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 | |
f793d64a KW |
127 | \ Quote the next metacharacter |
128 | ^ Match the beginning of the line | |
129 | . Match any character (except newline) | |
130 | $ Match the end of the line (or before newline at the end) | |
131 | | Alternation | |
132 | () Grouping | |
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 | 157 | |
f793d64a KW |
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 | |
a0d0e21e LW |
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 | |
f793d64a KW |
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 | ||
f793d64a KW |
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 | |
198 | {n}+ Match exactly n times and give nothing back (redundant) | |
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 | ||
f793d64a KW |
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) | |
f793d64a | 232 | \cK control char (example: VT) |
dc0d9c48 | 233 | \x{}, \x00 character whose ordinal is the given hexadecimal number |
f793d64a KW |
234 | \N{name} named Unicode character |
235 | \N{U+263D} Unicode character (example: FIRST QUARTER MOON) | |
f0a2b745 | 236 | \o{}, \000 character whose ordinal is the given octal number |
f793d64a KW |
237 | \l lowercase next char (think vi) |
238 | \u uppercase next char (think vi) | |
239 | \L lowercase till \E (think vi) | |
240 | \U uppercase till \E (think vi) | |
241 | \Q quote (disable) pattern metacharacters till \E | |
242 | \E end either case modification or quoted section, think vi | |
a0d0e21e | 243 | |
9bb1f947 | 244 | Details are in L<perlop/Quote and Quote-like Operators>. |
1d2dff63 | 245 | |
e1d1eefb | 246 | =head3 Character Classes and other Special Escapes |
04838cea | 247 | |
a0d0e21e | 248 | In addition, Perl defines the following: |
d0b16107 | 249 | X<\g> X<\k> X<\K> X<backreference> |
a0d0e21e | 250 | |
f793d64a KW |
251 | Sequence Note Description |
252 | [...] [1] Match a character according to the rules of the | |
253 | bracketed character class defined by the "...". | |
254 | Example: [a-z] matches "a" or "b" or "c" ... or "z" | |
255 | [[:...:]] [2] Match a character according to the rules of the POSIX | |
256 | character class "..." within the outer bracketed | |
257 | character class. Example: [[:upper:]] matches any | |
258 | uppercase character. | |
259 | \w [3] Match a "word" character (alphanumeric plus "_") | |
260 | \W [3] Match a non-"word" character | |
261 | \s [3] Match a whitespace character | |
262 | \S [3] Match a non-whitespace character | |
263 | \d [3] Match a decimal digit character | |
264 | \D [3] Match a non-digit character | |
265 | \pP [3] Match P, named property. Use \p{Prop} for longer names | |
266 | \PP [3] Match non-P | |
267 | \X [4] Match Unicode "eXtended grapheme cluster" | |
268 | \C Match a single C-language char (octet) even if that is | |
269 | part of a larger UTF-8 character. Thus it breaks up | |
270 | characters into their UTF-8 bytes, so you may end up | |
271 | with malformed pieces of UTF-8. Unsupported in | |
272 | lookbehind. | |
c27a5cfe | 273 | \1 [5] Backreference to a specific capture group or buffer. |
f793d64a KW |
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 | |
c27a5cfe | 277 | previous group and may optionally be wrapped in |
f793d64a KW |
278 | curly brackets for 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 | |
283 | /s 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 | |
c27a5cfe | 310 | See L</Capture groups> 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 | |
e761bb84 CO |
366 | my $string = 'ABC'; |
367 | pos($string) = 1; | |
368 | while ($string =~ /(.\G)/g) { | |
369 | print $1; | |
370 | } | |
58e23c8d YO |
371 | |
372 | It will print 'A' and then terminate, as it considers the match to | |
373 | be zero-width, and thus will not match at the same position twice in a | |
374 | row. | |
375 | ||
376 | It is worth noting that C<\G> improperly used can result in an infinite | |
377 | loop. Take care when using patterns that include C<\G> in an alternation. | |
378 | ||
c27a5cfe | 379 | =head3 Capture groups |
04838cea | 380 | |
c27a5cfe KW |
381 | The bracketing construct C<( ... )> creates capture groups (also referred to as |
382 | capture buffers). To refer to the current contents of a group later on, within | |
d8b950dc KW |
383 | the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>) |
384 | for the second, and so on. | |
385 | This is called a I<backreference>. | |
d74e8afc | 386 | X<regex, capture buffer> X<regexp, capture buffer> |
c27a5cfe | 387 | X<regex, capture group> X<regexp, capture group> |
d74e8afc | 388 | X<regular expression, capture buffer> X<backreference> |
c27a5cfe | 389 | X<regular expression, capture group> X<backreference> |
1f1031fe | 390 | X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference> |
d8b950dc KW |
391 | X<named capture buffer> X<regular expression, named capture buffer> |
392 | X<named capture group> X<regular expression, named capture group> | |
393 | X<%+> X<$+{name}> X<< \k<name> >> | |
394 | There is no limit to the number of captured substrings that you may use. | |
395 | Groups are numbered with the leftmost open parenthesis being number 1, etc. If | |
396 | a group did not match, the associated backreference won't match either. (This | |
397 | can happen if the group is optional, or in a different branch of an | |
398 | alternation.) | |
399 | You can omit the C<"g">, and write C<"\1">, etc, but there are some issues with | |
400 | this form, described below. | |
401 | ||
402 | You can also refer to capture groups relatively, by using a negative number, so | |
403 | that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture | |
404 | group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For | |
405 | example: | |
5624f11d YO |
406 | |
407 | / | |
c27a5cfe KW |
408 | (Y) # group 1 |
409 | ( # group 2 | |
410 | (X) # group 3 | |
411 | \g{-1} # backref to group 3 | |
412 | \g{-3} # backref to group 1 | |
5624f11d YO |
413 | ) |
414 | /x | |
415 | ||
d8b950dc KW |
416 | would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to |
417 | interpolate regexes into larger regexes and not have to worry about the | |
418 | capture groups being renumbered. | |
419 | ||
420 | You can dispense with numbers altogether and create named capture groups. | |
421 | The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to | |
422 | reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may | |
423 | also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.) | |
424 | I<name> must not begin with a number, nor contain hyphens. | |
425 | When different groups within the same pattern have the same name, any reference | |
426 | to that name assumes the leftmost defined group. Named groups count in | |
427 | absolute and relative numbering, and so can also be referred to by those | |
428 | numbers. | |
429 | (It's possible to do things with named capture groups that would otherwise | |
430 | require C<(??{})>.) | |
431 | ||
432 | Capture group contents are dynamically scoped and available to you outside the | |
433 | pattern until the end of the enclosing block or until the next successful | |
434 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
435 | You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">, | |
436 | etc); or by name via the C<%+> hash, using C<"$+{I<name>}">. | |
437 | ||
438 | Braces are required in referring to named capture groups, but are optional for | |
439 | absolute or relative numbered ones. Braces are safer when creating a regex by | |
440 | concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a> | |
441 | contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which | |
442 | is probably not what you intended. | |
443 | ||
444 | The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that | |
445 | there were no named nor relative numbered capture groups. Absolute numbered | |
446 | groups were referred to using C<\1>, C<\2>, etc, and this notation is still | |
447 | accepted (and likely always will be). But it leads to some ambiguities if | |
448 | there are more than 9 capture groups, as C<\10> could mean either the tenth | |
449 | capture group, or the character whose ordinal in octal is 010 (a backspace in | |
450 | ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference | |
451 | only if at least 10 left parentheses have opened before it. Likewise C<\11> is | |
452 | a backreference only if at least 11 left parentheses have opened before it. | |
e1f120a9 KW |
453 | And so on. C<\1> through C<\9> are always interpreted as backreferences. |
454 | There are several examples below that illustrate these perils. You can avoid | |
455 | the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups; | |
456 | and for octal constants always using C<\o{}>, or for C<\077> and below, using 3 | |
457 | digits padded with leading zeros, since a leading zero implies an octal | |
458 | constant. | |
d8b950dc KW |
459 | |
460 | The C<\I<digit>> notation also works in certain circumstances outside | |
461 | the pattern. See L</Warning on \1 Instead of $1> below for details.) | |
81714fb9 | 462 | |
14218588 | 463 | Examples: |
a0d0e21e LW |
464 | |
465 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
466 | ||
d8b950dc | 467 | /(.)\g1/ # find first doubled char |
81714fb9 YO |
468 | and print "'$1' is the first doubled character\n"; |
469 | ||
470 | /(?<char>.)\k<char>/ # ... a different way | |
471 | and print "'$+{char}' is the first doubled character\n"; | |
472 | ||
d8b950dc | 473 | /(?'char'.)\g1/ # ... mix and match |
81714fb9 | 474 | and print "'$1' is the first doubled character\n"; |
c47ff5f1 | 475 | |
14218588 | 476 | if (/Time: (..):(..):(..)/) { # parse out values |
f793d64a KW |
477 | $hours = $1; |
478 | $minutes = $2; | |
479 | $seconds = $3; | |
a0d0e21e | 480 | } |
c47ff5f1 | 481 | |
9d860678 KW |
482 | /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference |
483 | /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal | |
484 | /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference | |
485 | /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal | |
486 | ||
487 | $a = '(.)\1'; # Creates problems when concatenated. | |
488 | $b = '(.)\g{1}'; # Avoids the problems. | |
489 | "aa" =~ /${a}/; # True | |
490 | "aa" =~ /${b}/; # True | |
491 | "aa0" =~ /${a}0/; # False! | |
492 | "aa0" =~ /${b}0/; # True | |
dc0d9c48 KW |
493 | "aa\x08" =~ /${a}0/; # True! |
494 | "aa\x08" =~ /${b}0/; # False | |
9d860678 | 495 | |
14218588 GS |
496 | Several special variables also refer back to portions of the previous |
497 | match. C<$+> returns whatever the last bracket match matched. | |
498 | C<$&> returns the entire matched string. (At one point C<$0> did | |
499 | also, but now it returns the name of the program.) C<$`> returns | |
77ea4f6d JV |
500 | everything before the matched string. C<$'> returns everything |
501 | after the matched string. And C<$^N> contains whatever was matched by | |
502 | the most-recently closed group (submatch). C<$^N> can be used in | |
503 | extended patterns (see below), for example to assign a submatch to a | |
81714fb9 | 504 | variable. |
d74e8afc | 505 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 | 506 | |
d8b950dc KW |
507 | These special variables, like the C<%+> hash and the numbered match variables |
508 | (C<$1>, C<$2>, C<$3>, etc.) are dynamically scoped | |
