shorthand equivalent to C<d-imsx>. Any positive |
949 | flags (except C<"d">) may follow the caret, so | |
fb85c044 KW |
950 | |
951 | (?^x:foo) | |
952 | ||
953 | is equivalent to | |
954 | ||
955 | (?x-ims:foo) | |
956 | ||
957 | The caret tells Perl that this cluster doesn't inherit the flags of any | |
0b928c2f | 958 | surrounding pattern, but uses the system defaults (C<d-imsx>), |
fb85c044 KW |
959 | modified by any flags specified. |
960 | ||
961 | The caret allows for simpler stringification of compiled regular | |
962 | expressions. These look like | |
963 | ||
964 | (?^:pattern) | |
965 | ||
966 | with any non-default flags appearing between the caret and the colon. | |
967 | A test that looks at such stringification thus doesn't need to have the | |
968 | system default flags hard-coded in it, just the caret. If new flags are | |
969 | added to Perl, the meaning of the caret's expansion will change to include | |
970 | the default for those flags, so the test will still work, unchanged. | |
971 | ||
972 | Specifying a negative flag after the caret is an error, as the flag is | |
973 | redundant. | |
974 | ||
975 | Mnemonic for C<(?^...)>: A fresh beginning since the usual use of a caret is | |
976 | to match at the beginning. | |
977 | ||
594d7033 YO |
978 | =item C<(?|pattern)> |
979 | X<(?|)> X<Branch reset> | |
980 | ||
981 | This is the "branch reset" pattern, which has the special property | |
c27a5cfe | 982 | that the capture groups are numbered from the same starting point |
99d59c4d | 983 | in each alternation branch. It is available starting from perl 5.10.0. |
4deaaa80 | 984 | |
c27a5cfe | 985 | Capture groups are numbered from left to right, but inside this |
693596a8 | 986 | construct the numbering is restarted for each branch. |
4deaaa80 | 987 | |
c27a5cfe | 988 | The numbering within each branch will be as normal, and any groups |
4deaaa80 PJ |
989 | following this construct will be numbered as though the construct |
990 | contained only one branch, that being the one with the most capture | |
c27a5cfe | 991 | groups in it. |
4deaaa80 | 992 | |
0b928c2f | 993 | This construct is useful when you want to capture one of a |
4deaaa80 PJ |
994 | number of alternative matches. |
995 | ||
996 | Consider the following pattern. The numbers underneath show in | |
c27a5cfe | 997 | which group the captured content will be stored. |
594d7033 YO |
998 | |
999 | ||
1000 | # before ---------------branch-reset----------- after | |
1001 | / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x | |
1002 | # 1 2 2 3 2 3 4 | |
1003 | ||
ab106183 A |
1004 | Be careful when using the branch reset pattern in combination with |
1005 | named captures. Named captures are implemented as being aliases to | |
c27a5cfe | 1006 | numbered groups holding the captures, and that interferes with the |
ab106183 A |
1007 | implementation of the branch reset pattern. If you are using named |
1008 | captures in a branch reset pattern, it's best to use the same names, | |
1009 | in the same order, in each of the alternations: | |
1010 | ||
1011 | /(?| (?<a> x ) (?<b> y ) | |
1012 | | (?<a> z ) (?<b> w )) /x | |
1013 | ||
1014 | Not doing so may lead to surprises: | |
1015 | ||
1016 | "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x; | |
1017 | say $+ {a}; # Prints '12' | |
1018 | say $+ {b}; # *Also* prints '12'. | |
1019 | ||
c27a5cfe KW |
1020 | The problem here is that both the group named C<< a >> and the group |
1021 | named C<< b >> are aliases for the group belonging to C<< $1 >>. | |
90a18110 | 1022 | |
ee9b8eae YO |
1023 | =item Look-Around Assertions |
1024 | X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround> | |
1025 | ||
0b928c2f | 1026 | Look-around assertions are zero-width patterns which match a specific |
ee9b8eae YO |
1027 | pattern without including it in C<$&>. Positive assertions match when |
1028 | their subpattern matches, negative assertions match when their subpattern | |
1029 | fails. Look-behind matches text up to the current match position, | |
1030 | look-ahead matches text following the current match position. | |
1031 | ||
1032 | =over 4 | |
1033 | ||
5a964f20 | 1034 | =item C<(?=pattern)> |
d74e8afc | 1035 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> |
a0d0e21e | 1036 | |
19799a22 | 1037 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e LW |
1038 | matches a word followed by a tab, without including the tab in C<$&>. |
1039 | ||
5a964f20 | 1040 | =item C<(?!pattern)> |
d74e8afc | 1041 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> |
a0d0e21e | 1042 | |
19799a22 | 1043 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e | 1044 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 GS |
1045 | however that look-ahead and look-behind are NOT the same thing. You cannot |
1046 | use this for look-behind. | |
7b8d334a | 1047 | |
5a964f20 | 1048 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a GS |
1049 | will not do what you want. That's because the C<(?!foo)> is just saying that |
1050 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
0b928c2f | 1051 | match. Use look-behind instead (see below). |
c277df42 | 1052 | |
ee9b8eae YO |
1053 | =item C<(?<=pattern)> C<\K> |
1054 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K> | |
c277df42 | 1055 | |
c47ff5f1 | 1056 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 GS |
1057 | matches a word that follows a tab, without including the tab in C<$&>. |
1058 | Works only for fixed-width look-behind. | |
c277df42 | 1059 | |
ee9b8eae YO |
1060 | There is a special form of this construct, called C<\K>, which causes the |
1061 | regex engine to "keep" everything it had matched prior to the C<\K> and | |
0b928c2f | 1062 | not include it in C<$&>. This effectively provides variable-length |
ee9b8eae YO |
1063 | look-behind. The use of C<\K> inside of another look-around assertion |
1064 | is allowed, but the behaviour is currently not well defined. | |
1065 | ||
c62285ac | 1066 | For various reasons C<\K> may be significantly more efficient than the |
ee9b8eae YO |
1067 | equivalent C<< (?<=...) >> construct, and it is especially useful in |
1068 | situations where you want to efficiently remove something following | |
1069 | something else in a string. For instance | |
1070 | ||
1071 | s/(foo)bar/$1/g; | |
1072 | ||
1073 | can be rewritten as the much more efficient | |
1074 | ||
1075 | s/foo\Kbar//g; | |
1076 | ||
5a964f20 | 1077 | =item C<(?<!pattern)> |
d74e8afc | 1078 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> |
c277df42 | 1079 | |
19799a22 GS |
1080 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
1081 | matches any occurrence of "foo" that does not follow "bar". Works | |
1082 | only for fixed-width look-behind. | |
c277df42 | 1083 | |
ee9b8eae YO |
1084 | =back |
1085 | ||
81714fb9 YO |
1086 | =item C<(?'NAME'pattern)> |
1087 | ||
1088 | =item C<< (?<NAME>pattern) >> | |
1089 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> | |
1090 | ||
c27a5cfe | 1091 | A named capture group. Identical in every respect to normal capturing |
0b928c2f FC |
1092 | parentheses C<()> but for the additional fact that the group |
1093 | can be referred to by name in various regular expression | |
1094 | constructs (like C<\g{NAME}>) and can be accessed by name | |
1095 | after a successful match via C<%+> or C<%->. See L<perlvar> | |
90a18110 | 1096 | for more details on the C<%+> and C<%-> hashes. |
81714fb9 | 1097 | |
c27a5cfe KW |
1098 | If multiple distinct capture groups have the same name then the |
1099 | $+{NAME} will refer to the leftmost defined group in the match. | |
81714fb9 | 1100 | |
0d017f4d | 1101 | The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent. |
81714fb9 YO |
1102 | |
1103 | B<NOTE:> While the notation of this construct is the same as the similar | |
c27a5cfe | 1104 | function in .NET regexes, the behavior is not. In Perl the groups are |
81714fb9 YO |
1105 | numbered sequentially regardless of being named or not. Thus in the |
1106 | pattern | |
1107 | ||
1108 | /(x)(?<foo>y)(z)/ | |
1109 | ||
1110 | $+{foo} will be the same as $2, and $3 will contain 'z' instead of | |
1111 | the opposite which is what a .NET regex hacker might expect. | |
1112 | ||
1f1031fe YO |
1113 | Currently NAME is restricted to simple identifiers only. |
1114 | In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or | |
1115 | its Unicode extension (see L<utf8>), | |
1116 | though it isn't extended by the locale (see L<perllocale>). | |
81714fb9 | 1117 | |
1f1031fe | 1118 | B<NOTE:> In order to make things easier for programmers with experience |
ae5648b3 | 1119 | with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >> |
0d017f4d | 1120 | may be used instead of C<< (?<NAME>pattern) >>; however this form does not |
64c5a566 | 1121 | support the use of single quotes as a delimiter for the name. |
81714fb9 | 1122 | |
1f1031fe YO |
1123 | =item C<< \k<NAME> >> |
1124 | ||
1125 | =item C<< \k'NAME' >> | |
81714fb9 YO |
1126 | |
1127 | Named backreference. Similar to numeric backreferences, except that | |
1128 | the group is designated by name and not number. If multiple groups | |
1129 | have the same name then it refers to the leftmost defined group in | |
1130 | the current match. | |
1131 | ||
0d017f4d | 1132 | It is an error to refer to a name not defined by a C<< (?<NAME>) >> |
81714fb9 YO |
