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 | |
55 | locale. See L<perllocale>. | |
56 | ||
54310121 | 57 | =item x |
d74e8afc | 58 | X</x> |
55497cff | 59 | |
60 | Extend your pattern's legibility by permitting whitespace and comments. | |
61 | ||
87e95b7f YO |
62 | =item p |
63 | X</p> X<regex, preserve> X<regexp, preserve> | |
64 | ||
632a1772 | 65 | Preserve the string matched such that ${^PREMATCH}, ${^MATCH}, and |
87e95b7f YO |
66 | ${^POSTMATCH} are available for use after matching. |
67 | ||
e2e6bec7 DN |
68 | =item g and c |
69 | X</g> X</c> | |
70 | ||
71 | Global matching, and keep the Current position after failed matching. | |
72 | Unlike i, m, s and x, these two flags affect the way the regex is used | |
73 | rather than the regex itself. See | |
74 | L<perlretut/"Using regular expressions in Perl"> for further explanation | |
75 | of the g and c modifiers. | |
76 | ||
55497cff | 77 | =back |
a0d0e21e LW |
78 | |
79 | These are usually written as "the C</x> modifier", even though the delimiter | |
14218588 | 80 | in question might not really be a slash. Any of these |
a0d0e21e | 81 | modifiers may also be embedded within the regular expression itself using |
14218588 | 82 | the C<(?...)> construct. See below. |
a0d0e21e | 83 | |
4633a7c4 | 84 | The C</x> modifier itself needs a little more explanation. It tells |
7b059540 | 85 | the regular expression parser to ignore most whitespace that is neither |
55497cff | 86 | backslashed nor within a character class. You can use this to break up |
4633a7c4 | 87 | your regular expression into (slightly) more readable parts. The C<#> |
54310121 | 88 | character is also treated as a metacharacter introducing a comment, |
55497cff | 89 | just as in ordinary Perl code. This also means that if you want real |
14218588 | 90 | whitespace or C<#> characters in the pattern (outside a character |
f9a3ff1a | 91 | class, where they are unaffected by C</x>), then you'll either have to |
7b059540 KW |
92 | escape them (using backslashes or C<\Q...\E>) or encode them using octal, |
93 | hex, or C<\N{}> escapes. Taken together, these features go a long way towards | |
8933a740 RGS |
94 | making Perl's regular expressions more readable. Note that you have to |
95 | be careful not to include the pattern delimiter in the comment--perl has | |
96 | no way of knowing you did not intend to close the pattern early. See | |
97 | the C-comment deletion code in L<perlop>. Also note that anything inside | |
7651b971 KW |
98 | a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect |
99 | whether space interpretation within a single multi-character construct. For | |
100 | example in C<\x{...}>, regardless of the C</x> modifier, there can be no | |
9bb1f947 | 101 | spaces. Same for a L<quantifier|/Quantifiers> such as C<{3}> or |
f9e949fd KW |
102 | C<{5,}>. Similarly, C<(?:...)> can't have a space between the C<?> and C<:>, |
103 | but can between the C<(> and C<?>. Within any delimiters for such a | |
104 | construct, allowed spaces are not affected by C</x>, and depend on the | |
105 | construct. For example, C<\x{...}> can't have spaces because hexadecimal | |
106 | numbers don't have spaces in them. But, Unicode properties can have spaces, so | |
107 | in C<\p{...}> there can be spaces that follow the Unicode rules, for which see | |
9bb1f947 | 108 | L<perluniprops/Properties accessible through \p{} and \P{}>. |
d74e8afc | 109 | X</x> |
a0d0e21e LW |
110 | |
111 | =head2 Regular Expressions | |
112 | ||
04838cea RGS |
113 | =head3 Metacharacters |
114 | ||
384f06ae | 115 | The patterns used in Perl pattern matching evolved from those supplied in |
14218588 | 116 | the Version 8 regex routines. (The routines are derived |
19799a22 GS |
117 | (distantly) from Henry Spencer's freely redistributable reimplementation |
118 | of the V8 routines.) See L<Version 8 Regular Expressions> for | |
119 | details. | |
a0d0e21e LW |
120 | |
121 | In particular the following metacharacters have their standard I<egrep>-ish | |
122 | meanings: | |
d74e8afc ITB |
123 | X<metacharacter> |
124 | X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]> | |
125 | ||
a0d0e21e | 126 | |
54310121 | 127 | \ Quote the next metacharacter |
a0d0e21e LW |
128 | ^ Match the beginning of the line |
129 | . Match any character (except newline) | |
c07a80fd | 130 | $ Match the end of the line (or before newline at the end) |
a0d0e21e LW |
131 | | Alternation |
132 | () Grouping | |
9bb1f947 | 133 | [] Bracketed Character class |
a0d0e21e | 134 | |
14218588 GS |
135 | By default, the "^" character is guaranteed to match only the |
136 | beginning of the string, the "$" character only the end (or before the | |
137 | newline at the end), and Perl does certain optimizations with the | |
a0d0e21e LW |
138 | assumption that the string contains only one line. Embedded newlines |
139 | will not be matched by "^" or "$". You may, however, wish to treat a | |
4a6725af | 140 | string as a multi-line buffer, such that the "^" will match after any |
0d520e8e YO |
141 | newline within the string (except if the newline is the last character in |
142 | the string), and "$" will match before any newline. At the | |
a0d0e21e LW |
143 | cost of a little more overhead, you can do this by using the /m modifier |
144 | on the pattern match operator. (Older programs did this by setting C<$*>, | |
f02c194e | 145 | but this practice has been removed in perl 5.9.) |
d74e8afc | 146 | X<^> X<$> X</m> |
a0d0e21e | 147 | |
14218588 | 148 | To simplify multi-line substitutions, the "." character never matches a |
55497cff | 149 | newline unless you use the C</s> modifier, which in effect tells Perl to pretend |
f02c194e | 150 | the string is a single line--even if it isn't. |
d74e8afc | 151 | X<.> X</s> |
a0d0e21e | 152 | |
04838cea RGS |
153 | =head3 Quantifiers |
154 | ||
a0d0e21e | 155 | The following standard quantifiers are recognized: |
d74e8afc | 156 | X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}> |
a0d0e21e LW |
157 | |
158 | * Match 0 or more times | |
159 | + Match 1 or more times | |
160 | ? Match 1 or 0 times | |
161 | {n} Match exactly n times | |
162 | {n,} Match at least n times | |
163 | {n,m} Match at least n but not more than m times | |
164 | ||
165 | (If a curly bracket occurs in any other context, it is treated | |
b975c076 | 166 | as a regular character. In particular, the lower bound |
527e91da BB |
167 | is not optional.) The "*" quantifier is equivalent to C<{0,}>, the "+" |
168 | quantifier to C<{1,}>, and the "?" quantifier to C<{0,1}>. n and m are limited | |
d0b16107 | 169 | to non-negative integral values less than a preset limit defined when perl is built. |
9c79236d GS |
170 | This is usually 32766 on the most common platforms. The actual limit can |
171 | be seen in the error message generated by code such as this: | |
172 | ||
820475bd | 173 | $_ **= $_ , / {$_} / for 2 .. 42; |
a0d0e21e | 174 | |
54310121 | 175 | By default, a quantified subpattern is "greedy", that is, it will match as |
176 | many times as possible (given a particular starting location) while still | |
177 | allowing the rest of the pattern to match. If you want it to match the | |
178 | minimum number of times possible, follow the quantifier with a "?". Note | |
179 | that the meanings don't change, just the "greediness": | |
0d017f4d | 180 | X<metacharacter> X<greedy> X<greediness> |
d74e8afc | 181 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?> |
a0d0e21e | 182 | |
0d017f4d WL |
183 | *? Match 0 or more times, not greedily |
184 | +? Match 1 or more times, not greedily | |
185 | ?? Match 0 or 1 time, not greedily | |
186 | {n}? Match exactly n times, not greedily | |
187 | {n,}? Match at least n times, not greedily | |
188 | {n,m}? Match at least n but not more than m times, not greedily | |
a0d0e21e | 189 | |
b9b4dddf YO |
190 | By default, when a quantified subpattern does not allow the rest of the |
191 | overall pattern to match, Perl will backtrack. However, this behaviour is | |
0d017f4d | 192 | sometimes undesirable. Thus Perl provides the "possessive" quantifier form |
b9b4dddf YO |
193 | as well. |
194 | ||
0d017f4d WL |
195 | *+ Match 0 or more times and give nothing back |
196 | ++ Match 1 or more times and give nothing back | |
197 | ?+ Match 0 or 1 time and give nothing back | |
b9b4dddf | 198 | {n}+ Match exactly n times and give nothing back (redundant) |
04838cea RGS |
199 | {n,}+ Match at least n times and give nothing back |
200 | {n,m}+ Match at least n but not more than m times and give nothing back | |
b9b4dddf YO |
201 | |
202 | For instance, | |
203 | ||
204 | 'aaaa' =~ /a++a/ | |
205 | ||
206 | will never match, as the C<a++> will gobble up all the C<a>'s in the | |
207 | string and won't leave any for the remaining part of the pattern. This | |
208 | feature can be extremely useful to give perl hints about where it | |
209 | shouldn't backtrack. For instance, the typical "match a double-quoted | |
210 | string" problem can be most efficiently performed when written as: | |
211 | ||
212 | /"(?:[^"\\]++|\\.)*+"/ | |
213 | ||
0d017f4d | 214 | as we know that if the final quote does not match, backtracking will not |
b9b4dddf YO |
215 | help. See the independent subexpression C<< (?>...) >> for more details; |
216 | possessive quantifiers are just syntactic sugar for that construct. For | |
217 | instance the above example could also be written as follows: | |
218 | ||
219 | /"(?>(?:(?>[^"\\]+)|\\.)*)"/ | |
220 | ||
04838cea RGS |
221 | =head3 Escape sequences |
222 | ||
5f05dabc | 223 | Because patterns are processed as double quoted strings, the following |
a0d0e21e LW |
224 | also work: |
225 | ||
0f36ee90 | 226 | \t tab (HT, TAB) |
227 | \n newline (LF, NL) | |
228 | \r return (CR) | |
229 | \f form feed (FF) | |
230 | \a alarm (bell) (BEL) | |
231 | \e escape (think troff) (ESC) | |
0d017f4d WL |
232 | \033 octal char (example: ESC) |
233 | \x1B hex char (example: ESC) | |
196ac2fc | 234 | \x{263a} long hex char (example: Unicode SMILEY) |
0d017f4d | 235 | \cK control char (example: VT) |
196ac2fc | 236 | \N{name} named Unicode character |
e526e8bb | 237 | \N{U+263D} Unicode character (example: FIRST QUARTER MOON) |
cb1a09d0 AD |
238 | \l lowercase next char (think vi) |
239 | \u uppercase next char (think vi) | |
240 | \L lowercase till \E (think vi) | |
241 | \U uppercase till \E (think vi) | |
5a964f20 | 242 | \Q quote (disable) pattern metacharacters till \E |
9bb1f947 | 243 | \E end either case modification or quoted section (think vi) |
a0d0e21e | 244 | |
9bb1f947 | 245 | Details are in L<perlop/Quote and Quote-like Operators>. |
1d2dff63 | 246 | |
e1d1eefb | 247 | =head3 Character Classes and other Special Escapes |
04838cea | 248 | |
a0d0e21e | 249 | In addition, Perl defines the following: |
d0b16107 | 250 | X<\g> X<\k> X<\K> X<backreference> |
a0d0e21e | 251 | |
9bb1f947 KW |
252 | Sequence Note Description |
253 | [...] [1] Match a character according to the rules of the bracketed | |
254 | character class defined by the "...". Example: [a-z] | |
255 | matches "a" or "b" or "c" ... or "z" | |
256 | [[:...:]] [2] Match a character according to the rules of the POSIX | |
257 | character class "..." within the outer bracketed character | |
258 | class. Example: [[:upper:]] matches any uppercase | |
259 | character. | |
260 | \w [3] Match a "word" character (alphanumeric plus "_") | |
261 | \W [3] Match a non-"word" character | |
262 | \s [3] Match a whitespace character | |
263 | \S [3] Match a non-whitespace character | |
264 | \d [3] Match a decimal digit character | |
265 | \D [3] Match a non-digit character | |
266 | \pP [3] Match P, named property. Use \p{Prop} for longer names. | |
267 | \PP [3] Match non-P | |
268 | \X [4] Match Unicode "eXtended grapheme cluster" | |
269 | \C Match a single C-language char (octet) even if that is part | |
270 | of a larger UTF-8 character. Thus it breaks up characters | |
271 | into their UTF-8 bytes, so you may end up with malformed | |
272 | pieces of UTF-8. Unsupported in lookbehind. | |
273 | \1 [5] Backreference to a specific capture buffer or group. | |
274 | '1' may actually be any positive integer. | |
275 | \g1 [5] Backreference to a specific or previous group, | |
276 | \g{-1} [5] The number may be negative indicating a relative previous | |
277 | buffer and may optionally be wrapped in curly brackets for | |
278 | safer parsing. | |
279 | \g{name} [5] Named backreference | |
280 | \k<name> [5] Named backreference | |
281 | \K [6] Keep the stuff left of the \K, don't include it in $& | |
282 | \N [7] Any character but \n (experimental). Not affected by /s | |
283 | modifier | |
284 | \v [3] Vertical whitespace | |