14218588 GS |
509 | until the end of the enclosing block or until the next successful |
510 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
d74e8afc ITB |
511 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
512 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> | |
513 | ||
0d017f4d | 514 | B<NOTE>: Failed matches in Perl do not reset the match variables, |
5146ce24 | 515 | which makes it easier to write code that tests for a series of more |
665e98b9 JH |
516 | specific cases and remembers the best match. |
517 | ||
14218588 GS |
518 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or |
519 | C<$'> anywhere in the program, it has to provide them for every | |
520 | pattern match. This may substantially slow your program. Perl | |
d8b950dc | 521 | uses the same mechanism to produce C<$1>, C<$2>, etc, so you also pay a |
14218588 GS |
522 | price for each pattern that contains capturing parentheses. (To |
523 | avoid this cost while retaining the grouping behaviour, use the | |
524 | extended regular expression C<(?: ... )> instead.) But if you never | |
525 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
526 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
527 | if you can, but if you can't (and some algorithms really appreciate | |
528 | them), once you've used them once, use them at will, because you've | |
529 | already paid the price. As of 5.005, C<$&> is not so costly as the | |
530 | other two. | |
d74e8afc | 531 | X<$&> X<$`> X<$'> |
68dc0745 | 532 | |
99d59c4d | 533 | As a workaround for this problem, Perl 5.10.0 introduces C<${^PREMATCH}>, |
cde0cee5 YO |
534 | C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&> |
535 | and C<$'>, B<except> that they are only guaranteed to be defined after a | |
87e95b7f | 536 | successful match that was executed with the C</p> (preserve) modifier. |
cde0cee5 YO |
537 | The use of these variables incurs no global performance penalty, unlike |
538 | their punctuation char equivalents, however at the trade-off that you | |
539 | have to tell perl when you want to use them. | |
87e95b7f | 540 | X</p> X<p modifier> |
cde0cee5 | 541 | |
9d727203 KW |
542 | =head2 Quoting metacharacters |
543 | ||
19799a22 GS |
544 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
545 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
546 | are no backslashed symbols that aren't alphanumeric. So anything | |
c47ff5f1 | 547 | that looks like \\, \(, \), \<, \>, \{, or \} is always |
19799a22 GS |
548 | interpreted as a literal character, not a metacharacter. This was |
549 | once used in a common idiom to disable or quote the special meanings | |
550 | of regular expression metacharacters in a string that you want to | |
36bbe248 | 551 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e LW |
552 | |
553 | $pattern =~ s/(\W)/\\$1/g; | |
554 | ||
f1cbbd6e | 555 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 GS |
556 | Today it is more common to use the quotemeta() function or the C<\Q> |
557 | metaquoting escape sequence to disable all metacharacters' special | |
558 | meanings like this: | |
a0d0e21e LW |
559 | |
560 | /$unquoted\Q$quoted\E$unquoted/ | |
561 | ||
9da458fc IZ |
562 | Beware that if you put literal backslashes (those not inside |
563 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
564 | backslash interpolation may lead to confusing results. If you | |
565 | I<need> to use literal backslashes within C<\Q...\E>, | |
566 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
567 | ||
19799a22 GS |
568 | =head2 Extended Patterns |
569 | ||
14218588 GS |
570 | Perl also defines a consistent extension syntax for features not |
571 | found in standard tools like B<awk> and B<lex>. The syntax is a | |
572 | pair of parentheses with a question mark as the first thing within | |
573 | the parentheses. The character after the question mark indicates | |
574 | the extension. | |
19799a22 | 575 | |
14218588 GS |
576 | The stability of these extensions varies widely. Some have been |
577 | part of the core language for many years. Others are experimental | |
578 | and may change without warning or be completely removed. Check | |
579 | the documentation on an individual feature to verify its current | |
580 | status. | |
19799a22 | 581 | |
14218588 GS |
582 | A question mark was chosen for this and for the minimal-matching |
583 | construct because 1) question marks are rare in older regular | |
584 | expressions, and 2) whenever you see one, you should stop and | |
585 | "question" exactly what is going on. That's psychology... | |
a0d0e21e LW |
586 | |
587 | =over 10 | |
588 | ||
cc6b7395 | 589 | =item C<(?#text)> |
d74e8afc | 590 | X<(?#)> |
a0d0e21e | 591 | |
14218588 | 592 | A comment. The text is ignored. If the C</x> modifier enables |
19799a22 | 593 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes |
259138e3 GS |
594 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
595 | C<)> in the comment. | |
a0d0e21e | 596 | |
f7819f85 | 597 | =item C<(?pimsx-imsx)> |
d74e8afc | 598 | X<(?)> |
19799a22 | 599 | |
0b6d1084 JH |
600 | One or more embedded pattern-match modifiers, to be turned on (or |
601 | turned off, if preceded by C<->) for the remainder of the pattern or | |
602 | the remainder of the enclosing pattern group (if any). This is | |
603 | particularly useful for dynamic patterns, such as those read in from a | |
0d017f4d WL |
604 | configuration file, taken from an argument, or specified in a table |
605 | somewhere. Consider the case where some patterns want to be case | |
606 | sensitive and some do not: The case insensitive ones merely need to | |
607 | include C<(?i)> at the front of the pattern. For example: | |
19799a22 GS |
608 | |
609 | $pattern = "foobar"; | |
5d458dd8 | 610 | if ( /$pattern/i ) { } |
19799a22 GS |
611 | |
612 | # more flexible: | |
613 | ||
614 | $pattern = "(?i)foobar"; | |
5d458dd8 | 615 | if ( /$pattern/ ) { } |
19799a22 | 616 | |
0b6d1084 | 617 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 | 618 | |
d8b950dc | 619 | ( (?i) blah ) \s+ \g1 |
19799a22 | 620 | |
0d017f4d WL |
621 | will match C<blah> in any case, some spaces, and an exact (I<including the case>!) |
622 | repetition of the previous word, assuming the C</x> modifier, and no C</i> | |
623 | modifier outside this group. | |
19799a22 | 624 | |
8eb5594e DR |
625 | These modifiers do not carry over into named subpatterns called in the |
626 | enclosing group. In other words, a pattern such as C<((?i)(&NAME))> does not | |
627 | change the case-sensitivity of the "NAME" pattern. | |
628 | ||
5530442b | 629 | Note that the C<p> modifier is special in that it can only be enabled, |
cde0cee5 | 630 | not disabled, and that its presence anywhere in a pattern has a global |
5530442b | 631 | effect. Thus C<(?-p)> and C<(?-p:...)> are meaningless and will warn |
cde0cee5 YO |
632 | when executed under C<use warnings>. |
633 | ||
5a964f20 | 634 | =item C<(?:pattern)> |
d74e8afc | 635 | X<(?:)> |
a0d0e21e | 636 | |
ca9dfc88 IZ |
637 | =item C<(?imsx-imsx:pattern)> |
638 | ||
5a964f20 TC |
639 | This is for clustering, not capturing; it groups subexpressions like |
640 | "()", but doesn't make backreferences as "()" does. So | |
a0d0e21e | 641 | |
5a964f20 | 642 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e LW |
643 | |
644 | is like | |
645 | ||
5a964f20 | 646 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e | 647 | |
19799a22 GS |
648 | but doesn't spit out extra fields. It's also cheaper not to capture |
649 | characters if you don't need to. | |
a0d0e21e | 650 | |
19799a22 | 651 | Any letters between C<?> and C<:> act as flags modifiers as with |
5d458dd8 | 652 | C<(?imsx-imsx)>. For example, |
ca9dfc88 IZ |
653 | |
654 | /(?s-i:more.*than).*million/i | |
655 | ||
14218588 | 656 | is equivalent to the more verbose |
ca9dfc88 IZ |
657 | |
658 | /(?:(?s-i)more.*than).*million/i | |
659 | ||
594d7033 YO |
660 | =item C<(?|pattern)> |
661 | X<(?|)> X<Branch reset> | |
662 | ||
663 | This is the "branch reset" pattern, which has the special property | |
c27a5cfe | 664 | that the capture groups are numbered from the same starting point |
99d59c4d | 665 | in each alternation branch. It is available starting from perl 5.10.0. |
4deaaa80 | 666 | |
c27a5cfe | 667 | Capture groups are numbered from left to right, but inside this |
693596a8 | 668 | construct the numbering is restarted for each branch. |
4deaaa80 | 669 | |
c27a5cfe | 670 | The numbering within each branch will be as normal, and any groups |
4deaaa80 PJ |
671 | following this construct will be numbered as though the construct |
672 | contained only one branch, that being the one with the most capture | |
c27a5cfe | 673 | groups in it. |
4deaaa80 PJ |
674 | |
675 | This construct will be useful when you want to capture one of a | |
676 | number of alternative matches. | |
677 | ||
678 | Consider the following pattern. The numbers underneath show in | |
c27a5cfe | 679 | which group the captured content will be stored. |
594d7033 YO |
680 | |
681 | ||
682 | # before ---------------branch-reset----------- after | |
683 | / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x | |
684 | # 1 2 2 3 2 3 4 | |
685 | ||
ab106183 A |
686 | Be careful when using the branch reset pattern in combination with |
687 | named captures. Named captures are implemented as being aliases to | |
c27a5cfe | 688 | numbered groups holding the captures, and that interferes with the |
ab106183 A |
689 | implementation of the branch reset pattern. If you are using named |
690 | captures in a branch reset pattern, it's best to use the same names, | |
691 | in the same order, in each of the alternations: | |
692 | ||
693 | /(?| (?<a> x ) (?<b> y ) | |
694 | | (?<a> z ) (?<b> w )) /x | |
695 | ||
696 | Not doing so may lead to surprises: | |
697 | ||
698 | "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x; | |
699 | say $+ {a}; # Prints '12' | |
700 | say $+ {b}; # *Also* prints '12'. | |
701 | ||
c27a5cfe KW |
702 | The problem here is that both the group named C<< a >> and the group |
703 | named C<< b >> are aliases for the group belonging to C<< $1 >>. | |
90a18110 | 704 | |
ee9b8eae YO |
705 | =item Look-Around Assertions |
706 | X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround> | |
707 | ||
708 | Look-around assertions are zero width patterns which match a specific | |
709 | pattern without including it in C<$&>. Positive assertions match when | |
710 | their subpattern matches, negative assertions match when their subpattern | |
711 | fails. Look-behind matches text up to the current match position, | |
712 | look-ahead matches text following the current match position. | |
713 | ||
714 | =over 4 | |
715 | ||
5a964f20 | 716 | =item C<(?=pattern)> |
d74e8afc | 717 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> |
a0d0e21e | 718 | |
19799a22 | 719 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e LW |
720 | matches a word followed by a tab, without including the tab in C<$&>. |
721 | ||
5a964f20 | 722 | =item C<(?!pattern)> |
d74e8afc | 723 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> |
a0d0e21e | 724 | |
19799a22 | 725 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e | 726 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 GS |
727 | however that look-ahead and look-behind are NOT the same thing. You cannot |