1133 | earlier in the pattern. |
1134 | ||
1135 | Both forms are equivalent. | |
1136 | ||
1f1031fe | 1137 | B<NOTE:> In order to make things easier for programmers with experience |
0d017f4d | 1138 | with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >> |
64c5a566 | 1139 | may be used instead of C<< \k<NAME> >>. |
1f1031fe | 1140 | |
cc6b7395 | 1141 | =item C<(?{ code })> |
d74e8afc | 1142 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 | 1143 | |
19799a22 | 1144 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
1145 | experimental, and may be changed without notice. Code executed that |
1146 | has side effects may not perform identically from version to version | |
1147 | due to the effect of future optimisations in the regex engine. | |
c277df42 | 1148 | |
cc46d5f2 | 1149 | This zero-width assertion evaluates any embedded Perl code. It |
19799a22 GS |
1150 | always succeeds, and its C<code> is not interpolated. Currently, |
1151 | the rules to determine where the C<code> ends are somewhat convoluted. | |
1152 | ||
77ea4f6d JV |
1153 | This feature can be used together with the special variable C<$^N> to |
1154 | capture the results of submatches in variables without having to keep | |
1155 | track of the number of nested parentheses. For example: | |
1156 | ||
1157 | $_ = "The brown fox jumps over the lazy dog"; | |
1158 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
1159 | print "color = $color, animal = $animal\n"; | |
1160 | ||
754091cb RGS |
1161 | Inside the C<(?{...})> block, C<$_> refers to the string the regular |
1162 | expression is matching against. You can also use C<pos()> to know what is | |
fa11829f | 1163 | the current position of matching within this string. |
754091cb | 1164 | |
19799a22 GS |
1165 | The C<code> is properly scoped in the following sense: If the assertion |
1166 | is backtracked (compare L<"Backtracking">), all changes introduced after | |
1167 | C<local>ization are undone, so that | |
b9ac3b5b GS |
1168 | |
1169 | $_ = 'a' x 8; | |
5d458dd8 | 1170 | m< |
d1fbf752 | 1171 | (?{ $cnt = 0 }) # Initialize $cnt. |
b9ac3b5b | 1172 | ( |
5d458dd8 | 1173 | a |
b9ac3b5b | 1174 | (?{ |
d1fbf752 KW |
1175 | local $cnt = $cnt + 1; # Update $cnt, |
1176 | # backtracking-safe. | |
b9ac3b5b | 1177 | }) |
5d458dd8 | 1178 | )* |
b9ac3b5b | 1179 | aaaa |
d1fbf752 KW |
1180 | (?{ $res = $cnt }) # On success copy to |
1181 | # non-localized location. | |
b9ac3b5b GS |
1182 | >x; |
1183 | ||
0d017f4d | 1184 | will set C<$res = 4>. Note that after the match, C<$cnt> returns to the globally |
14218588 | 1185 | introduced value, because the scopes that restrict C<local> operators |
b9ac3b5b GS |
1186 | are unwound. |
1187 | ||
19799a22 GS |
1188 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> |
1189 | switch. If I<not> used in this way, the result of evaluation of | |
1190 | C<code> is put into the special variable C<$^R>. This happens | |
1191 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions | |
1192 | inside the same regular expression. | |
b9ac3b5b | 1193 | |
19799a22 GS |
1194 | The assignment to C<$^R> above is properly localized, so the old |
1195 | value of C<$^R> is restored if the assertion is backtracked; compare | |
1196 | L<"Backtracking">. | |
b9ac3b5b | 1197 | |
19799a22 GS |
1198 | For reasons of security, this construct is forbidden if the regular |
1199 | expression involves run-time interpolation of variables, unless the | |
1200 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
0b928c2f | 1201 | variables contain results of the C<qr//> operator (see |
b6fa137b | 1202 | L<perlop/"qr/STRINGE<sol>msixpodual">). |
871b0233 | 1203 | |
0d017f4d | 1204 | This restriction is due to the wide-spread and remarkably convenient |
19799a22 | 1205 | custom of using run-time determined strings as patterns. For example: |
871b0233 IZ |
1206 | |
1207 | $re = <>; | |
1208 | chomp $re; | |
1209 | $string =~ /$re/; | |
1210 | ||
14218588 GS |
1211 | Before Perl knew how to execute interpolated code within a pattern, |
1212 | this operation was completely safe from a security point of view, | |
1213 | although it could raise an exception from an illegal pattern. If | |
1214 | you turn on the C<use re 'eval'>, though, it is no longer secure, | |
1215 | so you should only do so if you are also using taint checking. | |
1216 | Better yet, use the carefully constrained evaluation within a Safe | |
cc46d5f2 | 1217 | compartment. See L<perlsec> for details about both these mechanisms. |
871b0233 | 1218 | |
e95d7314 GG |
1219 | B<WARNING>: Use of lexical (C<my>) variables in these blocks is |
1220 | broken. The result is unpredictable and will make perl unstable. The | |
1221 | workaround is to use global (C<our>) variables. | |
1222 | ||
8525cfae FC |
1223 | B<WARNING>: In perl 5.12.x and earlier, the regex engine |
1224 | was not re-entrant, so interpolated code could not | |
1225 | safely invoke the regex engine either directly with | |
e95d7314 | 1226 | C<m//> or C<s///>), or indirectly with functions such as |
8525cfae | 1227 | C<split>. Invoking the regex engine in these blocks would make perl |
e95d7314 | 1228 | unstable. |
8988a1bb | 1229 | |
14455d6c | 1230 | =item C<(??{ code })> |
d74e8afc ITB |
1231 | X<(??{})> |
1232 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> | |
0f5d15d6 | 1233 | |
19799a22 | 1234 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
1235 | experimental, and may be changed without notice. Code executed that |
1236 | has side effects may not perform identically from version to version | |
1237 | due to the effect of future optimisations in the regex engine. | |
0f5d15d6 | 1238 | |
19799a22 GS |
1239 | This is a "postponed" regular subexpression. The C<code> is evaluated |
1240 | at run time, at the moment this subexpression may match. The result | |
0b928c2f | 1241 | of evaluation is considered a regular expression and matched as |
61528107 | 1242 | if it were inserted instead of this construct. Note that this means |
c27a5cfe | 1243 | that the contents of capture groups defined inside an eval'ed pattern |
6bda09f9 | 1244 | are not available outside of the pattern, and vice versa, there is no |
c27a5cfe | 1245 | way for the inner pattern to refer to a capture group defined outside. |
6bda09f9 YO |
1246 | Thus, |
1247 | ||
1248 | ('a' x 100)=~/(??{'(.)' x 100})/ | |
1249 | ||
81714fb9 | 1250 | B<will> match, it will B<not> set $1. |
0f5d15d6 | 1251 | |
428594d9 | 1252 | The C<code> is not interpolated. As before, the rules to determine |
19799a22 GS |
1253 | where the C<code> ends are currently somewhat convoluted. |
1254 | ||
1255 | The following pattern matches a parenthesized group: | |
0f5d15d6 | 1256 | |
d1fbf752 KW |
1257 | $re = qr{ |
1258 | \( | |
1259 | (?: | |
1260 | (?> [^()]+ ) # Non-parens without backtracking | |
1261 | | | |
1262 | (??{ $re }) # Group with matching parens | |
1263 | )* | |
1264 | \) | |
1265 | }x; | |
0f5d15d6 | 1266 | |
6bda09f9 YO |
1267 | See also C<(?PARNO)> for a different, more efficient way to accomplish |
1268 | the same task. | |
1269 | ||
0b370c0a A |
1270 | For reasons of security, this construct is forbidden if the regular |
1271 | expression involves run-time interpolation of variables, unless the | |
1272 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
0b928c2f | 1273 | variables contain results of the C<qr//> operator (see |
b6fa137b | 1274 | L<perlop/"qrE<sol>STRINGE<sol>msixpodual">). |
0b370c0a | 1275 | |
8525cfae FC |
1276 | In perl 5.12.x and earlier, because the regex engine was not re-entrant, |
1277 | delayed code could not safely invoke the regex engine either directly with | |
1278 | C<m//> or C<s///>), or indirectly with functions such as C<split>. | |
8988a1bb | 1279 | |
5d458dd8 YO |
1280 | Recursing deeper than 50 times without consuming any input string will |
1281 | result in a fatal error. The maximum depth is compiled into perl, so | |
6bda09f9 YO |
1282 | changing it requires a custom build. |
1283 | ||
542fa716 YO |
1284 | =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)> |
1285 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)> | |
6bda09f9 | 1286 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
542fa716 | 1287 | X<regex, relative recursion> |
6bda09f9 | 1288 | |
81714fb9 | 1289 | Similar to C<(??{ code })> except it does not involve compiling any code, |
c27a5cfe KW |
1290 | instead it treats the contents of a capture group as an independent |
1291 | pattern that must match at the current position. Capture groups | |
81714fb9 | 1292 | contained by the pattern will have the value as determined by the |
6bda09f9 YO |
1293 | outermost recursion. |
1294 | ||
894be9b7 | 1295 | PARNO is a sequence of digits (not starting with 0) whose value reflects |
c27a5cfe | 1296 | the paren-number of the capture group to recurse to. C<(?R)> recurses to |
894be9b7 | 1297 | the beginning of the whole pattern. C<(?0)> is an alternate syntax for |
542fa716 | 1298 | C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed |
c27a5cfe | 1299 | to be relative, with negative numbers indicating preceding capture groups |
542fa716 | 1300 | and positive ones following. Thus C<(?-1)> refers to the most recently |
c27a5cfe | 1301 | declared group, and C<(?+1)> indicates the next group to be declared. |
c74340f9 | 1302 | Note that the counting for relative recursion differs from that of |