285 | \V [3] Not vertical whitespace | |
286 | \h [3] Horizontal whitespace | |
287 | \H [3] Not horizontal whitespace | |
288 | \R [4] Linebreak | |
e1d1eefb | 289 | |
9bb1f947 KW |
290 | =over 4 |
291 | ||
292 | =item [1] | |
293 | ||
294 | See L<perlrecharclass/Bracketed Character Classes> for details. | |
df225385 | 295 | |
9bb1f947 | 296 | =item [2] |
b8c5462f | 297 | |
9bb1f947 | 298 | See L<perlrecharclass/POSIX Character Classes> for details. |
b8c5462f | 299 | |
9bb1f947 | 300 | =item [3] |
5496314a | 301 | |
9bb1f947 | 302 | See L<perlrecharclass/Backslash sequences> for details. |
5496314a | 303 | |
9bb1f947 | 304 | =item [4] |
5496314a | 305 | |
9bb1f947 | 306 | See L<perlrebackslash/Misc> for details. |
d0b16107 | 307 | |
9bb1f947 | 308 | =item [5] |
b8c5462f | 309 | |
9bb1f947 | 310 | See L</Capture buffers> below for details. |
93733859 | 311 | |
9bb1f947 | 312 | =item [6] |
b8c5462f | 313 | |
9bb1f947 KW |
314 | See L</Extended Patterns> below for details. |
315 | ||
316 | =item [7] | |
317 | ||
318 | Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the | |
319 | character whose name is C<NAME>; and similarly when of the form | |
320 | C<\N{U+I<wide hex char>}>, it matches the character whose Unicode ordinal is | |
321 | I<wide hex char>. Otherwise it matches any character but C<\n>. | |
322 | ||
323 | =back | |
d0b16107 | 324 | |
04838cea RGS |
325 | =head3 Assertions |
326 | ||
a0d0e21e | 327 | Perl defines the following zero-width assertions: |
d74e8afc ITB |
328 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> |
329 | X<regexp, zero-width assertion> | |
330 | X<regular expression, zero-width assertion> | |
331 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> | |
a0d0e21e | 332 | |
9bb1f947 KW |
333 | \b Match a word boundary |
334 | \B Match except at a word boundary | |
335 | \A Match only at beginning of string | |
336 | \Z Match only at end of string, or before newline at the end | |
337 | \z Match only at end of string | |
338 | \G Match only at pos() (e.g. at the end-of-match position | |
9da458fc | 339 | of prior m//g) |
a0d0e21e | 340 | |
14218588 | 341 | A word boundary (C<\b>) is a spot between two characters |
19799a22 GS |
342 | that has a C<\w> on one side of it and a C<\W> on the other side |
343 | of it (in either order), counting the imaginary characters off the | |
344 | beginning and end of the string as matching a C<\W>. (Within | |
345 | character classes C<\b> represents backspace rather than a word | |
346 | boundary, just as it normally does in any double-quoted string.) | |
347 | The C<\A> and C<\Z> are just like "^" and "$", except that they | |
348 | won't match multiple times when the C</m> modifier is used, while | |
349 | "^" and "$" will match at every internal line boundary. To match | |
350 | the actual end of the string and not ignore an optional trailing | |
351 | newline, use C<\z>. | |
d74e8afc | 352 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 GS |
353 | |
354 | The C<\G> assertion can be used to chain global matches (using | |
355 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. | |
356 | It is also useful when writing C<lex>-like scanners, when you have | |
357 | several patterns that you want to match against consequent substrings | |
358 | of your string, see the previous reference. The actual location | |
359 | where C<\G> will match can also be influenced by using C<pos()> as | |
58e23c8d YO |
360 | an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length |
361 | matches is modified somewhat, in that contents to the left of C<\G> is | |
362 | not counted when determining the length of the match. Thus the following | |
363 | will not match forever: | |
d74e8afc | 364 | X<\G> |
c47ff5f1 | 365 | |
58e23c8d YO |
366 | $str = 'ABC'; |
367 | pos($str) = 1; | |
368 | while (/.\G/g) { | |
369 | print $&; | |
370 | } | |
371 | ||
372 | It will print 'A' and then terminate, as it considers the match to | |
373 | be zero-width, and thus will not match at the same position twice in a | |
374 | row. | |
375 | ||
376 | It is worth noting that C<\G> improperly used can result in an infinite | |
377 | loop. Take care when using patterns that include C<\G> in an alternation. | |
378 | ||
04838cea RGS |
379 | =head3 Capture buffers |
380 | ||
0d017f4d WL |
381 | The bracketing construct C<( ... )> creates capture buffers. To refer |
382 | to the current contents of a buffer later on, within the same pattern, | |
383 | use \1 for the first, \2 for the second, and so on. | |
384 | Outside the match use "$" instead of "\". (The | |
81714fb9 | 385 | \<digit> notation works in certain circumstances outside |
14218588 GS |
386 | the match. See the warning below about \1 vs $1 for details.) |
387 | Referring back to another part of the match is called a | |
388 | I<backreference>. | |
d74e8afc ITB |
389 | X<regex, capture buffer> X<regexp, capture buffer> |
390 | X<regular expression, capture buffer> X<backreference> | |
14218588 GS |
391 | |
392 | There is no limit to the number of captured substrings that you may | |
393 | use. However Perl also uses \10, \11, etc. as aliases for \010, | |
fb55449c JH |
394 | \011, etc. (Recall that 0 means octal, so \011 is the character at |
395 | number 9 in your coded character set; which would be the 10th character, | |
81714fb9 YO |
396 | a horizontal tab under ASCII.) Perl resolves this |
397 | ambiguity by interpreting \10 as a backreference only if at least 10 | |
398 | left parentheses have opened before it. Likewise \11 is a | |
399 | backreference only if at least 11 left parentheses have opened | |
400 | before it. And so on. \1 through \9 are always interpreted as | |
5624f11d | 401 | backreferences. |
40863337 ML |
402 | If the bracketing group did not match, the associated backreference won't |
403 | match either. (This can happen if the bracketing group is optional, or | |
404 | in a different branch of an alternation.) | |
405 | ||
1f1031fe | 406 | X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference> |
2bf803e2 | 407 | In order to provide a safer and easier way to construct patterns using |
99d59c4d RGS |
408 | backreferences, Perl provides the C<\g{N}> notation (starting with perl |
409 | 5.10.0). The curly brackets are optional, however omitting them is less | |
410 | safe as the meaning of the pattern can be changed by text (such as digits) | |
411 | following it. When N is a positive integer the C<\g{N}> notation is | |
412 | exactly equivalent to using normal backreferences. When N is a negative | |
413 | integer then it is a relative backreference referring to the previous N'th | |
414 | capturing group. When the bracket form is used and N is not an integer, it | |
415 | is treated as a reference to a named buffer. | |
2bf803e2 YO |
416 | |
417 | Thus C<\g{-1}> refers to the last buffer, C<\g{-2}> refers to the | |
418 | buffer before that. For example: | |
5624f11d YO |
419 | |
420 | / | |
421 | (Y) # buffer 1 | |
422 | ( # buffer 2 | |
423 | (X) # buffer 3 | |
2bf803e2 YO |
424 | \g{-1} # backref to buffer 3 |
425 | \g{-3} # backref to buffer 1 | |
5624f11d YO |
426 | ) |
427 | /x | |
428 | ||
2bf803e2 | 429 | and would match the same as C</(Y) ( (X) \3 \1 )/x>. |
14218588 | 430 | |
99d59c4d | 431 | Additionally, as of Perl 5.10.0 you may use named capture buffers and named |
1f1031fe | 432 | backreferences. The notation is C<< (?<name>...) >> to declare and C<< \k<name> >> |
0d017f4d WL |
433 | to reference. You may also use apostrophes instead of angle brackets to delimit the |
434 | name; and you may use the bracketed C<< \g{name} >> backreference syntax. | |
435 | It's possible to refer to a named capture buffer by absolute and relative number as well. | |
436 | Outside the pattern, a named capture buffer is available via the C<%+> hash. | |
437 | When different buffers within the same pattern have the same name, C<$+{name}> | |
438 | and C<< \k<name> >> refer to the leftmost defined group. (Thus it's possible | |
439 | to do things with named capture buffers that would otherwise require C<(??{})> | |
440 | code to accomplish.) | |
441 | X<named capture buffer> X<regular expression, named capture buffer> | |
64c5a566 | 442 | X<%+> X<$+{name}> X<< \k<name> >> |
81714fb9 | 443 | |
14218588 | 444 | Examples: |
a0d0e21e LW |
445 | |
446 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
447 | ||
81714fb9 YO |
448 | /(.)\1/ # find first doubled char |
449 | and print "'$1' is the first doubled character\n"; | |
450 | ||
451 | /(?<char>.)\k<char>/ # ... a different way | |
452 | and print "'$+{char}' is the first doubled character\n"; | |
453 | ||
0d017f4d | 454 | /(?'char'.)\1/ # ... mix and match |
81714fb9 | 455 | and print "'$1' is the first doubled character\n"; |
c47ff5f1 | 456 | |
14218588 | 457 | if (/Time: (..):(..):(..)/) { # parse out values |
a0d0e21e LW |
458 | $hours = $1; |
459 | $minutes = $2; | |
460 | $seconds = $3; | |
461 | } | |
c47ff5f1 | 462 | |
14218588 GS |
463 | Several special variables also refer back to portions of the previous |
464 | match. C<$+> returns whatever the last bracket match matched. | |
465 | C<$&> returns the entire matched string. (At one point C<$0> did | |
466 | also, but now it returns the name of the program.) C<$`> returns | |
77ea4f6d JV |
467 | everything before the matched string. C<$'> returns everything |
468 | after the matched string. And C<$^N> contains whatever was matched by | |
469 | the most-recently closed group (submatch). C<$^N> can be used in | |
470 | extended patterns (see below), for example to assign a submatch to a | |
81714fb9 | 471 | variable. |
d74e8afc | 472 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 | 473 | |
665e98b9 | 474 | The numbered match variables ($1, $2, $3, etc.) and the related punctuation |
77ea4f6d | 475 | set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped |
14218588 GS |
476 | until the end of the enclosing block or until the next successful |
477 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
d74e8afc ITB |
478 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
479 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> | |
480 | ||
14218588 | 481 | |
0d017f4d | 482 | B<NOTE>: Failed matches in Perl do not reset the match variables, |
5146ce24 | 483 | which makes it easier to write code that tests for a series of more |
665e98b9 JH |
484 | specific cases and remembers the best match. |
485 | ||
14218588 GS |
486 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or |
487 | C<$'> anywhere in the program, it has to provide them for every | |
488 | pattern match. This may substantially slow your program. Perl | |
489 | uses the same mechanism to produce $1, $2, etc, so you also pay a | |
490 | price for each pattern that contains capturing parentheses. (To | |
491 | avoid this cost while retaining the grouping behaviour, use the | |
492 | extended regular expression C<(?: ... )> instead.) But if you never | |
493 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
494 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
495 | if you can, but if you can't (and some algorithms really appreciate | |
496 | them), once you've used them once, use them at will, because you've | |
497 | already paid the price. As of 5.005, C<$&> is not so costly as the | |
498 | other two. | |
d74e8afc | 499 | X<$&> X<$`> X<$'> |
68dc0745 | 500 | |
99d59c4d | 501 | As a workaround for this problem, Perl 5.10.0 introduces C<${^PREMATCH}>, |
cde0cee5 YO |
502 | C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&> |
503 | and C<$'>, B<except> that they are only guaranteed to be defined after a | |
87e95b7f | 504 | successful match that was executed with the C</p> (preserve) modifier. |
cde0cee5 YO |
505 | The use of these variables incurs no global performance penalty, unlike |
506 | their punctuation char equivalents, however at the trade-off that you | |
507 | have to tell perl when you want to use them. | |
87e95b7f | 508 | X</p> X<p modifier> |
cde0cee5 | 509 | |
19799a22 GS |
510 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
511 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
512 | are no backslashed symbols that aren't alphanumeric. So anything | |
c47ff5f1 | 513 | that looks like \\, \(, \), \<, \>, \{, or \} is always |
19799a22 GS |
514 | interpreted as a literal character, not a metacharacter. This was |
515 | once used in a common idiom to disable or quote the special meanings | |
516 | of regular expression metacharacters in a string that you want to | |
36bbe248 | 517 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e LW |
518 | |
519 | $pattern =~ s/(\W)/\\$1/g; | |
520 | ||
f1cbbd6e | 521 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 GS |