728 | use this for look-behind. | |
7b8d334a | 729 | |
5a964f20 | 730 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a GS |
731 | will not do what you want. That's because the C<(?!foo)> is just saying that |
732 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
733 | match. You would have to do something like C</(?!foo)...bar/> for that. We | |
734 | say "like" because there's the case of your "bar" not having three characters | |
735 | before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>. | |
736 | Sometimes it's still easier just to say: | |
a0d0e21e | 737 | |
a3cb178b | 738 | if (/bar/ && $` !~ /foo$/) |
a0d0e21e | 739 | |
19799a22 | 740 | For look-behind see below. |
c277df42 | 741 | |
ee9b8eae YO |
742 | =item C<(?<=pattern)> C<\K> |
743 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K> | |
c277df42 | 744 | |
c47ff5f1 | 745 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 GS |
746 | matches a word that follows a tab, without including the tab in C<$&>. |
747 | Works only for fixed-width look-behind. | |
c277df42 | 748 | |
ee9b8eae YO |
749 | There is a special form of this construct, called C<\K>, which causes the |
750 | regex engine to "keep" everything it had matched prior to the C<\K> and | |
751 | not include it in C<$&>. This effectively provides variable length | |
752 | look-behind. The use of C<\K> inside of another look-around assertion | |
753 | is allowed, but the behaviour is currently not well defined. | |
754 | ||
c62285ac | 755 | For various reasons C<\K> may be significantly more efficient than the |
ee9b8eae YO |
756 | equivalent C<< (?<=...) >> construct, and it is especially useful in |
757 | situations where you want to efficiently remove something following | |
758 | something else in a string. For instance | |
759 | ||
760 | s/(foo)bar/$1/g; | |
761 | ||
762 | can be rewritten as the much more efficient | |
763 | ||
764 | s/foo\Kbar//g; | |
765 | ||
5a964f20 | 766 | =item C<(?<!pattern)> |
d74e8afc | 767 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> |
c277df42 | 768 | |
19799a22 GS |
769 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
770 | matches any occurrence of "foo" that does not follow "bar". Works | |
771 | only for fixed-width look-behind. | |
c277df42 | 772 | |
ee9b8eae YO |
773 | =back |
774 | ||
81714fb9 YO |
775 | =item C<(?'NAME'pattern)> |
776 | ||
777 | =item C<< (?<NAME>pattern) >> | |
778 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> | |
779 | ||
c27a5cfe | 780 | A named capture group. Identical in every respect to normal capturing |
90a18110 | 781 | parentheses C<()> but for the additional fact that C<%+> or C<%-> may be |
c27a5cfe | 782 | used after a successful match to refer to a named group. See C<perlvar> |
90a18110 | 783 | for more details on the C<%+> and C<%-> hashes. |
81714fb9 | 784 | |
c27a5cfe KW |
785 | If multiple distinct capture groups have the same name then the |
786 | $+{NAME} will refer to the leftmost defined group in the match. | |
81714fb9 | 787 | |
0d017f4d | 788 | The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent. |
81714fb9 YO |
789 | |
790 | B<NOTE:> While the notation of this construct is the same as the similar | |
c27a5cfe | 791 | function in .NET regexes, the behavior is not. In Perl the groups are |
81714fb9 YO |
792 | numbered sequentially regardless of being named or not. Thus in the |
793 | pattern | |
794 | ||
795 | /(x)(?<foo>y)(z)/ | |
796 | ||
797 | $+{foo} will be the same as $2, and $3 will contain 'z' instead of | |
798 | the opposite which is what a .NET regex hacker might expect. | |
799 | ||
1f1031fe YO |
800 | Currently NAME is restricted to simple identifiers only. |
801 | In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or | |
802 | its Unicode extension (see L<utf8>), | |
803 | though it isn't extended by the locale (see L<perllocale>). | |
81714fb9 | 804 | |
1f1031fe | 805 | B<NOTE:> In order to make things easier for programmers with experience |
ae5648b3 | 806 | with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >> |
0d017f4d | 807 | may be used instead of C<< (?<NAME>pattern) >>; however this form does not |
64c5a566 | 808 | support the use of single quotes as a delimiter for the name. |
81714fb9 | 809 | |
1f1031fe YO |
810 | =item C<< \k<NAME> >> |
811 | ||
812 | =item C<< \k'NAME' >> | |
81714fb9 YO |
813 | |
814 | Named backreference. Similar to numeric backreferences, except that | |
815 | the group is designated by name and not number. If multiple groups | |
816 | have the same name then it refers to the leftmost defined group in | |
817 | the current match. | |
818 | ||
0d017f4d | 819 | It is an error to refer to a name not defined by a C<< (?<NAME>) >> |
81714fb9 YO |
820 | earlier in the pattern. |
821 | ||
822 | Both forms are equivalent. | |
823 | ||
1f1031fe | 824 | B<NOTE:> In order to make things easier for programmers with experience |
0d017f4d | 825 | with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >> |
64c5a566 | 826 | may be used instead of C<< \k<NAME> >>. |
1f1031fe | 827 | |
cc6b7395 | 828 | =item C<(?{ code })> |
d74e8afc | 829 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 | 830 | |
19799a22 | 831 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
832 | experimental, and may be changed without notice. Code executed that |
833 | has side effects may not perform identically from version to version | |
834 | due to the effect of future optimisations in the regex engine. | |
c277df42 | 835 | |
cc46d5f2 | 836 | This zero-width assertion evaluates any embedded Perl code. It |
19799a22 GS |
837 | always succeeds, and its C<code> is not interpolated. Currently, |
838 | the rules to determine where the C<code> ends are somewhat convoluted. | |
839 | ||
77ea4f6d JV |
840 | This feature can be used together with the special variable C<$^N> to |
841 | capture the results of submatches in variables without having to keep | |
842 | track of the number of nested parentheses. For example: | |
843 | ||
844 | $_ = "The brown fox jumps over the lazy dog"; | |
845 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
846 | print "color = $color, animal = $animal\n"; | |
847 | ||
754091cb RGS |
848 | Inside the C<(?{...})> block, C<$_> refers to the string the regular |
849 | expression is matching against. You can also use C<pos()> to know what is | |
fa11829f | 850 | the current position of matching within this string. |
754091cb | 851 | |
19799a22 GS |
852 | The C<code> is properly scoped in the following sense: If the assertion |
853 | is backtracked (compare L<"Backtracking">), all changes introduced after | |
854 | C<local>ization are undone, so that | |
b9ac3b5b GS |
855 | |
856 | $_ = 'a' x 8; | |
5d458dd8 | 857 | m< |
f793d64a | 858 | (?{ $cnt = 0 }) # Initialize $cnt. |
b9ac3b5b | 859 | ( |
5d458dd8 | 860 | a |
b9ac3b5b | 861 | (?{ |
f793d64a | 862 | local $cnt = $cnt + 1; # Update $cnt, backtracking-safe. |
b9ac3b5b | 863 | }) |
5d458dd8 | 864 | )* |
b9ac3b5b | 865 | aaaa |
f793d64a KW |
866 | (?{ $res = $cnt }) # On success copy to |
867 | # non-localized location. | |
b9ac3b5b GS |
868 | >x; |
869 | ||
0d017f4d | 870 | will set C<$res = 4>. Note that after the match, C<$cnt> returns to the globally |
14218588 | 871 | introduced value, because the scopes that restrict C<local> operators |
b9ac3b5b GS |
872 | are unwound. |
873 | ||
19799a22 GS |
874 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> |
875 | switch. If I<not> used in this way, the result of evaluation of | |
876 | C<code> is put into the special variable C<$^R>. This happens | |
877 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions | |
878 | inside the same regular expression. | |
b9ac3b5b | 879 | |
19799a22 GS |
880 | The assignment to C<$^R> above is properly localized, so the old |
881 | value of C<$^R> is restored if the assertion is backtracked; compare | |
882 | L<"Backtracking">. | |
b9ac3b5b | 883 | |
19799a22 GS |
884 | For reasons of security, this construct is forbidden if the regular |
885 | expression involves run-time interpolation of variables, unless the | |
886 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
887 | variables contain results of C<qr//> operator (see | |
9bb1f947 | 888 | L<perlop/"qr/STRINGE<sol>msixpo">). |
871b0233 | 889 | |
0d017f4d | 890 | This restriction is due to the wide-spread and remarkably convenient |
19799a22 | 891 | custom of using run-time determined strings as patterns. For example: |
871b0233 IZ |
892 | |
893 | $re = <>; | |
894 | chomp $re; | |
895 | $string =~ /$re/; | |
896 | ||
14218588 GS |
897 | Before Perl knew how to execute interpolated code within a pattern, |
898 | this operation was completely safe from a security point of view, | |
899 | although it could raise an exception from an illegal pattern. If | |
900 | you turn on the C<use re 'eval'>, though, it is no longer secure, | |
901 | so you should only do so if you are also using taint checking. | |
902 | Better yet, use the carefully constrained evaluation within a Safe | |
cc46d5f2 | 903 | compartment. See L<perlsec> for details about both these mechanisms. |
871b0233 | 904 | |
e95d7314 GG |
905 | B<WARNING>: Use of lexical (C<my>) variables in these blocks is |
906 | broken. The result is unpredictable and will make perl unstable. The | |
907 | workaround is to use global (C<our>) variables. | |
908 | ||
909 | B<WARNING>: Because Perl's regex engine is currently not re-entrant, | |
910 | interpolated code may not invoke the regex engine either directly with | |
911 | C<m//> or C<s///>), or indirectly with functions such as | |
912 | C<split>. Invoking the regex engine in these blocks will make perl | |
913 | unstable. | |
8988a1bb | 914 | |
14455d6c | 915 | =item C<(??{ code })> |
d74e8afc ITB |
916 | X<(??{})> |
917 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> | |
0f5d15d6 | 918 | |
19799a22 | 919 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
920 | experimental, and may be changed without notice. Code executed that |
921 | has side effects may not perform identically from version to version | |
922 | due to the effect of future optimisations in the regex engine. | |
0f5d15d6 | 923 | |
19799a22 GS |
924 | This is a "postponed" regular subexpression. The C<code> is evaluated |
925 | at run time, at the moment this subexpression may match. The result | |
926 | of evaluation is considered as a regular expression and matched as | |
61528107 | 927 | if it were inserted instead of this construct. Note that this means |
c27a5cfe | 928 | that the contents of capture groups defined inside an eval'ed pattern |
6bda09f9 | 929 | are not available outside of the pattern, and vice versa, there is no |
c27a5cfe | 930 | way for the inner pattern to refer to a capture group defined outside. |
6bda09f9 YO |
931 | Thus, |
932 | ||
933 | ('a' x 100)=~/(??{'(.)' x 100})/ | |
934 | ||
81714fb9 | 935 | B<will> match, it will B<not> set $1. |
0f5d15d6 | 936 | |
428594d9 | 937 | The C<code> is not interpolated. As before, the rules to determine |
19799a22 GS |
938 | where the C<code> ends are currently somewhat convoluted. |
939 | ||
940 | The following pattern matches a parenthesized group: | |
0f5d15d6 IZ |
941 | |
942 | $re = qr{ | |
f793d64a KW |
943 | \( |
944 | (?: | |
945 | (?> [^()]+ ) # Non-parens without backtracking | |
946 | | | |
947 | (??{ $re }) # Group with matching parens | |
948 | )* | |
949 | \) | |
950 | }x; | |