c27a5cfe | 1303 | relative backreferences, in that with recursion unclosed groups B<are> |
c74340f9 | 1304 | included. |
6bda09f9 | 1305 | |
81714fb9 | 1306 | The following pattern matches a function foo() which may contain |
f145b7e9 | 1307 | balanced parentheses as the argument. |
6bda09f9 | 1308 | |
d1fbf752 | 1309 | $re = qr{ ( # paren group 1 (full function) |
81714fb9 | 1310 | foo |
d1fbf752 | 1311 | ( # paren group 2 (parens) |
6bda09f9 | 1312 | \( |
d1fbf752 | 1313 | ( # paren group 3 (contents of parens) |
6bda09f9 | 1314 | (?: |
d1fbf752 | 1315 | (?> [^()]+ ) # Non-parens without backtracking |
6bda09f9 | 1316 | | |
d1fbf752 | 1317 | (?2) # Recurse to start of paren group 2 |
6bda09f9 YO |
1318 | )* |
1319 | ) | |
1320 | \) | |
1321 | ) | |
1322 | ) | |
1323 | }x; | |
1324 | ||
1325 | If the pattern was used as follows | |
1326 | ||
1327 | 'foo(bar(baz)+baz(bop))'=~/$re/ | |
1328 | and print "\$1 = $1\n", | |
1329 | "\$2 = $2\n", | |
1330 | "\$3 = $3\n"; | |
1331 | ||
1332 | the output produced should be the following: | |
1333 | ||
1334 | $1 = foo(bar(baz)+baz(bop)) | |
1335 | $2 = (bar(baz)+baz(bop)) | |
81714fb9 | 1336 | $3 = bar(baz)+baz(bop) |
6bda09f9 | 1337 | |
c27a5cfe | 1338 | If there is no corresponding capture group defined, then it is a |
61528107 | 1339 | fatal error. Recursing deeper than 50 times without consuming any input |
81714fb9 | 1340 | string will also result in a fatal error. The maximum depth is compiled |
6bda09f9 YO |
1341 | into perl, so changing it requires a custom build. |
1342 | ||
542fa716 YO |
1343 | The following shows how using negative indexing can make it |
1344 | easier to embed recursive patterns inside of a C<qr//> construct | |
1345 | for later use: | |
1346 | ||
1347 | my $parens = qr/(\((?:[^()]++|(?-1))*+\))/; | |
1348 | if (/foo $parens \s+ + \s+ bar $parens/x) { | |
1349 | # do something here... | |
1350 | } | |
1351 | ||
81714fb9 | 1352 | B<Note> that this pattern does not behave the same way as the equivalent |
0d017f4d | 1353 | PCRE or Python construct of the same form. In Perl you can backtrack into |
6bda09f9 | 1354 | a recursed group, in PCRE and Python the recursed into group is treated |
542fa716 YO |
1355 | as atomic. Also, modifiers are resolved at compile time, so constructs |
1356 | like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will | |
1357 | be processed. | |
6bda09f9 | 1358 | |
894be9b7 YO |
1359 | =item C<(?&NAME)> |
1360 | X<(?&NAME)> | |
1361 | ||
0d017f4d WL |
1362 | Recurse to a named subpattern. Identical to C<(?PARNO)> except that the |
1363 | parenthesis to recurse to is determined by name. If multiple parentheses have | |
894be9b7 YO |
1364 | the same name, then it recurses to the leftmost. |
1365 | ||
1366 | It is an error to refer to a name that is not declared somewhere in the | |
1367 | pattern. | |
1368 | ||
1f1031fe YO |
1369 | B<NOTE:> In order to make things easier for programmers with experience |
1370 | with the Python or PCRE regex engines the pattern C<< (?P>NAME) >> | |
64c5a566 | 1371 | may be used instead of C<< (?&NAME) >>. |
1f1031fe | 1372 | |
e2e6a0f1 YO |
1373 | =item C<(?(condition)yes-pattern|no-pattern)> |
1374 | X<(?()> | |
286f584a | 1375 | |
e2e6a0f1 | 1376 | =item C<(?(condition)yes-pattern)> |
286f584a | 1377 | |
41ef34de ML |
1378 | Conditional expression. Matches C<yes-pattern> if C<condition> yields |
1379 | a true value, matches C<no-pattern> otherwise. A missing pattern always | |
1380 | matches. | |
1381 | ||
1382 | C<(condition)> should be either an integer in | |
e2e6a0f1 YO |
1383 | parentheses (which is valid if the corresponding pair of parentheses |
1384 | matched), a look-ahead/look-behind/evaluate zero-width assertion, a | |
c27a5cfe | 1385 | name in angle brackets or single quotes (which is valid if a group |
e2e6a0f1 YO |
1386 | with the given name matched), or the special symbol (R) (true when |
1387 | evaluated inside of recursion or eval). Additionally the R may be | |
1388 | followed by a number, (which will be true when evaluated when recursing | |
1389 | inside of the appropriate group), or by C<&NAME>, in which case it will | |
1390 | be true only when evaluated during recursion in the named group. | |
1391 | ||
1392 | Here's a summary of the possible predicates: | |
1393 | ||
1394 | =over 4 | |
1395 | ||
1396 | =item (1) (2) ... | |
1397 | ||
c27a5cfe | 1398 | Checks if the numbered capturing group has matched something. |
e2e6a0f1 YO |
1399 | |
1400 | =item (<NAME>) ('NAME') | |
1401 | ||
c27a5cfe | 1402 | Checks if a group with the given name has matched something. |
e2e6a0f1 | 1403 | |
f01cd190 FC |
1404 | =item (?=...) (?!...) (?<=...) (?<!...) |
1405 | ||
1406 | Checks whether the pattern matches (or does not match, for the '!' | |
1407 | variants). | |
1408 | ||
e2e6a0f1 YO |
1409 | =item (?{ CODE }) |
1410 | ||
f01cd190 | 1411 | Treats the return value of the code block as the condition. |
e2e6a0f1 YO |
1412 | |
1413 | =item (R) | |
1414 | ||
1415 | Checks if the expression has been evaluated inside of recursion. | |
1416 | ||
1417 | =item (R1) (R2) ... | |
1418 | ||
1419 | Checks if the expression has been evaluated while executing directly | |
1420 | inside of the n-th capture group. This check is the regex equivalent of | |
1421 | ||
1422 | if ((caller(0))[3] eq 'subname') { ... } | |
1423 | ||
1424 | In other words, it does not check the full recursion stack. | |
1425 | ||
1426 | =item (R&NAME) | |
1427 | ||
1428 | Similar to C<(R1)>, this predicate checks to see if we're executing | |
1429 | directly inside of the leftmost group with a given name (this is the same | |
1430 | logic used by C<(?&NAME)> to disambiguate). It does not check the full | |
1431 | stack, but only the name of the innermost active recursion. | |
1432 | ||
1433 | =item (DEFINE) | |
1434 | ||
1435 | In this case, the yes-pattern is never directly executed, and no | |
1436 | no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient. | |
1437 | See below for details. | |
1438 | ||
1439 | =back | |
1440 | ||
1441 | For example: | |
1442 | ||
1443 | m{ ( \( )? | |
1444 | [^()]+ | |
1445 | (?(1) \) ) | |
1446 | }x | |
1447 | ||
1448 | matches a chunk of non-parentheses, possibly included in parentheses | |
1449 | themselves. | |
1450 | ||
0b928c2f FC |
1451 | A special form is the C<(DEFINE)> predicate, which never executes its |
1452 | yes-pattern directly, and does not allow a no-pattern. This allows one to | |
1453 | define subpatterns which will be executed only by the recursion mechanism. | |
e2e6a0f1 YO |
1454 | This way, you can define a set of regular expression rules that can be |
1455 | bundled into any pattern you choose. | |
1456 | ||
1457 | It is recommended that for this usage you put the DEFINE block at the | |
1458 | end of the pattern, and that you name any subpatterns defined within it. | |
1459 | ||
1460 | Also, it's worth noting that patterns defined this way probably will | |
1461 | not be as efficient, as the optimiser is not very clever about | |
1462 | handling them. | |
1463 | ||
1464 | An example of how this might be used is as follows: | |
1465 | ||
2bf803e2 | 1466 | /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT)) |
e2e6a0f1 | 1467 | (?(DEFINE) |
2bf803e2 YO |
1468 | (?<NAME_PAT>....) |
1469 | (?<ADRESS_PAT>....) | |
e2e6a0f1 YO |
1470 | )/x |
1471 | ||
c27a5cfe KW |
1472 | Note that capture groups matched inside of recursion are not accessible |
1473 | after the recursion returns, so the extra layer of capturing groups is | |
e2e6a0f1 YO |
1474 | necessary. Thus C<$+{NAME_PAT}> would not be defined even though |
1475 | C<$+{NAME}> would be. | |
286f584a | 1476 | |
c47ff5f1 | 1477 | =item C<< (?>pattern) >> |
6bda09f9 | 1478 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 | 1479 | |
19799a22 GS |
1480 | An "independent" subexpression, one which matches the substring |
1481 | that a I<standalone> C<pattern> would match if anchored at the given | |
9da458fc | 1482 | position, and it matches I<nothing other than this substring>. This |
19799a22 GS |
1483 | construct is useful for optimizations of what would otherwise be |
1484 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). | |
9da458fc IZ |
1485 | It may also be useful in places where the "grab all you can, and do not |
1486 | give anything back" semantic is desirable. | |
19799a22 | 1487 | |
c47ff5f1 | 1488 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 GS |
1489 | (anchored at the beginning of string, as above) will match I<all> |
1490 | characters C<a> at the beginning of string, leaving no C<a> for | |
1491 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, | |
1492 | since the match of the subgroup C<a*> is influenced by the following | |
1493 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside | |
1494 | C<a*ab> will match fewer characters than a standalone C<a*>, since | |
1495 | this makes the tail match. | |
1496 | ||
0b928c2f FC |
1497 | C<< (?>pattern) >> does not disable backtracking altogether once it has |
1498 | matched. It is still possible to backtrack past the construct, but not | |
1499 | into it. So C<< ((?>a*)|(?>b*))ar >> will still match "bar". | |
1500 | ||
c47ff5f1 | 1501 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
0b928c2f FC |
1502 | C<(?=(pattern))\g{-1}>. This matches the same substring as a standalone |
1503 | C<a+>, and the following C<\g{-1}> eats the matched string; it therefore | |