522 | Today it is more common to use the quotemeta() function or the C<\Q> |
523 | metaquoting escape sequence to disable all metacharacters' special | |
524 | meanings like this: | |
a0d0e21e LW |
525 | |
526 | /$unquoted\Q$quoted\E$unquoted/ | |
527 | ||
9da458fc IZ |
528 | Beware that if you put literal backslashes (those not inside |
529 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
530 | backslash interpolation may lead to confusing results. If you | |
531 | I<need> to use literal backslashes within C<\Q...\E>, | |
532 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
533 | ||
19799a22 GS |
534 | =head2 Extended Patterns |
535 | ||
14218588 GS |
536 | Perl also defines a consistent extension syntax for features not |
537 | found in standard tools like B<awk> and B<lex>. The syntax is a | |
538 | pair of parentheses with a question mark as the first thing within | |
539 | the parentheses. The character after the question mark indicates | |
540 | the extension. | |
19799a22 | 541 | |
14218588 GS |
542 | The stability of these extensions varies widely. Some have been |
543 | part of the core language for many years. Others are experimental | |
544 | and may change without warning or be completely removed. Check | |
545 | the documentation on an individual feature to verify its current | |
546 | status. | |
19799a22 | 547 | |
14218588 GS |
548 | A question mark was chosen for this and for the minimal-matching |
549 | construct because 1) question marks are rare in older regular | |
550 | expressions, and 2) whenever you see one, you should stop and | |
551 | "question" exactly what is going on. That's psychology... | |
a0d0e21e LW |
552 | |
553 | =over 10 | |
554 | ||
cc6b7395 | 555 | =item C<(?#text)> |
d74e8afc | 556 | X<(?#)> |
a0d0e21e | 557 | |
14218588 | 558 | A comment. The text is ignored. If the C</x> modifier enables |
19799a22 | 559 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes |
259138e3 GS |
560 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
561 | C<)> in the comment. | |
a0d0e21e | 562 | |
f7819f85 | 563 | =item C<(?pimsx-imsx)> |
d74e8afc | 564 | X<(?)> |
19799a22 | 565 | |
0b6d1084 JH |
566 | One or more embedded pattern-match modifiers, to be turned on (or |
567 | turned off, if preceded by C<->) for the remainder of the pattern or | |
568 | the remainder of the enclosing pattern group (if any). This is | |
569 | particularly useful for dynamic patterns, such as those read in from a | |
0d017f4d WL |
570 | configuration file, taken from an argument, or specified in a table |
571 | somewhere. Consider the case where some patterns want to be case | |
572 | sensitive and some do not: The case insensitive ones merely need to | |
573 | include C<(?i)> at the front of the pattern. For example: | |
19799a22 GS |
574 | |
575 | $pattern = "foobar"; | |
5d458dd8 | 576 | if ( /$pattern/i ) { } |
19799a22 GS |
577 | |
578 | # more flexible: | |
579 | ||
580 | $pattern = "(?i)foobar"; | |
5d458dd8 | 581 | if ( /$pattern/ ) { } |
19799a22 | 582 | |
0b6d1084 | 583 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 GS |
584 | |
585 | ( (?i) blah ) \s+ \1 | |
586 | ||
0d017f4d WL |
587 | will match C<blah> in any case, some spaces, and an exact (I<including the case>!) |
588 | repetition of the previous word, assuming the C</x> modifier, and no C</i> | |
589 | modifier outside this group. | |
19799a22 | 590 | |
8eb5594e DR |
591 | These modifiers do not carry over into named subpatterns called in the |
592 | enclosing group. In other words, a pattern such as C<((?i)(&NAME))> does not | |
593 | change the case-sensitivity of the "NAME" pattern. | |
594 | ||
5530442b | 595 | Note that the C<p> modifier is special in that it can only be enabled, |
cde0cee5 | 596 | not disabled, and that its presence anywhere in a pattern has a global |
5530442b | 597 | effect. Thus C<(?-p)> and C<(?-p:...)> are meaningless and will warn |
cde0cee5 YO |
598 | when executed under C<use warnings>. |
599 | ||
5a964f20 | 600 | =item C<(?:pattern)> |
d74e8afc | 601 | X<(?:)> |
a0d0e21e | 602 | |
ca9dfc88 IZ |
603 | =item C<(?imsx-imsx:pattern)> |
604 | ||
5a964f20 TC |
605 | This is for clustering, not capturing; it groups subexpressions like |
606 | "()", but doesn't make backreferences as "()" does. So | |
a0d0e21e | 607 | |
5a964f20 | 608 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e LW |
609 | |
610 | is like | |
611 | ||
5a964f20 | 612 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e | 613 | |
19799a22 GS |
614 | but doesn't spit out extra fields. It's also cheaper not to capture |
615 | characters if you don't need to. | |
a0d0e21e | 616 | |
19799a22 | 617 | Any letters between C<?> and C<:> act as flags modifiers as with |
5d458dd8 | 618 | C<(?imsx-imsx)>. For example, |
ca9dfc88 IZ |
619 | |
620 | /(?s-i:more.*than).*million/i | |
621 | ||
14218588 | 622 | is equivalent to the more verbose |
ca9dfc88 IZ |
623 | |
624 | /(?:(?s-i)more.*than).*million/i | |
625 | ||
594d7033 YO |
626 | =item C<(?|pattern)> |
627 | X<(?|)> X<Branch reset> | |
628 | ||
629 | This is the "branch reset" pattern, which has the special property | |
630 | that the capture buffers are numbered from the same starting point | |
99d59c4d | 631 | in each alternation branch. It is available starting from perl 5.10.0. |
4deaaa80 | 632 | |
693596a8 RGS |
633 | Capture buffers are numbered from left to right, but inside this |
634 | construct the numbering is restarted for each branch. | |
4deaaa80 PJ |
635 | |
636 | The numbering within each branch will be as normal, and any buffers | |
637 | following this construct will be numbered as though the construct | |
638 | contained only one branch, that being the one with the most capture | |
639 | buffers in it. | |
640 | ||
641 | This construct will be useful when you want to capture one of a | |
642 | number of alternative matches. | |
643 | ||
644 | Consider the following pattern. The numbers underneath show in | |
645 | which buffer the captured content will be stored. | |
594d7033 YO |
646 | |
647 | ||
648 | # before ---------------branch-reset----------- after | |
649 | / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x | |
650 | # 1 2 2 3 2 3 4 | |
651 | ||
ab106183 A |
652 | Be careful when using the branch reset pattern in combination with |
653 | named captures. Named captures are implemented as being aliases to | |
654 | numbered buffers holding the captures, and that interferes with the | |
655 | implementation of the branch reset pattern. If you are using named | |
656 | captures in a branch reset pattern, it's best to use the same names, | |
657 | in the same order, in each of the alternations: | |
658 | ||
659 | /(?| (?<a> x ) (?<b> y ) | |
660 | | (?<a> z ) (?<b> w )) /x | |
661 | ||
662 | Not doing so may lead to surprises: | |
663 | ||
664 | "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x; | |
665 | say $+ {a}; # Prints '12' | |
666 | say $+ {b}; # *Also* prints '12'. | |
667 | ||
668 | The problem here is that both the buffer named C<< a >> and the buffer | |
669 | named C<< b >> are aliases for the buffer belonging to C<< $1 >>. | |
90a18110 | 670 | |
ee9b8eae YO |
671 | =item Look-Around Assertions |
672 | X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround> | |
673 | ||
674 | Look-around assertions are zero width patterns which match a specific | |
675 | pattern without including it in C<$&>. Positive assertions match when | |
676 | their subpattern matches, negative assertions match when their subpattern | |
677 | fails. Look-behind matches text up to the current match position, | |
678 | look-ahead matches text following the current match position. | |
679 | ||
680 | =over 4 | |
681 | ||
5a964f20 | 682 | =item C<(?=pattern)> |
d74e8afc | 683 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> |
a0d0e21e | 684 | |
19799a22 | 685 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e LW |
686 | matches a word followed by a tab, without including the tab in C<$&>. |
687 | ||
5a964f20 | 688 | =item C<(?!pattern)> |
d74e8afc | 689 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> |
a0d0e21e | 690 | |
19799a22 | 691 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e | 692 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 GS |
693 | however that look-ahead and look-behind are NOT the same thing. You cannot |
694 | use this for look-behind. | |
7b8d334a | 695 | |
5a964f20 | 696 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a GS |
697 | will not do what you want. That's because the C<(?!foo)> is just saying that |
698 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
699 | match. You would have to do something like C</(?!foo)...bar/> for that. We | |
700 | say "like" because there's the case of your "bar" not having three characters | |
701 | before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>. | |
702 | Sometimes it's still easier just to say: | |
a0d0e21e | 703 | |
a3cb178b | 704 | if (/bar/ && $` !~ /foo$/) |
a0d0e21e | 705 | |
19799a22 | 706 | For look-behind see below. |
c277df42 | 707 | |
ee9b8eae YO |
708 | =item C<(?<=pattern)> C<\K> |
709 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K> | |
c277df42 | 710 | |
c47ff5f1 | 711 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 GS |
712 | matches a word that follows a tab, without including the tab in C<$&>. |
713 | Works only for fixed-width look-behind. | |
c277df42 | 714 | |
ee9b8eae YO |
715 | There is a special form of this construct, called C<\K>, which causes the |
716 | regex engine to "keep" everything it had matched prior to the C<\K> and | |
717 | not include it in C<$&>. This effectively provides variable length | |
718 | look-behind. The use of C<\K> inside of another look-around assertion | |
719 | is allowed, but the behaviour is currently not well defined. | |
720 | ||
c62285ac | 721 | For various reasons C<\K> may be significantly more efficient than the |
ee9b8eae YO |
722 | equivalent C<< (?<=...) >> construct, and it is especially useful in |
723 | situations where you want to efficiently remove something following | |
724 | something else in a string. For instance | |
725 | ||
726 | s/(foo)bar/$1/g; | |
727 | ||
728 | can be rewritten as the much more efficient | |
729 | ||
730 | s/foo\Kbar//g; | |
731 | ||
5a964f20 | 732 | =item C<(?<!pattern)> |
d74e8afc | 733 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> |
c277df42 | 734 | |
19799a22 GS |
735 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
736 | matches any occurrence of "foo" that does not follow "bar". Works | |
737 | only for fixed-width look-behind. | |
c277df42 | 738 | |
ee9b8eae YO |
739 | =back |
740 | ||
81714fb9 YO |
741 | =item C<(?'NAME'pattern)> |
742 | ||
743 | =item C<< (?<NAME>pattern) >> | |
744 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> | |
745 | ||
746 | A named capture buffer. Identical in every respect to normal capturing | |
90a18110 RGS |
747 | parentheses C<()> but for the additional fact that C<%+> or C<%-> may be |
748 | used after a successful match to refer to a named buffer. See C<perlvar> | |
749 | for more details on the C<%+> and C<%-> hashes. | |
81714fb9 YO |
750 | |
751 | If multiple distinct capture buffers have the same name then the | |
752 | $+{NAME} will refer to the leftmost defined buffer in the match. | |
753 | ||
0d017f4d | 754 | The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent. |
81714fb9 YO |
755 | |
756 | B<NOTE:> While the notation of this construct is the same as the similar | |
0d017f4d | 757 | function in .NET regexes, the behavior is not. In Perl the buffers are |
81714fb9 YO |
758 | numbered sequentially regardless of being named or not. Thus in the |
759 | pattern | |
760 | ||
761 | /(x)(?<foo>y)(z)/ | |
762 | ||
763 | $+{foo} will be the same as $2, and $3 will contain 'z' instead of | |
764 | the opposite which is what a .NET regex hacker might expect. | |
765 | ||
1f1031fe YO |
766 | Currently NAME is restricted to simple identifiers only. |
767 | In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or | |
768 | its Unicode extension (see L<utf8>), | |
769 | though it isn't extended by the locale (see L<perllocale>). | |
81714fb9 | 770 | |
1f1031fe | 771 | B<NOTE:> In order to make things easier for programmers with experience |
ae5648b3 | 772 | with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >> |
0d017f4d | 773 | may be used instead of C<< (?<NAME>pattern) >>; however this form does not |
64c5a566 | 774 | support the use of single quotes as a delimiter for the name. |
81714fb9 | 775 | |
1f1031fe YO |
776 | =item C<< \k<NAME> >> |
777 | ||
778 | =item C<< \k'NAME' >> | |
81714fb9 YO |
779 | |