0f5d15d6 | 951 | |
6bda09f9 YO |
952 | See also C<(?PARNO)> for a different, more efficient way to accomplish |
953 | the same task. | |
954 | ||
0b370c0a A |
955 | For reasons of security, this construct is forbidden if the regular |
956 | expression involves run-time interpolation of variables, unless the | |
957 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
958 | variables contain results of C<qr//> operator (see | |
9bb1f947 | 959 | L<perlop/"qrE<sol>STRINGE<sol>msixpo">). |
0b370c0a | 960 | |
5d458dd8 | 961 | Because perl's regex engine is not currently re-entrant, delayed |
8988a1bb DD |
962 | code may not invoke the regex engine either directly with C<m//> or C<s///>), |
963 | or indirectly with functions such as C<split>. | |
964 | ||
5d458dd8 YO |
965 | Recursing deeper than 50 times without consuming any input string will |
966 | result in a fatal error. The maximum depth is compiled into perl, so | |
6bda09f9 YO |
967 | changing it requires a custom build. |
968 | ||
542fa716 YO |
969 | =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)> |
970 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)> | |
6bda09f9 | 971 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
542fa716 | 972 | X<regex, relative recursion> |
6bda09f9 | 973 | |
81714fb9 | 974 | Similar to C<(??{ code })> except it does not involve compiling any code, |
c27a5cfe KW |
975 | instead it treats the contents of a capture group as an independent |
976 | pattern that must match at the current position. Capture groups | |
81714fb9 | 977 | contained by the pattern will have the value as determined by the |
6bda09f9 YO |
978 | outermost recursion. |
979 | ||
894be9b7 | 980 | PARNO is a sequence of digits (not starting with 0) whose value reflects |
c27a5cfe | 981 | the paren-number of the capture group to recurse to. C<(?R)> recurses to |
894be9b7 | 982 | the beginning of the whole pattern. C<(?0)> is an alternate syntax for |
542fa716 | 983 | C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed |
c27a5cfe | 984 | to be relative, with negative numbers indicating preceding capture groups |
542fa716 | 985 | and positive ones following. Thus C<(?-1)> refers to the most recently |
c27a5cfe | 986 | declared group, and C<(?+1)> indicates the next group to be declared. |
c74340f9 | 987 | Note that the counting for relative recursion differs from that of |
c27a5cfe | 988 | relative backreferences, in that with recursion unclosed groups B<are> |
c74340f9 | 989 | included. |
6bda09f9 | 990 | |
81714fb9 | 991 | The following pattern matches a function foo() which may contain |
f145b7e9 | 992 | balanced parentheses as the argument. |
6bda09f9 YO |
993 | |
994 | $re = qr{ ( # paren group 1 (full function) | |
81714fb9 | 995 | foo |
6bda09f9 YO |
996 | ( # paren group 2 (parens) |
997 | \( | |
998 | ( # paren group 3 (contents of parens) | |
999 | (?: | |
1000 | (?> [^()]+ ) # Non-parens without backtracking | |
1001 | | | |
1002 | (?2) # Recurse to start of paren group 2 | |
1003 | )* | |
1004 | ) | |
1005 | \) | |
1006 | ) | |
1007 | ) | |
1008 | }x; | |
1009 | ||
1010 | If the pattern was used as follows | |
1011 | ||
1012 | 'foo(bar(baz)+baz(bop))'=~/$re/ | |
1013 | and print "\$1 = $1\n", | |
1014 | "\$2 = $2\n", | |
1015 | "\$3 = $3\n"; | |
1016 | ||
1017 | the output produced should be the following: | |
1018 | ||
1019 | $1 = foo(bar(baz)+baz(bop)) | |
1020 | $2 = (bar(baz)+baz(bop)) | |
81714fb9 | 1021 | $3 = bar(baz)+baz(bop) |
6bda09f9 | 1022 | |
c27a5cfe | 1023 | If there is no corresponding capture group defined, then it is a |
61528107 | 1024 | fatal error. Recursing deeper than 50 times without consuming any input |
81714fb9 | 1025 | string will also result in a fatal error. The maximum depth is compiled |
6bda09f9 YO |
1026 | into perl, so changing it requires a custom build. |
1027 | ||
542fa716 YO |
1028 | The following shows how using negative indexing can make it |
1029 | easier to embed recursive patterns inside of a C<qr//> construct | |
1030 | for later use: | |
1031 | ||
1032 | my $parens = qr/(\((?:[^()]++|(?-1))*+\))/; | |
1033 | if (/foo $parens \s+ + \s+ bar $parens/x) { | |
1034 | # do something here... | |
1035 | } | |
1036 | ||
81714fb9 | 1037 | B<Note> that this pattern does not behave the same way as the equivalent |
0d017f4d | 1038 | PCRE or Python construct of the same form. In Perl you can backtrack into |
6bda09f9 | 1039 | a recursed group, in PCRE and Python the recursed into group is treated |
542fa716 YO |
1040 | as atomic. Also, modifiers are resolved at compile time, so constructs |
1041 | like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will | |
1042 | be processed. | |
6bda09f9 | 1043 | |
894be9b7 YO |
1044 | =item C<(?&NAME)> |
1045 | X<(?&NAME)> | |
1046 | ||
0d017f4d WL |
1047 | Recurse to a named subpattern. Identical to C<(?PARNO)> except that the |
1048 | parenthesis to recurse to is determined by name. If multiple parentheses have | |
894be9b7 YO |
1049 | the same name, then it recurses to the leftmost. |
1050 | ||
1051 | It is an error to refer to a name that is not declared somewhere in the | |
1052 | pattern. | |
1053 | ||
1f1031fe YO |
1054 | B<NOTE:> In order to make things easier for programmers with experience |
1055 | with the Python or PCRE regex engines the pattern C<< (?P>NAME) >> | |
64c5a566 | 1056 | may be used instead of C<< (?&NAME) >>. |
1f1031fe | 1057 | |
e2e6a0f1 YO |
1058 | =item C<(?(condition)yes-pattern|no-pattern)> |
1059 | X<(?()> | |
286f584a | 1060 | |
e2e6a0f1 | 1061 | =item C<(?(condition)yes-pattern)> |
286f584a | 1062 | |
e2e6a0f1 YO |
1063 | Conditional expression. C<(condition)> should be either an integer in |
1064 | parentheses (which is valid if the corresponding pair of parentheses | |
1065 | matched), a look-ahead/look-behind/evaluate zero-width assertion, a | |
c27a5cfe | 1066 | name in angle brackets or single quotes (which is valid if a group |
e2e6a0f1 YO |
1067 | with the given name matched), or the special symbol (R) (true when |
1068 | evaluated inside of recursion or eval). Additionally the R may be | |
1069 | followed by a number, (which will be true when evaluated when recursing | |
1070 | inside of the appropriate group), or by C<&NAME>, in which case it will | |
1071 | be true only when evaluated during recursion in the named group. | |
1072 | ||
1073 | Here's a summary of the possible predicates: | |
1074 | ||
1075 | =over 4 | |
1076 | ||
1077 | =item (1) (2) ... | |
1078 | ||
c27a5cfe | 1079 | Checks if the numbered capturing group has matched something. |
e2e6a0f1 YO |
1080 | |
1081 | =item (<NAME>) ('NAME') | |
1082 | ||
c27a5cfe | 1083 | Checks if a group with the given name has matched something. |
e2e6a0f1 YO |
1084 | |
1085 | =item (?{ CODE }) | |
1086 | ||
1087 | Treats the code block as the condition. | |
1088 | ||
1089 | =item (R) | |
1090 | ||
1091 | Checks if the expression has been evaluated inside of recursion. | |
1092 | ||
1093 | =item (R1) (R2) ... | |
1094 | ||
1095 | Checks if the expression has been evaluated while executing directly | |
1096 | inside of the n-th capture group. This check is the regex equivalent of | |
1097 | ||
1098 | if ((caller(0))[3] eq 'subname') { ... } | |
1099 | ||
1100 | In other words, it does not check the full recursion stack. | |
1101 | ||
1102 | =item (R&NAME) | |
1103 | ||
1104 | Similar to C<(R1)>, this predicate checks to see if we're executing | |
1105 | directly inside of the leftmost group with a given name (this is the same | |
1106 | logic used by C<(?&NAME)> to disambiguate). It does not check the full | |
1107 | stack, but only the name of the innermost active recursion. | |
1108 | ||
1109 | =item (DEFINE) | |
1110 | ||
1111 | In this case, the yes-pattern is never directly executed, and no | |
1112 | no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient. | |
1113 | See below for details. | |
1114 | ||
1115 | =back | |
1116 | ||
1117 | For example: | |
1118 | ||
1119 | m{ ( \( )? | |
1120 | [^()]+ | |
1121 | (?(1) \) ) | |
1122 | }x | |
1123 | ||
1124 | matches a chunk of non-parentheses, possibly included in parentheses | |
1125 | themselves. | |
1126 | ||
1127 | A special form is the C<(DEFINE)> predicate, which never executes directly | |
1128 | its yes-pattern, and does not allow a no-pattern. This allows to define | |
1129 | subpatterns which will be executed only by using the recursion mechanism. | |
1130 | This way, you can define a set of regular expression rules that can be | |
1131 | bundled into any pattern you choose. | |
1132 | ||
1133 | It is recommended that for this usage you put the DEFINE block at the | |
1134 | end of the pattern, and that you name any subpatterns defined within it. | |
1135 | ||
1136 | Also, it's worth noting that patterns defined this way probably will | |
1137 | not be as efficient, as the optimiser is not very clever about | |
1138 | handling them. | |
1139 | ||
1140 | An example of how this might be used is as follows: | |
1141 | ||
2bf803e2 | 1142 | /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT)) |
e2e6a0f1 | 1143 | (?(DEFINE) |
2bf803e2 YO |
1144 | (?<NAME_PAT>....) |
1145 | (?<ADRESS_PAT>....) | |
e2e6a0f1 YO |
1146 | )/x |
1147 | ||
c27a5cfe KW |
1148 | Note that capture groups matched inside of recursion are not accessible |
1149 | after the recursion returns, so the extra layer of capturing groups is | |
e2e6a0f1 YO |
1150 | necessary. Thus C<$+{NAME_PAT}> would not be defined even though |
1151 | C<$+{NAME}> would be. | |
286f584a | 1152 | |
c47ff5f1 | 1153 | =item C<< (?>pattern) >> |
6bda09f9 | 1154 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 | 1155 | |
19799a22 GS |
1156 | An "independent" subexpression, one which matches the substring |
1157 | that a I<standalone> C<pattern> would match if anchored at the given | |
9da458fc | 1158 | position, and it matches I<nothing other than this substring>. This |
19799a22 GS |
1159 | construct is useful for optimizations of what would otherwise be |
1160 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). | |
9da458fc IZ |
1161 | It may also be useful in places where the "grab all you can, and do not |
1162 | give anything back" semantic is desirable. | |
19799a22 | 1163 | |
c47ff5f1 | 1164 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 GS |
1165 | (anchored at the beginning of string, as above) will match I<all> |
1166 | characters C<a> at the beginning of string, leaving no C<a> for | |
1167 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, | |
1168 | since the match of the subgroup C<a*> is influenced by the following | |
1169 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside | |
1170 | C<a*ab> will match fewer characters than a standalone C<a*>, since | |
1171 | this makes the tail match. | |
1172 | ||
c47ff5f1 | 1173 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
d8b950dc KW |
1174 | C<(?=(pattern))\g1>. This matches the same substring as a standalone |
1175 | C<a+>, and the following C<\g1> eats the matched string; it therefore | |
c47ff5f1 | 1176 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 GS |
1177 | (The difference between these two constructs is that the second one |
1178 | uses a capturing group, thus shifting ordinals of backreferences | |
1179 | in the rest of a regular expression.) | |
1180 | ||
1181 | Consider this pattern: | |
c277df42 | 1182 | |
871b0233 | 1183 | m{ \( |
e2e6a0f1 | 1184 | ( |