c47ff5f1 | 1504 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 GS |
1505 | (The difference between these two constructs is that the second one |
1506 | uses a capturing group, thus shifting ordinals of backreferences | |
1507 | in the rest of a regular expression.) | |
1508 | ||
1509 | Consider this pattern: | |
c277df42 | 1510 | |
871b0233 | 1511 | m{ \( |
e2e6a0f1 | 1512 | ( |
f793d64a | 1513 | [^()]+ # x+ |
e2e6a0f1 | 1514 | | |
871b0233 IZ |
1515 | \( [^()]* \) |
1516 | )+ | |
e2e6a0f1 | 1517 | \) |
871b0233 | 1518 | }x |
5a964f20 | 1519 | |
19799a22 GS |
1520 | That will efficiently match a nonempty group with matching parentheses |
1521 | two levels deep or less. However, if there is no such group, it | |
1522 | will take virtually forever on a long string. That's because there | |
1523 | are so many different ways to split a long string into several | |
1524 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
1525 | to a subpattern of the above pattern. Consider how the pattern | |
1526 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
1527 | seconds, but that each extra letter doubles this time. This | |
1528 | exponential performance will make it appear that your program has | |
14218588 | 1529 | hung. However, a tiny change to this pattern |
5a964f20 | 1530 | |
e2e6a0f1 YO |
1531 | m{ \( |
1532 | ( | |
f793d64a | 1533 | (?> [^()]+ ) # change x+ above to (?> x+ ) |
e2e6a0f1 | 1534 | | |
871b0233 IZ |
1535 | \( [^()]* \) |
1536 | )+ | |
e2e6a0f1 | 1537 | \) |
871b0233 | 1538 | }x |
c277df42 | 1539 | |
c47ff5f1 | 1540 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 TC |
1541 | this yourself would be a productive exercise), but finishes in a fourth |
1542 | the time when used on a similar string with 1000000 C<a>s. Be aware, | |
0b928c2f FC |
1543 | however, that, when this construct is followed by a |
1544 | quantifier, it currently triggers a warning message under | |
9f1b1f2d | 1545 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b | 1546 | C<"matches null string many times in regex">. |
c277df42 | 1547 | |
c47ff5f1 | 1548 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 | 1549 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 IZ |
1550 | This was only 4 times slower on a string with 1000000 C<a>s. |
1551 | ||
9da458fc IZ |
1552 | The "grab all you can, and do not give anything back" semantic is desirable |
1553 | in many situations where on the first sight a simple C<()*> looks like | |
1554 | the correct solution. Suppose we parse text with comments being delimited | |
1555 | by C<#> followed by some optional (horizontal) whitespace. Contrary to | |
4375e838 | 1556 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc IZ |
1557 | the comment delimiter, because it may "give up" some whitespace if |
1558 | the remainder of the pattern can be made to match that way. The correct | |
1559 | answer is either one of these: | |
1560 | ||
1561 | (?>#[ \t]*) | |
1562 | #[ \t]*(?![ \t]) | |
1563 | ||
1564 | For example, to grab non-empty comments into $1, one should use either | |
1565 | one of these: | |
1566 | ||
1567 | / (?> \# [ \t]* ) ( .+ ) /x; | |
1568 | / \# [ \t]* ( [^ \t] .* ) /x; | |
1569 | ||
1570 | Which one you pick depends on which of these expressions better reflects | |
1571 | the above specification of comments. | |
1572 | ||
6bda09f9 YO |
1573 | In some literature this construct is called "atomic matching" or |
1574 | "possessive matching". | |
1575 | ||
b9b4dddf YO |
1576 | Possessive quantifiers are equivalent to putting the item they are applied |
1577 | to inside of one of these constructs. The following equivalences apply: | |
1578 | ||
1579 | Quantifier Form Bracketing Form | |
1580 | --------------- --------------- | |
1581 | PAT*+ (?>PAT*) | |
1582 | PAT++ (?>PAT+) | |
1583 | PAT?+ (?>PAT?) | |
1584 | PAT{min,max}+ (?>PAT{min,max}) | |
1585 | ||
e2e6a0f1 YO |
1586 | =back |
1587 | ||
1588 | =head2 Special Backtracking Control Verbs | |
1589 | ||
1590 | B<WARNING:> These patterns are experimental and subject to change or | |
0d017f4d | 1591 | removal in a future version of Perl. Their usage in production code should |
e2e6a0f1 YO |
1592 | be noted to avoid problems during upgrades. |
1593 | ||
1594 | These special patterns are generally of the form C<(*VERB:ARG)>. Unless | |
1595 | otherwise stated the ARG argument is optional; in some cases, it is | |
1596 | forbidden. | |
1597 | ||
1598 | Any pattern containing a special backtracking verb that allows an argument | |
e1020413 | 1599 | has the special behaviour that when executed it sets the current package's |
5d458dd8 YO |
1600 | C<$REGERROR> and C<$REGMARK> variables. When doing so the following |
1601 | rules apply: | |
e2e6a0f1 | 1602 | |
5d458dd8 YO |
1603 | On failure, the C<$REGERROR> variable will be set to the ARG value of the |
1604 | verb pattern, if the verb was involved in the failure of the match. If the | |
1605 | ARG part of the pattern was omitted, then C<$REGERROR> will be set to the | |
1606 | name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was | |
1607 | none. Also, the C<$REGMARK> variable will be set to FALSE. | |
e2e6a0f1 | 1608 | |
5d458dd8 YO |
1609 | On a successful match, the C<$REGERROR> variable will be set to FALSE, and |
1610 | the C<$REGMARK> variable will be set to the name of the last | |
1611 | C<(*MARK:NAME)> pattern executed. See the explanation for the | |
1612 | C<(*MARK:NAME)> verb below for more details. | |
e2e6a0f1 | 1613 | |
5d458dd8 | 1614 | B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1> |
0b928c2f | 1615 | and most other regex-related variables. They are not local to a scope, nor |
5d458dd8 YO |
1616 | readonly, but instead are volatile package variables similar to C<$AUTOLOAD>. |
1617 | Use C<local> to localize changes to them to a specific scope if necessary. | |
e2e6a0f1 YO |
1618 | |
1619 | If a pattern does not contain a special backtracking verb that allows an | |
5d458dd8 | 1620 | argument, then C<$REGERROR> and C<$REGMARK> are not touched at all. |
e2e6a0f1 | 1621 | |
70ca8714 | 1622 | =over 3 |
e2e6a0f1 YO |
1623 | |
1624 | =item Verbs that take an argument | |
1625 | ||
1626 | =over 4 | |
1627 | ||
5d458dd8 | 1628 | =item C<(*PRUNE)> C<(*PRUNE:NAME)> |
f7819f85 | 1629 | X<(*PRUNE)> X<(*PRUNE:NAME)> |
54612592 | 1630 | |
5d458dd8 YO |
1631 | This zero-width pattern prunes the backtracking tree at the current point |
1632 | when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>, | |
1633 | where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached, | |
1634 | A may backtrack as necessary to match. Once it is reached, matching | |
1635 | continues in B, which may also backtrack as necessary; however, should B | |
1636 | not match, then no further backtracking will take place, and the pattern | |
1637 | will fail outright at the current starting position. | |
54612592 YO |
1638 | |
1639 | The following example counts all the possible matching strings in a | |
1640 | pattern (without actually matching any of them). | |
1641 | ||
e2e6a0f1 | 1642 | 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1643 | print "Count=$count\n"; |
1644 | ||
1645 | which produces: | |
1646 | ||
1647 | aaab | |
1648 | aaa | |
1649 | aa | |
1650 | a | |
1651 | aab | |
1652 | aa | |
1653 | a | |
1654 | ab | |
1655 | a | |
1656 | Count=9 | |
1657 | ||
5d458dd8 | 1658 | If we add a C<(*PRUNE)> before the count like the following |
54612592 | 1659 | |
5d458dd8 | 1660 | 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1661 | print "Count=$count\n"; |
1662 | ||
0b928c2f | 1663 | we prevent backtracking and find the count of the longest matching string |
353c6505 | 1664 | at each matching starting point like so: |
54612592 YO |
1665 | |
1666 | aaab | |
1667 | aab | |
1668 | ab | |
1669 | Count=3 | |
1670 | ||
5d458dd8 | 1671 | Any number of C<(*PRUNE)> assertions may be used in a pattern. |
54612592 | 1672 | |
5d458dd8 YO |
1673 | See also C<< (?>pattern) >> and possessive quantifiers for other ways to |
1674 | control backtracking. In some cases, the use of C<(*PRUNE)> can be | |
1675 | replaced with a C<< (?>pattern) >> with no functional difference; however, | |
1676 | C<(*PRUNE)> can be used to handle cases that cannot be expressed using a | |
1677 | C<< (?>pattern) >> alone. | |
54612592 | 1678 | |
e2e6a0f1 | 1679 | |
5d458dd8 YO |
1680 | =item C<(*SKIP)> C<(*SKIP:NAME)> |
1681 | X<(*SKIP)> | |
e2e6a0f1 | 1682 | |
5d458dd8 | 1683 | This zero-width pattern is similar to C<(*PRUNE)>, except that on |
e2e6a0f1 | 1684 | failure it also signifies that whatever text that was matched leading up |
5d458dd8 YO |
1685 | to the C<(*SKIP)> pattern being executed cannot be part of I<any> match |
1686 | of this pattern. This effectively means that the regex engine "skips" forward | |
1687 | to this position on failure and tries to match again, (assuming that | |
1688 | there is sufficient room to match). | |
1689 | ||
1690 | The name of the C<(*SKIP:NAME)> pattern has special significance. If a | |
1691 | C<(*MARK:NAME)> was encountered while matching, then it is that position | |
1692 | which is used as the "skip point". If no C<(*MARK)> of that name was | |
1693 | encountered, then the C<(*SKIP)> operator has no effect. When used | |
1694 | without a name the "skip point" is where the match point was when | |
1695 | executing the (*SKIP) pattern. | |