780 | Named backreference. Similar to numeric backreferences, except that | |
781 | the group is designated by name and not number. If multiple groups | |
782 | have the same name then it refers to the leftmost defined group in | |
783 | the current match. | |
784 | ||
0d017f4d | 785 | It is an error to refer to a name not defined by a C<< (?<NAME>) >> |
81714fb9 YO |
786 | earlier in the pattern. |
787 | ||
788 | Both forms are equivalent. | |
789 | ||
1f1031fe | 790 | B<NOTE:> In order to make things easier for programmers with experience |
0d017f4d | 791 | with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >> |
64c5a566 | 792 | may be used instead of C<< \k<NAME> >>. |
1f1031fe | 793 | |
cc6b7395 | 794 | =item C<(?{ code })> |
d74e8afc | 795 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 | 796 | |
19799a22 | 797 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
798 | experimental, and may be changed without notice. Code executed that |
799 | has side effects may not perform identically from version to version | |
800 | due to the effect of future optimisations in the regex engine. | |
c277df42 | 801 | |
cc46d5f2 | 802 | This zero-width assertion evaluates any embedded Perl code. It |
19799a22 GS |
803 | always succeeds, and its C<code> is not interpolated. Currently, |
804 | the rules to determine where the C<code> ends are somewhat convoluted. | |
805 | ||
77ea4f6d JV |
806 | This feature can be used together with the special variable C<$^N> to |
807 | capture the results of submatches in variables without having to keep | |
808 | track of the number of nested parentheses. For example: | |
809 | ||
810 | $_ = "The brown fox jumps over the lazy dog"; | |
811 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
812 | print "color = $color, animal = $animal\n"; | |
813 | ||
754091cb RGS |
814 | Inside the C<(?{...})> block, C<$_> refers to the string the regular |
815 | expression is matching against. You can also use C<pos()> to know what is | |
fa11829f | 816 | the current position of matching within this string. |
754091cb | 817 | |
19799a22 GS |
818 | The C<code> is properly scoped in the following sense: If the assertion |
819 | is backtracked (compare L<"Backtracking">), all changes introduced after | |
820 | C<local>ization are undone, so that | |
b9ac3b5b GS |
821 | |
822 | $_ = 'a' x 8; | |
5d458dd8 | 823 | m< |
b9ac3b5b GS |
824 | (?{ $cnt = 0 }) # Initialize $cnt. |
825 | ( | |
5d458dd8 | 826 | a |
b9ac3b5b GS |
827 | (?{ |
828 | local $cnt = $cnt + 1; # Update $cnt, backtracking-safe. | |
829 | }) | |
5d458dd8 | 830 | )* |
b9ac3b5b GS |
831 | aaaa |
832 | (?{ $res = $cnt }) # On success copy to non-localized | |
833 | # location. | |
834 | >x; | |
835 | ||
0d017f4d | 836 | will set C<$res = 4>. Note that after the match, C<$cnt> returns to the globally |
14218588 | 837 | introduced value, because the scopes that restrict C<local> operators |
b9ac3b5b GS |
838 | are unwound. |
839 | ||
19799a22 GS |
840 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> |
841 | switch. If I<not> used in this way, the result of evaluation of | |
842 | C<code> is put into the special variable C<$^R>. This happens | |
843 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions | |
844 | inside the same regular expression. | |
b9ac3b5b | 845 | |
19799a22 GS |
846 | The assignment to C<$^R> above is properly localized, so the old |
847 | value of C<$^R> is restored if the assertion is backtracked; compare | |
848 | L<"Backtracking">. | |
b9ac3b5b | 849 | |
19799a22 GS |
850 | For reasons of security, this construct is forbidden if the regular |
851 | expression involves run-time interpolation of variables, unless the | |
852 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
853 | variables contain results of C<qr//> operator (see | |
9bb1f947 | 854 | L<perlop/"qr/STRINGE<sol>msixpo">). |
871b0233 | 855 | |
0d017f4d | 856 | This restriction is due to the wide-spread and remarkably convenient |
19799a22 | 857 | custom of using run-time determined strings as patterns. For example: |
871b0233 IZ |
858 | |
859 | $re = <>; | |
860 | chomp $re; | |
861 | $string =~ /$re/; | |
862 | ||
14218588 GS |
863 | Before Perl knew how to execute interpolated code within a pattern, |
864 | this operation was completely safe from a security point of view, | |
865 | although it could raise an exception from an illegal pattern. If | |
866 | you turn on the C<use re 'eval'>, though, it is no longer secure, | |
867 | so you should only do so if you are also using taint checking. | |
868 | Better yet, use the carefully constrained evaluation within a Safe | |
cc46d5f2 | 869 | compartment. See L<perlsec> for details about both these mechanisms. |
871b0233 | 870 | |
e95d7314 GG |
871 | B<WARNING>: Use of lexical (C<my>) variables in these blocks is |
872 | broken. The result is unpredictable and will make perl unstable. The | |
873 | workaround is to use global (C<our>) variables. | |
874 | ||
875 | B<WARNING>: Because Perl's regex engine is currently not re-entrant, | |
876 | interpolated code may not invoke the regex engine either directly with | |
877 | C<m//> or C<s///>), or indirectly with functions such as | |
878 | C<split>. Invoking the regex engine in these blocks will make perl | |
879 | unstable. | |
8988a1bb | 880 | |
14455d6c | 881 | =item C<(??{ code })> |
d74e8afc ITB |
882 | X<(??{})> |
883 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> | |
0f5d15d6 | 884 | |
19799a22 | 885 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
886 | experimental, and may be changed without notice. Code executed that |
887 | has side effects may not perform identically from version to version | |
888 | due to the effect of future optimisations in the regex engine. | |
0f5d15d6 | 889 | |
19799a22 GS |
890 | This is a "postponed" regular subexpression. The C<code> is evaluated |
891 | at run time, at the moment this subexpression may match. The result | |
892 | of evaluation is considered as a regular expression and matched as | |
61528107 | 893 | if it were inserted instead of this construct. Note that this means |
6bda09f9 YO |
894 | that the contents of capture buffers defined inside an eval'ed pattern |
895 | are not available outside of the pattern, and vice versa, there is no | |
896 | way for the inner pattern to refer to a capture buffer defined outside. | |
897 | Thus, | |
898 | ||
899 | ('a' x 100)=~/(??{'(.)' x 100})/ | |
900 | ||
81714fb9 | 901 | B<will> match, it will B<not> set $1. |
0f5d15d6 | 902 | |
428594d9 | 903 | The C<code> is not interpolated. As before, the rules to determine |
19799a22 GS |
904 | where the C<code> ends are currently somewhat convoluted. |
905 | ||
906 | The following pattern matches a parenthesized group: | |
0f5d15d6 IZ |
907 | |
908 | $re = qr{ | |
909 | \( | |
910 | (?: | |
911 | (?> [^()]+ ) # Non-parens without backtracking | |
912 | | | |
14455d6c | 913 | (??{ $re }) # Group with matching parens |
0f5d15d6 IZ |
914 | )* |
915 | \) | |
916 | }x; | |
917 | ||
6bda09f9 YO |
918 | See also C<(?PARNO)> for a different, more efficient way to accomplish |
919 | the same task. | |
920 | ||
0b370c0a A |
921 | For reasons of security, this construct is forbidden if the regular |
922 | expression involves run-time interpolation of variables, unless the | |
923 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
924 | variables contain results of C<qr//> operator (see | |
9bb1f947 | 925 | L<perlop/"qrE<sol>STRINGE<sol>msixpo">). |
0b370c0a | 926 | |
5d458dd8 | 927 | Because perl's regex engine is not currently re-entrant, delayed |
8988a1bb DD |
928 | code may not invoke the regex engine either directly with C<m//> or C<s///>), |
929 | or indirectly with functions such as C<split>. | |
930 | ||
5d458dd8 YO |
931 | Recursing deeper than 50 times without consuming any input string will |
932 | result in a fatal error. The maximum depth is compiled into perl, so | |
6bda09f9 YO |
933 | changing it requires a custom build. |
934 | ||
542fa716 YO |
935 | =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)> |
936 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)> | |
6bda09f9 | 937 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
542fa716 | 938 | X<regex, relative recursion> |
6bda09f9 | 939 | |
81714fb9 YO |
940 | Similar to C<(??{ code })> except it does not involve compiling any code, |
941 | instead it treats the contents of a capture buffer as an independent | |
61528107 | 942 | pattern that must match at the current position. Capture buffers |
81714fb9 | 943 | contained by the pattern will have the value as determined by the |
6bda09f9 YO |
944 | outermost recursion. |
945 | ||
894be9b7 YO |
946 | PARNO is a sequence of digits (not starting with 0) whose value reflects |
947 | the paren-number of the capture buffer to recurse to. C<(?R)> recurses to | |
948 | the beginning of the whole pattern. C<(?0)> is an alternate syntax for | |
542fa716 YO |
949 | C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed |
950 | to be relative, with negative numbers indicating preceding capture buffers | |
951 | and positive ones following. Thus C<(?-1)> refers to the most recently | |
952 | declared buffer, and C<(?+1)> indicates the next buffer to be declared. | |
c74340f9 YO |
953 | Note that the counting for relative recursion differs from that of |
954 | relative backreferences, in that with recursion unclosed buffers B<are> | |
955 | included. | |
6bda09f9 | 956 | |
81714fb9 | 957 | The following pattern matches a function foo() which may contain |
f145b7e9 | 958 | balanced parentheses as the argument. |
6bda09f9 YO |
959 | |
960 | $re = qr{ ( # paren group 1 (full function) | |
81714fb9 | 961 | foo |
6bda09f9 YO |
962 | ( # paren group 2 (parens) |
963 | \( | |
964 | ( # paren group 3 (contents of parens) | |
965 | (?: | |
966 | (?> [^()]+ ) # Non-parens without backtracking | |
967 | | | |
968 | (?2) # Recurse to start of paren group 2 | |
969 | )* | |
970 | ) | |
971 | \) | |
972 | ) | |
973 | ) | |
974 | }x; | |
975 | ||
976 | If the pattern was used as follows | |
977 | ||
978 | 'foo(bar(baz)+baz(bop))'=~/$re/ | |
979 | and print "\$1 = $1\n", | |
980 | "\$2 = $2\n", | |
981 | "\$3 = $3\n"; | |
982 | ||
983 | the output produced should be the following: | |
984 | ||
985 | $1 = foo(bar(baz)+baz(bop)) | |
986 | $2 = (bar(baz)+baz(bop)) | |
81714fb9 | 987 | $3 = bar(baz)+baz(bop) |
6bda09f9 | 988 | |
81714fb9 | 989 | If there is no corresponding capture buffer defined, then it is a |
61528107 | 990 | fatal error. Recursing deeper than 50 times without consuming any input |
81714fb9 | 991 | string will also result in a fatal error. The maximum depth is compiled |
6bda09f9 YO |
992 | into perl, so changing it requires a custom build. |
993 | ||
542fa716 YO |
994 | The following shows how using negative indexing can make it |
995 | easier to embed recursive patterns inside of a C<qr//> construct | |
996 | for later use: | |
997 | ||
998 | my $parens = qr/(\((?:[^()]++|(?-1))*+\))/; | |
999 | if (/foo $parens \s+ + \s+ bar $parens/x) { | |
1000 | # do something here... | |
1001 | } | |
1002 | ||
81714fb9 | 1003 | B<Note> that this pattern does not behave the same way as the equivalent |
0d017f4d | 1004 | PCRE or Python construct of the same form. In Perl you can backtrack into |
6bda09f9 | 1005 | a recursed group, in PCRE and Python the recursed into group is treated |
542fa716 YO |
1006 | as atomic. Also, modifiers are resolved at compile time, so constructs |
1007 | like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will | |
1008 | be processed. | |
6bda09f9 | 1009 | |
894be9b7 YO |
1010 | =item C<(?&NAME)> |
1011 | X<(?&NAME)> | |
1012 | ||
0d017f4d WL |
1013 | Recurse to a named subpattern. Identical to C<(?PARNO)> except that the |
1014 | parenthesis to recurse to is determined by name. If multiple parentheses have | |
894be9b7 YO |
1015 | the same name, then it recurses to the leftmost. |
1016 | ||
1017 | It is an error to refer to a name that is not declared somewhere in the | |
1018 | pattern. | |
1019 | ||
1f1031fe YO |
1020 | B<NOTE:> In order to make things easier for programmers with experience |
1021 | with the Python or PCRE regex engines the pattern C<< (?P>NAME) >> | |
64c5a566 | 1022 | may be used instead of C<< (?&NAME) >>. |
1f1031fe | 1023 | |
e2e6a0f1 YO |
1024 | =item C<(?(condition)yes-pattern|no-pattern)> |
1025 | X<(?()> | |
286f584a | 1026 | |
e2e6a0f1 | 1027 | =item C<(?(condition)yes-pattern)> |
286f584a | 1028 | |
e2e6a0f1 YO |
1029 | Conditional expression. C<(condition)> should be either an integer in |
1030 | parentheses (which is valid if the corresponding pair of parentheses | |
1031 | matched), a look-ahead/look-behind/evaluate zero-width assertion, a | |