f793d64a | 1185 | [^()]+ # x+ |
e2e6a0f1 | 1186 | | |
871b0233 IZ |
1187 | \( [^()]* \) |
1188 | )+ | |
e2e6a0f1 | 1189 | \) |
871b0233 | 1190 | }x |
5a964f20 | 1191 | |
19799a22 GS |
1192 | That will efficiently match a nonempty group with matching parentheses |
1193 | two levels deep or less. However, if there is no such group, it | |
1194 | will take virtually forever on a long string. That's because there | |
1195 | are so many different ways to split a long string into several | |
1196 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
1197 | to a subpattern of the above pattern. Consider how the pattern | |
1198 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
1199 | seconds, but that each extra letter doubles this time. This | |
1200 | exponential performance will make it appear that your program has | |
14218588 | 1201 | hung. However, a tiny change to this pattern |
5a964f20 | 1202 | |
e2e6a0f1 YO |
1203 | m{ \( |
1204 | ( | |
f793d64a | 1205 | (?> [^()]+ ) # change x+ above to (?> x+ ) |
e2e6a0f1 | 1206 | | |
871b0233 IZ |
1207 | \( [^()]* \) |
1208 | )+ | |
e2e6a0f1 | 1209 | \) |
871b0233 | 1210 | }x |
c277df42 | 1211 | |
c47ff5f1 | 1212 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 TC |
1213 | this yourself would be a productive exercise), but finishes in a fourth |
1214 | the time when used on a similar string with 1000000 C<a>s. Be aware, | |
1215 | however, that this pattern currently triggers a warning message under | |
9f1b1f2d | 1216 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b | 1217 | C<"matches null string many times in regex">. |
c277df42 | 1218 | |
c47ff5f1 | 1219 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 | 1220 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 IZ |
1221 | This was only 4 times slower on a string with 1000000 C<a>s. |
1222 | ||
9da458fc IZ |
1223 | The "grab all you can, and do not give anything back" semantic is desirable |
1224 | in many situations where on the first sight a simple C<()*> looks like | |
1225 | the correct solution. Suppose we parse text with comments being delimited | |
1226 | by C<#> followed by some optional (horizontal) whitespace. Contrary to | |
4375e838 | 1227 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc IZ |
1228 | the comment delimiter, because it may "give up" some whitespace if |
1229 | the remainder of the pattern can be made to match that way. The correct | |
1230 | answer is either one of these: | |
1231 | ||
1232 | (?>#[ \t]*) | |
1233 | #[ \t]*(?![ \t]) | |
1234 | ||
1235 | For example, to grab non-empty comments into $1, one should use either | |
1236 | one of these: | |
1237 | ||
1238 | / (?> \# [ \t]* ) ( .+ ) /x; | |
1239 | / \# [ \t]* ( [^ \t] .* ) /x; | |
1240 | ||
1241 | Which one you pick depends on which of these expressions better reflects | |
1242 | the above specification of comments. | |
1243 | ||
6bda09f9 YO |
1244 | In some literature this construct is called "atomic matching" or |
1245 | "possessive matching". | |
1246 | ||
b9b4dddf YO |
1247 | Possessive quantifiers are equivalent to putting the item they are applied |
1248 | to inside of one of these constructs. The following equivalences apply: | |
1249 | ||
1250 | Quantifier Form Bracketing Form | |
1251 | --------------- --------------- | |
1252 | PAT*+ (?>PAT*) | |
1253 | PAT++ (?>PAT+) | |
1254 | PAT?+ (?>PAT?) | |
1255 | PAT{min,max}+ (?>PAT{min,max}) | |
1256 | ||
e2e6a0f1 YO |
1257 | =back |
1258 | ||
1259 | =head2 Special Backtracking Control Verbs | |
1260 | ||
1261 | B<WARNING:> These patterns are experimental and subject to change or | |
0d017f4d | 1262 | removal in a future version of Perl. Their usage in production code should |
e2e6a0f1 YO |
1263 | be noted to avoid problems during upgrades. |
1264 | ||
1265 | These special patterns are generally of the form C<(*VERB:ARG)>. Unless | |
1266 | otherwise stated the ARG argument is optional; in some cases, it is | |
1267 | forbidden. | |
1268 | ||
1269 | Any pattern containing a special backtracking verb that allows an argument | |
e1020413 | 1270 | has the special behaviour that when executed it sets the current package's |
5d458dd8 YO |
1271 | C<$REGERROR> and C<$REGMARK> variables. When doing so the following |
1272 | rules apply: | |
e2e6a0f1 | 1273 | |
5d458dd8 YO |
1274 | On failure, the C<$REGERROR> variable will be set to the ARG value of the |
1275 | verb pattern, if the verb was involved in the failure of the match. If the | |
1276 | ARG part of the pattern was omitted, then C<$REGERROR> will be set to the | |
1277 | name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was | |
1278 | none. Also, the C<$REGMARK> variable will be set to FALSE. | |
e2e6a0f1 | 1279 | |
5d458dd8 YO |
1280 | On a successful match, the C<$REGERROR> variable will be set to FALSE, and |
1281 | the C<$REGMARK> variable will be set to the name of the last | |
1282 | C<(*MARK:NAME)> pattern executed. See the explanation for the | |
1283 | C<(*MARK:NAME)> verb below for more details. | |
e2e6a0f1 | 1284 | |
5d458dd8 YO |
1285 | B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1> |
1286 | and most other regex related variables. They are not local to a scope, nor | |
1287 | readonly, but instead are volatile package variables similar to C<$AUTOLOAD>. | |
1288 | Use C<local> to localize changes to them to a specific scope if necessary. | |
e2e6a0f1 YO |
1289 | |
1290 | If a pattern does not contain a special backtracking verb that allows an | |
5d458dd8 | 1291 | argument, then C<$REGERROR> and C<$REGMARK> are not touched at all. |
e2e6a0f1 YO |
1292 | |
1293 | =over 4 | |
1294 | ||
1295 | =item Verbs that take an argument | |
1296 | ||
1297 | =over 4 | |
1298 | ||
5d458dd8 | 1299 | =item C<(*PRUNE)> C<(*PRUNE:NAME)> |
f7819f85 | 1300 | X<(*PRUNE)> X<(*PRUNE:NAME)> |
54612592 | 1301 | |
5d458dd8 YO |
1302 | This zero-width pattern prunes the backtracking tree at the current point |
1303 | when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>, | |
1304 | where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached, | |
1305 | A may backtrack as necessary to match. Once it is reached, matching | |
1306 | continues in B, which may also backtrack as necessary; however, should B | |
1307 | not match, then no further backtracking will take place, and the pattern | |
1308 | will fail outright at the current starting position. | |
54612592 YO |
1309 | |
1310 | The following example counts all the possible matching strings in a | |
1311 | pattern (without actually matching any of them). | |
1312 | ||
e2e6a0f1 | 1313 | 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1314 | print "Count=$count\n"; |
1315 | ||
1316 | which produces: | |
1317 | ||
1318 | aaab | |
1319 | aaa | |
1320 | aa | |
1321 | a | |
1322 | aab | |
1323 | aa | |
1324 | a | |
1325 | ab | |
1326 | a | |
1327 | Count=9 | |
1328 | ||
5d458dd8 | 1329 | If we add a C<(*PRUNE)> before the count like the following |
54612592 | 1330 | |
5d458dd8 | 1331 | 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1332 | print "Count=$count\n"; |
1333 | ||
1334 | we prevent backtracking and find the count of the longest matching | |
353c6505 | 1335 | at each matching starting point like so: |
54612592 YO |
1336 | |
1337 | aaab | |
1338 | aab | |
1339 | ab | |
1340 | Count=3 | |
1341 | ||
5d458dd8 | 1342 | Any number of C<(*PRUNE)> assertions may be used in a pattern. |
54612592 | 1343 | |
5d458dd8 YO |
1344 | See also C<< (?>pattern) >> and possessive quantifiers for other ways to |
1345 | control backtracking. In some cases, the use of C<(*PRUNE)> can be | |
1346 | replaced with a C<< (?>pattern) >> with no functional difference; however, | |
1347 | C<(*PRUNE)> can be used to handle cases that cannot be expressed using a | |
1348 | C<< (?>pattern) >> alone. | |
54612592 | 1349 | |
e2e6a0f1 | 1350 | |
5d458dd8 YO |
1351 | =item C<(*SKIP)> C<(*SKIP:NAME)> |
1352 | X<(*SKIP)> | |
e2e6a0f1 | 1353 | |
5d458dd8 | 1354 | This zero-width pattern is similar to C<(*PRUNE)>, except that on |
e2e6a0f1 | 1355 | failure it also signifies that whatever text that was matched leading up |
5d458dd8 YO |
1356 | to the C<(*SKIP)> pattern being executed cannot be part of I<any> match |
1357 | of this pattern. This effectively means that the regex engine "skips" forward | |
1358 | to this position on failure and tries to match again, (assuming that | |
1359 | there is sufficient room to match). | |
1360 | ||
1361 | The name of the C<(*SKIP:NAME)> pattern has special significance. If a | |
1362 | C<(*MARK:NAME)> was encountered while matching, then it is that position | |
1363 | which is used as the "skip point". If no C<(*MARK)> of that name was | |
1364 | encountered, then the C<(*SKIP)> operator has no effect. When used | |
1365 | without a name the "skip point" is where the match point was when | |
1366 | executing the (*SKIP) pattern. | |
1367 | ||
1368 | Compare the following to the examples in C<(*PRUNE)>, note the string | |
24b23f37 YO |
1369 | is twice as long: |
1370 | ||
5d458dd8 | 1371 | 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 YO |
1372 | print "Count=$count\n"; |
1373 | ||
1374 | outputs | |
1375 | ||
1376 | aaab | |
1377 | aaab | |
1378 | Count=2 | |
1379 | ||
5d458dd8 | 1380 | Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)> |
353c6505 | 1381 | executed, the next starting point will be where the cursor was when the |
5d458dd8 YO |
1382 | C<(*SKIP)> was executed. |
1383 | ||
5d458dd8 YO |
1384 | =item C<(*MARK:NAME)> C<(*:NAME)> |
1385 | X<(*MARK)> C<(*MARK:NAME)> C<(*:NAME)> | |
1386 | ||
1387 | This zero-width pattern can be used to mark the point reached in a string | |
1388 | when a certain part of the pattern has been successfully matched. This | |
1389 | mark may be given a name. A later C<(*SKIP)> pattern will then skip | |
1390 | forward to that point if backtracked into on failure. Any number of | |
b4222fa9 | 1391 | C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated. |
5d458dd8 YO |
1392 | |
1393 | In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)> | |
1394 | can be used to "label" a pattern branch, so that after matching, the | |
1395 | program can determine which branches of the pattern were involved in the | |
1396 | match. | |
1397 | ||
1398 | When a match is successful, the C<$REGMARK> variable will be set to the | |
1399 | name of the most recently executed C<(*MARK:NAME)> that was involved | |
1400 | in the match. | |
1401 | ||
1402 | This can be used to determine which branch of a pattern was matched | |
c27a5cfe | 1403 | without using a separate capture group for each branch, which in turn |
5d458dd8 YO |
1404 | can result in a performance improvement, as perl cannot optimize |
1405 | C</(?:(x)|(y)|(z))/> as efficiently as something like | |
1406 | C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>. | |
1407 | ||
1408 | When a match has failed, and unless another verb has been involved in | |
1409 | failing the match and has provided its own name to use, the C<$REGERROR> | |
1410 | variable will be set to the name of the most recently executed | |
1411 | C<(*MARK:NAME)>. | |
1412 | ||
1413 | See C<(*SKIP)> for more details. | |
1414 | ||
b62d2d15 YO |
1415 | As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>. |
1416 | ||
5d458dd8 YO |
1417 | =item C<(*THEN)> C<(*THEN:NAME)> |
1418 | ||
241e7389 | 1419 | This is similar to the "cut group" operator C<::> from Perl 6. Like |