1696 | ||
0b928c2f | 1697 | Compare the following to the examples in C<(*PRUNE)>; note the string |
24b23f37 YO |
1698 | is twice as long: |
1699 | ||
d1fbf752 KW |
1700 | 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/; |
1701 | print "Count=$count\n"; | |
24b23f37 YO |
1702 | |
1703 | outputs | |
1704 | ||
1705 | aaab | |
1706 | aaab | |
1707 | Count=2 | |
1708 | ||
5d458dd8 | 1709 | Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)> |
353c6505 | 1710 | executed, the next starting point will be where the cursor was when the |
5d458dd8 YO |
1711 | C<(*SKIP)> was executed. |
1712 | ||
5d458dd8 | 1713 | =item C<(*MARK:NAME)> C<(*:NAME)> |
b16db30f | 1714 | X<(*MARK)> X<(*MARK:NAME)> X<(*:NAME)> |
5d458dd8 YO |
1715 | |
1716 | This zero-width pattern can be used to mark the point reached in a string | |
1717 | when a certain part of the pattern has been successfully matched. This | |
1718 | mark may be given a name. A later C<(*SKIP)> pattern will then skip | |
1719 | forward to that point if backtracked into on failure. Any number of | |
b4222fa9 | 1720 | C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated. |
5d458dd8 YO |
1721 | |
1722 | In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)> | |
1723 | can be used to "label" a pattern branch, so that after matching, the | |
1724 | program can determine which branches of the pattern were involved in the | |
1725 | match. | |
1726 | ||
1727 | When a match is successful, the C<$REGMARK> variable will be set to the | |
1728 | name of the most recently executed C<(*MARK:NAME)> that was involved | |
1729 | in the match. | |
1730 | ||
1731 | This can be used to determine which branch of a pattern was matched | |
c27a5cfe | 1732 | without using a separate capture group for each branch, which in turn |
5d458dd8 YO |
1733 | can result in a performance improvement, as perl cannot optimize |
1734 | C</(?:(x)|(y)|(z))/> as efficiently as something like | |
1735 | C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>. | |
1736 | ||
1737 | When a match has failed, and unless another verb has been involved in | |
1738 | failing the match and has provided its own name to use, the C<$REGERROR> | |
1739 | variable will be set to the name of the most recently executed | |
1740 | C<(*MARK:NAME)>. | |
1741 | ||
42ac7c82 | 1742 | See L</(*SKIP)> for more details. |
5d458dd8 | 1743 | |
b62d2d15 YO |
1744 | As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>. |
1745 | ||
5d458dd8 YO |
1746 | =item C<(*THEN)> C<(*THEN:NAME)> |
1747 | ||
241e7389 | 1748 | This is similar to the "cut group" operator C<::> from Perl 6. Like |
5d458dd8 YO |
1749 | C<(*PRUNE)>, this verb always matches, and when backtracked into on |
1750 | failure, it causes the regex engine to try the next alternation in the | |
1751 | innermost enclosing group (capturing or otherwise). | |
1752 | ||
1753 | Its name comes from the observation that this operation combined with the | |
1754 | alternation operator (C<|>) can be used to create what is essentially a | |
1755 | pattern-based if/then/else block: | |
1756 | ||
1757 | ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) | |
1758 | ||
1759 | Note that if this operator is used and NOT inside of an alternation then | |
1760 | it acts exactly like the C<(*PRUNE)> operator. | |
1761 | ||
1762 | / A (*PRUNE) B / | |
1763 | ||
1764 | is the same as | |
1765 | ||
1766 | / A (*THEN) B / | |
1767 | ||
1768 | but | |
1769 | ||
1770 | / ( A (*THEN) B | C (*THEN) D ) / | |
1771 | ||
1772 | is not the same as | |
1773 | ||
1774 | / ( A (*PRUNE) B | C (*PRUNE) D ) / | |
1775 | ||
1776 | as after matching the A but failing on the B the C<(*THEN)> verb will | |
1777 | backtrack and try C; but the C<(*PRUNE)> verb will simply fail. | |
24b23f37 | 1778 | |
cbeadc21 JV |
1779 | =back |
1780 | ||
1781 | =item Verbs without an argument | |
1782 | ||
1783 | =over 4 | |
1784 | ||
e2e6a0f1 YO |
1785 | =item C<(*COMMIT)> |
1786 | X<(*COMMIT)> | |
24b23f37 | 1787 | |
241e7389 | 1788 | This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a |
5d458dd8 YO |
1789 | zero-width pattern similar to C<(*SKIP)>, except that when backtracked |
1790 | into on failure it causes the match to fail outright. No further attempts | |
1791 | to find a valid match by advancing the start pointer will occur again. | |
1792 | For example, | |
24b23f37 | 1793 | |
d1fbf752 KW |
1794 | 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/; |
1795 | print "Count=$count\n"; | |
24b23f37 YO |
1796 | |
1797 | outputs | |
1798 | ||
1799 | aaab | |
1800 | Count=1 | |
1801 | ||
e2e6a0f1 YO |
1802 | In other words, once the C<(*COMMIT)> has been entered, and if the pattern |
1803 | does not match, the regex engine will not try any further matching on the | |
1804 | rest of the string. | |
c277df42 | 1805 | |
e2e6a0f1 YO |
1806 | =item C<(*FAIL)> C<(*F)> |
1807 | X<(*FAIL)> X<(*F)> | |
9af228c6 | 1808 | |
e2e6a0f1 YO |
1809 | This pattern matches nothing and always fails. It can be used to force the |
1810 | engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In | |
1811 | fact, C<(?!)> gets optimised into C<(*FAIL)> internally. | |
9af228c6 | 1812 | |
e2e6a0f1 | 1813 | It is probably useful only when combined with C<(?{})> or C<(??{})>. |
9af228c6 | 1814 | |
e2e6a0f1 YO |
1815 | =item C<(*ACCEPT)> |
1816 | X<(*ACCEPT)> | |
9af228c6 | 1817 | |
e2e6a0f1 YO |
1818 | B<WARNING:> This feature is highly experimental. It is not recommended |
1819 | for production code. | |
9af228c6 | 1820 | |
e2e6a0f1 YO |
1821 | This pattern matches nothing and causes the end of successful matching at |
1822 | the point at which the C<(*ACCEPT)> pattern was encountered, regardless of | |
1823 | whether there is actually more to match in the string. When inside of a | |
0d017f4d | 1824 | nested pattern, such as recursion, or in a subpattern dynamically generated |
e2e6a0f1 | 1825 | via C<(??{})>, only the innermost pattern is ended immediately. |
9af228c6 | 1826 | |
c27a5cfe | 1827 | If the C<(*ACCEPT)> is inside of capturing groups then the groups are |
e2e6a0f1 YO |
1828 | marked as ended at the point at which the C<(*ACCEPT)> was encountered. |
1829 | For instance: | |
9af228c6 | 1830 | |
e2e6a0f1 | 1831 | 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x; |
9af228c6 | 1832 | |
e2e6a0f1 | 1833 | will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not |
0b928c2f | 1834 | be set. If another branch in the inner parentheses was matched, such as in the |
e2e6a0f1 | 1835 | string 'ACDE', then the C<D> and C<E> would have to be matched as well. |
9af228c6 YO |
1836 | |
1837 | =back | |
c277df42 | 1838 | |
a0d0e21e LW |
1839 | =back |
1840 | ||
c07a80fd | 1841 | =head2 Backtracking |
d74e8afc | 1842 | X<backtrack> X<backtracking> |
c07a80fd | 1843 | |
35a734be IZ |
1844 | NOTE: This section presents an abstract approximation of regular |
1845 | expression behavior. For a more rigorous (and complicated) view of | |
1846 | the rules involved in selecting a match among possible alternatives, | |
0d017f4d | 1847 | see L<Combining RE Pieces>. |
35a734be | 1848 | |
c277df42 | 1849 | A fundamental feature of regular expression matching involves the |
5a964f20 | 1850 | notion called I<backtracking>, which is currently used (when needed) |
0d017f4d | 1851 | by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc IZ |
1852 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
1853 | internally, but the general principle outlined here is valid. | |
c07a80fd | 1854 | |
1855 | For a regular expression to match, the I<entire> regular expression must | |
1856 | match, not just part of it. So if the beginning of a pattern containing a | |
1857 | quantifier succeeds in a way that causes later parts in the pattern to | |
1858 | fail, the matching engine backs up and recalculates the beginning | |
1859 | part--that's why it's called backtracking. | |
1860 | ||
1861 | Here is an example of backtracking: Let's say you want to find the | |
1862 | word following "foo" in the string "Food is on the foo table.": | |
1863 | ||
1864 | $_ = "Food is on the foo table."; | |
1865 | if ( /\b(foo)\s+(\w+)/i ) { | |
f793d64a | 1866 | print "$2 follows $1.\n"; |
c07a80fd | 1867 | } |
1868 | ||
1869 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
1870 | finds a possible match right at the beginning of the string, and loads up | |
1871 | $1 with "Foo". However, as soon as the matching engine sees that there's | |
1872 | no whitespace following the "Foo" that it had saved in $1, it realizes its | |
68dc0745 | 1873 | mistake and starts over again one character after where it had the |
c07a80fd | 1874 | tentative match. This time it goes all the way until the next occurrence |
1875 | of "foo". The complete regular expression matches this time, and you get | |
1876 | the expected output of "table follows foo." | |
1877 | ||
1878 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
1879 | everything between "foo" and "bar". Initially, you write something | |
1880 | like this: | |
1881 | ||
1882 | $_ = "The food is under the bar in the barn."; | |
1883 | if ( /foo(.*)bar/ ) { | |
f793d64a | 1884 | print "got <$1>\n"; |
c07a80fd | 1885 | } |
1886 | ||
1887 | Which perhaps unexpectedly yields: | |
1888 | ||
1889 | got <d is under the bar in the > | |
1890 | ||
1891 | That's because C<.*> was greedy, so you get everything between the | |