1032 | name in angle brackets or single quotes (which is valid if a buffer | |
1033 | with the given name matched), or the special symbol (R) (true when | |
1034 | evaluated inside of recursion or eval). Additionally the R may be | |
1035 | followed by a number, (which will be true when evaluated when recursing | |
1036 | inside of the appropriate group), or by C<&NAME>, in which case it will | |
1037 | be true only when evaluated during recursion in the named group. | |
1038 | ||
1039 | Here's a summary of the possible predicates: | |
1040 | ||
1041 | =over 4 | |
1042 | ||
1043 | =item (1) (2) ... | |
1044 | ||
1045 | Checks if the numbered capturing buffer has matched something. | |
1046 | ||
1047 | =item (<NAME>) ('NAME') | |
1048 | ||
1049 | Checks if a buffer with the given name has matched something. | |
1050 | ||
1051 | =item (?{ CODE }) | |
1052 | ||
1053 | Treats the code block as the condition. | |
1054 | ||
1055 | =item (R) | |
1056 | ||
1057 | Checks if the expression has been evaluated inside of recursion. | |
1058 | ||
1059 | =item (R1) (R2) ... | |
1060 | ||
1061 | Checks if the expression has been evaluated while executing directly | |
1062 | inside of the n-th capture group. This check is the regex equivalent of | |
1063 | ||
1064 | if ((caller(0))[3] eq 'subname') { ... } | |
1065 | ||
1066 | In other words, it does not check the full recursion stack. | |
1067 | ||
1068 | =item (R&NAME) | |
1069 | ||
1070 | Similar to C<(R1)>, this predicate checks to see if we're executing | |
1071 | directly inside of the leftmost group with a given name (this is the same | |
1072 | logic used by C<(?&NAME)> to disambiguate). It does not check the full | |
1073 | stack, but only the name of the innermost active recursion. | |
1074 | ||
1075 | =item (DEFINE) | |
1076 | ||
1077 | In this case, the yes-pattern is never directly executed, and no | |
1078 | no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient. | |
1079 | See below for details. | |
1080 | ||
1081 | =back | |
1082 | ||
1083 | For example: | |
1084 | ||
1085 | m{ ( \( )? | |
1086 | [^()]+ | |
1087 | (?(1) \) ) | |
1088 | }x | |
1089 | ||
1090 | matches a chunk of non-parentheses, possibly included in parentheses | |
1091 | themselves. | |
1092 | ||
1093 | A special form is the C<(DEFINE)> predicate, which never executes directly | |
1094 | its yes-pattern, and does not allow a no-pattern. This allows to define | |
1095 | subpatterns which will be executed only by using the recursion mechanism. | |
1096 | This way, you can define a set of regular expression rules that can be | |
1097 | bundled into any pattern you choose. | |
1098 | ||
1099 | It is recommended that for this usage you put the DEFINE block at the | |
1100 | end of the pattern, and that you name any subpatterns defined within it. | |
1101 | ||
1102 | Also, it's worth noting that patterns defined this way probably will | |
1103 | not be as efficient, as the optimiser is not very clever about | |
1104 | handling them. | |
1105 | ||
1106 | An example of how this might be used is as follows: | |
1107 | ||
2bf803e2 | 1108 | /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT)) |
e2e6a0f1 | 1109 | (?(DEFINE) |
2bf803e2 YO |
1110 | (?<NAME_PAT>....) |
1111 | (?<ADRESS_PAT>....) | |
e2e6a0f1 YO |
1112 | )/x |
1113 | ||
1114 | Note that capture buffers matched inside of recursion are not accessible | |
0d017f4d | 1115 | after the recursion returns, so the extra layer of capturing buffers is |
e2e6a0f1 YO |
1116 | necessary. Thus C<$+{NAME_PAT}> would not be defined even though |
1117 | C<$+{NAME}> would be. | |
286f584a | 1118 | |
c47ff5f1 | 1119 | =item C<< (?>pattern) >> |
6bda09f9 | 1120 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 | 1121 | |
19799a22 GS |
1122 | An "independent" subexpression, one which matches the substring |
1123 | that a I<standalone> C<pattern> would match if anchored at the given | |
9da458fc | 1124 | position, and it matches I<nothing other than this substring>. This |
19799a22 GS |
1125 | construct is useful for optimizations of what would otherwise be |
1126 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). | |
9da458fc IZ |
1127 | It may also be useful in places where the "grab all you can, and do not |
1128 | give anything back" semantic is desirable. | |
19799a22 | 1129 | |
c47ff5f1 | 1130 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 GS |
1131 | (anchored at the beginning of string, as above) will match I<all> |
1132 | characters C<a> at the beginning of string, leaving no C<a> for | |
1133 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, | |
1134 | since the match of the subgroup C<a*> is influenced by the following | |
1135 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside | |
1136 | C<a*ab> will match fewer characters than a standalone C<a*>, since | |
1137 | this makes the tail match. | |
1138 | ||
c47ff5f1 | 1139 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
19799a22 GS |
1140 | C<(?=(pattern))\1>. This matches the same substring as a standalone |
1141 | C<a+>, and the following C<\1> eats the matched string; it therefore | |
c47ff5f1 | 1142 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 GS |
1143 | (The difference between these two constructs is that the second one |
1144 | uses a capturing group, thus shifting ordinals of backreferences | |
1145 | in the rest of a regular expression.) | |
1146 | ||
1147 | Consider this pattern: | |
c277df42 | 1148 | |
871b0233 | 1149 | m{ \( |
e2e6a0f1 YO |
1150 | ( |
1151 | [^()]+ # x+ | |
1152 | | | |
871b0233 IZ |
1153 | \( [^()]* \) |
1154 | )+ | |
e2e6a0f1 | 1155 | \) |
871b0233 | 1156 | }x |
5a964f20 | 1157 | |
19799a22 GS |
1158 | That will efficiently match a nonempty group with matching parentheses |
1159 | two levels deep or less. However, if there is no such group, it | |
1160 | will take virtually forever on a long string. That's because there | |
1161 | are so many different ways to split a long string into several | |
1162 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
1163 | to a subpattern of the above pattern. Consider how the pattern | |
1164 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
1165 | seconds, but that each extra letter doubles this time. This | |
1166 | exponential performance will make it appear that your program has | |
14218588 | 1167 | hung. However, a tiny change to this pattern |
5a964f20 | 1168 | |
e2e6a0f1 YO |
1169 | m{ \( |
1170 | ( | |
1171 | (?> [^()]+ ) # change x+ above to (?> x+ ) | |
1172 | | | |
871b0233 IZ |
1173 | \( [^()]* \) |
1174 | )+ | |
e2e6a0f1 | 1175 | \) |
871b0233 | 1176 | }x |
c277df42 | 1177 | |
c47ff5f1 | 1178 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 TC |
1179 | this yourself would be a productive exercise), but finishes in a fourth |
1180 | the time when used on a similar string with 1000000 C<a>s. Be aware, | |
1181 | however, that this pattern currently triggers a warning message under | |
9f1b1f2d | 1182 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b | 1183 | C<"matches null string many times in regex">. |
c277df42 | 1184 | |
c47ff5f1 | 1185 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 | 1186 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 IZ |
1187 | This was only 4 times slower on a string with 1000000 C<a>s. |
1188 | ||
9da458fc IZ |
1189 | The "grab all you can, and do not give anything back" semantic is desirable |
1190 | in many situations where on the first sight a simple C<()*> looks like | |
1191 | the correct solution. Suppose we parse text with comments being delimited | |
1192 | by C<#> followed by some optional (horizontal) whitespace. Contrary to | |
4375e838 | 1193 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc IZ |
1194 | the comment delimiter, because it may "give up" some whitespace if |
1195 | the remainder of the pattern can be made to match that way. The correct | |
1196 | answer is either one of these: | |
1197 | ||
1198 | (?>#[ \t]*) | |
1199 | #[ \t]*(?![ \t]) | |
1200 | ||
1201 | For example, to grab non-empty comments into $1, one should use either | |
1202 | one of these: | |
1203 | ||
1204 | / (?> \# [ \t]* ) ( .+ ) /x; | |
1205 | / \# [ \t]* ( [^ \t] .* ) /x; | |
1206 | ||
1207 | Which one you pick depends on which of these expressions better reflects | |
1208 | the above specification of comments. | |
1209 | ||
6bda09f9 YO |
1210 | In some literature this construct is called "atomic matching" or |
1211 | "possessive matching". | |
1212 | ||
b9b4dddf YO |
1213 | Possessive quantifiers are equivalent to putting the item they are applied |
1214 | to inside of one of these constructs. The following equivalences apply: | |
1215 | ||
1216 | Quantifier Form Bracketing Form | |
1217 | --------------- --------------- | |
1218 | PAT*+ (?>PAT*) | |
1219 | PAT++ (?>PAT+) | |
1220 | PAT?+ (?>PAT?) | |
1221 | PAT{min,max}+ (?>PAT{min,max}) | |
1222 | ||
e2e6a0f1 YO |
1223 | =back |
1224 | ||
1225 | =head2 Special Backtracking Control Verbs | |
1226 | ||
1227 | B<WARNING:> These patterns are experimental and subject to change or | |
0d017f4d | 1228 | removal in a future version of Perl. Their usage in production code should |
e2e6a0f1 YO |
1229 | be noted to avoid problems during upgrades. |
1230 | ||
1231 | These special patterns are generally of the form C<(*VERB:ARG)>. Unless | |
1232 | otherwise stated the ARG argument is optional; in some cases, it is | |
1233 | forbidden. | |
1234 | ||
1235 | Any pattern containing a special backtracking verb that allows an argument | |
e1020413 | 1236 | has the special behaviour that when executed it sets the current package's |
5d458dd8 YO |
1237 | C<$REGERROR> and C<$REGMARK> variables. When doing so the following |
1238 | rules apply: | |
e2e6a0f1 | 1239 | |
5d458dd8 YO |
1240 | On failure, the C<$REGERROR> variable will be set to the ARG value of the |
1241 | verb pattern, if the verb was involved in the failure of the match. If the | |
1242 | ARG part of the pattern was omitted, then C<$REGERROR> will be set to the | |
1243 | name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was | |
1244 | none. Also, the C<$REGMARK> variable will be set to FALSE. | |
e2e6a0f1 | 1245 | |
5d458dd8 YO |
1246 | On a successful match, the C<$REGERROR> variable will be set to FALSE, and |
1247 | the C<$REGMARK> variable will be set to the name of the last | |
1248 | C<(*MARK:NAME)> pattern executed. See the explanation for the | |
1249 | C<(*MARK:NAME)> verb below for more details. | |
e2e6a0f1 | 1250 | |
5d458dd8 YO |
1251 | B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1> |
1252 | and most other regex related variables. They are not local to a scope, nor | |
1253 | readonly, but instead are volatile package variables similar to C<$AUTOLOAD>. | |
1254 | Use C<local> to localize changes to them to a specific scope if necessary. | |
e2e6a0f1 YO |
1255 | |
1256 | If a pattern does not contain a special backtracking verb that allows an | |
5d458dd8 | 1257 | argument, then C<$REGERROR> and C<$REGMARK> are not touched at all. |
e2e6a0f1 YO |
1258 | |
1259 | =over 4 | |
1260 | ||
1261 | =item Verbs that take an argument | |
1262 | ||
1263 | =over 4 | |
1264 | ||
5d458dd8 | 1265 | =item C<(*PRUNE)> C<(*PRUNE:NAME)> |
f7819f85 | 1266 | X<(*PRUNE)> X<(*PRUNE:NAME)> |
54612592 | 1267 | |
5d458dd8 YO |
1268 | This zero-width pattern prunes the backtracking tree at the current point |
1269 | when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>, | |
1270 | where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached, | |
1271 | A may backtrack as necessary to match. Once it is reached, matching | |
1272 | continues in B, which may also backtrack as necessary; however, should B | |
1273 | not match, then no further backtracking will take place, and the pattern | |
1274 | will fail outright at the current starting position. | |
54612592 YO |
1275 | |
1276 | The following example counts all the possible matching strings in a | |
1277 | pattern (without actually matching any of them). | |
1278 | ||
e2e6a0f1 | 1279 | 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1280 | print "Count=$count\n"; |
1281 | ||
1282 | which produces: | |
1283 | ||
1284 | aaab | |
1285 | aaa | |
1286 | aa | |
1287 | a | |
1288 | aab | |
1289 | aa | |
1290 | a | |
1291 | ab | |
1292 | a | |
1293 | Count=9 | |
1294 | ||
5d458dd8 | 1295 | If we add a C<(*PRUNE)> before the count like the following |
54612592 | 1296 | |
5d458dd8 | 1297 | 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1298 | print "Count=$count\n"; |
1299 | ||
1300 | we prevent backtracking and find the count of the longest matching | |
353c6505 | 1301 | at each matching starting point like so: |
54612592 YO |