5d458dd8 YO |
1420 | C<(*PRUNE)>, this verb always matches, and when backtracked into on |
1421 | failure, it causes the regex engine to try the next alternation in the | |
1422 | innermost enclosing group (capturing or otherwise). | |
1423 | ||
1424 | Its name comes from the observation that this operation combined with the | |
1425 | alternation operator (C<|>) can be used to create what is essentially a | |
1426 | pattern-based if/then/else block: | |
1427 | ||
1428 | ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) | |
1429 | ||
1430 | Note that if this operator is used and NOT inside of an alternation then | |
1431 | it acts exactly like the C<(*PRUNE)> operator. | |
1432 | ||
1433 | / A (*PRUNE) B / | |
1434 | ||
1435 | is the same as | |
1436 | ||
1437 | / A (*THEN) B / | |
1438 | ||
1439 | but | |
1440 | ||
1441 | / ( A (*THEN) B | C (*THEN) D ) / | |
1442 | ||
1443 | is not the same as | |
1444 | ||
1445 | / ( A (*PRUNE) B | C (*PRUNE) D ) / | |
1446 | ||
1447 | as after matching the A but failing on the B the C<(*THEN)> verb will | |
1448 | backtrack and try C; but the C<(*PRUNE)> verb will simply fail. | |
24b23f37 | 1449 | |
e2e6a0f1 YO |
1450 | =item C<(*COMMIT)> |
1451 | X<(*COMMIT)> | |
24b23f37 | 1452 | |
241e7389 | 1453 | This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a |
5d458dd8 YO |
1454 | zero-width pattern similar to C<(*SKIP)>, except that when backtracked |
1455 | into on failure it causes the match to fail outright. No further attempts | |
1456 | to find a valid match by advancing the start pointer will occur again. | |
1457 | For example, | |
24b23f37 | 1458 | |
e2e6a0f1 | 1459 | 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/; |
24b23f37 YO |
1460 | print "Count=$count\n"; |
1461 | ||
1462 | outputs | |
1463 | ||
1464 | aaab | |
1465 | Count=1 | |
1466 | ||
e2e6a0f1 YO |
1467 | In other words, once the C<(*COMMIT)> has been entered, and if the pattern |
1468 | does not match, the regex engine will not try any further matching on the | |
1469 | rest of the string. | |
c277df42 | 1470 | |
e2e6a0f1 | 1471 | =back |
9af228c6 | 1472 | |
e2e6a0f1 | 1473 | =item Verbs without an argument |
9af228c6 YO |
1474 | |
1475 | =over 4 | |
1476 | ||
e2e6a0f1 YO |
1477 | =item C<(*FAIL)> C<(*F)> |
1478 | X<(*FAIL)> X<(*F)> | |
9af228c6 | 1479 | |
e2e6a0f1 YO |
1480 | This pattern matches nothing and always fails. It can be used to force the |
1481 | engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In | |
1482 | fact, C<(?!)> gets optimised into C<(*FAIL)> internally. | |
9af228c6 | 1483 | |
e2e6a0f1 | 1484 | It is probably useful only when combined with C<(?{})> or C<(??{})>. |
9af228c6 | 1485 | |
e2e6a0f1 YO |
1486 | =item C<(*ACCEPT)> |
1487 | X<(*ACCEPT)> | |
9af228c6 | 1488 | |
e2e6a0f1 YO |
1489 | B<WARNING:> This feature is highly experimental. It is not recommended |
1490 | for production code. | |
9af228c6 | 1491 | |
e2e6a0f1 YO |
1492 | This pattern matches nothing and causes the end of successful matching at |
1493 | the point at which the C<(*ACCEPT)> pattern was encountered, regardless of | |
1494 | whether there is actually more to match in the string. When inside of a | |
0d017f4d | 1495 | nested pattern, such as recursion, or in a subpattern dynamically generated |
e2e6a0f1 | 1496 | via C<(??{})>, only the innermost pattern is ended immediately. |
9af228c6 | 1497 | |
c27a5cfe | 1498 | If the C<(*ACCEPT)> is inside of capturing groups then the groups are |
e2e6a0f1 YO |
1499 | marked as ended at the point at which the C<(*ACCEPT)> was encountered. |
1500 | For instance: | |
9af228c6 | 1501 | |
e2e6a0f1 | 1502 | 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x; |
9af228c6 | 1503 | |
e2e6a0f1 | 1504 | will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not |
0d017f4d | 1505 | be set. If another branch in the inner parentheses were matched, such as in the |
e2e6a0f1 | 1506 | string 'ACDE', then the C<D> and C<E> would have to be matched as well. |
9af228c6 YO |
1507 | |
1508 | =back | |
c277df42 | 1509 | |
a0d0e21e LW |
1510 | =back |
1511 | ||
c07a80fd | 1512 | =head2 Backtracking |
d74e8afc | 1513 | X<backtrack> X<backtracking> |
c07a80fd | 1514 | |
35a734be IZ |
1515 | NOTE: This section presents an abstract approximation of regular |
1516 | expression behavior. For a more rigorous (and complicated) view of | |
1517 | the rules involved in selecting a match among possible alternatives, | |
0d017f4d | 1518 | see L<Combining RE Pieces>. |
35a734be | 1519 | |
c277df42 | 1520 | A fundamental feature of regular expression matching involves the |
5a964f20 | 1521 | notion called I<backtracking>, which is currently used (when needed) |
0d017f4d | 1522 | by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc IZ |
1523 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
1524 | internally, but the general principle outlined here is valid. | |
c07a80fd | 1525 | |
1526 | For a regular expression to match, the I<entire> regular expression must | |
1527 | match, not just part of it. So if the beginning of a pattern containing a | |
1528 | quantifier succeeds in a way that causes later parts in the pattern to | |
1529 | fail, the matching engine backs up and recalculates the beginning | |
1530 | part--that's why it's called backtracking. | |
1531 | ||
1532 | Here is an example of backtracking: Let's say you want to find the | |
1533 | word following "foo" in the string "Food is on the foo table.": | |
1534 | ||
1535 | $_ = "Food is on the foo table."; | |
1536 | if ( /\b(foo)\s+(\w+)/i ) { | |
f793d64a | 1537 | print "$2 follows $1.\n"; |
c07a80fd | 1538 | } |
1539 | ||
1540 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
1541 | finds a possible match right at the beginning of the string, and loads up | |
1542 | $1 with "Foo". However, as soon as the matching engine sees that there's | |
1543 | no whitespace following the "Foo" that it had saved in $1, it realizes its | |
68dc0745 | 1544 | mistake and starts over again one character after where it had the |
c07a80fd | 1545 | tentative match. This time it goes all the way until the next occurrence |
1546 | of "foo". The complete regular expression matches this time, and you get | |
1547 | the expected output of "table follows foo." | |
1548 | ||
1549 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
1550 | everything between "foo" and "bar". Initially, you write something | |
1551 | like this: | |
1552 | ||
1553 | $_ = "The food is under the bar in the barn."; | |
1554 | if ( /foo(.*)bar/ ) { | |
f793d64a | 1555 | print "got <$1>\n"; |
c07a80fd | 1556 | } |
1557 | ||
1558 | Which perhaps unexpectedly yields: | |
1559 | ||
1560 | got <d is under the bar in the > | |
1561 | ||
1562 | That's because C<.*> was greedy, so you get everything between the | |
14218588 | 1563 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd | 1564 | to use minimal matching to make sure you get the text between a "foo" |
1565 | and the first "bar" thereafter. | |
1566 | ||
1567 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
1568 | got <d is under the > | |
1569 | ||
0d017f4d | 1570 | Here's another example. Let's say you'd like to match a number at the end |
b6e13d97 | 1571 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd | 1572 | So you write this: |
1573 | ||
1574 | $_ = "I have 2 numbers: 53147"; | |
f793d64a KW |
1575 | if ( /(.*)(\d*)/ ) { # Wrong! |
1576 | print "Beginning is <$1>, number is <$2>.\n"; | |
c07a80fd | 1577 | } |
1578 | ||
1579 | That won't work at all, because C<.*> was greedy and gobbled up the | |
1580 | whole string. As C<\d*> can match on an empty string the complete | |
1581 | regular expression matched successfully. | |
1582 | ||
8e1088bc | 1583 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd | 1584 | |
1585 | Here are some variants, most of which don't work: | |
1586 | ||
1587 | $_ = "I have 2 numbers: 53147"; | |
1588 | @pats = qw{ | |
f793d64a KW |
1589 | (.*)(\d*) |
1590 | (.*)(\d+) | |
1591 | (.*?)(\d*) | |
1592 | (.*?)(\d+) | |
1593 | (.*)(\d+)$ | |
1594 | (.*?)(\d+)$ | |
1595 | (.*)\b(\d+)$ | |
1596 | (.*\D)(\d+)$ | |
c07a80fd | 1597 | }; |
1598 | ||
1599 | for $pat (@pats) { | |
f793d64a KW |
1600 | printf "%-12s ", $pat; |
1601 | if ( /$pat/ ) { | |
1602 | print "<$1> <$2>\n"; | |
1603 | } else { | |
1604 | print "FAIL\n"; | |
1605 | } | |
c07a80fd | 1606 | } |
1607 | ||
1608 | That will print out: | |
1609 | ||
1610 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
1611 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
1612 | (.*?)(\d*) <> <> | |
1613 | (.*?)(\d+) <I have > <2> | |
1614 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
1615 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
1616 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
1617 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
1618 | ||
1619 | As you see, this can be a bit tricky. It's important to realize that a | |
1620 | regular expression is merely a set of assertions that gives a definition | |
1621 | of success. There may be 0, 1, or several different ways that the | |
1622 | definition might succeed against a particular string. And if there are | |
5a964f20 TC |
1623 | multiple ways it might succeed, you need to understand backtracking to |
1624 | know which variety of success you will achieve. | |
c07a80fd | 1625 | |
19799a22 | 1626 | When using look-ahead assertions and negations, this can all get even |
8b19b778 | 1627 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd | 1628 | followed by "123". You might try to write that as |
1629 | ||
871b0233 | 1630 | $_ = "ABC123"; |
f793d64a KW |
1631 | if ( /^\D*(?!123)/ ) { # Wrong! |
1632 | print "Yup, no 123 in $_\n"; | |
871b0233 | 1633 | } |
c07a80fd | 1634 | |
1635 | But that isn't going to match; at least, not the way you're hoping. It | |
1636 | claims that there is no 123 in the string. Here's a clearer picture of | |
9b9391b2 | 1637 | why that pattern matches, contrary to popular expectations: |
c07a80fd | 1638 | |
4358a253 SS |
1639 | $x = 'ABC123'; |
1640 | $y = 'ABC445'; | |
c07a80fd | 1641 | |
4358a253 SS |
1642 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
1643 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; | |
c07a80fd | 1644 | |
4358a253 SS |
1645 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; |
1646 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; | |
c07a80fd | 1647 | |
1648 | This prints | |
1649 | ||
1650 | 2: got ABC | |
1651 | 3: got AB | |
1652 | 4: got ABC | |
1653 | ||
5f05dabc | 1654 | You might have expected test 3 to fail because it seems to a more |
c07a80fd | 1655 | general purpose version of test 1. The important difference between |
1656 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
1657 | backtracking, whereas test 1 will not. What's happening is | |
1658 | that you've asked "Is it true that at the start of $x, following 0 or more | |
5f05dabc | 1659 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd | 1660 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 | 1661 | fail. |
14218588 | 1662 | |
c07a80fd | 1663 | The search engine will initially match C<\D*> with "ABC". Then it will |
14218588 | 1664 | try to match C<(?!123> with "123", which fails. But because |
c07a80fd | 1665 | a quantifier (C<\D*>) has been used in the regular expression, the |
1666 | search engine can backtrack and retry the match differently | |