14218588 | 1892 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd | 1893 | to use minimal matching to make sure you get the text between a "foo" |
1894 | and the first "bar" thereafter. | |
1895 | ||
1896 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
1897 | got <d is under the > | |
1898 | ||
0d017f4d | 1899 | Here's another example. Let's say you'd like to match a number at the end |
b6e13d97 | 1900 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd | 1901 | So you write this: |
1902 | ||
1903 | $_ = "I have 2 numbers: 53147"; | |
f793d64a KW |
1904 | if ( /(.*)(\d*)/ ) { # Wrong! |
1905 | print "Beginning is <$1>, number is <$2>.\n"; | |
c07a80fd | 1906 | } |
1907 | ||
1908 | That won't work at all, because C<.*> was greedy and gobbled up the | |
1909 | whole string. As C<\d*> can match on an empty string the complete | |
1910 | regular expression matched successfully. | |
1911 | ||
8e1088bc | 1912 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd | 1913 | |
1914 | Here are some variants, most of which don't work: | |
1915 | ||
1916 | $_ = "I have 2 numbers: 53147"; | |
1917 | @pats = qw{ | |
f793d64a KW |
1918 | (.*)(\d*) |
1919 | (.*)(\d+) | |
1920 | (.*?)(\d*) | |
1921 | (.*?)(\d+) | |
1922 | (.*)(\d+)$ | |
1923 | (.*?)(\d+)$ | |
1924 | (.*)\b(\d+)$ | |
1925 | (.*\D)(\d+)$ | |
c07a80fd | 1926 | }; |
1927 | ||
1928 | for $pat (@pats) { | |
f793d64a KW |
1929 | printf "%-12s ", $pat; |
1930 | if ( /$pat/ ) { | |
1931 | print "<$1> <$2>\n"; | |
1932 | } else { | |
1933 | print "FAIL\n"; | |
1934 | } | |
c07a80fd | 1935 | } |
1936 | ||
1937 | That will print out: | |
1938 | ||
1939 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
1940 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
1941 | (.*?)(\d*) <> <> | |
1942 | (.*?)(\d+) <I have > <2> | |
1943 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
1944 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
1945 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
1946 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
1947 | ||
1948 | As you see, this can be a bit tricky. It's important to realize that a | |
1949 | regular expression is merely a set of assertions that gives a definition | |
1950 | of success. There may be 0, 1, or several different ways that the | |
1951 | definition might succeed against a particular string. And if there are | |
5a964f20 TC |
1952 | multiple ways it might succeed, you need to understand backtracking to |
1953 | know which variety of success you will achieve. | |
c07a80fd | 1954 | |
19799a22 | 1955 | When using look-ahead assertions and negations, this can all get even |
8b19b778 | 1956 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd | 1957 | followed by "123". You might try to write that as |
1958 | ||
871b0233 | 1959 | $_ = "ABC123"; |
f793d64a KW |
1960 | if ( /^\D*(?!123)/ ) { # Wrong! |
1961 | print "Yup, no 123 in $_\n"; | |
871b0233 | 1962 | } |
c07a80fd | 1963 | |
1964 | But that isn't going to match; at least, not the way you're hoping. It | |
1965 | claims that there is no 123 in the string. Here's a clearer picture of | |
9b9391b2 | 1966 | why that pattern matches, contrary to popular expectations: |
c07a80fd | 1967 | |
4358a253 SS |
1968 | $x = 'ABC123'; |
1969 | $y = 'ABC445'; | |
c07a80fd | 1970 | |
4358a253 SS |
1971 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
1972 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; | |
c07a80fd | 1973 | |
4358a253 SS |
1974 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; |
1975 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; | |
c07a80fd | 1976 | |
1977 | This prints | |
1978 | ||
1979 | 2: got ABC | |
1980 | 3: got AB | |
1981 | 4: got ABC | |
1982 | ||
5f05dabc | 1983 | You might have expected test 3 to fail because it seems to a more |
c07a80fd | 1984 | general purpose version of test 1. The important difference between |
1985 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
1986 | backtracking, whereas test 1 will not. What's happening is | |
1987 | that you've asked "Is it true that at the start of $x, following 0 or more | |
5f05dabc | 1988 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd | 1989 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 | 1990 | fail. |
14218588 | 1991 | |
c07a80fd | 1992 | The search engine will initially match C<\D*> with "ABC". Then it will |
0b928c2f | 1993 | try to match C<(?!123)> with "123", which fails. But because |
c07a80fd | 1994 | a quantifier (C<\D*>) has been used in the regular expression, the |
1995 | search engine can backtrack and retry the match differently | |
54310121 | 1996 | in the hope of matching the complete regular expression. |
c07a80fd | 1997 | |
5a964f20 TC |
1998 | The pattern really, I<really> wants to succeed, so it uses the |
1999 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
c07a80fd | 2000 | time. Now there's indeed something following "AB" that is not |
14218588 | 2001 | "123". It's "C123", which suffices. |
c07a80fd | 2002 | |
14218588 GS |
2003 | We can deal with this by using both an assertion and a negation. |
2004 | We'll say that the first part in $1 must be followed both by a digit | |
2005 | and by something that's not "123". Remember that the look-aheads | |
2006 | are zero-width expressions--they only look, but don't consume any | |
2007 | of the string in their match. So rewriting this way produces what | |
c07a80fd | 2008 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
2009 | ||
4358a253 SS |
2010 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; |
2011 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; | |
c07a80fd | 2012 | |
2013 | 6: got ABC | |
2014 | ||
5a964f20 | 2015 | In other words, the two zero-width assertions next to each other work as though |
19799a22 | 2016 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd | 2017 | matches only if you're at the beginning of the line AND the end of the |
2018 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
2019 | regular expressions always means AND, except when you write an explicit OR | |
2020 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
2021 | although the attempted matches are made at different positions because "a" | |
2022 | is not a zero-width assertion, but a one-width assertion. | |
2023 | ||
0d017f4d | 2024 | B<WARNING>: Particularly complicated regular expressions can take |
14218588 | 2025 | exponential time to solve because of the immense number of possible |
0d017f4d | 2026 | ways they can use backtracking to try for a match. For example, without |
9da458fc IZ |
2027 | internal optimizations done by the regular expression engine, this will |
2028 | take a painfully long time to run: | |
c07a80fd | 2029 | |
e1901655 IZ |
2030 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
2031 | ||
2032 | And if you used C<*>'s in the internal groups instead of limiting them | |
2033 | to 0 through 5 matches, then it would take forever--or until you ran | |
2034 | out of stack space. Moreover, these internal optimizations are not | |
2035 | always applicable. For example, if you put C<{0,5}> instead of C<*> | |
2036 | on the external group, no current optimization is applicable, and the | |
2037 | match takes a long time to finish. | |
c07a80fd | 2038 | |
9da458fc IZ |
2039 | A powerful tool for optimizing such beasts is what is known as an |
2040 | "independent group", | |
96090e4f | 2041 | which does not backtrack (see L</C<< (?>pattern) >>>). Note also that |
9da458fc | 2042 | zero-length look-ahead/look-behind assertions will not backtrack to make |
5d458dd8 | 2043 | the tail match, since they are in "logical" context: only |
14218588 | 2044 | whether they match is considered relevant. For an example |
9da458fc | 2045 | where side-effects of look-ahead I<might> have influenced the |
96090e4f | 2046 | following match, see L</C<< (?>pattern) >>>. |
c277df42 | 2047 | |
a0d0e21e | 2048 | =head2 Version 8 Regular Expressions |
d74e8afc | 2049 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> |
a0d0e21e | 2050 | |
5a964f20 | 2051 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e LW |
2052 | routines, here are the pattern-matching rules not described above. |
2053 | ||
54310121 | 2054 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e | 2055 | with a special meaning described here or above. You can cause |
5a964f20 | 2056 | characters that normally function as metacharacters to be interpreted |
5f05dabc | 2057 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
0d017f4d WL |
2058 | character; "\\" matches a "\"). This escape mechanism is also required |
2059 | for the character used as the pattern delimiter. | |
2060 | ||
2061 | A series of characters matches that series of characters in the target | |
0b928c2f | 2062 | string, so the pattern C<blurfl> would match "blurfl" in the target |
0d017f4d | 2063 | string. |
a0d0e21e LW |
2064 | |
2065 | You can specify a character class, by enclosing a list of characters | |
5d458dd8 | 2066 | in C<[]>, which will match any character from the list. If the |
a0d0e21e | 2067 | first character after the "[" is "^", the class matches any character not |
14218588 | 2068 | in the list. Within a list, the "-" character specifies a |
5a964f20 | 2069 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 GS |
2070 | inclusive. If you want either "-" or "]" itself to be a member of a |