1302 | |
1303 | aaab | |
1304 | aab | |
1305 | ab | |
1306 | Count=3 | |
1307 | ||
5d458dd8 | 1308 | Any number of C<(*PRUNE)> assertions may be used in a pattern. |
54612592 | 1309 | |
5d458dd8 YO |
1310 | See also C<< (?>pattern) >> and possessive quantifiers for other ways to |
1311 | control backtracking. In some cases, the use of C<(*PRUNE)> can be | |
1312 | replaced with a C<< (?>pattern) >> with no functional difference; however, | |
1313 | C<(*PRUNE)> can be used to handle cases that cannot be expressed using a | |
1314 | C<< (?>pattern) >> alone. | |
54612592 | 1315 | |
e2e6a0f1 | 1316 | |
5d458dd8 YO |
1317 | =item C<(*SKIP)> C<(*SKIP:NAME)> |
1318 | X<(*SKIP)> | |
e2e6a0f1 | 1319 | |
5d458dd8 | 1320 | This zero-width pattern is similar to C<(*PRUNE)>, except that on |
e2e6a0f1 | 1321 | failure it also signifies that whatever text that was matched leading up |
5d458dd8 YO |
1322 | to the C<(*SKIP)> pattern being executed cannot be part of I<any> match |
1323 | of this pattern. This effectively means that the regex engine "skips" forward | |
1324 | to this position on failure and tries to match again, (assuming that | |
1325 | there is sufficient room to match). | |
1326 | ||
1327 | The name of the C<(*SKIP:NAME)> pattern has special significance. If a | |
1328 | C<(*MARK:NAME)> was encountered while matching, then it is that position | |
1329 | which is used as the "skip point". If no C<(*MARK)> of that name was | |
1330 | encountered, then the C<(*SKIP)> operator has no effect. When used | |
1331 | without a name the "skip point" is where the match point was when | |
1332 | executing the (*SKIP) pattern. | |
1333 | ||
1334 | Compare the following to the examples in C<(*PRUNE)>, note the string | |
24b23f37 YO |
1335 | is twice as long: |
1336 | ||
5d458dd8 | 1337 | 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 YO |
1338 | print "Count=$count\n"; |
1339 | ||
1340 | outputs | |
1341 | ||
1342 | aaab | |
1343 | aaab | |
1344 | Count=2 | |
1345 | ||
5d458dd8 | 1346 | Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)> |
353c6505 | 1347 | executed, the next starting point will be where the cursor was when the |
5d458dd8 YO |
1348 | C<(*SKIP)> was executed. |
1349 | ||
5d458dd8 YO |
1350 | =item C<(*MARK:NAME)> C<(*:NAME)> |
1351 | X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)> | |
1352 | ||
1353 | This zero-width pattern can be used to mark the point reached in a string | |
1354 | when a certain part of the pattern has been successfully matched. This | |
1355 | mark may be given a name. A later C<(*SKIP)> pattern will then skip | |
1356 | forward to that point if backtracked into on failure. Any number of | |
b4222fa9 | 1357 | C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated. |
5d458dd8 YO |
1358 | |
1359 | In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)> | |
1360 | can be used to "label" a pattern branch, so that after matching, the | |
1361 | program can determine which branches of the pattern were involved in the | |
1362 | match. | |
1363 | ||
1364 | When a match is successful, the C<$REGMARK> variable will be set to the | |
1365 | name of the most recently executed C<(*MARK:NAME)> that was involved | |
1366 | in the match. | |
1367 | ||
1368 | This can be used to determine which branch of a pattern was matched | |
c62285ac | 1369 | without using a separate capture buffer for each branch, which in turn |
5d458dd8 YO |
1370 | can result in a performance improvement, as perl cannot optimize |
1371 | C</(?:(x)|(y)|(z))/> as efficiently as something like | |
1372 | C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>. | |
1373 | ||
1374 | When a match has failed, and unless another verb has been involved in | |
1375 | failing the match and has provided its own name to use, the C<$REGERROR> | |
1376 | variable will be set to the name of the most recently executed | |
1377 | C<(*MARK:NAME)>. | |
1378 | ||
1379 | See C<(*SKIP)> for more details. | |
1380 | ||
b62d2d15 YO |
1381 | As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>. |
1382 | ||
5d458dd8 YO |
1383 | =item C<(*THEN)> C<(*THEN:NAME)> |
1384 | ||
241e7389 | 1385 | This is similar to the "cut group" operator C<::> from Perl 6. Like |
5d458dd8 YO |
1386 | C<(*PRUNE)>, this verb always matches, and when backtracked into on |
1387 | failure, it causes the regex engine to try the next alternation in the | |
1388 | innermost enclosing group (capturing or otherwise). | |
1389 | ||
1390 | Its name comes from the observation that this operation combined with the | |
1391 | alternation operator (C<|>) can be used to create what is essentially a | |
1392 | pattern-based if/then/else block: | |
1393 | ||
1394 | ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) | |
1395 | ||
1396 | Note that if this operator is used and NOT inside of an alternation then | |
1397 | it acts exactly like the C<(*PRUNE)> operator. | |
1398 | ||
1399 | / A (*PRUNE) B / | |
1400 | ||
1401 | is the same as | |
1402 | ||
1403 | / A (*THEN) B / | |
1404 | ||
1405 | but | |
1406 | ||
1407 | / ( A (*THEN) B | C (*THEN) D ) / | |
1408 | ||
1409 | is not the same as | |
1410 | ||
1411 | / ( A (*PRUNE) B | C (*PRUNE) D ) / | |
1412 | ||
1413 | as after matching the A but failing on the B the C<(*THEN)> verb will | |
1414 | backtrack and try C; but the C<(*PRUNE)> verb will simply fail. | |
24b23f37 | 1415 | |
e2e6a0f1 YO |
1416 | =item C<(*COMMIT)> |
1417 | X<(*COMMIT)> | |
24b23f37 | 1418 | |
241e7389 | 1419 | This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a |
5d458dd8 YO |
1420 | zero-width pattern similar to C<(*SKIP)>, except that when backtracked |
1421 | into on failure it causes the match to fail outright. No further attempts | |
1422 | to find a valid match by advancing the start pointer will occur again. | |
1423 | For example, | |
24b23f37 | 1424 | |
e2e6a0f1 | 1425 | 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 YO |
1426 | print "Count=$count\n"; |
1427 | ||
1428 | outputs | |
1429 | ||
1430 | aaab | |
1431 | Count=1 | |
1432 | ||
e2e6a0f1 YO |
1433 | In other words, once the C<(*COMMIT)> has been entered, and if the pattern |
1434 | does not match, the regex engine will not try any further matching on the | |
1435 | rest of the string. | |
c277df42 | 1436 | |
e2e6a0f1 | 1437 | =back |
9af228c6 | 1438 | |
e2e6a0f1 | 1439 | =item Verbs without an argument |
9af228c6 YO |
1440 | |
1441 | =over 4 | |
1442 | ||
e2e6a0f1 YO |
1443 | =item C<(*FAIL)> C<(*F)> |
1444 | X<(*FAIL)> X<(*F)> | |
9af228c6 | 1445 | |
e2e6a0f1 YO |
1446 | This pattern matches nothing and always fails. It can be used to force the |
1447 | engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In | |
1448 | fact, C<(?!)> gets optimised into C<(*FAIL)> internally. | |
9af228c6 | 1449 | |
e2e6a0f1 | 1450 | It is probably useful only when combined with C<(?{})> or C<(??{})>. |
9af228c6 | 1451 | |
e2e6a0f1 YO |
1452 | =item C<(*ACCEPT)> |
1453 | X<(*ACCEPT)> | |
9af228c6 | 1454 | |
e2e6a0f1 YO |
1455 | B<WARNING:> This feature is highly experimental. It is not recommended |
1456 | for production code. | |
9af228c6 | 1457 | |
e2e6a0f1 YO |
1458 | This pattern matches nothing and causes the end of successful matching at |
1459 | the point at which the C<(*ACCEPT)> pattern was encountered, regardless of | |
1460 | whether there is actually more to match in the string. When inside of a | |
0d017f4d | 1461 | nested pattern, such as recursion, or in a subpattern dynamically generated |
e2e6a0f1 | 1462 | via C<(??{})>, only the innermost pattern is ended immediately. |
9af228c6 | 1463 | |
e2e6a0f1 YO |
1464 | If the C<(*ACCEPT)> is inside of capturing buffers then the buffers are |
1465 | marked as ended at the point at which the C<(*ACCEPT)> was encountered. | |
1466 | For instance: | |
9af228c6 | 1467 | |
e2e6a0f1 | 1468 | 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x; |
9af228c6 | 1469 | |
e2e6a0f1 | 1470 | will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not |
0d017f4d | 1471 | be set. If another branch in the inner parentheses were matched, such as in the |
e2e6a0f1 | 1472 | string 'ACDE', then the C<D> and C<E> would have to be matched as well. |
9af228c6 YO |
1473 | |
1474 | =back | |
c277df42 | 1475 | |
a0d0e21e LW |
1476 | =back |
1477 | ||
c07a80fd | 1478 | =head2 Backtracking |
d74e8afc | 1479 | X<backtrack> X<backtracking> |
c07a80fd | 1480 | |
35a734be IZ |
1481 | NOTE: This section presents an abstract approximation of regular |
1482 | expression behavior. For a more rigorous (and complicated) view of | |
1483 | the rules involved in selecting a match among possible alternatives, | |
0d017f4d | 1484 | see L<Combining RE Pieces>. |
35a734be | 1485 | |
c277df42 | 1486 | A fundamental feature of regular expression matching involves the |
5a964f20 | 1487 | notion called I<backtracking>, which is currently used (when needed) |
0d017f4d | 1488 | by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc IZ |
1489 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
1490 | internally, but the general principle outlined here is valid. | |
c07a80fd | 1491 | |
1492 | For a regular expression to match, the I<entire> regular expression must | |
1493 | match, not just part of it. So if the beginning of a pattern containing a | |
1494 | quantifier succeeds in a way that causes later parts in the pattern to | |
1495 | fail, the matching engine backs up and recalculates the beginning | |
1496 | part--that's why it's called backtracking. | |
1497 | ||
1498 | Here is an example of backtracking: Let's say you want to find the | |
1499 | word following "foo" in the string "Food is on the foo table.": | |
1500 | ||
1501 | $_ = "Food is on the foo table."; | |
1502 | if ( /\b(foo)\s+(\w+)/i ) { | |
1503 | print "$2 follows $1.\n"; | |
1504 | } | |
1505 | ||
1506 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
1507 | finds a possible match right at the beginning of the string, and loads up | |
1508 | $1 with "Foo". However, as soon as the matching engine sees that there's | |
1509 | no whitespace following the "Foo" that it had saved in $1, it realizes its | |
68dc0745 | 1510 | mistake and starts over again one character after where it had the |
c07a80fd | 1511 | tentative match. This time it goes all the way until the next occurrence |
1512 | of "foo". The complete regular expression matches this time, and you get | |
1513 | the expected output of "table follows foo." | |
1514 | ||
1515 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
1516 | everything between "foo" and "bar". Initially, you write something | |
1517 | like this: | |
1518 | ||
1519 | $_ = "The food is under the bar in the barn."; | |
1520 | if ( /foo(.*)bar/ ) { | |
1521 | print "got <$1>\n"; | |
1522 | } | |
1523 | ||
1524 | Which perhaps unexpectedly yields: | |
1525 | ||
1526 | got <d is under the bar in the > | |
1527 | ||
1528 | That's because C<.*> was greedy, so you get everything between the | |
14218588 | 1529 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd | 1530 | to use minimal matching to make sure you get the text between a "foo" |
1531 | and the first "bar" thereafter. | |
1532 | ||
1533 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
1534 | got <d is under the > | |
1535 | ||
0d017f4d | 1536 | Here's another example. Let's say you'd like to match a number at the end |
b6e13d97 | 1537 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd | 1538 | So you write this: |
1539 | ||
1540 | $_ = "I have 2 numbers: 53147"; | |
1541 | if ( /(.*)(\d*)/ ) { # Wrong! | |
1542 | print "Beginning is <$1>, number is <$2>.\n"; | |
1543 | } | |
1544 | ||
1545 | That won't work at all, because C<.*> was greedy and gobbled up the | |
1546 | whole string. As C<\d*> can match on an empty string the complete | |
1547 | regular expression matched successfully. | |
1548 | ||
8e1088bc | 1549 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd | 1550 | |
1551 | Here are some variants, most of which don't work: | |
1552 | ||
1553 | $_ = "I have 2 numbers: 53147"; | |
1554 | @pats = qw{ | |
1555 | (.*)(\d*) | |
1556 | (.*)(\d+) | |
1557 | (.*?)(\d*) | |
1558 | (.*?)(\d+) | |
1559 | (.*)(\d+)$ | |
1560 | (.*?)(\d+)$ | |
1561 | (.*)\b(\d+)$ | |
1562 | (.*\D)(\d+)$ | |
1563 | }; | |
1564 | ||
1565 | for $pat (@pats) { | |
1566 | printf "%-12s ", $pat; | |
1567 | if ( /$pat/ ) { | |
1568 | print "<$1> <$2>\n"; | |
1569 | } else { | |
1570 | print "FAIL\n"; | |
1571 | } | |
1572 | } | |
1573 | ||
1574 | That will print out: | |
1575 | ||
1576 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
1577 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
1578 | (.*?)(\d*) <> <> | |
1579 | (.*?)(\d+) <I have > <2> | |
1580 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