54310121 | 1667 | in the hope of matching the complete regular expression. |
c07a80fd | 1668 | |
5a964f20 TC |
1669 | The pattern really, I<really> wants to succeed, so it uses the |
1670 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
c07a80fd | 1671 | time. Now there's indeed something following "AB" that is not |
14218588 | 1672 | "123". It's "C123", which suffices. |
c07a80fd | 1673 | |
14218588 GS |
1674 | We can deal with this by using both an assertion and a negation. |
1675 | We'll say that the first part in $1 must be followed both by a digit | |
1676 | and by something that's not "123". Remember that the look-aheads | |
1677 | are zero-width expressions--they only look, but don't consume any | |
1678 | of the string in their match. So rewriting this way produces what | |
c07a80fd | 1679 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
1680 | ||
4358a253 SS |
1681 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; |
1682 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; | |
c07a80fd | 1683 | |
1684 | 6: got ABC | |
1685 | ||
5a964f20 | 1686 | In other words, the two zero-width assertions next to each other work as though |
19799a22 | 1687 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd | 1688 | matches only if you're at the beginning of the line AND the end of the |
1689 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
1690 | regular expressions always means AND, except when you write an explicit OR | |
1691 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
1692 | although the attempted matches are made at different positions because "a" | |
1693 | is not a zero-width assertion, but a one-width assertion. | |
1694 | ||
0d017f4d | 1695 | B<WARNING>: Particularly complicated regular expressions can take |
14218588 | 1696 | exponential time to solve because of the immense number of possible |
0d017f4d | 1697 | ways they can use backtracking to try for a match. For example, without |
9da458fc IZ |
1698 | internal optimizations done by the regular expression engine, this will |
1699 | take a painfully long time to run: | |
c07a80fd | 1700 | |
e1901655 IZ |
1701 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
1702 | ||
1703 | And if you used C<*>'s in the internal groups instead of limiting them | |
1704 | to 0 through 5 matches, then it would take forever--or until you ran | |
1705 | out of stack space. Moreover, these internal optimizations are not | |
1706 | always applicable. For example, if you put C<{0,5}> instead of C<*> | |
1707 | on the external group, no current optimization is applicable, and the | |
1708 | match takes a long time to finish. | |
c07a80fd | 1709 | |
9da458fc IZ |
1710 | A powerful tool for optimizing such beasts is what is known as an |
1711 | "independent group", | |
c47ff5f1 | 1712 | which does not backtrack (see L<C<< (?>pattern) >>>). Note also that |
9da458fc | 1713 | zero-length look-ahead/look-behind assertions will not backtrack to make |
5d458dd8 | 1714 | the tail match, since they are in "logical" context: only |
14218588 | 1715 | whether they match is considered relevant. For an example |
9da458fc | 1716 | where side-effects of look-ahead I<might> have influenced the |
c47ff5f1 | 1717 | following match, see L<C<< (?>pattern) >>>. |
c277df42 | 1718 | |
a0d0e21e | 1719 | =head2 Version 8 Regular Expressions |
d74e8afc | 1720 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> |
a0d0e21e | 1721 | |
5a964f20 | 1722 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e LW |
1723 | routines, here are the pattern-matching rules not described above. |
1724 | ||
54310121 | 1725 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e | 1726 | with a special meaning described here or above. You can cause |
5a964f20 | 1727 | characters that normally function as metacharacters to be interpreted |
5f05dabc | 1728 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
0d017f4d WL |
1729 | character; "\\" matches a "\"). This escape mechanism is also required |
1730 | for the character used as the pattern delimiter. | |
1731 | ||
1732 | A series of characters matches that series of characters in the target | |
1733 | string, so the pattern C<blurfl> would match "blurfl" in the target | |
1734 | string. | |
a0d0e21e LW |
1735 | |
1736 | You can specify a character class, by enclosing a list of characters | |
5d458dd8 | 1737 | in C<[]>, which will match any character from the list. If the |
a0d0e21e | 1738 | first character after the "[" is "^", the class matches any character not |
14218588 | 1739 | in the list. Within a list, the "-" character specifies a |
5a964f20 | 1740 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 GS |
1741 | inclusive. If you want either "-" or "]" itself to be a member of a |
1742 | class, put it at the start of the list (possibly after a "^"), or | |
1743 | escape it with a backslash. "-" is also taken literally when it is | |
1744 | at the end of the list, just before the closing "]". (The | |
84850974 DD |
1745 | following all specify the same class of three characters: C<[-az]>, |
1746 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which | |
5d458dd8 YO |
1747 | specifies a class containing twenty-six characters, even on EBCDIC-based |
1748 | character sets.) Also, if you try to use the character | |
1749 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of | |
1750 | a range, the "-" is understood literally. | |
a0d0e21e | 1751 | |
8ada0baa JH |
1752 | Note also that the whole range idea is rather unportable between |
1753 | character sets--and even within character sets they may cause results | |
1754 | you probably didn't expect. A sound principle is to use only ranges | |
0d017f4d | 1755 | that begin from and end at either alphabetics of equal case ([a-e], |
8ada0baa JH |
1756 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, |
1757 | spell out the character sets in full. | |
1758 | ||
54310121 | 1759 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e LW |
1760 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
1761 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string | |
dc0d9c48 | 1762 | of three octal digits, matches the character whose coded character set value |
5d458dd8 | 1763 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, |
dc0d9c48 | 1764 | matches the character whose ordinal is I<nn>. The expression \cI<x> |
5d458dd8 | 1765 | matches the character control-I<x>. Finally, the "." metacharacter |
fb55449c | 1766 | matches any character except "\n" (unless you use C</s>). |
a0d0e21e LW |
1767 | |
1768 | You can specify a series of alternatives for a pattern using "|" to | |
1769 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", | |
5a964f20 | 1770 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e LW |
1771 | first alternative includes everything from the last pattern delimiter |
1772 | ("(", "[", or the beginning of the pattern) up to the first "|", and | |
1773 | the last alternative contains everything from the last "|" to the next | |
14218588 GS |
1774 | pattern delimiter. That's why it's common practice to include |
1775 | alternatives in parentheses: to minimize confusion about where they | |
a3cb178b GS |
1776 | start and end. |
1777 | ||
5a964f20 | 1778 | Alternatives are tried from left to right, so the first |
a3cb178b GS |
1779 | alternative found for which the entire expression matches, is the one that |
1780 | is chosen. This means that alternatives are not necessarily greedy. For | |
628afcb5 | 1781 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b GS |
1782 | part will match, as that is the first alternative tried, and it successfully |
1783 | matches the target string. (This might not seem important, but it is | |
1784 | important when you are capturing matched text using parentheses.) | |
1785 | ||
5a964f20 | 1786 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b | 1787 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e | 1788 | |
14218588 GS |
1789 | Within a pattern, you may designate subpatterns for later reference |
1790 | by enclosing them in parentheses, and you may refer back to the | |
1791 | I<n>th subpattern later in the pattern using the metacharacter | |
1792 | \I<n>. Subpatterns are numbered based on the left to right order | |
1793 | of their opening parenthesis. A backreference matches whatever | |
1794 | actually matched the subpattern in the string being examined, not | |
d8b950dc | 1795 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\g1\d*> will |
14218588 GS |
1796 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern |
1797 | 1 matched "0x", even though the rule C<0|0x> could potentially match | |
1798 | the leading 0 in the second number. | |
cb1a09d0 | 1799 | |
0d017f4d | 1800 | =head2 Warning on \1 Instead of $1 |
cb1a09d0 | 1801 | |
5a964f20 | 1802 | Some people get too used to writing things like: |
cb1a09d0 AD |
1803 | |
1804 | $pattern =~ s/(\W)/\\\1/g; | |
1805 | ||
3ff1c45a KW |
1806 | This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid |
1807 | shocking the | |
cb1a09d0 | 1808 | B<sed> addicts, but it's a dirty habit to get into. That's because in |
d1be9408 | 1809 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 AD |
1810 | the usual double-quoted string means a control-A. The customary Unix |
1811 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
1812 | of doing that, you get yourself into trouble if you then add an C</e> | |
1813 | modifier. | |
1814 | ||
f793d64a | 1815 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 AD |
1816 | |
1817 | Or if you try to do | |
1818 | ||
1819 | s/(\d+)/\1000/; | |
1820 | ||
1821 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
14218588 | 1822 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 AD |
1823 | with the operation of matching a backreference. Certainly they mean two |
1824 | different things on the I<left> side of the C<s///>. | |
9fa51da4 | 1825 | |
0d017f4d | 1826 | =head2 Repeated Patterns Matching a Zero-length Substring |
c84d73f1 | 1827 | |
19799a22 | 1828 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 IZ |
1829 | |
1830 | Regular expressions provide a terse and powerful programming language. As | |
1831 | with most other power tools, power comes together with the ability | |
1832 | to wreak havoc. | |
1833 | ||
1834 | A common abuse of this power stems from the ability to make infinite | |
628afcb5 | 1835 | loops using regular expressions, with something as innocuous as: |
c84d73f1 IZ |
1836 | |
1837 | 'foo' =~ m{ ( o? )* }x; | |
1838 | ||
0d017f4d | 1839 | The C<o?> matches at the beginning of C<'foo'>, and since the position |
c84d73f1 | 1840 | in the string is not moved by the match, C<o?> would match again and again |
527e91da | 1841 | because of the C<*> quantifier. Another common way to create a similar cycle |
c84d73f1 IZ |
1842 | is with the looping modifier C<//g>: |
1843 | ||
1844 | @matches = ( 'foo' =~ m{ o? }xg ); | |
1845 | ||
1846 | or | |
1847 | ||
1848 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
1849 | ||
1850 | or the loop implied by split(). | |
1851 | ||
1852 | However, long experience has shown that many programming tasks may | |
14218588 GS |
1853 | be significantly simplified by using repeated subexpressions that |
1854 | may match zero-length substrings. Here's a simple example being: | |
c84d73f1 | 1855 | |
f793d64a | 1856 | @chars = split //, $string; # // is not magic in split |
c84d73f1 IZ |