2071 | class, put it at the start of the list (possibly after a "^"), or | |
2072 | escape it with a backslash. "-" is also taken literally when it is | |
2073 | at the end of the list, just before the closing "]". (The | |
84850974 DD |
2074 | following all specify the same class of three characters: C<[-az]>, |
2075 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which | |
5d458dd8 YO |
2076 | specifies a class containing twenty-six characters, even on EBCDIC-based |
2077 | character sets.) Also, if you try to use the character | |
2078 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of | |
2079 | a range, the "-" is understood literally. | |
a0d0e21e | 2080 | |
8ada0baa JH |
2081 | Note also that the whole range idea is rather unportable between |
2082 | character sets--and even within character sets they may cause results | |
2083 | you probably didn't expect. A sound principle is to use only ranges | |
0d017f4d | 2084 | that begin from and end at either alphabetics of equal case ([a-e], |
8ada0baa JH |
2085 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, |
2086 | spell out the character sets in full. | |
2087 | ||
54310121 | 2088 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e LW |
2089 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
2090 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string | |
dc0d9c48 | 2091 | of three octal digits, matches the character whose coded character set value |
5d458dd8 | 2092 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, |
dc0d9c48 | 2093 | matches the character whose ordinal is I<nn>. The expression \cI<x> |
5d458dd8 | 2094 | matches the character control-I<x>. Finally, the "." metacharacter |
fb55449c | 2095 | matches any character except "\n" (unless you use C</s>). |
a0d0e21e LW |
2096 | |
2097 | You can specify a series of alternatives for a pattern using "|" to | |
2098 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", | |
5a964f20 | 2099 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e | 2100 | first alternative includes everything from the last pattern delimiter |
0b928c2f | 2101 | ("(", "(?:", etc. or the beginning of the pattern) up to the first "|", and |
a0d0e21e | 2102 | the last alternative contains everything from the last "|" to the next |
0b928c2f | 2103 | closing pattern delimiter. That's why it's common practice to include |
14218588 | 2104 | alternatives in parentheses: to minimize confusion about where they |
a3cb178b GS |
2105 | start and end. |
2106 | ||
5a964f20 | 2107 | Alternatives are tried from left to right, so the first |
a3cb178b GS |
2108 | alternative found for which the entire expression matches, is the one that |
2109 | is chosen. This means that alternatives are not necessarily greedy. For | |
628afcb5 | 2110 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b GS |
2111 | part will match, as that is the first alternative tried, and it successfully |
2112 | matches the target string. (This might not seem important, but it is | |
2113 | important when you are capturing matched text using parentheses.) | |
2114 | ||
5a964f20 | 2115 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b | 2116 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e | 2117 | |
14218588 GS |
2118 | Within a pattern, you may designate subpatterns for later reference |
2119 | by enclosing them in parentheses, and you may refer back to the | |
2120 | I<n>th subpattern later in the pattern using the metacharacter | |
0b928c2f | 2121 | \I<n> or \gI<n>. Subpatterns are numbered based on the left to right order |
14218588 GS |
2122 | of their opening parenthesis. A backreference matches whatever |
2123 | actually matched the subpattern in the string being examined, not | |
d8b950dc | 2124 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\g1\d*> will |
14218588 GS |
2125 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern |
2126 | 1 matched "0x", even though the rule C<0|0x> could potentially match | |
2127 | the leading 0 in the second number. | |
cb1a09d0 | 2128 | |
0d017f4d | 2129 | =head2 Warning on \1 Instead of $1 |
cb1a09d0 | 2130 | |
5a964f20 | 2131 | Some people get too used to writing things like: |
cb1a09d0 AD |
2132 | |
2133 | $pattern =~ s/(\W)/\\\1/g; | |
2134 | ||
3ff1c45a KW |
2135 | This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid |
2136 | shocking the | |
cb1a09d0 | 2137 | B<sed> addicts, but it's a dirty habit to get into. That's because in |
d1be9408 | 2138 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 AD |
2139 | the usual double-quoted string means a control-A. The customary Unix |
2140 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
2141 | of doing that, you get yourself into trouble if you then add an C</e> | |
2142 | modifier. | |
2143 | ||
f793d64a | 2144 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 AD |
2145 | |
2146 | Or if you try to do | |
2147 | ||
2148 | s/(\d+)/\1000/; | |
2149 | ||
2150 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
14218588 | 2151 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 AD |
2152 | with the operation of matching a backreference. Certainly they mean two |
2153 | different things on the I<left> side of the C<s///>. | |
9fa51da4 | 2154 | |
0d017f4d | 2155 | =head2 Repeated Patterns Matching a Zero-length Substring |
c84d73f1 | 2156 | |
19799a22 | 2157 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 IZ |
2158 | |
2159 | Regular expressions provide a terse and powerful programming language. As | |
2160 | with most other power tools, power comes together with the ability | |
2161 | to wreak havoc. | |
2162 | ||
2163 | A common abuse of this power stems from the ability to make infinite | |
628afcb5 | 2164 | loops using regular expressions, with something as innocuous as: |
c84d73f1 IZ |
2165 | |
2166 | 'foo' =~ m{ ( o? )* }x; | |
2167 | ||
0d017f4d | 2168 | The C<o?> matches at the beginning of C<'foo'>, and since the position |
c84d73f1 | 2169 | in the string is not moved by the match, C<o?> would match again and again |
527e91da | 2170 | because of the C<*> quantifier. Another common way to create a similar cycle |
c84d73f1 IZ |
2171 | is with the looping modifier C<//g>: |
2172 | ||
2173 | @matches = ( 'foo' =~ m{ o? }xg ); | |
2174 | ||
2175 | or | |
2176 | ||
2177 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
2178 | ||
2179 | or the loop implied by split(). | |
2180 | ||
2181 | However, long experience has shown that many programming tasks may | |
14218588 GS |
2182 | be significantly simplified by using repeated subexpressions that |
2183 | may match zero-length substrings. Here's a simple example being: | |
c84d73f1 | 2184 | |
d1fbf752 | 2185 | @chars = split //, $string; # // is not magic in split |
c84d73f1 IZ |
2186 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / |
2187 | ||
9da458fc | 2188 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 | 2189 | the infinite loop>. The rules for this are different for lower-level |
527e91da | 2190 | loops given by the greedy quantifiers C<*+{}>, and for higher-level |
c84d73f1 IZ |
2191 | ones like the C</g> modifier or split() operator. |
2192 | ||
19799a22 GS |
2193 | The lower-level loops are I<interrupted> (that is, the loop is |
2194 | broken) when Perl detects that a repeated expression matched a | |
2195 | zero-length substring. Thus | |
c84d73f1 IZ |
2196 | |
2197 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
2198 | ||
5d458dd8 | 2199 | is made equivalent to |
c84d73f1 | 2200 | |
0b928c2f FC |
2201 | m{ (?: NON_ZERO_LENGTH )* (?: ZERO_LENGTH )? }x; |
2202 | ||
2203 | For example, this program | |
2204 | ||
2205 | #!perl -l | |
2206 | "aaaaab" =~ / | |
2207 | (?: | |
2208 | a # non-zero | |
2209 | | # or | |
2210 | (?{print "hello"}) # print hello whenever this | |
2211 | # branch is tried | |
2212 | (?=(b)) # zero-width assertion | |
2213 | )* # any number of times | |
2214 | /x; | |
2215 | print $&; | |
2216 | print $1; | |
c84d73f1 | 2217 | |
0b928c2f FC |
2218 | prints |
2219 | ||
2220 | hello | |
2221 | aaaaa | |
2222 | b | |
2223 | ||
2224 | Notice that "hello" is only printed once, as when Perl sees that the sixth | |
2225 | iteration of the outermost C<(?:)*> matches a zero-length string, it stops | |
2226 | the C<*>. | |
2227 | ||
2228 | The higher-level loops preserve an additional state between iterations: | |
5d458dd8 | 2229 | whether the last match was zero-length. To break the loop, the following |
c84d73f1 | 2230 | match after a zero-length match is prohibited to have a length of zero. |
5d458dd8 | 2231 | This prohibition interacts with backtracking (see L<"Backtracking">), |
c84d73f1 IZ |
2232 | and so the I<second best> match is chosen if the I<best> match is of |
2233 | zero length. | |
2234 | ||
19799a22 | 2235 | For example: |
c84d73f1 IZ |
2236 | |
2237 | $_ = 'bar'; | |
2238 | s/\w??/<$&>/g; | |
2239 | ||
20fb949f | 2240 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
5d458dd8 | 2241 | match given by non-greedy C<??> is the zero-length match, and the I<second |
c84d73f1 IZ |
2242 | best> match is what is matched by C<\w>. Thus zero-length matches |
2243 | alternate with one-character-long matches. | |
2244 | ||
5d458dd8 | 2245 | Similarly, for repeated C<m/()/g> the second-best match is the match at the |
c84d73f1 IZ |
2246 | position one notch further in the string. |
2247 | ||
