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