1857 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / |
1858 | ||
9da458fc | 1859 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 | 1860 | the infinite loop>. The rules for this are different for lower-level |
527e91da | 1861 | loops given by the greedy quantifiers C<*+{}>, and for higher-level |
c84d73f1 IZ |
1862 | ones like the C</g> modifier or split() operator. |
1863 | ||
19799a22 GS |
1864 | The lower-level loops are I<interrupted> (that is, the loop is |
1865 | broken) when Perl detects that a repeated expression matched a | |
1866 | zero-length substring. Thus | |
c84d73f1 IZ |
1867 | |
1868 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
1869 | ||
5d458dd8 | 1870 | is made equivalent to |
c84d73f1 | 1871 | |
5d458dd8 YO |
1872 | m{ (?: NON_ZERO_LENGTH )* |
1873 | | | |
1874 | (?: ZERO_LENGTH )? | |
c84d73f1 IZ |
1875 | }x; |
1876 | ||
1877 | The higher level-loops preserve an additional state between iterations: | |
5d458dd8 | 1878 | whether the last match was zero-length. To break the loop, the following |
c84d73f1 | 1879 | match after a zero-length match is prohibited to have a length of zero. |
5d458dd8 | 1880 | This prohibition interacts with backtracking (see L<"Backtracking">), |
c84d73f1 IZ |
1881 | and so the I<second best> match is chosen if the I<best> match is of |
1882 | zero length. | |
1883 | ||
19799a22 | 1884 | For example: |
c84d73f1 IZ |
1885 | |
1886 | $_ = 'bar'; | |
1887 | s/\w??/<$&>/g; | |
1888 | ||
20fb949f | 1889 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
5d458dd8 | 1890 | match given by non-greedy C<??> is the zero-length match, and the I<second |
c84d73f1 IZ |
1891 | best> match is what is matched by C<\w>. Thus zero-length matches |
1892 | alternate with one-character-long matches. | |
1893 | ||
5d458dd8 | 1894 | Similarly, for repeated C<m/()/g> the second-best match is the match at the |
c84d73f1 IZ |
1895 | position one notch further in the string. |
1896 | ||
19799a22 | 1897 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 | 1898 | the matched string, and is reset by each assignment to pos(). |
9da458fc IZ |
1899 | Zero-length matches at the end of the previous match are ignored |
1900 | during C<split>. | |
c84d73f1 | 1901 | |
0d017f4d | 1902 | =head2 Combining RE Pieces |
35a734be IZ |
1903 | |
1904 | Each of the elementary pieces of regular expressions which were described | |
1905 | before (such as C<ab> or C<\Z>) could match at most one substring | |
1906 | at the given position of the input string. However, in a typical regular | |
1907 | expression these elementary pieces are combined into more complicated | |
1908 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc | |
1909 | (in these examples C<S> and C<T> are regular subexpressions). | |
1910 | ||
1911 | Such combinations can include alternatives, leading to a problem of choice: | |
1912 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
1913 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
1914 | actually matched is the concept of backtracking (see L<"Backtracking">). | |
1915 | However, this description is too low-level and makes you think | |
1916 | in terms of a particular implementation. | |
1917 | ||
1918 | Another description starts with notions of "better"/"worse". All the | |
1919 | substrings which may be matched by the given regular expression can be | |
1920 | sorted from the "best" match to the "worst" match, and it is the "best" | |
1921 | match which is chosen. This substitutes the question of "what is chosen?" | |
1922 | by the question of "which matches are better, and which are worse?". | |
1923 | ||
1924 | Again, for elementary pieces there is no such question, since at most | |
1925 | one match at a given position is possible. This section describes the | |
1926 | notion of better/worse for combining operators. In the description | |
1927 | below C<S> and C<T> are regular subexpressions. | |
1928 | ||
13a2d996 | 1929 | =over 4 |
35a734be IZ |
1930 | |
1931 | =item C<ST> | |
1932 | ||
1933 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are | |
1934 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings | |
5d458dd8 | 1935 | which can be matched by C<T>. |
35a734be IZ |
1936 | |
1937 | If C<A> is better match for C<S> than C<A'>, C<AB> is a better | |
1938 | match than C<A'B'>. | |
1939 | ||
1940 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if | |
1941 | C<B> is better match for C<T> than C<B'>. | |
1942 | ||
1943 | =item C<S|T> | |
1944 | ||
1945 | When C<S> can match, it is a better match than when only C<T> can match. | |
1946 | ||
1947 | Ordering of two matches for C<S> is the same as for C<S>. Similar for | |
1948 | two matches for C<T>. | |
1949 | ||
1950 | =item C<S{REPEAT_COUNT}> | |
1951 | ||
1952 | Matches as C<SSS...S> (repeated as many times as necessary). | |
1953 | ||
1954 | =item C<S{min,max}> | |
1955 | ||
1956 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
1957 | ||
1958 | =item C<S{min,max}?> | |
1959 | ||
1960 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
1961 | ||
1962 | =item C<S?>, C<S*>, C<S+> | |
1963 | ||
1964 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
1965 | ||
1966 | =item C<S??>, C<S*?>, C<S+?> | |
1967 | ||
1968 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
1969 | ||
c47ff5f1 | 1970 | =item C<< (?>S) >> |
35a734be IZ |
1971 | |
1972 | Matches the best match for C<S> and only that. | |
1973 | ||
1974 | =item C<(?=S)>, C<(?<=S)> | |
1975 | ||
1976 | Only the best match for C<S> is considered. (This is important only if | |
1977 | C<S> has capturing parentheses, and backreferences are used somewhere | |
1978 | else in the whole regular expression.) | |
1979 | ||
1980 | =item C<(?!S)>, C<(?<!S)> | |
1981 | ||
1982 | For this grouping operator there is no need to describe the ordering, since | |
1983 | only whether or not C<S> can match is important. | |
1984 | ||
6bda09f9 | 1985 | =item C<(??{ EXPR })>, C<(?PARNO)> |
35a734be IZ |
1986 | |
1987 | The ordering is the same as for the regular expression which is | |
c27a5cfe | 1988 | the result of EXPR, or the pattern contained by capture group PARNO. |
35a734be IZ |
1989 | |
1990 | =item C<(?(condition)yes-pattern|no-pattern)> | |
1991 | ||
1992 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is | |
1993 | already determined. The ordering of the matches is the same as for the | |
1994 | chosen subexpression. | |
1995 | ||
1996 | =back | |
1997 | ||
1998 | The above recipes describe the ordering of matches I<at a given position>. | |
1999 | One more rule is needed to understand how a match is determined for the | |
2000 | whole regular expression: a match at an earlier position is always better | |
2001 | than a match at a later position. | |
2002 | ||
0d017f4d | 2003 | =head2 Creating Custom RE Engines |
c84d73f1 IZ |
2004 | |
2005 | Overloaded constants (see L<overload>) provide a simple way to extend | |
2006 | the functionality of the RE engine. | |
2007 | ||
2008 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
0d017f4d | 2009 | matches at a boundary between whitespace characters and non-whitespace |
c84d73f1 IZ |
2010 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
2011 | at these positions, so we want to have each C<\Y|> in the place of the | |
2012 | more complicated version. We can create a module C<customre> to do | |
2013 | this: | |
2014 | ||
2015 | package customre; | |
2016 | use overload; | |
2017 | ||
2018 | sub import { | |
2019 | shift; | |
2020 | die "No argument to customre::import allowed" if @_; | |
2021 | overload::constant 'qr' => \&convert; | |
2022 | } | |
2023 | ||
2024 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
2025 | ||
580a9fe1 RGS |
2026 | # We must also take care of not escaping the legitimate \\Y| |
2027 | # sequence, hence the presence of '\\' in the conversion rules. | |
5d458dd8 | 2028 | my %rules = ( '\\' => '\\\\', |
f793d64a | 2029 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
c84d73f1 IZ |
2030 | sub convert { |
2031 | my $re = shift; | |
5d458dd8 | 2032 | $re =~ s{ |
c84d73f1 IZ |
2033 | \\ ( \\ | Y . ) |
2034 | } | |
5d458dd8 | 2035 | { $rules{$1} or invalid($re,$1) }sgex; |
c84d73f1 IZ |
2036 | return $re; |
2037 | } | |
2038 | ||
2039 | Now C<use customre> enables the new escape in constant regular | |
2040 | expressions, i.e., those without any runtime variable interpolations. | |
2041 | As documented in L<overload>, this conversion will work only over | |
2042 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
2043 | part of this regular expression needs to be converted explicitly | |
2044 | (but only if the special meaning of C<\Y|> should be enabled inside $re): | |
2045 | ||
2046 | use customre; | |
2047 | $re = <>; | |
2048 | chomp $re; | |
2049 | $re = customre::convert $re; | |
2050 | /\Y|$re\Y|/; | |
2051 | ||
1f1031fe YO |
2052 | =head1 PCRE/Python Support |
2053 | ||
99d59c4d | 2054 | As of Perl 5.10.0, Perl supports several Python/PCRE specific extensions |
1f1031fe | 2055 | to the regex syntax. While Perl programmers are encouraged to use the |
99d59c4d | 2056 | Perl specific syntax, the following are also accepted: |
1f1031fe YO |
2057 | |
2058 | =over 4 | |
2059 | ||
ae5648b3 | 2060 | =item C<< (?PE<lt>NAMEE<gt>pattern) >> |
1f1031fe | 2061 | |
c27a5cfe | 2062 | Define a named capture group. Equivalent to C<< (?<NAME>pattern) >>. |
1f1031fe YO |
2063 | |
2064 | =item C<< (?P=NAME) >> | |
2065 | ||
c27a5cfe | 2066 | Backreference to a named capture group. Equivalent to C<< \g{NAME} >>. |
1f1031fe YO |
2067 | |
2068 | =item C<< (?P>NAME) >> | |
2069 | ||
c27a5cfe | 2070 | Subroutine call to a named capture group. Equivalent to C<< (?&NAME) >>. |
1f1031fe | 2071 | |
ee9b8eae | 2072 | =back |
1f1031fe | 2073 | |
19799a22 GS |
2074 | =head1 BUGS |
2075 | ||
78288b8e KW |
2076 | There are numerous problems with case insensitive matching of characters |
2077 | outside the ASCII range, especially with those whose folds are multiple | |
2078 | characters, such as ligatures like C<LATIN SMALL LIGATURE FF>. | |
2079 | ||
f253210b KW |
2080 | In a bracketed character class with case insensitive matching, ranges only work |
2081 | for ASCII characters. For example, | |
2082 | C<m/[\N{CYRILLIC CAPITAL LETTER A}-\N{CYRILLIC CAPITAL LETTER YA}]/i> | |
2083 | doesn't match all the Russian upper and lower case letters. | |
2084 | ||
88c9975e KW |
2085 | Many regular expression constructs don't work on EBCDIC platforms. |
2086 | ||
9da458fc IZ |
2087 | This document varies from difficult to understand to completely |
2088 | and utterly opaque. The wandering prose riddled with jargon is | |
2089 | hard to fathom in several places. | |
2090 | ||
2091 | This document needs a rewrite that separates the tutorial content | |
2092 | from the reference content. | |
19799a22 GS |
2093 | |
2094 | =head1 SEE ALSO | |
9fa51da4 | 2095 | |
91e0c79e MJD |
2096 | L<perlrequick>. |
2097 | ||
2098 | L<perlretut>. | |
2099 | ||
9b599b2a GS |
2100 | L<perlop/"Regexp Quote-Like Operators">. |
2101 | ||
1e66bd83 PP |
2102 | L<perlop/"Gory details of parsing quoted constructs">. |
2103 | ||
14218588 GS |
2104 | L<perlfaq6>. |
2105 | ||
9b599b2a GS |
2106 | L<perlfunc/pos>. |
2107 | ||
2108 | L<perllocale>. | |
2109 | ||
fb55449c JH |
2110 | L<perlebcdic>. |
2111 | ||
14218588 GS |
2112 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
2113 | by O'Reilly and Associates. |