1581 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
1582 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
1583 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
1584 | ||
1585 | As you see, this can be a bit tricky. It's important to realize that a | |
1586 | regular expression is merely a set of assertions that gives a definition | |
1587 | of success. There may be 0, 1, or several different ways that the | |
1588 | definition might succeed against a particular string. And if there are | |
5a964f20 TC |
1589 | multiple ways it might succeed, you need to understand backtracking to |
1590 | know which variety of success you will achieve. | |
c07a80fd | 1591 | |
19799a22 | 1592 | When using look-ahead assertions and negations, this can all get even |
8b19b778 | 1593 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd | 1594 | followed by "123". You might try to write that as |
1595 | ||
871b0233 IZ |
1596 | $_ = "ABC123"; |
1597 | if ( /^\D*(?!123)/ ) { # Wrong! | |
1598 | print "Yup, no 123 in $_\n"; | |
1599 | } | |
c07a80fd | 1600 | |
1601 | But that isn't going to match; at least, not the way you're hoping. It | |
1602 | claims that there is no 123 in the string. Here's a clearer picture of | |
9b9391b2 | 1603 | why that pattern matches, contrary to popular expectations: |
c07a80fd | 1604 | |
4358a253 SS |
1605 | $x = 'ABC123'; |
1606 | $y = 'ABC445'; | |
c07a80fd | 1607 | |
4358a253 SS |
1608 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
1609 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; | |
c07a80fd | 1610 | |
4358a253 SS |
1611 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; |
1612 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; | |
c07a80fd | 1613 | |
1614 | This prints | |
1615 | ||
1616 | 2: got ABC | |
1617 | 3: got AB | |
1618 | 4: got ABC | |
1619 | ||
5f05dabc | 1620 | You might have expected test 3 to fail because it seems to a more |
c07a80fd | 1621 | general purpose version of test 1. The important difference between |
1622 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
1623 | backtracking, whereas test 1 will not. What's happening is | |
1624 | that you've asked "Is it true that at the start of $x, following 0 or more | |
5f05dabc | 1625 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd | 1626 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 | 1627 | fail. |
14218588 | 1628 | |
c07a80fd | 1629 | The search engine will initially match C<\D*> with "ABC". Then it will |
14218588 | 1630 | try to match C<(?!123> with "123", which fails. But because |
c07a80fd | 1631 | a quantifier (C<\D*>) has been used in the regular expression, the |
1632 | search engine can backtrack and retry the match differently | |
54310121 | 1633 | in the hope of matching the complete regular expression. |
c07a80fd | 1634 | |
5a964f20 TC |
1635 | The pattern really, I<really> wants to succeed, so it uses the |
1636 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
c07a80fd | 1637 | time. Now there's indeed something following "AB" that is not |
14218588 | 1638 | "123". It's "C123", which suffices. |
c07a80fd | 1639 | |
14218588 GS |
1640 | We can deal with this by using both an assertion and a negation. |
1641 | We'll say that the first part in $1 must be followed both by a digit | |
1642 | and by something that's not "123". Remember that the look-aheads | |
1643 | are zero-width expressions--they only look, but don't consume any | |
1644 | of the string in their match. So rewriting this way produces what | |
c07a80fd | 1645 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
1646 | ||
4358a253 SS |
1647 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; |
1648 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; | |
c07a80fd | 1649 | |
1650 | 6: got ABC | |
1651 | ||
5a964f20 | 1652 | In other words, the two zero-width assertions next to each other work as though |
19799a22 | 1653 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd | 1654 | matches only if you're at the beginning of the line AND the end of the |
1655 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
1656 | regular expressions always means AND, except when you write an explicit OR | |
1657 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
1658 | although the attempted matches are made at different positions because "a" | |
1659 | is not a zero-width assertion, but a one-width assertion. | |
1660 | ||
0d017f4d | 1661 | B<WARNING>: Particularly complicated regular expressions can take |
14218588 | 1662 | exponential time to solve because of the immense number of possible |
0d017f4d | 1663 | ways they can use backtracking to try for a match. For example, without |
9da458fc IZ |
1664 | internal optimizations done by the regular expression engine, this will |
1665 | take a painfully long time to run: | |
c07a80fd | 1666 | |
e1901655 IZ |
1667 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
1668 | ||
1669 | And if you used C<*>'s in the internal groups instead of limiting them | |
1670 | to 0 through 5 matches, then it would take forever--or until you ran | |
1671 | out of stack space. Moreover, these internal optimizations are not | |
1672 | always applicable. For example, if you put C<{0,5}> instead of C<*> | |
1673 | on the external group, no current optimization is applicable, and the | |
1674 | match takes a long time to finish. | |
c07a80fd | 1675 | |
9da458fc IZ |
1676 | A powerful tool for optimizing such beasts is what is known as an |
1677 | "independent group", | |
c47ff5f1 | 1678 | which does not backtrack (see L<C<< (?>pattern) >>>). Note also that |
9da458fc | 1679 | zero-length look-ahead/look-behind assertions will not backtrack to make |
5d458dd8 | 1680 | the tail match, since they are in "logical" context: only |
14218588 | 1681 | whether they match is considered relevant. For an example |
9da458fc | 1682 | where side-effects of look-ahead I<might> have influenced the |
c47ff5f1 | 1683 | following match, see L<C<< (?>pattern) >>>. |
c277df42 | 1684 | |
a0d0e21e | 1685 | =head2 Version 8 Regular Expressions |
d74e8afc | 1686 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> |
a0d0e21e | 1687 | |
5a964f20 | 1688 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e LW |
1689 | routines, here are the pattern-matching rules not described above. |
1690 | ||
54310121 | 1691 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e | 1692 | with a special meaning described here or above. You can cause |
5a964f20 | 1693 | characters that normally function as metacharacters to be interpreted |
5f05dabc | 1694 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
0d017f4d WL |
1695 | character; "\\" matches a "\"). This escape mechanism is also required |
1696 | for the character used as the pattern delimiter. | |
1697 | ||
1698 | A series of characters matches that series of characters in the target | |
1699 | string, so the pattern C<blurfl> would match "blurfl" in the target | |
1700 | string. | |
a0d0e21e LW |
1701 | |
1702 | You can specify a character class, by enclosing a list of characters | |
5d458dd8 | 1703 | in C<[]>, which will match any character from the list. If the |
a0d0e21e | 1704 | first character after the "[" is "^", the class matches any character not |
14218588 | 1705 | in the list. Within a list, the "-" character specifies a |
5a964f20 | 1706 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 GS |
1707 | inclusive. If you want either "-" or "]" itself to be a member of a |
1708 | class, put it at the start of the list (possibly after a "^"), or | |
1709 | escape it with a backslash. "-" is also taken literally when it is | |
1710 | at the end of the list, just before the closing "]". (The | |
84850974 DD |
1711 | following all specify the same class of three characters: C<[-az]>, |
1712 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which | |
5d458dd8 YO |
1713 | specifies a class containing twenty-six characters, even on EBCDIC-based |
1714 | character sets.) Also, if you try to use the character | |
1715 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of | |
1716 | a range, the "-" is understood literally. | |
a0d0e21e | 1717 | |
8ada0baa JH |
1718 | Note also that the whole range idea is rather unportable between |
1719 | character sets--and even within character sets they may cause results | |
1720 | you probably didn't expect. A sound principle is to use only ranges | |
0d017f4d | 1721 | that begin from and end at either alphabetics of equal case ([a-e], |
8ada0baa JH |
1722 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, |
1723 | spell out the character sets in full. | |
1724 | ||
54310121 | 1725 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e LW |
1726 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
1727 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string | |
5d458dd8 YO |
1728 | of octal digits, matches the character whose coded character set value |
1729 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, | |
1730 | matches the character whose numeric value is I<nn>. The expression \cI<x> | |
1731 | matches the character control-I<x>. Finally, the "." metacharacter | |
fb55449c | 1732 | matches any character except "\n" (unless you use C</s>). |
a0d0e21e LW |
1733 | |
1734 | You can specify a series of alternatives for a pattern using "|" to | |
1735 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", | |
5a964f20 | 1736 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e LW |
1737 | first alternative includes everything from the last pattern delimiter |
1738 | ("(", "[", or the beginning of the pattern) up to the first "|", and | |
1739 | the last alternative contains everything from the last "|" to the next | |
14218588 GS |
1740 | pattern delimiter. That's why it's common practice to include |
1741 | alternatives in parentheses: to minimize confusion about where they | |
a3cb178b GS |
1742 | start and end. |
1743 | ||
5a964f20 | 1744 | Alternatives are tried from left to right, so the first |
a3cb178b GS |
1745 | alternative found for which the entire expression matches, is the one that |
1746 | is chosen. This means that alternatives are not necessarily greedy. For | |
628afcb5 | 1747 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b GS |
1748 | part will match, as that is the first alternative tried, and it successfully |
1749 | matches the target string. (This might not seem important, but it is | |
1750 | important when you are capturing matched text using parentheses.) | |
1751 | ||
5a964f20 | 1752 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b | 1753 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e | 1754 | |
14218588 GS |
1755 | Within a pattern, you may designate subpatterns for later reference |
1756 | by enclosing them in parentheses, and you may refer back to the | |
1757 | I<n>th subpattern later in the pattern using the metacharacter | |
1758 | \I<n>. Subpatterns are numbered based on the left to right order | |
1759 | of their opening parenthesis. A backreference matches whatever | |
1760 | actually matched the subpattern in the string being examined, not | |
1761 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will | |
1762 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern | |
1763 | 1 matched "0x", even though the rule C<0|0x> could potentially match | |
1764 | the leading 0 in the second number. | |
cb1a09d0 | 1765 | |
0d017f4d | 1766 | =head2 Warning on \1 Instead of $1 |
cb1a09d0 | 1767 | |
5a964f20 | 1768 | Some people get too used to writing things like: |
cb1a09d0 AD |
1769 | |
1770 | $pattern =~ s/(\W)/\\\1/g; | |
1771 | ||
1772 | This is grandfathered for the RHS of a substitute to avoid shocking the | |
1773 | B<sed> addicts, but it's a dirty habit to get into. That's because in | |
d1be9408 | 1774 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 AD |
1775 | the usual double-quoted string means a control-A. The customary Unix |
1776 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
1777 | of doing that, you get yourself into trouble if you then add an C</e> | |
1778 | modifier. | |
1779 | ||
5a964f20 | 1780 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 AD |
1781 | |
1782 | Or if you try to do | |
1783 | ||
1784 | s/(\d+)/\1000/; | |
1785 | ||
1786 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
14218588 | 1787 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 AD |
1788 | with the operation of matching a backreference. Certainly they mean two |
1789 | different things on the I<left> side of the C<s///>. | |
9fa51da4 | 1790 | |
0d017f4d | 1791 | =head2 Repeated Patterns Matching a Zero-length Substring |
c84d73f1 | 1792 | |