19799a22 | 2248 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 | 2249 | the matched string, and is reset by each assignment to pos(). |
9da458fc IZ |
2250 | Zero-length matches at the end of the previous match are ignored |
2251 | during C<split>. | |
c84d73f1 | 2252 | |
0d017f4d | 2253 | =head2 Combining RE Pieces |
35a734be IZ |
2254 | |
2255 | Each of the elementary pieces of regular expressions which were described | |
2256 | before (such as C<ab> or C<\Z>) could match at most one substring | |
2257 | at the given position of the input string. However, in a typical regular | |
2258 | expression these elementary pieces are combined into more complicated | |
0b928c2f | 2259 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc. |
35a734be IZ |
2260 | (in these examples C<S> and C<T> are regular subexpressions). |
2261 | ||
2262 | Such combinations can include alternatives, leading to a problem of choice: | |
2263 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
2264 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
2265 | actually matched is the concept of backtracking (see L<"Backtracking">). | |
2266 | However, this description is too low-level and makes you think | |
2267 | in terms of a particular implementation. | |
2268 | ||
2269 | Another description starts with notions of "better"/"worse". All the | |
2270 | substrings which may be matched by the given regular expression can be | |
2271 | sorted from the "best" match to the "worst" match, and it is the "best" | |
2272 | match which is chosen. This substitutes the question of "what is chosen?" | |
2273 | by the question of "which matches are better, and which are worse?". | |
2274 | ||
2275 | Again, for elementary pieces there is no such question, since at most | |
2276 | one match at a given position is possible. This section describes the | |
2277 | notion of better/worse for combining operators. In the description | |
2278 | below C<S> and C<T> are regular subexpressions. | |
2279 | ||
13a2d996 | 2280 | =over 4 |
35a734be IZ |
2281 | |
2282 | =item C<ST> | |
2283 | ||
2284 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are | |
2285 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings | |
5d458dd8 | 2286 | which can be matched by C<T>. |
35a734be | 2287 | |
0b928c2f | 2288 | If C<A> is a better match for C<S> than C<A'>, C<AB> is a better |
35a734be IZ |
2289 | match than C<A'B'>. |
2290 | ||
2291 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if | |
0b928c2f | 2292 | C<B> is a better match for C<T> than C<B'>. |
35a734be IZ |
2293 | |
2294 | =item C<S|T> | |
2295 | ||
2296 | When C<S> can match, it is a better match than when only C<T> can match. | |
2297 | ||
2298 | Ordering of two matches for C<S> is the same as for C<S>. Similar for | |
2299 | two matches for C<T>. | |
2300 | ||
2301 | =item C<S{REPEAT_COUNT}> | |
2302 | ||
2303 | Matches as C<SSS...S> (repeated as many times as necessary). | |
2304 | ||
2305 | =item C<S{min,max}> | |
2306 | ||
2307 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
2308 | ||
2309 | =item C<S{min,max}?> | |
2310 | ||
2311 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
2312 | ||
2313 | =item C<S?>, C<S*>, C<S+> | |
2314 | ||
2315 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
2316 | ||
2317 | =item C<S??>, C<S*?>, C<S+?> | |
2318 | ||
2319 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
2320 | ||
c47ff5f1 | 2321 | =item C<< (?>S) >> |
35a734be IZ |
2322 | |
2323 | Matches the best match for C<S> and only that. | |
2324 | ||
2325 | =item C<(?=S)>, C<(?<=S)> | |
2326 | ||
2327 | Only the best match for C<S> is considered. (This is important only if | |
2328 | C<S> has capturing parentheses, and backreferences are used somewhere | |
2329 | else in the whole regular expression.) | |
2330 | ||
2331 | =item C<(?!S)>, C<(?<!S)> | |
2332 | ||
2333 | For this grouping operator there is no need to describe the ordering, since | |
2334 | only whether or not C<S> can match is important. | |
2335 | ||
6bda09f9 | 2336 | =item C<(??{ EXPR })>, C<(?PARNO)> |
35a734be IZ |
2337 | |
2338 | The ordering is the same as for the regular expression which is | |
c27a5cfe | 2339 | the result of EXPR, or the pattern contained by capture group PARNO. |
35a734be IZ |
2340 | |
2341 | =item C<(?(condition)yes-pattern|no-pattern)> | |
2342 | ||
2343 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is | |
2344 | already determined. The ordering of the matches is the same as for the | |
2345 | chosen subexpression. | |
2346 | ||
2347 | =back | |
2348 | ||
2349 | The above recipes describe the ordering of matches I<at a given position>. | |
2350 | One more rule is needed to understand how a match is determined for the | |
2351 | whole regular expression: a match at an earlier position is always better | |
2352 | than a match at a later position. | |
2353 | ||
0d017f4d | 2354 | =head2 Creating Custom RE Engines |
c84d73f1 | 2355 | |
0b928c2f FC |
2356 | As of Perl 5.10.0, one can create custom regular expression engines. This |
2357 | is not for the faint of heart, as they have to plug in at the C level. See | |
2358 | L<perlreapi> for more details. | |
2359 | ||
2360 | As an alternative, overloaded constants (see L<overload>) provide a simple | |
2361 | way to extend the functionality of the RE engine, by substituting one | |
2362 | pattern for another. | |
c84d73f1 IZ |
2363 | |
2364 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
0d017f4d | 2365 | matches at a boundary between whitespace characters and non-whitespace |
c84d73f1 IZ |
2366 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
2367 | at these positions, so we want to have each C<\Y|> in the place of the | |
2368 | more complicated version. We can create a module C<customre> to do | |
2369 | this: | |
2370 | ||
2371 | package customre; | |
2372 | use overload; | |
2373 | ||
2374 | sub import { | |
2375 | shift; | |
2376 | die "No argument to customre::import allowed" if @_; | |
2377 | overload::constant 'qr' => \&convert; | |
2378 | } | |
2379 | ||
2380 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
2381 | ||
580a9fe1 RGS |
2382 | # We must also take care of not escaping the legitimate \\Y| |
2383 | # sequence, hence the presence of '\\' in the conversion rules. | |
5d458dd8 | 2384 | my %rules = ( '\\' => '\\\\', |
f793d64a | 2385 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
c84d73f1 IZ |
2386 | sub convert { |
2387 | my $re = shift; | |
5d458dd8 | 2388 | $re =~ s{ |
c84d73f1 IZ |
2389 | \\ ( \\ | Y . ) |
2390 | } | |
5d458dd8 | 2391 | { $rules{$1} or invalid($re,$1) }sgex; |
c84d73f1 IZ |
2392 | return $re; |
2393 | } | |
2394 | ||
2395 | Now C<use customre> enables the new escape in constant regular | |
2396 | expressions, i.e., those without any runtime variable interpolations. | |
2397 | As documented in L<overload>, this conversion will work only over | |
2398 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
2399 | part of this regular expression needs to be converted explicitly | |
2400 | (but only if the special meaning of C<\Y|> should be enabled inside $re): | |
2401 | ||
2402 | use customre; | |
2403 | $re = <>; | |
2404 | chomp $re; | |
2405 | $re = customre::convert $re; | |
2406 | /\Y|$re\Y|/; | |
2407 | ||
0b928c2f | 2408 | =head2 PCRE/Python Support |
1f1031fe | 2409 | |
0b928c2f | 2410 | As of Perl 5.10.0, Perl supports several Python/PCRE-specific extensions |
1f1031fe | 2411 | to the regex syntax. While Perl programmers are encouraged to use the |
0b928c2f | 2412 | Perl-specific syntax, the following are also accepted: |
1f1031fe YO |
2413 | |
2414 | =over 4 | |
2415 | ||
ae5648b3 | 2416 | =item C<< (?PE<lt>NAMEE<gt>pattern) >> |
1f1031fe | 2417 | |
c27a5cfe | 2418 | Define a named capture group. Equivalent to C<< (?<NAME>pattern) >>. |
1f1031fe YO |
2419 | |
2420 | =item C<< (?P=NAME) >> | |
2421 | ||
c27a5cfe | 2422 | Backreference to a named capture group. Equivalent to C<< \g{NAME} >>. |
1f1031fe YO |
2423 | |
2424 | =item C<< (?P>NAME) >> | |
2425 | ||
c27a5cfe | 2426 | Subroutine call to a named capture group. Equivalent to C<< (?&NAME) >>. |
1f1031fe | 2427 | |
ee9b8eae | 2428 | =back |
1f1031fe | 2429 | |
19799a22 GS |
2430 | =head1 BUGS |
2431 | ||
88c9975e KW |
2432 | Many regular expression constructs don't work on EBCDIC platforms. |
2433 | ||
ed7efc79 KW |
2434 | There are a number of issues with regard to case-insensitive matching |
2435 | in Unicode rules. See C<i> under L</Modifiers> above. | |
2436 | ||
9da458fc IZ |
2437 | This document varies from difficult to understand to completely |
2438 | and utterly opaque. The wandering prose riddled with jargon is | |
2439 | hard to fathom in several places. | |
2440 | ||
2441 | This document needs a rewrite that separates the tutorial content | |
2442 | from the reference content. | |
19799a22 GS |
2443 | |
2444 | =head1 SEE ALSO | |
9fa51da4 | 2445 | |
91e0c79e MJD |
2446 | L<perlrequick>. |
2447 | ||
2448 | L<perlretut>. | |
2449 | ||
9b599b2a GS |
2450 | L<perlop/"Regexp Quote-Like Operators">. |
2451 | ||
1e66bd83 PP |
2452 | L<perlop/"Gory details of parsing quoted constructs">. |
2453 | ||
14218588 GS |
2454 | L<perlfaq6>. |
2455 | ||
9b599b2a GS |
2456 | L<perlfunc/pos>. |
2457 | ||
2458 | L<perllocale>. | |
2459 | ||
fb55449c JH |
2460 | L<perlebcdic>. |
2461 | ||
14218588 GS |
2462 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
2463 | by O'Reilly and Associates. |