19799a22 | 1793 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 IZ |
1794 | |
1795 | Regular expressions provide a terse and powerful programming language. As | |
1796 | with most other power tools, power comes together with the ability | |
1797 | to wreak havoc. | |
1798 | ||
1799 | A common abuse of this power stems from the ability to make infinite | |
628afcb5 | 1800 | loops using regular expressions, with something as innocuous as: |
c84d73f1 IZ |
1801 | |
1802 | 'foo' =~ m{ ( o? )* }x; | |
1803 | ||
0d017f4d | 1804 | The C<o?> matches at the beginning of C<'foo'>, and since the position |
c84d73f1 | 1805 | in the string is not moved by the match, C<o?> would match again and again |
527e91da | 1806 | because of the C<*> quantifier. Another common way to create a similar cycle |
c84d73f1 IZ |
1807 | is with the looping modifier C<//g>: |
1808 | ||
1809 | @matches = ( 'foo' =~ m{ o? }xg ); | |
1810 | ||
1811 | or | |
1812 | ||
1813 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
1814 | ||
1815 | or the loop implied by split(). | |
1816 | ||
1817 | However, long experience has shown that many programming tasks may | |
14218588 GS |
1818 | be significantly simplified by using repeated subexpressions that |
1819 | may match zero-length substrings. Here's a simple example being: | |
c84d73f1 IZ |
1820 | |
1821 | @chars = split //, $string; # // is not magic in split | |
1822 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / | |
1823 | ||
9da458fc | 1824 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 | 1825 | the infinite loop>. The rules for this are different for lower-level |
527e91da | 1826 | loops given by the greedy quantifiers C<*+{}>, and for higher-level |
c84d73f1 IZ |
1827 | ones like the C</g> modifier or split() operator. |
1828 | ||
19799a22 GS |
1829 | The lower-level loops are I<interrupted> (that is, the loop is |
1830 | broken) when Perl detects that a repeated expression matched a | |
1831 | zero-length substring. Thus | |
c84d73f1 IZ |
1832 | |
1833 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
1834 | ||
5d458dd8 | 1835 | is made equivalent to |
c84d73f1 | 1836 | |
5d458dd8 YO |
1837 | m{ (?: NON_ZERO_LENGTH )* |
1838 | | | |
1839 | (?: ZERO_LENGTH )? | |
c84d73f1 IZ |
1840 | }x; |
1841 | ||
1842 | The higher level-loops preserve an additional state between iterations: | |
5d458dd8 | 1843 | whether the last match was zero-length. To break the loop, the following |
c84d73f1 | 1844 | match after a zero-length match is prohibited to have a length of zero. |
5d458dd8 | 1845 | This prohibition interacts with backtracking (see L<"Backtracking">), |
c84d73f1 IZ |
1846 | and so the I<second best> match is chosen if the I<best> match is of |
1847 | zero length. | |
1848 | ||
19799a22 | 1849 | For example: |
c84d73f1 IZ |
1850 | |
1851 | $_ = 'bar'; | |
1852 | s/\w??/<$&>/g; | |
1853 | ||
20fb949f | 1854 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
5d458dd8 | 1855 | match given by non-greedy C<??> is the zero-length match, and the I<second |
c84d73f1 IZ |
1856 | best> match is what is matched by C<\w>. Thus zero-length matches |
1857 | alternate with one-character-long matches. | |
1858 | ||
5d458dd8 | 1859 | Similarly, for repeated C<m/()/g> the second-best match is the match at the |
c84d73f1 IZ |
1860 | position one notch further in the string. |
1861 | ||
19799a22 | 1862 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 | 1863 | the matched string, and is reset by each assignment to pos(). |
9da458fc IZ |
1864 | Zero-length matches at the end of the previous match are ignored |
1865 | during C<split>. | |
c84d73f1 | 1866 | |
0d017f4d | 1867 | =head2 Combining RE Pieces |
35a734be IZ |
1868 | |
1869 | Each of the elementary pieces of regular expressions which were described | |
1870 | before (such as C<ab> or C<\Z>) could match at most one substring | |
1871 | at the given position of the input string. However, in a typical regular | |
1872 | expression these elementary pieces are combined into more complicated | |
1873 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc | |
1874 | (in these examples C<S> and C<T> are regular subexpressions). | |
1875 | ||
1876 | Such combinations can include alternatives, leading to a problem of choice: | |
1877 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
1878 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
1879 | actually matched is the concept of backtracking (see L<"Backtracking">). | |
1880 | However, this description is too low-level and makes you think | |
1881 | in terms of a particular implementation. | |
1882 | ||
1883 | Another description starts with notions of "better"/"worse". All the | |
1884 | substrings which may be matched by the given regular expression can be | |
1885 | sorted from the "best" match to the "worst" match, and it is the "best" | |
1886 | match which is chosen. This substitutes the question of "what is chosen?" | |
1887 | by the question of "which matches are better, and which are worse?". | |
1888 | ||
1889 | Again, for elementary pieces there is no such question, since at most | |
1890 | one match at a given position is possible. This section describes the | |
1891 | notion of better/worse for combining operators. In the description | |
1892 | below C<S> and C<T> are regular subexpressions. | |
1893 | ||
13a2d996 | 1894 | =over 4 |
35a734be IZ |
1895 | |
1896 | =item C<ST> | |
1897 | ||
1898 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are | |
1899 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings | |
5d458dd8 | 1900 | which can be matched by C<T>. |
35a734be IZ |
1901 | |
1902 | If C<A> is better match for C<S> than C<A'>, C<AB> is a better | |
1903 | match than C<A'B'>. | |
1904 | ||
1905 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if | |
1906 | C<B> is better match for C<T> than C<B'>. | |
1907 | ||
1908 | =item C<S|T> | |
1909 | ||
1910 | When C<S> can match, it is a better match than when only C<T> can match. | |
1911 | ||
1912 | Ordering of two matches for C<S> is the same as for C<S>. Similar for | |
1913 | two matches for C<T>. | |
1914 | ||
1915 | =item C<S{REPEAT_COUNT}> | |
1916 | ||
1917 | Matches as C<SSS...S> (repeated as many times as necessary). | |
1918 | ||
1919 | =item C<S{min,max}> | |
1920 | ||
1921 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
1922 | ||
1923 | =item C<S{min,max}?> | |
1924 | ||
1925 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
1926 | ||
1927 | =item C<S?>, C<S*>, C<S+> | |
1928 | ||
1929 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
1930 | ||
1931 | =item C<S??>, C<S*?>, C<S+?> | |
1932 | ||
1933 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
1934 | ||
c47ff5f1 | 1935 | =item C<< (?>S) >> |
35a734be IZ |
1936 | |
1937 | Matches the best match for C<S> and only that. | |
1938 | ||
1939 | =item C<(?=S)>, C<(?<=S)> | |
1940 | ||
1941 | Only the best match for C<S> is considered. (This is important only if | |
1942 | C<S> has capturing parentheses, and backreferences are used somewhere | |
1943 | else in the whole regular expression.) | |
1944 | ||
1945 | =item C<(?!S)>, C<(?<!S)> | |
1946 | ||
1947 | For this grouping operator there is no need to describe the ordering, since | |
1948 | only whether or not C<S> can match is important. | |
1949 | ||
6bda09f9 | 1950 | =item C<(??{ EXPR })>, C<(?PARNO)> |
35a734be IZ |
1951 | |
1952 | The ordering is the same as for the regular expression which is | |
6bda09f9 | 1953 | the result of EXPR, or the pattern contained by capture buffer PARNO. |
35a734be IZ |
1954 | |
1955 | =item C<(?(condition)yes-pattern|no-pattern)> | |
1956 | ||
1957 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is | |
1958 | already determined. The ordering of the matches is the same as for the | |
1959 | chosen subexpression. | |
1960 | ||
1961 | =back | |
1962 | ||
1963 | The above recipes describe the ordering of matches I<at a given position>. | |
1964 | One more rule is needed to understand how a match is determined for the | |
1965 | whole regular expression: a match at an earlier position is always better | |
1966 | than a match at a later position. | |
1967 | ||
0d017f4d | 1968 | =head2 Creating Custom RE Engines |
c84d73f1 IZ |
1969 | |
1970 | Overloaded constants (see L<overload>) provide a simple way to extend | |
1971 | the functionality of the RE engine. | |
1972 | ||
1973 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
0d017f4d | 1974 | matches at a boundary between whitespace characters and non-whitespace |
c84d73f1 IZ |
1975 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
1976 | at these positions, so we want to have each C<\Y|> in the place of the | |
1977 | more complicated version. We can create a module C<customre> to do | |
1978 | this: | |
1979 | ||
1980 | package customre; | |
1981 | use overload; | |
1982 | ||
1983 | sub import { | |
1984 | shift; | |
1985 | die "No argument to customre::import allowed" if @_; | |
1986 | overload::constant 'qr' => \&convert; | |
1987 | } | |
1988 | ||
1989 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
1990 | ||
580a9fe1 RGS |
1991 | # We must also take care of not escaping the legitimate \\Y| |
1992 | # sequence, hence the presence of '\\' in the conversion rules. | |
5d458dd8 | 1993 | my %rules = ( '\\' => '\\\\', |
c84d73f1 IZ |
1994 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
1995 | sub convert { | |
1996 | my $re = shift; | |
5d458dd8 | 1997 | $re =~ s{ |
c84d73f1 IZ |
1998 | \\ ( \\ | Y . ) |
1999 | } | |
5d458dd8 | 2000 | { $rules{$1} or invalid($re,$1) }sgex; |
c84d73f1 IZ |
2001 | return $re; |
2002 | } | |
2003 | ||
2004 | Now C<use customre> enables the new escape in constant regular | |
2005 | expressions, i.e., those without any runtime variable interpolations. | |
2006 | As documented in L<overload>, this conversion will work only over | |
2007 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
2008 | part of this regular expression needs to be converted explicitly | |
2009 | (but only if the special meaning of C<\Y|> should be enabled inside $re): | |
2010 | ||
2011 | use customre; | |
2012 | $re = <>; | |
2013 | chomp $re; | |
2014 | $re = customre::convert $re; | |
2015 | /\Y|$re\Y|/; | |
2016 | ||
1f1031fe YO |
2017 | =head1 PCRE/Python Support |
2018 | ||
99d59c4d | 2019 | As of Perl 5.10.0, Perl supports several Python/PCRE specific extensions |
1f1031fe | 2020 | to the regex syntax. While Perl programmers are encouraged to use the |
99d59c4d | 2021 | Perl specific syntax, the following are also accepted: |
1f1031fe YO |
2022 | |
2023 | =over 4 | |
2024 | ||
ae5648b3 | 2025 | =item C<< (?PE<lt>NAMEE<gt>pattern) >> |
1f1031fe YO |
2026 | |
2027 | Define a named capture buffer. Equivalent to C<< (?<NAME>pattern) >>. | |
2028 | ||
2029 | =item C<< (?P=NAME) >> | |
2030 | ||
2031 | Backreference to a named capture buffer. Equivalent to C<< \g{NAME} >>. | |
2032 | ||
2033 | =item C<< (?P>NAME) >> | |
2034 | ||
2035 | Subroutine call to a named capture buffer. Equivalent to C<< (?&NAME) >>. | |
2036 | ||
ee9b8eae | 2037 | =back |
1f1031fe | 2038 | |
19799a22 GS |
2039 | =head1 BUGS |
2040 | ||
78288b8e KW |
2041 | There are numerous problems with case insensitive matching of characters |
2042 | outside the ASCII range, especially with those whose folds are multiple | |
2043 | characters, such as ligatures like C<LATIN SMALL LIGATURE FF>. | |
2044 | ||
f253210b KW |
2045 | In a bracketed character class with case insensitive matching, ranges only work |
2046 | for ASCII characters. For example, | |
2047 | C<m/[\N{CYRILLIC CAPITAL LETTER A}-\N{CYRILLIC CAPITAL LETTER YA}]/i> | |
2048 | doesn't match all the Russian upper and lower case letters. | |
2049 | ||
88c9975e KW |
2050 | Many regular expression constructs don't work on EBCDIC platforms. |
2051 | ||
9da458fc IZ |
2052 | This document varies from difficult to understand to completely |
2053 | and utterly opaque. The wandering prose riddled with jargon is | |
2054 | hard to fathom in several places. | |
2055 | ||
2056 | This document needs a rewrite that separates the tutorial content | |
2057 | from the reference content. | |
19799a22 GS |
2058 | |
2059 | =head1 SEE ALSO | |
9fa51da4 | 2060 | |
91e0c79e MJD |
2061 | L<perlrequick>. |
2062 | ||
2063 | L<perlretut>. | |
2064 | ||
9b599b2a GS |
2065 | L<perlop/"Regexp Quote-Like Operators">. |
2066 | ||
1e66bd83 PP |
2067 | L<perlop/"Gory details of parsing quoted constructs">. |
2068 | ||
14218588 GS |
2069 | L<perlfaq6>. |
2070 | ||
9b599b2a GS |
2071 | L<perlfunc/pos>. |
2072 | ||
2073 | L<perllocale>. | |
2074 | ||
fb55449c JH |
2075 | L<perlebcdic>. |
2076 | ||
14218588 GS |
2077 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
2078 | by O'Reilly and Associates. |